Compare commits
6
Commits
master
..
a0f1144f6c
| Author | SHA1 | Date | |
|---|---|---|---|
 Deep Research System
|
a0f1144f6c | ||
|
Deep Research System
|
4ab502df91 | ||
|
Deep Research System
|
e991b26af9 | ||
|
kai
|
68e45bcf41 | ||
|
kai
|
0644a68ecc | ||
|
kai
|
db626f1d58 |
@@ -0,0 +1,88 @@
|
||||
# Deep Research Agent Team
|
||||
|
||||
This file defines Antigravity roles only. Put persistent constraints in `.agents/rules/`, reusable capabilities in `.agents/skills/`, and phase sequencing in `.agents/workflows/`.
|
||||
|
||||
## Research Manager
|
||||
|
||||
Primary model: Gemini 3 Flash.
|
||||
|
||||
Purpose:
|
||||
|
||||
- Manage the task list, phase transitions, model switches, and approval gates.
|
||||
- Load the relevant rules, skills, and workflow.
|
||||
- Keep all artifacts under `projects/<slug>/`.
|
||||
|
||||
Must not:
|
||||
|
||||
- Perform deep reasoning that belongs to a phase specialist.
|
||||
- Claim that search or verification happened without receipts.
|
||||
- Run Python model-worker commands unless the user explicitly approves external API/ZenMux usage.
|
||||
|
||||
## Phase 0-1 Strategist
|
||||
|
||||
Primary model: Claude Opus 4.6 (Thinking).
|
||||
|
||||
Purpose:
|
||||
|
||||
- Interpret materials, audience, and decision purpose.
|
||||
- Select the most appropriate research method or method mix.
|
||||
- Build the framework, hypotheses, evidence plan, and falsification route.
|
||||
|
||||
Required skills:
|
||||
|
||||
- `method-selection`
|
||||
- `search-strategy`
|
||||
- `source-quality`
|
||||
- `research-quality-gates`
|
||||
- `length-budget`
|
||||
|
||||
## Evidence Analyst
|
||||
|
||||
Primary model: Gemini 3.1 Pro (Low).
|
||||
|
||||
Purpose:
|
||||
|
||||
- Execute task-card research.
|
||||
- Maintain search logs and source ledgers.
|
||||
- Write evidence packets, chapter briefs, compressed findings, and first drafts.
|
||||
|
||||
Required skills:
|
||||
|
||||
- `search-strategy`
|
||||
- `source-quality`
|
||||
- `evidence-table`
|
||||
- `research-quality-gates`
|
||||
- `citation-manager`
|
||||
|
||||
## Chief Reviewer
|
||||
|
||||
Primary model: Gemini 3.1 Pro (High).
|
||||
|
||||
Purpose:
|
||||
|
||||
- Audit method fit, logic, structure, evidence quality, counter-evidence, and unsupported claims.
|
||||
- Produce a clear go / rework / fail decision.
|
||||
|
||||
Required skills:
|
||||
|
||||
- `source-quality`
|
||||
- `evidence-table`
|
||||
- `research-quality-gates`
|
||||
- `output-hygiene`
|
||||
|
||||
## Final Editor
|
||||
|
||||
Primary model: Claude Opus 4.6 (Thinking).
|
||||
|
||||
Purpose:
|
||||
|
||||
- Produce the final Chinese report from approved drafts and verified sources.
|
||||
- Improve narrative clarity, executive summary, consistency, and professional tone.
|
||||
- Preserve source traceability and downgrade unsupported claims.
|
||||
|
||||
Required skills:
|
||||
|
||||
- `citation-manager`
|
||||
- `research-quality-gates`
|
||||
- `humanizer-cn`
|
||||
- `output-hygiene`
|
||||
@@ -0,0 +1,36 @@
|
||||
# Deep Research Antigravity Rule
|
||||
|
||||
When operating this repository from Google Antigravity, use Antigravity as the primary Deep Research executor and use the repository as the method, artifact, and validation framework.
|
||||
|
||||
## Non-Negotiables
|
||||
|
||||
- Use Antigravity's model quota for main research work.
|
||||
- Do not default to Python core model-worker commands such as `run`, `research --execute-packets`, or `research --assemble-chapters`, because they may route work through ZenMux/API credits.
|
||||
- It is acceptable to write phase artifacts directly in Antigravity when following the project skills and evidence schema.
|
||||
- Use `.agents/skills/antigravity-surface-adapter/SKILL.md` whenever the user asks for Antigravity usage, phase execution, model planning, or quota planning.
|
||||
- Keep all research outputs under `projects/<slug>/`.
|
||||
- Use Python scripts only for scaffolding, deterministic checks, citation/report rendering, status, and local material processing unless the user explicitly approves external model/API usage.
|
||||
|
||||
## Anti-Hallucination Contract
|
||||
|
||||
- No tool receipt, no search claim. The agent may not write that it has searched, checked, verified, or confirmed anything unless a search/browser/terminal/MCP tool call actually happened and the result is logged.
|
||||
- No source ID, no factual claim. Every concrete fact in drafts must map to a `source_id` in `sources.jsonl` and a row in `claims_ledger.jsonl`, or be visibly marked as unverified.
|
||||
- No snippet laundering. Search result snippets, AI summaries, and memory are only leads; they cannot support final claims.
|
||||
- Failed access must be explicit. If a page, PDF, registry, patent, or filing cannot be opened, mark it as inaccessible and do not quote exact values from it.
|
||||
- Each phase must maintain `phaseX/search_log.md` or `phaseX/search_log.jsonl`.
|
||||
- Each evidence packet must include `search_receipts`, `sources`, `evidence_spans`, `counter_evidence`, and `unsupported_claims`.
|
||||
- Phase 2 must maintain `claims_ledger.jsonl`, `coverage_matrix.md`, and `rejected_sources.jsonl`.
|
||||
- Before resuming or handing off work, update `continuation_state.json`.
|
||||
- When evidence is thin, write "该观点证据不足，待进一步验证" rather than smoothing over the gap.
|
||||
|
||||
## Quota Discipline
|
||||
|
||||
- Surface manager: Gemini 3 Flash.
|
||||
- Phase 0-1: Claude Opus 4.6 (Thinking).
|
||||
- Phase 2: Gemini 3.1 Pro (Low).
|
||||
- Phase 3: Gemini 3.1 Pro (High), with fallback to another model only if quality is inadequate.
|
||||
- Phase 4: Claude Opus 4.6 (Thinking).
|
||||
|
||||
## Pause Points
|
||||
|
||||
Pause after Phase 1 and Phase 3. Ask the user to confirm the framework or critique decision before moving to the next major phase.
|
||||
@@ -0,0 +1,148 @@
|
||||
---
|
||||
name: antigravity-surface-adapter
|
||||
description: Guides Google Antigravity to run Deep Research natively with Antigravity model quotas, while using this repository's skills, source standards, artifacts, and optional Python utilities for validation/rendering.
|
||||
---
|
||||
|
||||
# Antigravity Deep Research Adapter
|
||||
|
||||
Use this skill when the user asks to run, plan, review, or finalize a Deep Research project from Google Antigravity.
|
||||
|
||||
## Core Rule
|
||||
|
||||
Antigravity is the research operator. Use Antigravity's own model selector and task execution to perform the research phases. Python core is an auxiliary toolbox for project scaffolding, deterministic checks, citation/report rendering, and status inspection; do not let it silently route the main research through ZenMux unless the user explicitly asks.
|
||||
|
||||
The Deep Research method still comes from this repo:
|
||||
|
||||
- Follow `AGENTS.md` for evidence, source tiers, word budgets, pause points, and output hygiene.
|
||||
- Load project skills from `.agents/skills/*` as needed.
|
||||
- Write all artifacts under `projects/<slug>/`.
|
||||
- Keep claims traceable through search receipts, source IDs, and `claims_ledger.jsonl`.
|
||||
|
||||
## Anti-Hallucination Contract
|
||||
|
||||
This project treats Gemini-family research as high risk for false confidence. Follow these rules mechanically:
|
||||
|
||||
1. No tool receipt, no search claim. Do not write "I searched", "confirmed", "official data shows", or similar unless the tool call happened and is recorded.
|
||||
2. No source ID, no factual claim. Every concrete fact needs a `source_id` and claim-ledger entry.
|
||||
3. Search snippets are leads, not evidence. Open the original paper, filing, registry, patent, guideline, or company disclosure before citing.
|
||||
4. Failed access is a finding. Write that the source could not be opened; do not infer exact numbers from inaccessible pages.
|
||||
5. Log all searches in `phaseX/search_log.md` or `phaseX/search_log.jsonl`.
|
||||
6. Keep an `unsupported_claims` list. It is better to expose uncertainty than to make the report look complete.
|
||||
7. Keep `continuation_state.json` current so a resumed Antigravity session can continue from artifacts, not chat memory.
|
||||
8. Before finishing each phase, sample at least 10 factual claims and verify that each has a source ID, opened source, evidence locator, and claim-ledger record.
|
||||
|
||||
## Antigravity Model Plan
|
||||
|
||||
Roles are defined in `.agents/agents.md`. This skill only maps phase work to models:
|
||||
|
||||
| Phase | Model | Purpose |
|
||||
|---|---|
|
||||
| Surface management | Gemini 3 Flash | Read skills, maintain task list, move phase by phase, run light shell commands, collect artifacts. |
|
||||
| Phase 0-1 | Claude Opus 4.6 (Thinking) | Material interpretation, strategic framing, research method choice, bold hypotheses, chapter architecture. |
|
||||
| Phase 2 | Gemini 3.1 Pro (Low) | Evidence collection, chapter research packets, first-pass chapter drafts. |
|
||||
| Phase 3 | Gemini 3.1 Pro (High) | Independent critique, counter-evidence review, logic and evidence gap analysis. |
|
||||
| Phase 4 | Claude Opus 4.6 (Thinking) | Final Chinese editorial integration, executive summary, wording quality, final consistency pass. |
|
||||
|
||||
Gemini 3 Flash should remain the phase manager. It can ask the user to switch to the phase model before heavy thinking, then resume management after the phase artifact is written.
|
||||
|
||||
## Python Core Boundary
|
||||
|
||||
Use Python core only when it helps without taking over the research model budget:
|
||||
|
||||
- Good: scaffold folders, parse local materials, validate sources/citations, render PDF/DOCX, show status.
|
||||
- Risky: `research --execute-packets`, `run`, or any command that calls model workers through `configs/models.yaml`.
|
||||
- If a command would call model workers, ask the user first and state that it may use ZenMux/API credits instead of Antigravity quota.
|
||||
|
||||
Safe utility examples:
|
||||
|
||||
```bash
|
||||
uv run python scripts/dr.py init "研究主题" --slug <slug> --method <method>
|
||||
uv run python scripts/dr.py status <slug>
|
||||
uv run python scripts/dr.py skills validate
|
||||
uv run python scripts/dr.py finalize <slug> --no-pdf --no-docx
|
||||
```
|
||||
|
||||
## Native Runbook
|
||||
|
||||
1. Open the repository root in Antigravity.
|
||||
2. Start with Gemini 3 Flash as Research Manager. Load `.agents/agents.md`, this skill, `AGENTS.md`, `method-selection`, `search-strategy`, `source-quality`, `evidence-table`, `research-quality-gates`, `citation-manager`, `length-budget`, and `output-hygiene`.
|
||||
3. Create or confirm `projects/<slug>/` and the phase folders. Use Python `init` only for scaffolding if helpful.
|
||||
4. Phase 0-1: ask the user to switch to Claude Opus 4.6 (Thinking). Produce:
|
||||
|
||||
- `phase1/material_brief.md`
|
||||
- `phase1/framework.md`
|
||||
- `phase1/method_decision.md`
|
||||
- `phase1/research_brief.md`
|
||||
- `phase1/research_brief.json`
|
||||
- `phase1/search_log.md`
|
||||
- `phase1/assumptions.md`
|
||||
- `phase1/unsupported_claims.md`
|
||||
|
||||
5. Pause for user confirmation.
|
||||
6. Phase 2: ask the user to switch to Gemini 3.1 Pro (Low). Build evidence packets and chapter drafts without Python model workers:
|
||||
|
||||
- `phase2/task_cards.json`
|
||||
- `phase2/search_log.jsonl`
|
||||
- `phase2/sources.jsonl`
|
||||
- `phase2/rejected_sources.jsonl`
|
||||
- `phase2/claims_ledger.jsonl`
|
||||
- `phase2/coverage_matrix.md`
|
||||
- `phase2/packets/*.json`
|
||||
- `phase2/chapter_briefs/*.json`
|
||||
- `phase2/compressed_findings/*.json`
|
||||
- `phase2/drafts/chXX.md`
|
||||
- `phase2/unsupported_claims.md`
|
||||
|
||||
If a claim lacks support, use delta retrieval: write the gap, run targeted searches, update sources and the claim ledger, then either verify or visibly caveat the claim.
|
||||
|
||||
7. Phase 3: ask the user to switch to Gemini 3.1 Pro (High). Produce `phase3/critique.md`, with explicit go/rework/fail decision, claim-ledger audit, contradiction audit, and a source-audit table for at least 10 core facts.
|
||||
8. Pause for user decision.
|
||||
9. Phase 4: ask the user to switch to Claude Opus 4.6 (Thinking). Produce `phase4/final_zh.md`, `phase4/editorial_notes.md`, and `phase4/final_fact_check.md`. New facts require new receipts and claim-ledger rows. Use deterministic renderers afterward for PDF/DOCX.
|
||||
|
||||
## Search Strategy
|
||||
|
||||
If the `search` CLI from `paperfoot/search-cli` is installed, prefer it as the Antigravity-native search front door:
|
||||
|
||||
```bash
|
||||
search agent-info
|
||||
search search -q "CRISPR clinical trial 2025" -m academic --json
|
||||
search search -q "BRCA1 gene patent" -m patents --json
|
||||
search search -q "FDA warning letter CAPA pharmaceutical" -m general --json
|
||||
search search -q "biotech market update 2026" -m news --json
|
||||
```
|
||||
|
||||
Useful modes for this repo: `academic`, `scholar`, `patents`, `news`, `deep`, `extract`, `similar`, and `general`.
|
||||
|
||||
Do not treat synthesized search answers as evidence. Every important claim still needs source IDs, tier rating, source score, and preferably original PDFs/pages or official records.
|
||||
|
||||
If `search` is not installed, use the existing project search tools or Antigravity browser, but keep the same source-tier discipline.
|
||||
|
||||
## Suggested Antigravity Permissions
|
||||
|
||||
Within this workspace, the user may allow:
|
||||
|
||||
- `read_file(<workspace>)`
|
||||
- `write_file(<workspace>)`
|
||||
- `command(uv run python scripts/)`
|
||||
- `command(search)`
|
||||
- `command(pandoc)`
|
||||
|
||||
Keep these as Ask/Deny unless the user explicitly allows them:
|
||||
|
||||
- writes outside the workspace
|
||||
- global Antigravity/Gemini settings
|
||||
- destructive git commands
|
||||
- reading `secrets.env`
|
||||
- broad downloads or package installs
|
||||
|
||||
## What To Report Back
|
||||
|
||||
Always report concrete artifact paths:
|
||||
|
||||
- `projects/<slug>/phase1/research_brief.md`
|
||||
- `projects/<slug>/phase2/task_cards.json`
|
||||
- `projects/<slug>/phase2/packet_errors/` if present
|
||||
- `projects/<slug>/phase3/critique.md`
|
||||
- `projects/<slug>/phase4/*.pdf` and `*.docx`
|
||||
|
||||
If a phase fails, report the failing artifact and the rework plan. Do not hide model switches, quota-sensitive steps, or missing evidence.
|
||||
@@ -9,6 +9,51 @@ description: 证据矩阵规范。规定每条核心结论必须有对应的证
|
||||
|
||||
**每条结论必须可追溯**。报告中每一个有 [src_xxx] 标注的观点，都必须在对应章节的 evidence 文件中有一行记录。
|
||||
|
||||
Antigravity native 模式下，证据不能只停留在模型上下文或章节草稿里。除章节 evidence markdown 外，还必须维护机器可审计的 claim ledger，让 Phase 3/4 可以逐条核验。
|
||||
|
||||
---
|
||||
|
||||
## Antigravity Native 证据落盘
|
||||
|
||||
Phase 2 必须至少维护以下文件：
|
||||
|
||||
- `projects/<slug>/phase2/search_log.jsonl`：检索、打开、抽取的回执
|
||||
- `projects/<slug>/phase2/sources.jsonl`：稳定 source registry
|
||||
- `projects/<slug>/phase2/claims_ledger.jsonl`：原子 claim 台账
|
||||
- `projects/<slug>/phase2/coverage_matrix.md`：章节/任务覆盖矩阵
|
||||
- `projects/<slug>/phase2/unsupported_claims.md`：未证实或仅弱证据支持的观点
|
||||
|
||||
`claims_ledger.jsonl` 每行必须是一个原子事实或判断，字段至少包括：
|
||||
|
||||
```json
|
||||
{
|
||||
"claim_id": "ch03_C012",
|
||||
"chapter": "ch03",
|
||||
"claim_type": "fact|trend|comparison|causal|forecast|recommendation",
|
||||
"claim": "...",
|
||||
"supporting_source_ids": ["src_042", "src_087"],
|
||||
"counter_source_ids": ["src_103"],
|
||||
"evidence_spans": [
|
||||
{
|
||||
"source_id": "src_042",
|
||||
"locator": "Table 2",
|
||||
"summary": "..."
|
||||
}
|
||||
],
|
||||
"confidence": "high|medium|low|unsupported",
|
||||
"verification_status": "verified|partially_verified|conflicted|unsupported",
|
||||
"needs_delta_retrieve": false,
|
||||
"visible_caveat_required": false
|
||||
}
|
||||
```
|
||||
|
||||
正文、章节 evidence markdown、`claims_ledger.jsonl` 三者必须能互相对上：
|
||||
|
||||
- 正文的 `Cxx/Fxx/Txx` ID 必须在 evidence markdown 和 `claims_ledger.jsonl` 中存在；
|
||||
- `claims_ledger.jsonl` 中 `verification_status=unsupported` 的 claim 必须进入 `unsupported_claims.md`；
|
||||
- `needs_delta_retrieve=true` 的 claim 必须补做定向检索，或在正文显式降级；
|
||||
- Phase 4 不得新增未进入 claim ledger 的事实。
|
||||
|
||||
---
|
||||
|
||||
## 证据矩阵文件格式
|
||||
|
||||
@@ -0,0 +1,157 @@
|
||||
---
|
||||
name: method-selection
|
||||
description: Selects the appropriate research framework and analytical tools for biomedical, market, investment, CMC/GMP, clinical, R&D, BD, and management consulting topics instead of defaulting to MECE or McKinsey-style framing.
|
||||
---
|
||||
|
||||
# Method Selection
|
||||
|
||||
Use this skill before Phase 0-1 framing and whenever the user changes the report purpose.
|
||||
|
||||
## Principle
|
||||
|
||||
Do not default to McKinsey/MECE. MECE, SCQA, and pyramid writing are useful communication tools, not universal research methods.
|
||||
|
||||
Select the method based on:
|
||||
|
||||
- report purpose
|
||||
- audience
|
||||
- decision to support
|
||||
- available materials
|
||||
- evidence type
|
||||
- regulatory or scientific risk
|
||||
- output format
|
||||
|
||||
## Method Router
|
||||
|
||||
### Market, investment, or commercial strategy
|
||||
|
||||
Use when the user asks about market opportunity, competition, investment thesis, BD opportunity, pricing, or commercialization.
|
||||
|
||||
Methods and tools:
|
||||
|
||||
- market sizing: TAM/SAM/SOM, bottom-up demand model, volume-price bridge
|
||||
- competitive landscape: segmentation, product differentiation, launch timing, share shifts
|
||||
- value chain: upstream supply, manufacturing bottlenecks, channel economics
|
||||
- investment: thesis tree, scenario analysis, sensitivity analysis, risk-adjusted NPV where feasible
|
||||
- transaction/BD: deal comps, licensing terms, patent cliff, strategic fit
|
||||
|
||||
Outputs:
|
||||
|
||||
- market map
|
||||
- competitor table
|
||||
- assumptions ledger
|
||||
- upside/base/downside scenarios
|
||||
|
||||
### Clinical, translational, or evidence review
|
||||
|
||||
Use when the topic is efficacy, safety, trial design, disease area, standard of care, biomarker, or clinical development.
|
||||
|
||||
Methods and tools:
|
||||
|
||||
- PICO or PICOTS framing
|
||||
- evidence hierarchy
|
||||
- endpoint and comparator analysis
|
||||
- trial design critique
|
||||
- TPP and go/no-go criteria
|
||||
- benefit-risk assessment
|
||||
|
||||
Outputs:
|
||||
|
||||
- clinical evidence matrix
|
||||
- trial landscape
|
||||
- endpoint comparability table
|
||||
- evidence gaps and patient-selection risks
|
||||
|
||||
### CMC, GMP, quality, or manufacturing operations
|
||||
|
||||
Use when the topic involves process development, tech transfer, validation, inspection readiness, deviations, CAPA, sterility assurance, supply chain, or manufacturing risk.
|
||||
|
||||
Methods and tools:
|
||||
|
||||
- ICH / FDA / EMA / NMPA regulatory baseline mapping
|
||||
- gap assessment
|
||||
- FMEA or risk matrix
|
||||
- process validation lifecycle
|
||||
- control strategy mapping
|
||||
- deviation/CAPA effectiveness review
|
||||
- data integrity assessment
|
||||
|
||||
Outputs:
|
||||
|
||||
- gap register
|
||||
- risk heatmap
|
||||
- CAPA prioritization
|
||||
- control strategy map
|
||||
- readiness scorecard
|
||||
|
||||
### R&D feasibility or technology assessment
|
||||
|
||||
Use when the topic asks whether a technology, target, modality, platform, assay, or process is feasible.
|
||||
|
||||
Methods and tools:
|
||||
|
||||
- technology readiness level
|
||||
- critical path analysis
|
||||
- mechanism and modality fit
|
||||
- experiment roadmap
|
||||
- scale-up risk assessment
|
||||
- IP and freedom-to-operate screening
|
||||
|
||||
Outputs:
|
||||
|
||||
- feasibility scorecard
|
||||
- key experiments
|
||||
- kill criteria
|
||||
- technical risk register
|
||||
|
||||
### Management, organization, or operating model
|
||||
|
||||
Use when the topic involves organization, governance, operating cadence, capability building, performance management, or transformation.
|
||||
|
||||
Methods and tools:
|
||||
|
||||
- issue tree
|
||||
- operating model design
|
||||
- RACI / decision rights
|
||||
- KPI tree
|
||||
- process maturity assessment
|
||||
- change management roadmap
|
||||
|
||||
Outputs:
|
||||
|
||||
- operating model options
|
||||
- capability gap assessment
|
||||
- implementation roadmap
|
||||
- KPI and governance model
|
||||
|
||||
### Policy, regulation, or access environment
|
||||
|
||||
Use when the topic asks about policy, reimbursement, regulatory pathways, pricing access, or regional rules.
|
||||
|
||||
Methods and tools:
|
||||
|
||||
- policy timeline
|
||||
- stakeholder map
|
||||
- regulatory pathway comparison
|
||||
- access and reimbursement framework
|
||||
- scenario analysis by jurisdiction
|
||||
|
||||
Outputs:
|
||||
|
||||
- jurisdiction comparison
|
||||
- policy impact map
|
||||
- regulatory milestone table
|
||||
- access risks
|
||||
|
||||
## Phase 0-1 Output Requirements
|
||||
|
||||
Write a method decision note in `phase1/research_brief.md`:
|
||||
|
||||
- selected method
|
||||
- why it fits
|
||||
- methods rejected and why
|
||||
- evidence types needed
|
||||
- search routes
|
||||
- expected output artifacts
|
||||
|
||||
If the topic spans multiple scenarios, combine methods explicitly and define which chapters use which method.
|
||||
@@ -0,0 +1,212 @@
|
||||
---
|
||||
name: research-quality-gates
|
||||
description: Applies stage gates for Antigravity-native Deep Research: search receipts, source access, claims ledger, triangulation, contradiction search, continuation state, and final assembly checks.
|
||||
---
|
||||
|
||||
# Research Quality Gates
|
||||
|
||||
Use this skill after every phase boundary and before any user-facing "done" claim. It is designed for Antigravity-native runs where model work happens inside Antigravity, while this repository provides the evidence discipline.
|
||||
|
||||
## Core Principle
|
||||
|
||||
Evidence must persist outside the model context. A claim is not verified because the model remembers it, summarized it, or saw it in a search snippet. It is verified only when the artifact trail contains:
|
||||
|
||||
1. a search receipt,
|
||||
2. an opened original source or explicit access failure,
|
||||
3. a registered `source_id`,
|
||||
4. an evidence span or data locator,
|
||||
5. a claim-ledger record linking the claim to supporting and counter evidence.
|
||||
|
||||
## Required Artifacts
|
||||
|
||||
Create these files when the relevant phase begins. Empty files are acceptable at creation time, but they must be populated before the phase gate passes.
|
||||
|
||||
| Phase | Artifact | Purpose |
|
||||
|---|---|---|
|
||||
| Phase 1 | `phase1/method_decision.md` | Selected method, rejected methods, evidence types, and search routes. |
|
||||
| Phase 1 | `phase1/assumptions.md` | Material assumptions and scope assumptions that need validation. |
|
||||
| Phase 2 | `phase2/search_log.jsonl` | One row per search/open/extract action. |
|
||||
| Phase 2 | `phase2/sources.jsonl` | Stable source registry with scores and tiers. |
|
||||
| Phase 2 | `phase2/rejected_sources.jsonl` | Rejected or low-quality sources with reasons. |
|
||||
| Phase 2 | `phase2/claims_ledger.jsonl` | Atomic factual and analytical claims with verification status. |
|
||||
| Phase 2 | `phase2/coverage_matrix.md` | Coverage of questions, chapters, methods, and evidence gaps. |
|
||||
| Phase 2 | `phase2/unsupported_claims.md` | Claims that could not be verified or need user-visible caveats. |
|
||||
| All phases | `continuation_state.json` | Current phase, completed sections, artifact paths, open gaps, and next actions. |
|
||||
| Phase 4 | `phase4/final_fact_check.md` | Final audit showing each core fact exists in the verified claim ledger. |
|
||||
|
||||
## Search Receipt Gate
|
||||
|
||||
Every search or retrieval action must write a row to `search_log.jsonl`.
|
||||
|
||||
Required fields:
|
||||
|
||||
```json
|
||||
{
|
||||
"receipt_id": "srch_0001",
|
||||
"timestamp": "YYYY-MM-DDTHH:MM:SSZ",
|
||||
"tool": "search-cli|browser|project-search|other",
|
||||
"mode": "academic|scholar|patents|news|extract|general|deep",
|
||||
"query_or_url": "...",
|
||||
"purpose": "which task/card/chapter this supports",
|
||||
"result_count": 10,
|
||||
"opened_urls": ["https://..."],
|
||||
"status": "ok|partial|failed",
|
||||
"failure_reason": null
|
||||
}
|
||||
```
|
||||
|
||||
Gate fails if a model says it searched, confirmed, checked, extracted, opened, or verified a fact without a corresponding receipt.
|
||||
|
||||
## Source Access Gate
|
||||
|
||||
For any source used as evidence, `sources.jsonl` must record access status.
|
||||
|
||||
Required additional fields:
|
||||
|
||||
```json
|
||||
{
|
||||
"id": "src_042",
|
||||
"search_receipt_ids": ["srch_0001"],
|
||||
"raw_url": "https://...",
|
||||
"canonical_url": "https://...",
|
||||
"title": "...",
|
||||
"publication_date": "YYYY-MM-DD",
|
||||
"source_type": "paper|regulator|trial_registry|patent|filing|database|consulting|media|other",
|
||||
"tier": 1,
|
||||
"score": 8.6,
|
||||
"access_status": "opened_original|opened_abstract|paywalled|failed|snippet_only",
|
||||
"evidence_locator": "page/table/section/trial id/patent claim",
|
||||
"independence_cluster": "cluster_...",
|
||||
"notes": "..."
|
||||
}
|
||||
```
|
||||
|
||||
`snippet_only` sources cannot support final claims. Paywalled or abstract-only sources can be used only with a visible caveat or a second opened source that carries the same fact.
|
||||
|
||||
## Claim Ledger Gate
|
||||
|
||||
Each chapter draft must be backed by `claims_ledger.jsonl`. Store atomic claims, not paragraphs.
|
||||
|
||||
Required fields:
|
||||
|
||||
```json
|
||||
{
|
||||
"claim_id": "ch03_C012",
|
||||
"chapter": "ch03",
|
||||
"claim_type": "fact|trend|comparison|causal|forecast|recommendation",
|
||||
"claim": "...",
|
||||
"supporting_source_ids": ["src_042", "src_087"],
|
||||
"counter_source_ids": ["src_103"],
|
||||
"evidence_spans": [
|
||||
{
|
||||
"source_id": "src_042",
|
||||
"locator": "Table 2",
|
||||
"summary": "..."
|
||||
}
|
||||
],
|
||||
"independence_clusters": ["cluster_a", "cluster_b"],
|
||||
"confidence": "high|medium|low|unsupported",
|
||||
"verification_status": "verified|partially_verified|conflicted|unsupported",
|
||||
"needs_delta_retrieve": false,
|
||||
"visible_caveat_required": false
|
||||
}
|
||||
```
|
||||
|
||||
Gate fails if:
|
||||
|
||||
- a core claim has fewer than 2 independent Tier 1-2 supporting sources;
|
||||
- a claim has source IDs but no evidence locator;
|
||||
- supporting sources all come from the same independence cluster;
|
||||
- a conflicted or unsupported claim is written as a settled conclusion.
|
||||
|
||||
## Triangulation Gate
|
||||
|
||||
Counting URLs is not enough. Sources are independent only when they do not trace back to the same press release, same trial registry entry, same company deck, same sell-side note, or same syndicated article.
|
||||
|
||||
For core conclusions, require at least two of these source families where available:
|
||||
|
||||
- regulator / official registry,
|
||||
- peer-reviewed paper or systematic review,
|
||||
- company filing / annual report / exchange disclosure,
|
||||
- patent or legal record,
|
||||
- independent database,
|
||||
- reputable industry or consulting report,
|
||||
- independent expert or professional media analysis.
|
||||
|
||||
If a field has only one source family available, mark the limitation explicitly in `unsupported_claims.md` and in the report limitations section.
|
||||
|
||||
## Contradiction Gate
|
||||
|
||||
Every chapter requires at least one deliberate counter-search or falsification pass. Record it in `search_log.jsonl` and summarize it in the chapter brief.
|
||||
|
||||
Minimum counter-evidence record:
|
||||
|
||||
```json
|
||||
{
|
||||
"claim_id": "ch03_C012",
|
||||
"counter_query": "...",
|
||||
"counter_source_ids": ["src_103"],
|
||||
"result": "none_found|weak_counter|material_counter",
|
||||
"impact": "keep|qualify|rewrite|delete"
|
||||
}
|
||||
```
|
||||
|
||||
Gate fails if the chapter contains no counter-evidence section, no counter-search receipt, or no decision about how contradictions affected the draft.
|
||||
|
||||
## Coverage Gate
|
||||
|
||||
Before drafting, create `phase2/coverage_matrix.md` with one row per chapter/task axis:
|
||||
|
||||
| Chapter | Method | Core Question | Required Evidence | Found Evidence | Gaps | Decision |
|
||||
|---|---|---|---|---|---|---|
|
||||
|
||||
Gate fails if a chapter is drafted while its method, required evidence, or gaps are blank.
|
||||
|
||||
## Delta-Retrieve Protocol
|
||||
|
||||
When a gap appears, do not patch it from memory. Create a targeted delta task:
|
||||
|
||||
1. write the gap in `coverage_matrix.md` or `claims_ledger.jsonl`;
|
||||
2. run 1-3 targeted searches with explicit query purpose;
|
||||
3. register any usable sources;
|
||||
4. update the relevant claim record;
|
||||
5. if still unsupported, keep the caveat visible.
|
||||
|
||||
Delta retrieval is mandatory when Phase 3 finds a critical evidence gap.
|
||||
|
||||
## Continuation Protocol
|
||||
|
||||
For long reports or interrupted runs, keep `projects/<slug>/continuation_state.json` current.
|
||||
|
||||
Minimum fields:
|
||||
|
||||
```json
|
||||
{
|
||||
"version": "0.20",
|
||||
"slug": "...",
|
||||
"current_phase": "phase2",
|
||||
"active_model": "Gemini 3.1 Pro (Low)",
|
||||
"completed_artifacts": ["phase2/task_cards.json"],
|
||||
"open_gates": ["claim_ledger_gate"],
|
||||
"artifact_paths": {
|
||||
"sources": "phase2/sources.jsonl",
|
||||
"claims": "phase2/claims_ledger.jsonl",
|
||||
"search_log": "phase2/search_log.jsonl"
|
||||
},
|
||||
"open_questions": [],
|
||||
"next_actions": []
|
||||
}
|
||||
```
|
||||
|
||||
Before resuming a project, read this state and the latest artifacts. Do not rely on chat history alone.
|
||||
|
||||
## Final Assembly Gate
|
||||
|
||||
Phase 4 must prove that final facts are a subset of verified claims:
|
||||
|
||||
- Sample at least 20 high-impact facts or all core claims, whichever is smaller.
|
||||
- For each sampled fact, record `claim_id`, `source_ids`, verification status, and final wording decision in `phase4/final_fact_check.md`.
|
||||
- New facts introduced during editing require new search receipts and claim-ledger records.
|
||||
- Delete or caveat any unsupported factual claim before rendering PDF/DOCX.
|
||||
|
||||
Do not render final deliverables until the final assembly gate passes.
|
||||
@@ -78,7 +78,52 @@ description: 信源质量评分系统（0-10 分制）与黑名单机制。规
|
||||
|
||||
---
|
||||
|
||||
## 三、黑名单（直接拒绝入库）
|
||||
## 三、检索回执与原文访问
|
||||
|
||||
Gemini / Antigravity 模式下，信源质量评分必须和检索回执绑定。任何来源进入 `sources.jsonl` 前，必须能追溯到 `search_log.jsonl` 中的至少一条记录。
|
||||
|
||||
### search_log.jsonl 最低字段
|
||||
|
||||
```json
|
||||
{
|
||||
"receipt_id": "srch_0001",
|
||||
"timestamp": "YYYY-MM-DDTHH:MM:SSZ",
|
||||
"tool": "search-cli|browser|project-search|other",
|
||||
"mode": "academic|scholar|patents|news|extract|general|deep",
|
||||
"query_or_url": "...",
|
||||
"purpose": "which task/card/chapter this supports",
|
||||
"result_count": 10,
|
||||
"opened_urls": ["https://..."],
|
||||
"status": "ok|partial|failed",
|
||||
"failure_reason": null
|
||||
}
|
||||
```
|
||||
|
||||
### sources.jsonl 追加字段
|
||||
|
||||
```json
|
||||
{
|
||||
"id": "src_042",
|
||||
"search_receipt_ids": ["srch_0001"],
|
||||
"raw_url": "https://...",
|
||||
"canonical_url": "https://...",
|
||||
"source_type": "paper|regulator|trial_registry|patent|filing|database|consulting|media|other",
|
||||
"access_status": "opened_original|opened_abstract|paywalled|failed|snippet_only",
|
||||
"evidence_locator": "page/table/section/trial id/patent claim",
|
||||
"independence_cluster": "cluster_..."
|
||||
}
|
||||
```
|
||||
|
||||
硬规则：
|
||||
|
||||
- `snippet_only` 只能作为发现线索，不能支撑正文事实。
|
||||
- `failed` 必须保留失败原因，不能根据标题或搜索摘要推断具体数据。
|
||||
- `paywalled` / `opened_abstract` 可以用于发现方向，但核心结论还需要至少一个可打开原文或官方记录支撑。
|
||||
- 同一公司新闻稿、同一试验登记、同一专利族、同一券商报告转载链，必须归入同一个 `independence_cluster`。
|
||||
|
||||
---
|
||||
|
||||
## 四、黑名单（直接拒绝入库）
|
||||
|
||||
以下信源**无论评分多少都禁用**：
|
||||
|
||||
@@ -112,7 +157,7 @@ description: 信源质量评分系统（0-10 分制）与黑名单机制。规
|
||||
|
||||
---
|
||||
|
||||
## 四、利益冲突检测要点
|
||||
## 五、利益冲突检测要点
|
||||
|
||||
### 常见利益冲突场景
|
||||
- 药企赞助的 RCT 对自家产品评价极高 → 查对比剂、盲法、样本量
|
||||
@@ -130,7 +175,7 @@ description: 信源质量评分系统（0-10 分制）与黑名单机制。规
|
||||
|
||||
---
|
||||
|
||||
## 五、评分执行流程（伪代码）
|
||||
## 六、评分执行流程（伪代码）
|
||||
|
||||
```
|
||||
for each candidate_source in search_results:
|
||||
@@ -162,7 +207,7 @@ for each candidate_source in search_results:
|
||||
|
||||
---
|
||||
|
||||
## 六、输出字段（写入 sources.jsonl）
|
||||
## 七、输出字段（写入 sources.jsonl）
|
||||
|
||||
```jsonl
|
||||
{
|
||||
@@ -177,12 +222,16 @@ for each candidate_source in search_results:
|
||||
"conflict_of_interest": null,
|
||||
"blacklist_checked": true,
|
||||
"retraction_checked": true,
|
||||
"search_receipt_ids": ["srch_0001"],
|
||||
"access_status": "opened_original",
|
||||
"evidence_locator": "Table 2",
|
||||
"independence_cluster": "cluster_trial_001",
|
||||
"notes": "NEJM 2025 原文，RCT 独立研究"
|
||||
}
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
## 七、审计留痕
|
||||
## 八、审计留痕
|
||||
|
||||
所有被**拒绝**的信源，也要写入 `projects/<slug>/phase2/rejected-sources.jsonl`（注明原因）。这是事后复盘的关键，不要静默丢弃。
|
||||
所有被**拒绝**的信源，也要写入 `projects/<slug>/phase2/rejected_sources.jsonl`（注明原因、对应 `receipt_id`、拒绝规则）。这是事后复盘的关键，不要静默丢弃。
|
||||
|
||||
@@ -0,0 +1,154 @@
|
||||
---
|
||||
description: Run native biomedical Deep Research in Antigravity using roles, rules, skills, workflows, source receipts, and explicit approval gates.
|
||||
---
|
||||
|
||||
# Deep Research Native Workflow
|
||||
|
||||
Invoke as `/deep-research-native <topic or slug>`.
|
||||
|
||||
## Step 0: Load Operating Context
|
||||
|
||||
- Load `AGENTS.md`.
|
||||
- Load `GEMINI.md`.
|
||||
- Load `.agents/agents.md`.
|
||||
- Load `.agents/rules/deep-research-antigravity.md`.
|
||||
- Load `.agents/skills/antigravity-surface-adapter/SKILL.md`.
|
||||
- Load `.agents/skills/method-selection/SKILL.md`.
|
||||
- Load relevant quality skills: `search-strategy`, `source-quality`, `evidence-table`, `research-quality-gates`, `citation-manager`, `length-budget`, `output-hygiene`.
|
||||
|
||||
Gate: if the active model has not loaded the above files, stop and ask the user to restart or explicitly mention them.
|
||||
|
||||
## Step 1: Define The Research Problem
|
||||
|
||||
Act as Research Manager with Gemini 3 Flash.
|
||||
|
||||
Confirm:
|
||||
|
||||
- topic and slug
|
||||
- report purpose
|
||||
- target reader
|
||||
- decision the report supports
|
||||
- report type and expected length
|
||||
- available input materials
|
||||
- allowed search tools
|
||||
- whether Python model-worker commands are forbidden or allowed
|
||||
|
||||
Create or confirm `projects/<slug>/` and phase folders. Create or update `projects/<slug>/continuation_state.json`. Use Python only for scaffolding if helpful.
|
||||
|
||||
Gate: do not continue if purpose, audience, and decision use are unclear.
|
||||
|
||||
## Step 2: Select Method
|
||||
|
||||
Ask the user to switch to Claude Opus 4.6 (Thinking).
|
||||
|
||||
Act as Phase 0-1 Strategist. Use `method-selection`.
|
||||
|
||||
Write a method decision note covering:
|
||||
|
||||
- selected method or method mix
|
||||
- why it fits the scenario
|
||||
- rejected methods and why
|
||||
- evidence types required
|
||||
- search routes by chapter or task axis
|
||||
- expected artifacts
|
||||
|
||||
Save it as `phase1/method_decision.md` or embed the same content in `phase1/research_brief.md` with a clear heading.
|
||||
|
||||
Gate: do not default to McKinsey, MECE, or SCQA. Use them only when they fit the decision problem.
|
||||
|
||||
## Step 3: Phase 0-1 Framing
|
||||
|
||||
Still using Claude Opus 4.6 (Thinking), produce:
|
||||
|
||||
- `phase1/material_brief.md`
|
||||
- `phase1/search_log.md`
|
||||
- `phase1/method_decision.md`
|
||||
- `phase1/assumptions.md`
|
||||
- `phase1/framework.md`
|
||||
- `phase1/research_brief.md`
|
||||
- `phase1/research_brief.json`
|
||||
- `phase1/unsupported_claims.md`
|
||||
|
||||
Rules:
|
||||
|
||||
- Hypotheses without evidence must be labeled as hypotheses.
|
||||
- Every searched claim must have a search receipt.
|
||||
- Each chapter must state method, core question, likely evidence, and falsification route.
|
||||
|
||||
Gate: pause for user approval before Phase 2.
|
||||
|
||||
## Step 4: Phase 2 Evidence And Drafting
|
||||
|
||||
Ask the user to switch to Gemini 3.1 Pro (Low).
|
||||
|
||||
Act as Evidence Analyst.
|
||||
|
||||
Produce:
|
||||
|
||||
- `phase2/task_cards.json`
|
||||
- `phase2/search_log.jsonl`
|
||||
- `phase2/sources.jsonl`
|
||||
- `phase2/rejected_sources.jsonl`
|
||||
- `phase2/claims_ledger.jsonl`
|
||||
- `phase2/coverage_matrix.md`
|
||||
- `phase2/packets/*.json`
|
||||
- `phase2/chapter_briefs/*.json`
|
||||
- `phase2/compressed_findings/*.json`
|
||||
- `phase2/drafts/chXX.md`
|
||||
- `phase2/unsupported_claims.md`
|
||||
|
||||
Rules:
|
||||
|
||||
- No tool receipt, no search claim.
|
||||
- No source ID, no factual claim.
|
||||
- Search snippets and AI summaries are leads only.
|
||||
- Every packet must include `search_receipts`, `sources`, `evidence_spans`, `counter_evidence`, and `unsupported_claims`.
|
||||
- Every core claim must be represented in `claims_ledger.jsonl`.
|
||||
- Every chapter must pass a counter-search or falsification pass.
|
||||
- Evidence gaps trigger delta retrieval before drafting or visible caveats if still unresolved.
|
||||
- Draft chapters only from approved chapter briefs and compressed findings.
|
||||
|
||||
Gate: run `research-quality-gates`. Do not move to Phase 3 if packet evidence is missing, claim-ledger records are incomplete, unsupported claims are hidden, source independence is not tracked, or counter-evidence is absent.
|
||||
|
||||
## Step 5: Phase 3 Review
|
||||
|
||||
Ask the user to switch to Gemini 3.1 Pro (High).
|
||||
|
||||
Act as Chief Reviewer.
|
||||
|
||||
Produce `phase3/critique.md` with:
|
||||
|
||||
- go / rework / fail decision
|
||||
- structural critique
|
||||
- method fit critique
|
||||
- evidence gap list
|
||||
- counter-evidence critique
|
||||
- claim-ledger audit
|
||||
- coverage matrix audit
|
||||
- source-audit table for at least 10 core facts
|
||||
- rework task list if needed
|
||||
|
||||
If the critique finds a critical evidence gap, create delta-retrieve tasks instead of asking Phase 4 to paper over the gap.
|
||||
|
||||
Gate: pause for user decision after critique.
|
||||
|
||||
## Step 6: Phase 4 Finalization
|
||||
|
||||
Ask the user to switch to Claude Opus 4.6 (Thinking).
|
||||
|
||||
Act as Final Editor.
|
||||
|
||||
Produce:
|
||||
|
||||
- `phase4/final_zh.md`
|
||||
- `phase4/editorial_notes.md`
|
||||
- `phase4/final_fact_check.md`
|
||||
|
||||
Rules:
|
||||
|
||||
- Do not introduce new facts unless new sources and search logs are added first.
|
||||
- Final facts must be a subset of verified or explicitly caveated `claims_ledger.jsonl` rows.
|
||||
- Downgrade or mark claims that remain unsupported.
|
||||
- Use deterministic rendering tools afterward for PDF/DOCX.
|
||||
|
||||
Gate: final output must pass citation and unsupported-claim review before rendering.
|
||||
@@ -0,0 +1,20 @@
|
||||
---
|
||||
name: dr-finalize
|
||||
description: Surface adapter command for Chinese-native Phase 4 finalization.
|
||||
---
|
||||
|
||||
Run the project-owned Python core finalization. Do not hand-translate the report in Claude Code.
|
||||
|
||||
Command:
|
||||
|
||||
```bash
|
||||
uv run python scripts/dr.py finalize $ARGUMENTS
|
||||
```
|
||||
|
||||
For old projects only, use:
|
||||
|
||||
```bash
|
||||
uv run python scripts/dr.py finalize <slug> --legacy-translate
|
||||
```
|
||||
|
||||
Report PDF/DOCX paths and any citation/rendering warnings.
|
||||
@@ -0,0 +1,14 @@
|
||||
---
|
||||
name: dr-frame
|
||||
description: Surface adapter command for generating Phase 1 framework.md through the Python core.
|
||||
---
|
||||
|
||||
Run the project-owned Python core framework generator. Do not perform Phase 1 orchestration in Claude Code.
|
||||
|
||||
Command:
|
||||
|
||||
```bash
|
||||
uv run python scripts/dr.py frame $ARGUMENTS
|
||||
```
|
||||
|
||||
Report the framework path and remind the user to approve it before Phase 2.
|
||||
@@ -0,0 +1,14 @@
|
||||
---
|
||||
name: dr-init
|
||||
description: Surface adapter command for initializing a Deep Research v0.20 project.
|
||||
---
|
||||
|
||||
Run the project-owned Python core initialization. Do not create manifest files manually.
|
||||
|
||||
Command:
|
||||
|
||||
```bash
|
||||
uv run python scripts/dr.py init $ARGUMENTS
|
||||
```
|
||||
|
||||
Report the project slug, manifest path, and next command.
|
||||
@@ -0,0 +1,22 @@
|
||||
---
|
||||
name: dr-research
|
||||
description: Surface adapter command for Deep Research v0.20 Phase 2 task-card research.
|
||||
---
|
||||
|
||||
Run the project-owned Python core Phase 2 command. Do not spawn Claude Code subagents for chapter research.
|
||||
|
||||
Command:
|
||||
|
||||
```bash
|
||||
uv run python scripts/dr.py research $ARGUMENTS
|
||||
```
|
||||
|
||||
Useful follow-ups:
|
||||
|
||||
```bash
|
||||
uv run python scripts/dr.py research <slug> --workers 6 --execute-packets
|
||||
uv run python scripts/dr.py research <slug> --workers 6 --build-briefs
|
||||
uv run python scripts/dr.py research <slug> --workers 6 --assemble-chapters
|
||||
```
|
||||
|
||||
Report packet/chapter error files if present.
|
||||
@@ -0,0 +1,14 @@
|
||||
---
|
||||
name: dr-review
|
||||
description: Surface adapter command for deterministic Phase 3 review.
|
||||
---
|
||||
|
||||
Run the project-owned Python core review. Claude Code may explain the critique afterwards, but should not overwrite it unless asked.
|
||||
|
||||
Command:
|
||||
|
||||
```bash
|
||||
uv run python scripts/dr.py review $ARGUMENTS
|
||||
```
|
||||
|
||||
Report the critique path and pause for user decision.
|
||||
@@ -0,0 +1,14 @@
|
||||
---
|
||||
name: dr-run
|
||||
description: Surface adapter command for Deep Research v0.20. Use when the user asks Claude Code to run or continue a Deep Research project.
|
||||
---
|
||||
|
||||
Run the project-owned Python core. Do not perform core orchestration in Claude Code.
|
||||
|
||||
Command:
|
||||
|
||||
```bash
|
||||
uv run python scripts/dr.py run $ARGUMENTS
|
||||
```
|
||||
|
||||
Report only the project path, generated files, failures, and next command.
|
||||
@@ -49,6 +49,9 @@ env_vars = ["TAVILY_API_KEY"]
|
||||
enabled = true
|
||||
required = false
|
||||
|
||||
[mcp_servers.tavily.tools.tavily_search]
|
||||
approval_mode = "approve"
|
||||
|
||||
[mcp_servers.brave_search]
|
||||
command = "npx"
|
||||
args = ["-y", "@modelcontextprotocol/server-brave-search"]
|
||||
|
||||
@@ -0,0 +1,9 @@
|
||||
description = "Run Chinese-native Phase 4 finalization through Python core."
|
||||
prompt = """
|
||||
Run the Deep Research Python core finalize command. Do not hand-translate the report in Gemini CLI.
|
||||
|
||||
Command:
|
||||
!{uv run python scripts/dr.py finalize {{args}}}
|
||||
|
||||
Report PDF/DOCX paths and any citation/rendering warnings.
|
||||
"""
|
||||
@@ -0,0 +1,9 @@
|
||||
description = "Generate Phase 1 framework.md through Python core."
|
||||
prompt = """
|
||||
Run the Deep Research Python core frame command.
|
||||
|
||||
Command:
|
||||
!{uv run python scripts/dr.py frame {{args}}}
|
||||
|
||||
Report the framework path and pause for user approval before Phase 2.
|
||||
"""
|
||||
@@ -0,0 +1,9 @@
|
||||
description = "Initialize a Deep Research v0.20 project through Python core."
|
||||
prompt = """
|
||||
Run the Deep Research Python core init command. Do not create project files manually.
|
||||
|
||||
Command:
|
||||
!{uv run python scripts/dr.py init {{args}}}
|
||||
|
||||
Report the project slug, manifest path, and next command.
|
||||
"""
|
||||
@@ -0,0 +1,9 @@
|
||||
description = "Run Phase 2 task-card research through Python core."
|
||||
prompt = """
|
||||
Run the Deep Research Python core research command. Do not spawn Gemini CLI agents for chapter research.
|
||||
|
||||
Command:
|
||||
!{uv run python scripts/dr.py research {{args}}}
|
||||
|
||||
Report task cards, packets, briefs, drafts, and any error files.
|
||||
"""
|
||||
@@ -0,0 +1,9 @@
|
||||
description = "Run deterministic Phase 3 review through Python core."
|
||||
prompt = """
|
||||
Run the Deep Research Python core review command.
|
||||
|
||||
Command:
|
||||
!{uv run python scripts/dr.py review {{args}}}
|
||||
|
||||
Report the critique path and pause for user decision.
|
||||
"""
|
||||
@@ -0,0 +1,9 @@
|
||||
description = "Run or initialize a Deep Research v0.20 project through Python core."
|
||||
prompt = """
|
||||
Run the Deep Research Python core. Do not orchestrate the workflow in Gemini CLI.
|
||||
|
||||
Command:
|
||||
!{uv run python scripts/dr.py run {{args}}}
|
||||
|
||||
Summarize only the project path, generated artifacts, failures, and next command.
|
||||
"""
|
||||
@@ -31,6 +31,7 @@ env/
|
||||
.python-version
|
||||
pip-log.txt
|
||||
pip-wheel-log/
|
||||
.worktrees/
|
||||
|
||||
# ============ 系统 ============
|
||||
.DS_Store
|
||||
@@ -48,6 +49,11 @@ Thumbs.db
|
||||
.opencode/log/
|
||||
.opencode/cache/
|
||||
|
||||
# ============ Codex adapter deployment target ============
|
||||
# v0.20 keeps Codex templates in codex_adapter_templates/ and deploys usable
|
||||
# files to $CODEX_HOME or ~/.codex via scripts/deploy_adapters.py.
|
||||
.codex/
|
||||
|
||||
# ============ 归档（不纳入版本控制）============
|
||||
archive/*
|
||||
!archive/.gitkeep
|
||||
|
||||
@@ -1,193 +1,30 @@
|
||||
---
|
||||
description: 章节深度研究 agent（英文工作语言）。负责对单个 chapter 进行多轮联网检索、证据收集、英文初稿撰写，产出符合麦肯锡方法论的章节草稿与证据矩阵。由 dr-pm 通过 Task 工具调度。
|
||||
description: "[COMPAT v0.20] analyst 兼容层。默认证据包与章节组装由 Python core 执行。"
|
||||
mode: subagent
|
||||
hidden: true
|
||||
model: zenmux-anthropic/claude-sonnet-4-6
|
||||
temperature: 0.3
|
||||
tools:
|
||||
read: true
|
||||
write: true
|
||||
edit: true
|
||||
webfetch: true
|
||||
bash: true
|
||||
skill: true
|
||||
permission:
|
||||
edit: allow
|
||||
bash:
|
||||
"*": deny
|
||||
"wc *": allow
|
||||
"python3 *": allow
|
||||
"uv run python scripts/search.py *": allow
|
||||
"uv run python scripts/ground.py *": allow
|
||||
"mkdir *": allow
|
||||
"grep *": allow
|
||||
"cat *": allow
|
||||
webfetch: allow
|
||||
"uv run python scripts/dr.py research *": allow
|
||||
edit: deny
|
||||
task:
|
||||
"*": deny
|
||||
---
|
||||
|
||||
# 角色：dr-analyst — 章节深度研究（English Writer）
|
||||
# dr-analyst Compatibility Role
|
||||
|
||||
You are the core researcher of the Deep Research system. Your job is to thoroughly investigate a single chapter assigned by dr-pm and produce a high-quality English draft + evidence matrix.
|
||||
|
||||
## Working Language: English
|
||||
|
||||
**All output (chapter draft, evidence matrix, source summaries) is in English.**
|
||||
|
||||
Reasons:
|
||||
- English training corpus is >80% of LLM training data; English generation has higher precision and better concept networks
|
||||
- Biomedical terminology is native to English (CMC, CQA, GH101, endoglycosidase, etc.)
|
||||
- dr-chief-editor reviews in English; dr-translator handles final Chinese output in Phase 4
|
||||
|
||||
## Required Skills (load at startup)
|
||||
|
||||
Load in order:
|
||||
1. `search-strategy` — Source prioritization and search rounds
|
||||
2. `source-quality` — Source scoring and blacklist
|
||||
3. `length-budget` — Word count budget (use English word count, not Chinese characters)
|
||||
4. `evidence-table` — Evidence matrix format
|
||||
5. `mckinsey-method` — Writing methodology (crucial: SCQA is only for Executive Summary, NOT per-chapter)
|
||||
6. `humanizer-cn` — English-side rules (§1-26) for avoiding AI patterns
|
||||
|
||||
## Core Workflow
|
||||
|
||||
dr-pm assigns you a chapter with:
|
||||
- Chapter number, title, English word quota
|
||||
- Research thinking (from framework.md)
|
||||
- Output paths (draft, evidence, sources)
|
||||
|
||||
### Step 1: Read Framework
|
||||
|
||||
Read `projects/<slug>/phase1/framework.md` to understand the chapter's positioning and section-level research questions.
|
||||
|
||||
### Step 2: Multi-Round Search (minimum 4 rounds per `search-strategy`)
|
||||
|
||||
- Round 1: PubMed / ClinicalTrials / openFDA / Patent DBs (Tier 1 precise queries)
|
||||
- Round 2: Consulting reports / systematic reviews (Tier 2)
|
||||
- Round 3: Counter-evidence (search for limitations, failures, controversies)
|
||||
- Round 4: Tavily/Exa/Brave for gap-filling, trace back to Tier 1-2 originals
|
||||
|
||||
Mandatory project search gateway:
|
||||
- Literature / reviews: `uv run python scripts/search.py "<query>" --route scholar --num-results 10 --year-low 2023`
|
||||
- Patents / FTO: `uv run python scripts/search.py "<query>" --route patents --num-results 10`
|
||||
- News / transactions: `uv run python scripts/search.py "<query>" --route news --num-results 10 --time-range m`
|
||||
- Generic gap-fill: `uv run python scripts/search.py "<query>" --route general --num-results 10`
|
||||
- Fast grounded fact-check (native model web search): `uv run python scripts/ground.py "<query>" --json`
|
||||
|
||||
Record the routes used in the evidence file. Do not use Tavily / Exa / Brave MCP as the primary path for literature or patent searches.
|
||||
|
||||
Search in **both English and Chinese** for each direction (Chinese sources critical for China market / NMPA / CSRC disclosures).
|
||||
|
||||
### Step 3: Source Scoring
|
||||
|
||||
Every source scored per `skill:source-quality`. Filter out score <5 and blacklist. Add to `projects/<slug>/phase2/sources.jsonl`.
|
||||
|
||||
### Step 4: Write Chapter Draft (English)
|
||||
|
||||
Follow `skill:mckinsey-method` strictly:
|
||||
|
||||
- Chapter title = a judgment/opinion, NOT "Overview" or "Current state"
|
||||
- Opening paragraph: give the conclusion first (pyramid principle)
|
||||
- Each section title = sub-judgment
|
||||
- Each paragraph structure: claim → evidence 1 → evidence 2 → So What
|
||||
- Every number/fact followed by `[src_xxx]`
|
||||
- If <2 independent Tier 1-2 sources: mark `[Unverified: only X source(s) support this]` explicitly
|
||||
|
||||
**DO NOT do** (per v0.4 lessons):
|
||||
- Put explicit `**Situation**:` / `**Complication**:` / `**Question**:` / `**Answer**:` labels
|
||||
- Write SCQA for every section (SCQA is for Executive Summary only)
|
||||
- Include metadata like "Chapter position: P0 Core" / "Word quota: 4,200" / "Researcher: dr-analyst"
|
||||
- Add `⚠️ To be verified` stylistic flags in body text (use formal language if flagging: "This data point has only one supporting source")
|
||||
|
||||
### Step 5: Word Count Self-Check
|
||||
v0.20 不再使用平台 analyst 做整章英文深研。默认 analyst 工作由 Python task workers 完成：
|
||||
|
||||
```bash
|
||||
wc -w projects/<slug>/phase2/drafts/chXX.md
|
||||
uv run python scripts/dr.py research <slug> --execute-packets
|
||||
uv run python scripts/dr.py research <slug> --build-briefs
|
||||
uv run python scripts/dr.py research <slug> --assemble-chapters
|
||||
```
|
||||
|
||||
Per `skill:length-budget`:
|
||||
- Actual/Quota < 0.7 → insufficient, keep digging
|
||||
- 0.7 ≤ ratio < 0.85 → warning, prefer to expand
|
||||
- 0.85 ≤ ratio ≤ 1.3 → pass
|
||||
- ratio > 1.3 → over-budget, consider trimming
|
||||
|
||||
### Step 6: Build Evidence Matrix
|
||||
|
||||
Per `skill:evidence-table`, for every core claim create a row with:
|
||||
- Claim ID (C01-C99)
|
||||
- Claim summary (≤30 English words)
|
||||
- Supporting Evidence 1 & 2 (with src_id, tier, score)
|
||||
- Confidence: High / Medium / Low / Unverified
|
||||
- Notes
|
||||
|
||||
Write to `projects/<slug>/phase2/evidence/chXX-evidence.md` (English).
|
||||
|
||||
### Step 7: Write to Files
|
||||
|
||||
**File writing protocol (v0.5.1)** — prefer `write` over `edit`/`apply_patch` for these files, because they are created fresh by you:
|
||||
|
||||
- Draft: `projects/<slug>/phase2/drafts/chXX.md` (English) — use `write` to create
|
||||
- Evidence matrix: `projects/<slug>/phase2/evidence/chXX-evidence.md` (English) — use `write` to create
|
||||
- Sources: `projects/<slug>/phase2/sources.jsonl` — read current content, append new source lines in memory, then `write` the full new content (do NOT use `apply_patch` to append JSONL lines — it often fails on whitespace matching)
|
||||
|
||||
**If you need to revise a file you already wrote in this session** (e.g., after a self-check you want to extend a section):
|
||||
1. `read` the file to get current content
|
||||
2. Compose the new full content in memory
|
||||
3. `write` the full content (overwrites atomically)
|
||||
|
||||
Do NOT use `apply_patch` to append content. This has caused task stalls in production (v0.4 lessons).
|
||||
|
||||
### Step 8: Report Back
|
||||
|
||||
Return to dr-pm:
|
||||
```
|
||||
Chapter: Ch X - <title>
|
||||
Actual words: X / quota X (XX%)
|
||||
Sources: X total (Tier1: X, Tier2: X)
|
||||
Unverified claims: X
|
||||
Files written:
|
||||
- phase2/drafts/chXX.md
|
||||
- phase2/evidence/chXX-evidence.md
|
||||
- phase2/sources.jsonl (appended)
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
## Style Requirements (English Writing)
|
||||
|
||||
Follow `skill:humanizer-cn` §1-26 strictly:
|
||||
|
||||
**Avoid**:
|
||||
- AI vocabulary: additionally, crucial, delve, emphasizing, enduring, enhance, fostering, pivotal, showcase, testament, underscore, valuable, vibrant
|
||||
- Copula avoidance: "X serves as Y" → "X is Y"
|
||||
- -ing phrase pile-up: "highlighting...", "reflecting...", "contributing to..."
|
||||
- Negative parallelism: "not just X, but Y"
|
||||
- Rule of three: don't force 3-item lists
|
||||
- False ranges: "from X to Y" where X and Y aren't on a scale
|
||||
- Vague attributions: "Industry observers", "Experts believe"
|
||||
- Em-dash overuse: ≤3 per chapter
|
||||
- Empty adjectives without data: "significant" must have a number
|
||||
- Chatbot artifacts: "Of course!", "I hope this helps"
|
||||
|
||||
**Prefer**:
|
||||
- Specific data over abstractions
|
||||
- Active voice
|
||||
- Short-long sentence rhythm mix
|
||||
- "If X, then Y" conditional judgments
|
||||
- Direct claims with supporting numbers
|
||||
|
||||
---
|
||||
|
||||
## Hard Rules
|
||||
|
||||
1. MUST: Every claim has `[src_xxx]` citation
|
||||
2. MUST: Every numerical fact has a source
|
||||
3. MUST: Counter-evidence paragraph is mandatory at chapter end. Per skill:evidence-table §"正文中反方证据段落的写作规范", the heading must express a concrete opinion (e.g., "反例：Codexis ECO 并非所有情境都优于 SPOS" or "值得警惕：临床前到 IND 的衰减率"), NOT a mechanical label like "Counter-Evidence" / "反驳证据". Use H2 or H3 heading level consistently; never use bold text as pseudo-heading.
|
||||
4. MUST: Word count ≥85% of quota, or continue searching
|
||||
5. MUST: No scheduling metadata in body text (no "P0 core", "quota: X", "researcher: dr-analyst")
|
||||
6. MUST: No SCQA labels (not even implicitly suggested by structure)
|
||||
7. MUST NOT: Fabricate data, URLs, DOIs
|
||||
8. MUST NOT: Use Chinese words for claims (English working language)
|
||||
9. MUST NOT: Delegate to other agents
|
||||
10. MUST NOT: **Use emoji anywhere in the draft** (no ✅ ❌ 🔶 🔷 ⭐ 🟢 🔴 ⚠️ 💡 📌 🔑 📊 etc.). The PDF font has no glyphs for colored emoji; they render as empty boxes. Use plain text equivalents (e.g., "✓", "×", "注:", "警告:", or descriptive words like "advantages / limitations / example").
|
||||
本 agent 只可解释失败包或辅助人工诊断，不得直接写 `phase2/drafts/chXX.md`。
|
||||
|
||||
@@ -1,201 +1,28 @@
|
||||
---
|
||||
description: 总编审校（Phase 3 only）。用超长上下文通读全部英文章节草稿，从逻辑自洽、证据充分、观点高度、金字塔原理等维度出具审校报告。仅产出 critique.md，不参与 Phase 4 的任何写作动作。
|
||||
description: "[COMPAT v0.20] Phase 3 审校兼容层。默认审校由 Python core deterministic review 执行。"
|
||||
mode: primary
|
||||
model: zenmux/google/gemini-3.1-pro-preview
|
||||
temperature: 0.3
|
||||
tools:
|
||||
read: true
|
||||
write: true
|
||||
webfetch: true
|
||||
bash: true
|
||||
skill: true
|
||||
permission:
|
||||
edit:
|
||||
"projects/*/phase3/**": allow
|
||||
"projects/*/phase1/**": deny
|
||||
"projects/*/phase2/**": deny
|
||||
"projects/*/phase4/**": deny
|
||||
"*": deny
|
||||
bash:
|
||||
"*": deny
|
||||
"wc *": allow
|
||||
"ls *": allow
|
||||
"cat *": allow
|
||||
"grep *": allow
|
||||
webfetch: allow
|
||||
"uv run python scripts/dr.py review *": allow
|
||||
edit: deny
|
||||
task:
|
||||
"*": deny
|
||||
color: "#10b981"
|
||||
---
|
||||
|
||||
# 角色：dr-chief-editor — Phase 3 审校官（只读角色）
|
||||
# dr-chief-editor Compatibility Role
|
||||
|
||||
你是 Deep Research 系统 Phase 3 的**唯一审校官**。你的职责**仅限于审校**，不参与 Phase 4 的任何写作、合并、润色、出稿动作。
|
||||
v0.20 的默认 Phase 3 审校入口是：
|
||||
|
||||
## 职责边界（硬规则）
|
||||
|
||||
- ✅ 读 `phase2/drafts/` 所有英文章节草稿
|
||||
- ✅ 读 `phase2/evidence/` 所有证据矩阵
|
||||
- ✅ 读 `phase1/framework.md` 对照原设计
|
||||
- ✅ 写 `phase3/critique.md`（审校报告）
|
||||
- ❌ 不得修改任何 phase1/phase2/phase4 文件
|
||||
- ❌ 不得合并章节、写摘要、生成术语表、出稿
|
||||
- ❌ 不得触发任何子 agent
|
||||
|
||||
---
|
||||
|
||||
## 你在什么时候被调度
|
||||
|
||||
用户执行 `/dr-review` 时，由命令直接触发你进入工作。
|
||||
|
||||
## Phase 3 审校工作流
|
||||
|
||||
### Step 1: 加载上下文
|
||||
|
||||
加载 skills：
|
||||
- `skill:mckinsey-method`（评判标准）
|
||||
- `skill:evidence-table`（证据核验标准）
|
||||
- `skill:length-budget`（字数核验）
|
||||
- `skill:output-hygiene`（格式规范）
|
||||
|
||||
读取：
|
||||
- `projects/<slug>/phase1/framework.md`（原始设计）
|
||||
- `projects/<slug>/phase2/drafts/ch*.md`（全部英文草稿）
|
||||
- `projects/<slug>/phase2/evidence/ch*-evidence.md`（证据矩阵，重点看 CRITICAL 标注）
|
||||
- `projects/<slug>/phase2/sources.jsonl`（信源库）
|
||||
- `projects/<slug>/manifest.json`（目标字数与元信息）
|
||||
|
||||
### Step 2: 八维审校
|
||||
|
||||
1. **全局论点一致性**：各章结论是否共同支撑 framework.md 的 Central Thesis？有无章节与总论点相悖？
|
||||
2. **逻辑链完整性**：章节间是否有跳跃？章内逻辑是否自洽？
|
||||
3. **MECE 验证**：各章节划分是否互斥且穷尽？有无遗漏重要维度？
|
||||
4. **证据充分性**：是否有章节缺乏 Tier 1-2 支撑？`[待验证]` 标注比例 <20%？
|
||||
5. **CRITICAL 反方证据处理**：dr-verifier 标注的 CRITICAL 问题是否在草稿中已有回应？
|
||||
6. **字数达标**：各章实际英文词数 vs 配额 ≥0.85？总字数达 `manifest.min_words_en`？
|
||||
7. **观点高度**：结论是否鲜明？有无升华空间未被利用？
|
||||
8. **AI 味检查**（新增）：草稿是否有明显 AI 套路（空泛形容词、三段式堆砌、negative parallelism、-ing 短语）？对比 `skill:mckinsey-method` §8
|
||||
|
||||
### Step 3: 出具审校报告（英文）
|
||||
|
||||
审校报告用**英文**撰写（因为草稿是英文，审校也应用英文保持一致性）。
|
||||
|
||||
写入 `projects/<slug>/phase3/critique.md`：
|
||||
|
||||
```markdown
|
||||
# Phase 3 Editorial Review
|
||||
|
||||
Generated: <datetime>
|
||||
Reviewer: dr-chief-editor (Gemini 3.1 Pro Preview)
|
||||
Total word count: X words / target X (XX%)
|
||||
Word language: English
|
||||
Final output will be translated to Chinese in Phase 4.
|
||||
|
||||
## Overall Rating
|
||||
A (ready for finalize) / B (minor revisions) / C (needs rework) / D (restart framework)
|
||||
|
||||
## Rating Rationale
|
||||
<1-3 sentences on the core judgment>
|
||||
|
||||
## Eight-Dimension Assessment
|
||||
|
||||
### 1. Central Thesis Coherence
|
||||
- Status: Strong / Adequate / Weak
|
||||
- Findings: ...
|
||||
|
||||
### 2. Logical Flow
|
||||
- Status: ...
|
||||
- Findings: ...
|
||||
|
||||
### 3. MECE Validation
|
||||
- Status: ...
|
||||
- Findings: ...
|
||||
|
||||
### 4. Evidence Sufficiency
|
||||
- Status: ...
|
||||
- [Unverified] markers: X chapters, Y total instances
|
||||
- Findings: ...
|
||||
|
||||
### 5. CRITICAL Counter-evidence Handling
|
||||
- CRITICAL flags raised by dr-verifier: X
|
||||
- Addressed in drafts: Y
|
||||
- Unaddressed (requires revision): Z
|
||||
|
||||
### 6. Word Count Audit
|
||||
| Chapter | Quota (EN) | Actual (EN) | Ratio | Status |
|
||||
|---|---|---|---|---|
|
||||
| 1 | 1260 | 1340 | 106% | OK |
|
||||
|
||||
### 7. Point-of-View Strength
|
||||
- Sharp judgments: Y
|
||||
- Neutral descriptions that should be sharpened: Z
|
||||
|
||||
### 8. AI-Pattern Scan
|
||||
- "-ing phrase pile-up": X instances
|
||||
- "Negative parallelism": X instances
|
||||
- Empty adjectives without data: X instances
|
||||
- SCQA over-labeling: X instances
|
||||
(These will be cleaned by dr-polisher in Phase 4; flag here for visibility)
|
||||
|
||||
## Must-Fix Issues (before finalize)
|
||||
|
||||
| # | Chapter | Type | Description | Suggested Action |
|
||||
|---|---|---|---|---|
|
||||
| 1 | ch03 | Logic gap | Chapter 3 jumps from mechanism to market without transition | Add a paragraph in §3.2 bridging the two |
|
||||
|
||||
## Recommended Improvements (optional)
|
||||
|
||||
| # | Chapter | Type | Description |
|
||||
|---|---|---|---|
|
||||
|
||||
## Highlights (preserve)
|
||||
|
||||
- ...
|
||||
|
||||
## Decision Guidance for User
|
||||
|
||||
- If rating A/B: proceed to /dr-finalize
|
||||
- If rating C: return specific chapters to Phase 2 for rework
|
||||
- If rating D: restart from Phase 1
|
||||
```bash
|
||||
uv run python scripts/dr.py review <slug>
|
||||
```
|
||||
|
||||
### Step 4: 暂停
|
||||
|
||||
审校报告写入 phase3/critique.md 后，**停下来等用户决策**。不要自动进入 Phase 4。
|
||||
|
||||
向用户汇报：
|
||||
```
|
||||
Phase 3 审校完成
|
||||
|
||||
审校报告：projects/<slug>/phase3/critique.md
|
||||
总体评级：<A/B/C/D>
|
||||
必修问题：X 项
|
||||
字数状态：X 字 / 目标 X 字 (XX%)
|
||||
|
||||
下一步请选择：
|
||||
- 评级 A/B：运行 /dr-finalize 进入成稿
|
||||
- 评级 C：告诉我哪些章节回炉，我会标记它们重新跑 Phase 2
|
||||
- 评级 D：运行 /dr-frame 重新规划框架
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
## 关键原则
|
||||
|
||||
1. **只读**：永远不修改草稿，永远不参与 Phase 4
|
||||
2. **严格**：发现问题必须指出，不做"过得去"的让步
|
||||
3. **英文对齐**：草稿是英文，审校也用英文
|
||||
4. **具体**：每个 Must-Fix 要具体到章节和段落，不能说"需要改进"
|
||||
5. **信任 dr-verifier**：反方证据已由 dr-verifier 核验，你重点看"章节是否响应了 CRITICAL 标注"
|
||||
|
||||
---
|
||||
|
||||
## 你不做的事（重要）
|
||||
|
||||
- ❌ 不写 Executive Summary 或 Abstract（那是 dr-editor-in-chief 在 Phase 4 做的）
|
||||
- ❌ 不合并 final_en.md（dr-editor-in-chief 做）
|
||||
- ❌ 不翻译成中文（dr-translator 做）
|
||||
- ❌ 不做润色（dr-polisher 做）
|
||||
- ❌ 不出 PDF/DOCX（dr-reporter 做）
|
||||
- ❌ 不修改任何 phase2 的章节草稿
|
||||
|
||||
你的输出只有一份：`phase3/critique.md`。
|
||||
Gemini 长上下文能力可用于解释或补充 `phase3/critique.md`，但不得默认覆盖 deterministic review，不得进入 Phase 4 写作。
|
||||
|
||||
@@ -1,307 +1,28 @@
|
||||
---
|
||||
description: 主编辑（Phase 4 总体）。只做创作性工作（Executive Summary / Abstract / Glossary / 章节合并）。翻译/润色/成稿全部委派给 Python 脚本（v0.6 架构）。
|
||||
description: "[COMPAT v0.20] Phase 4 兼容层。默认中文原生成稿由 Python core finalize 执行。"
|
||||
mode: primary
|
||||
model: zenmux-anthropic/claude-opus-4-7
|
||||
temperature: 0.4
|
||||
tools:
|
||||
read: true
|
||||
write: true
|
||||
edit: true
|
||||
apply_patch: false
|
||||
bash: true
|
||||
skill: true
|
||||
task: true
|
||||
permission:
|
||||
edit: allow
|
||||
bash:
|
||||
"*": deny
|
||||
"wc *": allow
|
||||
"ls *": allow
|
||||
"cat *": allow
|
||||
"head *": allow
|
||||
"tail *": allow
|
||||
"grep *": allow
|
||||
"mkdir *": allow
|
||||
"python3 *": allow
|
||||
"uv run *": allow
|
||||
"bash scripts/*": allow
|
||||
webfetch: deny
|
||||
"uv run python scripts/dr.py finalize *": allow
|
||||
edit: deny
|
||||
task:
|
||||
"*": deny
|
||||
color: "#9333ea"
|
||||
---
|
||||
|
||||
# 角色：dr-editor-in-chief — Phase 4 主编辑
|
||||
# dr-editor-in-chief Compatibility Role
|
||||
|
||||
你是 Deep Research 系统 Phase 4 的**总体执行者**。你决定报告最终长什么样：从章节组装到 Executive Summary 再到 Citations 回填，都由你把控。
|
||||
|
||||
## 为什么由 Opus 4-7 来做
|
||||
|
||||
- dr-analyst（Sonnet 4-6）写了正文；由同家族的 Opus 整合，保证风格连续性
|
||||
- Phase 3 的 Gemini 审校完成后，写作权交回 Anthropic 家族
|
||||
- Opus 的长上下文（1M）和综合判断力适合跨 12-15 章统一叙事
|
||||
|
||||
---
|
||||
|
||||
## 你的核心职责
|
||||
|
||||
当用户执行 `/dr-finalize` 时，**dr-editor-in-chief 是 Phase 4 的入口**。
|
||||
|
||||
### Step 1: 健康检查
|
||||
|
||||
读取 `projects/<slug>/manifest.json`，确认：
|
||||
- `phase2.status == "completed"`
|
||||
- `phase3.approved == true`（已通过审校）
|
||||
|
||||
读取 `projects/<slug>/phase3/critique.md`，确认：
|
||||
- Must-Fix 问题已清空（由 Phase 2 回炉解决）或用户明确接受
|
||||
|
||||
如果前置条件不满足，告知用户并停止。
|
||||
|
||||
### Step 2: 加载 Skills
|
||||
|
||||
必读：
|
||||
- `skill:mckinsey-method`（整体风格标准）
|
||||
- `skill:output-hygiene`（元数据黑名单）
|
||||
- `skill:length-budget`（字数校验）
|
||||
- `skill:humanizer-cn`（写作规则，即使写英文也应遵循 §英文部分）
|
||||
|
||||
### Step 3: 合并英文终稿 final_en.md
|
||||
|
||||
按以下结构组装 `projects/<slug>/phase4/final_en.md`：
|
||||
|
||||
```markdown
|
||||
# <Report Title (English)>
|
||||
|
||||
**<Subtitle (English)>**
|
||||
|
||||
Confidentiality: <from manifest.confidentiality>
|
||||
Date: <YYYY-MM>
|
||||
Version: <X.Y>
|
||||
|
||||
---
|
||||
|
||||
## Disclaimer
|
||||
|
||||
<from manifest.disclaimer, translated to English if needed>
|
||||
|
||||
---
|
||||
|
||||
## Executive Summary
|
||||
|
||||
<You write this, 800-1000 words, using implicit SCQA structure>
|
||||
<NEVER label S/C/Q/A explicitly>
|
||||
<4 core conclusions + key action priorities, similar to 9MW1911>
|
||||
|
||||
---
|
||||
|
||||
## Abstract
|
||||
|
||||
<You write this, 500-600 words, narrative style for broader readership>
|
||||
|
||||
---
|
||||
|
||||
## Glossary
|
||||
|
||||
<You extract all in-text abbreviations and generate bilingual table>
|
||||
<Format: Term | Full name (English) | Chinese equivalent | Brief explanation>
|
||||
|
||||
---
|
||||
|
||||
## Table of Contents
|
||||
|
||||
[Auto-generated by dr-reporter]
|
||||
|
||||
---
|
||||
|
||||
<All chapters from phase2/drafts/ch01.md, ch02.md, ..., concatenated in order>
|
||||
<Do NOT modify chapter content; only ensure transitions are smooth>
|
||||
<Fix any obvious typos or formatting inconsistencies>
|
||||
<Remove any leaked metadata (per skill:output-hygiene)>
|
||||
|
||||
---
|
||||
|
||||
## References
|
||||
|
||||
[Auto-filled by dr-reporter with content from citations.md]
|
||||
|
||||
---
|
||||
|
||||
## Appendix
|
||||
|
||||
<If framework.md listed appendices, aggregate them here>
|
||||
<If none, omit this section>
|
||||
|
||||
---
|
||||
|
||||
## Version History
|
||||
|
||||
- Generated: <datetime>
|
||||
- Report version: <X.Y>
|
||||
- System: Deep Research v0.5
|
||||
- Language workflow: English (drafts) → Chinese (final)
|
||||
```
|
||||
|
||||
### Step 4: Executive Summary 写作（关键）
|
||||
|
||||
Executive Summary 是整份报告最重要的章节。你要按 9MW1911 综合战略报告的风格写：
|
||||
|
||||
**结构模板**（800-1000 词英文）：
|
||||
|
||||
```
|
||||
Opening paragraph (80-120 words):
|
||||
- SCQA structure, implicit (no labels)
|
||||
- Sets up the core problem and report's answer
|
||||
|
||||
Core conclusions (4 numbered items, each 80-120 words):
|
||||
1. [Main conclusion 1, with key data point]
|
||||
2. [Main conclusion 2, with key data point]
|
||||
3. [Main conclusion 3, with key data point]
|
||||
4. [Action priorities / timing / risk summary]
|
||||
|
||||
Closing paragraph (40-60 words):
|
||||
- What happens if conditions met vs not met
|
||||
- Decision call to action
|
||||
```
|
||||
|
||||
**禁止**：
|
||||
- 显式标注 "Situation:", "Complication:", "Question:", "Answer:"
|
||||
- 空泛开头如 "In today's rapidly evolving landscape..."
|
||||
- 结尾泛泛的 "Exciting times lie ahead"
|
||||
|
||||
**推荐**：
|
||||
- 数据支撑每个判断
|
||||
- 每个结论都有 So What
|
||||
- 用 "If X happens, then Y" 表达条件性判断
|
||||
|
||||
### Step 5: Abstract 写作
|
||||
|
||||
Abstract 面向更广泛读者（500-600 词），叙事风格，不分条。内容：
|
||||
|
||||
- 背景（行业/疾病/技术的现状）
|
||||
- 核心挑战与机遇
|
||||
- 本报告分析的六个维度（或你的章节数）
|
||||
- 核心结论一句话
|
||||
- 报告的定位（谁会看，怎么用）
|
||||
|
||||
### Step 6: Glossary 写作
|
||||
|
||||
扫描所有章节的正文，提取出专业缩写和术语（首次出现时应有定义）。按字母序排列：
|
||||
|
||||
```markdown
|
||||
## Glossary
|
||||
|
||||
| Abbr. | Full Name (English) | Chinese | Notes |
|
||||
|---|---|---|---|
|
||||
| ADC | Antibody-Drug Conjugate | 抗体偶联药物 | 2024 年全球 ADC 销售额 100+ 亿美元 |
|
||||
| BEC | Blood Eosinophil Count | 血嗜酸性粒细胞计数 | COPD 生物制剂的常用生物标志物 |
|
||||
| ... | ... | ... | ... |
|
||||
```
|
||||
|
||||
### Step 7: 合并章节（禁止改写）
|
||||
|
||||
逐一读取 `projects/<slug>/phase2/drafts/chXX.md`，**直接拼接**到 final_en.md。
|
||||
|
||||
**你只能做**：
|
||||
- 添加/调整章节之间的过渡句（最多每章 1-2 句）
|
||||
- 修复格式不一致（如标题层级）
|
||||
- 清除 skill:output-hygiene 列出的元数据泄漏
|
||||
- 统一引用格式（[src_xxx] 三位数字）
|
||||
|
||||
**你不能做**：
|
||||
- 改写章节正文
|
||||
- 删除或大幅重组章节内容
|
||||
- 给每章强加 SCQA 开头（这是 v0.4 的错误做法）
|
||||
- 添加"章节定位/字数配额/研究员"等调度元数据
|
||||
|
||||
### Step 8: 翻译 — 调用 Python 脚本（v0.6 新）
|
||||
|
||||
final_en.md 写完后，直接 bash 调 translate.py。**不再使用 dr-translator agent**（v0.6 已废弃，原因：LLM 一次性处理整篇无法稳定）。
|
||||
v0.20 的默认 Phase 4 入口是中文原生成稿：
|
||||
|
||||
```bash
|
||||
uv run python scripts/translate.py <slug>
|
||||
uv run python scripts/dr.py finalize <slug>
|
||||
```
|
||||
|
||||
这个脚本会：
|
||||
- 按 H1/H2 切块（每块 <600 词）
|
||||
- 逐块调 Sonnet 4.6 翻译，断点续传
|
||||
- 累积术语表到 `phase4/glossary.json`
|
||||
- 合并输出 `phase4/final_zh.md`
|
||||
|
||||
典型耗时：17 分钟 / 19k 英文词，约 $1.70。
|
||||
|
||||
### Step 8.5: 术语表核查（强烈推荐，v0.6 新）
|
||||
|
||||
```bash
|
||||
uv run python scripts/build_glossary.py <slug> --workers 4
|
||||
uv run python scripts/apply_glossary.py <slug> --dry-run # 先预览
|
||||
uv run python scripts/apply_glossary.py <slug> # 确认后应用
|
||||
```
|
||||
|
||||
`build_glossary` 用 Haiku + Exa 搜索逐条核查术语中文译名与英文全称，发现拼写错误（如 Maywavee → Mabwell）与误译（如 Beyotime → '碧云天' 实应为 '必贝特医药'）。
|
||||
`apply_glossary` 把高置信度修正直接字面替换到 `final_zh.md`。
|
||||
|
||||
### Step 9: 润色 — 调用 Python 脚本
|
||||
|
||||
```bash
|
||||
uv run python scripts/polish.py <slug>
|
||||
```
|
||||
|
||||
这会按 H2 section 循环润色 `final_zh.md`，输出 `final_zh_polished.md`。单块 <2500 字，不会爆 output token。约 10 分钟 / $1.20。
|
||||
|
||||
### Step 10: 出稿 — 调用 Python 脚本
|
||||
|
||||
```bash
|
||||
uv run python scripts/build_report.py <slug>
|
||||
```
|
||||
|
||||
自动完成：
|
||||
- 按 `manifest.report_title` 命名输出文件（中文标题文件名）
|
||||
- ReportLab 生成 PDF（自动插入 TOC、从 `phase2/sources.jsonl` 生成 GB/T 7714 参考文献）
|
||||
- Pandoc 生成 DOCX
|
||||
|
||||
### Step 11: 收官汇报
|
||||
|
||||
所有脚本跑完后，更新 `manifest.phase4.status = "completed"` 并汇报：
|
||||
|
||||
```
|
||||
Phase 4 成稿完成
|
||||
|
||||
产出文件：
|
||||
- projects/<slug>/phase4/final_en.md (英文源稿)
|
||||
- projects/<slug>/phase4/final_zh.md (中文翻译初稿)
|
||||
- projects/<slug>/phase4/final_zh_polished.md (中文润色稿)
|
||||
- projects/<slug>/phase4/<Title>.pdf (中文 PDF，按标题命名)
|
||||
- projects/<slug>/phase4/<Title>.docx (中文 DOCX，按标题命名)
|
||||
- projects/<slug>/phase4/glossary.json (双语术语表，已核查)
|
||||
|
||||
统计：
|
||||
英文源：X words
|
||||
中文稿：X 字 (膨胀率 X%)
|
||||
信源：X 条
|
||||
页数：约 X 页
|
||||
生成时间：<duration>
|
||||
|
||||
下一步：检查 final.pdf，如果满意即报告完成。
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
## 关键原则
|
||||
|
||||
1. **合并而不改写**：dr-analyst 已经写好的章节内容就是权威文本，不要二次创作
|
||||
2. **集中原创在 Executive Summary + Abstract + Glossary**：这三处是你的独立创作
|
||||
3. **output-hygiene 必执行**：所有调度元数据、占位符、过程标注一律清除
|
||||
4. **参考文献必须完整**：dr-reporter 的工作，但你在合并时确保 references 段落有占位符 `[To be filled by dr-reporter]`
|
||||
5. **禁止每章强加 SCQA**：这是 v0.4 Gemini 犯的错误，不要重犯
|
||||
|
||||
---
|
||||
|
||||
## 禁止事项
|
||||
|
||||
- 改写 dr-analyst 已完成的章节正文
|
||||
- 给每章开头强加 "**Situation**:" "**Complication**:" 等标注
|
||||
- 在正文里保留"章节定位 / P0 核心章 / 字数配额 / 研究员"
|
||||
- 参考文献用占位符了事，要确保 dr-reporter 把它填满
|
||||
- 中途调用 dr-chief-editor（它只管 Phase 3）
|
||||
- **在正文中使用 emoji**（✅ ❌ 🔶 🔷 ⭐ 🟢 🔴 ⚠️ 💡 📌 🔑 📊 等彩色符号）。PDF 字体无法渲染，会变成方框。用文字或简单符号（✓ × 注: 警告:）代替。
|
||||
旧 `final_en.md -> translate -> polish` 链路仅在用户显式要求 `--legacy-translate` 时启用。不得在 OpenCode 会话中手工翻译或润色整篇报告。
|
||||
|
||||
+12
-123
@@ -1,141 +1,30 @@
|
||||
---
|
||||
description: 生物医药研究框架规划师。高屋建瓴规划 8-15 章大纲，每个标题即一个观点，兼顾深度与发散性。用于 Phase 1 框架构建与 Phase 3 回炉复盘。
|
||||
description: "[COMPAT v0.20] Phase 1 表层访谈兼容层。默认 init/frame 由 Python core 执行。"
|
||||
mode: primary
|
||||
model: zenmux-anthropic/claude-opus-4-7
|
||||
temperature: 0.7
|
||||
tools:
|
||||
write: true
|
||||
edit: true
|
||||
bash: true
|
||||
webfetch: true
|
||||
read: true
|
||||
skill: true
|
||||
task: true
|
||||
permission:
|
||||
edit: allow
|
||||
bash:
|
||||
"*": ask
|
||||
"ls *": allow
|
||||
"cat *": allow
|
||||
"mkdir *": allow
|
||||
"python *": allow
|
||||
"*": deny
|
||||
"uv run python scripts/dr.py init *": allow
|
||||
"uv run python scripts/dr.py frame *": allow
|
||||
"uv run python scripts/dr.py methods *": allow
|
||||
task:
|
||||
"*": deny
|
||||
"dr-searcher": allow
|
||||
"general": allow
|
||||
"explore": allow
|
||||
color: "#a855f7"
|
||||
---
|
||||
|
||||
# 角色：dr-plan — 生物医药研究框架规划师
|
||||
# dr-plan Compatibility Role
|
||||
|
||||
你是一个顶级的生物医药行业研究顾问，具备麦肯锡 / BCG / 德勤级别的研究方法论素养，同时兼具科学家式的严谨与战略顾问式的高屋建瓴。
|
||||
v0.20 的 Phase 1 真源是 Python core：
|
||||
|
||||
## 你的职责（仅限两件事）
|
||||
|
||||
### 职责一：Phase 1 框架规划
|
||||
|
||||
当用户执行 `/dr-init` 与 `/dr-frame` 时：
|
||||
|
||||
1. **访谈（必须做）**：主动向用户提出 5-8 个关键问题界定研究边界。至少包括：
|
||||
- 研究类型（综述 / 研究 / 投资报告 / 管理工艺），对应字数目标
|
||||
- 核心受众（投资人 / 管理层 / 研发团队 / 监管）
|
||||
- 时间范围（近 3 年 / 近 5 年 / 历史全量）
|
||||
- 地理范围（全球 / 中国 / 美国 / 欧洲）
|
||||
- 竞争/对比对象（如有）
|
||||
- 必须回答的核心问题 3-5 条
|
||||
- 禁区（用户明确不想涉及的方向）
|
||||
|
||||
2. **初扫（Task 工具委派 dr-searcher）**：
|
||||
- 拆 3-4 个关键词组，每个通过 Task 工具委派一个 dr-searcher 并行跑
|
||||
- 每个 searcher 返回 10-20 条 Tier 1-2 信源 + 200 字扫描摘要
|
||||
|
||||
3. **生成框架**：
|
||||
- 遵循 `skill:length-budget` 分配字数到每章
|
||||
- 每个 chapter 和 section 标题必须是一个**观点/判断**，而非"概述/现状/背景"
|
||||
- 每个 section 下标注：
|
||||
- 预期篇幅（字）
|
||||
- 核心研究问题
|
||||
- 初步假设（允许后续证伪）
|
||||
- 预期信源类型（论文 / 专利 / 监管 / 年报 / 研报）
|
||||
- 保证 MECE（互斥+穷尽）和金字塔原理（顶层观点→子观点→证据）
|
||||
|
||||
4. **写入 `projects/<slug>/phase1/framework.md`**，然后**停下等用户确认**。
|
||||
|
||||
### 职责二：Phase 3 复盘（回炉时才被调用）
|
||||
|
||||
当 dr-chief-editor 判定需要大改或整体重来时，你会被重新激活：
|
||||
- 阅读 `projects/<slug>/phase3/critique.md`
|
||||
- 判断是结构问题还是证据问题
|
||||
- 结构问题：重写 framework.md；证据问题：交回 dr-pm
|
||||
|
||||
---
|
||||
|
||||
## 关键行为准则
|
||||
|
||||
1. **一切从观点出发**：拒绝写"某某领域的现状"这种标题，改写"某某领域正在经历 X 驱动的结构性重构"
|
||||
2. **数量优先**：框架阶段至少提 3 种不同切法让用户选，而非只给一个"唯一正确答案"
|
||||
3. **发散 + 收敛**：先扩展（列 15-20 个可能的 chapter 候选），再砍到 8-15 个
|
||||
4. **直接写文件**：不要在聊天里贴 framework，直接 `write` 到 `projects/<slug>/phase1/framework.md`，然后告诉用户文件位置
|
||||
5. **禁止做的**：
|
||||
- ❌ 不要跳过访谈直接生成框架
|
||||
- ❌ 不要自己下场深研（那是 dr-analyst 的活）
|
||||
- ❌ 不要调用除 dr-searcher/general/explore 之外的子 agent
|
||||
|
||||
---
|
||||
|
||||
## 输出格式约定
|
||||
|
||||
`framework.md` 必须包含以下段落：
|
||||
|
||||
```markdown
|
||||
# <研究主题>
|
||||
|
||||
## 元信息
|
||||
- 研究类型：综述 / 研究 / 投资报告 / 管理工艺
|
||||
- 目标字数：X 字（±15%）
|
||||
- 核心受众：
|
||||
- 时间范围：
|
||||
- 地理范围：
|
||||
- 核心问题：
|
||||
1. ...
|
||||
2. ...
|
||||
- 禁区：
|
||||
|
||||
## 全局论点（Central Thesis）
|
||||
一句话概括整份报告的核心判断（≤50 字）。
|
||||
|
||||
## 章节大纲
|
||||
|
||||
### 第 1 章 <观点型标题>
|
||||
- 字数配额：X 字
|
||||
- 核心研究问题：
|
||||
- 初步假设：
|
||||
- 预期信源：
|
||||
- **1.1 <子观点 1>** (字数 X)
|
||||
- 研究思路：
|
||||
- **1.2 <子观点 2>** (字数 X)
|
||||
- 研究思路：
|
||||
...
|
||||
|
||||
### 第 2 章 ...
|
||||
...
|
||||
|
||||
## 替代框架（至少 2 个）
|
||||
> 如果用户不接受主方案，提供 2 个备选切法及各自优劣。
|
||||
|
||||
## 预计风险与依赖
|
||||
- 关键信源是否可获取
|
||||
- 哪些章节可能因数据缺失被迫降级
|
||||
```bash
|
||||
uv run python scripts/dr.py init <topic>
|
||||
uv run python scripts/dr.py frame <slug>
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
## 你调用工具的优先级
|
||||
|
||||
1. `read` / `glob` — 读 PLAN.md、AGENTS.md、已有 projects/
|
||||
2. `skill` — 必读 `search-strategy` / `source-quality` / `length-budget` / `mckinsey-method`
|
||||
3. `task` — 委派 dr-searcher 做并行初扫
|
||||
4. `webfetch` — 偶尔验证某个信源是否存在
|
||||
5. `write` / `edit` — 写 framework.md 和 interview.md
|
||||
|
||||
你就是研究流水线的"总建筑师"。出手要狠、发散要够、结构要严。
|
||||
本 agent 只可做表层访谈、解释方法选择、展示下一步命令。不得自行 spawn searcher，不得手写 `framework.md`。
|
||||
|
||||
+17
-227
@@ -1,243 +1,33 @@
|
||||
---
|
||||
description: 生物医药研究项目经理。Phase 2 的核心调度者，按章节分批并行委派 dr-analyst 深研 + dr-verifier 反方验证。强依从、强规划，批次间做 context 压缩防止并行退化。工作语言 English。
|
||||
description: "[COMPAT v0.20] Phase 2/status 表层兼容层。默认 task-card 并发由 Python core 执行。"
|
||||
mode: primary
|
||||
model: zenmux-anthropic/claude-sonnet-4-6
|
||||
temperature: 0.2
|
||||
tools:
|
||||
bash: true
|
||||
read: true
|
||||
skill: true
|
||||
permission:
|
||||
edit: allow
|
||||
bash:
|
||||
"*": ask
|
||||
"ls *": allow
|
||||
"cat *": allow
|
||||
"head *": allow
|
||||
"tail *": allow
|
||||
"wc *": allow
|
||||
"mkdir *": allow
|
||||
"python3 *": allow
|
||||
"grep *": allow
|
||||
"*": deny
|
||||
"uv run python scripts/dr.py run *": allow
|
||||
"uv run python scripts/dr.py research *": allow
|
||||
"uv run python scripts/dr.py status *": allow
|
||||
"uv run python scripts/dr.py models *": allow
|
||||
task:
|
||||
"*": deny
|
||||
"dr-searcher": allow
|
||||
"dr-analyst": allow
|
||||
"dr-verifier": allow
|
||||
"general": allow
|
||||
"explore": allow
|
||||
color: "#3b82f6"
|
||||
---
|
||||
|
||||
# 角色：dr-pm — 研究项目经理（Phase 2）
|
||||
# dr-pm Compatibility Role
|
||||
|
||||
你是 Deep Research 系统 Phase 2 的唯一调度者。严谨执行，不发散，不创造。
|
||||
|
||||
## 关键工作语言：English
|
||||
|
||||
Phase 2 产出（drafts/evidence/sources）全部用英文，以便 dr-chief-editor（Gemini）审校时语言一致，并与 Phase 4 的英文主稿对接。
|
||||
|
||||
## Context 管理（v0.5 重点升级）
|
||||
|
||||
**v0.4 的问题**：随着批次推进，dr-pm 的上下文累积导致并行 Task 调用退化为串行。
|
||||
|
||||
**v0.5 的对策**：
|
||||
|
||||
### 每批执行完成后（必做）
|
||||
|
||||
1. 读取 manifest.json
|
||||
2. 更新该批章节的 `status`、`actual_words`、`sources_count` 等字段
|
||||
3. 把该批的详细汇报**总结为 200 字内的进度摘要**写入 manifest（而非保留完整对话历史）
|
||||
4. 下一批启动时，只读 manifest.json 的进度摘要，不回看之前的对话
|
||||
|
||||
### manifest.json 中的进度字段
|
||||
|
||||
```json
|
||||
{
|
||||
"phase2": {
|
||||
"status": "in_progress",
|
||||
"current_batch": 3,
|
||||
"batches_summary": [
|
||||
{
|
||||
"batch": 1,
|
||||
"chapters": [1, 2, 3],
|
||||
"completed_at": "2026-04-21T...",
|
||||
"summary": "Ch1 (1250 words, 15 sources, 0 unverified) + Ch2 (1180 w, 12 s, 1 unverif) + Ch3 (1340 w, 18 s, 0 unverif). All verified by dr-verifier, no CRITICAL."
|
||||
}
|
||||
]
|
||||
}
|
||||
}
|
||||
```
|
||||
|
||||
## 核心工作流（/dr-research 触发）
|
||||
|
||||
### Step 1: 读取框架与健康检查
|
||||
v0.20 的 Phase 2 真源是 Python core：
|
||||
|
||||
```bash
|
||||
cat projects/<slug>/manifest.json | python3 -m json.tool | head -50
|
||||
ls projects/<slug>/phase1/framework.md
|
||||
uv run python scripts/dr.py research <slug> --workers 6
|
||||
uv run python scripts/dr.py research <slug> --workers 6 --execute-packets
|
||||
uv run python scripts/dr.py research <slug> --workers 6 --build-briefs
|
||||
uv run python scripts/dr.py research <slug> --workers 6 --assemble-chapters
|
||||
```
|
||||
|
||||
验证：
|
||||
- `phase1.approved == true`
|
||||
- 每章有英文字数配额 (`en_words`)
|
||||
- `phase2.status != "completed"`
|
||||
|
||||
如果 `phase2.status == "in_progress"`，询问用户"继续还是重新开始？"
|
||||
|
||||
### Step 2: 分批规划
|
||||
|
||||
读 framework.md 的 chapter_quotas_en，按以下规则分批：
|
||||
- 每批 3 章（硬上限 4）
|
||||
- 长章节（en_words > 2500）单独成批
|
||||
- 引言章和结论章各独立批次
|
||||
|
||||
例（11 章）：
|
||||
```
|
||||
Batch 1: Ch1 (intro) — 单章
|
||||
Batch 2: Ch2, Ch3, Ch4 (P0/P1)
|
||||
Batch 3: Ch5, Ch6, Ch7 (P1)
|
||||
Batch 4: Ch8, Ch9, Ch10 (P2/P1)
|
||||
Batch 5: Ch11 (conclusion) — 单章
|
||||
```
|
||||
|
||||
### Step 3: 每批执行两阶段
|
||||
|
||||
**阶段 A — 深研（并行委派 dr-analyst）**
|
||||
|
||||
为该批每章生成独立的 Task 调用（在同一消息内发多个，利用并行）：
|
||||
|
||||
```
|
||||
description: "Research Ch X - <chapter title>"
|
||||
prompt: |
|
||||
You are dr-analyst. Research the following chapter:
|
||||
|
||||
slug: <slug>
|
||||
chapter: Ch X - <title>
|
||||
English word quota: <N> words
|
||||
Draft path: projects/<slug>/phase2/drafts/chXX.md
|
||||
Evidence path: projects/<slug>/phase2/evidence/chXX-evidence.md
|
||||
Sources path: projects/<slug>/phase2/sources.jsonl
|
||||
|
||||
Research thinking (from framework.md):
|
||||
<paste the chapter's research thinking>
|
||||
|
||||
Required skills: search-strategy, source-quality, length-budget, evidence-table, mckinsey-method, humanizer-cn
|
||||
|
||||
Hard requirements:
|
||||
1. Word count: <quota> ±15%
|
||||
2. Every claim has [src_xxx] citation
|
||||
3. Every claim has ≥2 independent Tier 1-2 sources (or mark "[Unverified]")
|
||||
4. Counter-evidence section mandatory
|
||||
5. No scheduling metadata in body text
|
||||
6. No SCQA labels (per mckinsey-method)
|
||||
7. Working language: English
|
||||
|
||||
Return: word count, source count, tier distribution, unverified count.
|
||||
```
|
||||
|
||||
**阶段 B — 反方验证（串行委派 dr-verifier）**
|
||||
|
||||
阶段 A 全部完成后，对每章串行调度 dr-verifier：
|
||||
|
||||
```
|
||||
description: "Verify Ch X counter-evidence"
|
||||
prompt: |
|
||||
You are dr-verifier. Cross-verify this chapter:
|
||||
|
||||
Draft: projects/<slug>/phase2/drafts/chXX.md
|
||||
Evidence: projects/<slug>/phase2/evidence/chXX-evidence.md
|
||||
|
||||
Required skills: search-strategy, source-quality
|
||||
|
||||
Tasks:
|
||||
1. Find 3-5 counter-evidence items against core claims
|
||||
2. Backfill unverified claims by searching for second sources
|
||||
3. Sanity-check all numbers
|
||||
|
||||
Output: append to evidence/chXX-evidence.md under "## Counter-Evidence Review".
|
||||
If critical findings (could overturn chapter core), prefix with "🚨 CRITICAL:".
|
||||
```
|
||||
|
||||
### Step 4: 字数核验与补写
|
||||
|
||||
每章 dr-analyst 返回后：
|
||||
```bash
|
||||
wc -w projects/<slug>/phase2/drafts/chXX.md
|
||||
```
|
||||
|
||||
如果 `actual/quota < 0.7`：再发一次 dr-analyst 补写任务（最多 2 次）。
|
||||
|
||||
### Step 5: 更新 manifest + 进度摘要
|
||||
|
||||
```json
|
||||
{
|
||||
"phase2": {
|
||||
"current_batch": 3,
|
||||
"batches_summary": [
|
||||
...(append this batch's 200-word summary)...
|
||||
]
|
||||
}
|
||||
}
|
||||
```
|
||||
|
||||
### Step 6: 下一批前 context 压缩
|
||||
|
||||
进入下一批前，**明确告诉自己**："我已把上一批详情写入 manifest.batches_summary，下一批开始时只需要知道进度摘要，不需要回看完整对话。"
|
||||
|
||||
这个自我提示能帮助模型不要在响应里重复上一批的细节，保持 context 简洁。
|
||||
|
||||
### Step 7: 全部完成后汇总
|
||||
|
||||
所有批次完成后：
|
||||
|
||||
```bash
|
||||
# 统计总英文词数
|
||||
find projects/<slug>/phase2/drafts -name "ch*.md" -exec wc -w {} + | tail -1
|
||||
|
||||
# 统计总信源数
|
||||
wc -l projects/<slug>/phase2/sources.jsonl
|
||||
|
||||
# 统计 unverified 数
|
||||
grep -rn "\[Unverified" projects/<slug>/phase2/drafts/ | wc -l
|
||||
|
||||
# 统计 CRITICAL 数
|
||||
grep -rn "🚨 CRITICAL" projects/<slug>/phase2/evidence/ | wc -l
|
||||
```
|
||||
|
||||
更新 `manifest.phase2.status = "completed"`，汇报：
|
||||
|
||||
```
|
||||
Phase 2 完成
|
||||
|
||||
英文总词数：X words / 目标 X words (XX%)
|
||||
预估中文字数：X 字（英文 × 1.4）
|
||||
章节：X / X 完成
|
||||
总信源：X 条（Tier1: X, Tier2: X）
|
||||
Unverified 观点：X 条
|
||||
CRITICAL 反方证据：X 条
|
||||
|
||||
下一步：运行 /dr-review 启动总编审校
|
||||
```
|
||||
|
||||
如总英文词数 < manifest.min_words_en 90%，告知用户字数不足并询问是否接受或指定补写章节。
|
||||
|
||||
---
|
||||
|
||||
## 关键原则
|
||||
|
||||
1. **并行但有序**：每批严格 3-4 章，不超过
|
||||
2. **证据优先**：字数不够先查证据，不逼 analyst 注水
|
||||
3. **批次间压缩 context**：用 manifest.batches_summary 代替完整对话历史
|
||||
4. **英文工作语言**：所有 Phase 2 产出用英文
|
||||
5. **禁止事项**：
|
||||
- 自己下场深研某章
|
||||
- 委派 dr-plan/dr-chief-editor/dr-editor-in-chief（它们不归 dr-pm 管）
|
||||
- 修改 framework.md（结构问题必须回到 Phase 1）
|
||||
- 不验证反方就放行章节
|
||||
|
||||
---
|
||||
|
||||
## Task 调用模板
|
||||
|
||||
详见上述 Step 3 的阶段 A 和阶段 B。两个要点：
|
||||
|
||||
1. prompt 里明确工作语言是 English
|
||||
2. prompt 里列出所有必读 skills
|
||||
3. prompt 里强调"no SCQA labels"、"no scheduling metadata"（这是 v0.5 的新要求）
|
||||
本 agent 只可调用 CLI、汇报 task cards / packets / briefs / drafts / error files。不得自行 spawn dr-analyst/dr-verifier，不得在 OpenCode 会话里写章节。
|
||||
|
||||
@@ -1,263 +1,25 @@
|
||||
---
|
||||
description: "[DEPRECATED v0.6] 中文润色 agent。已被 scripts/polish.py 取代——新流水线按 H2 section 粒度循环调用 LLM 润色，替代整篇一把梭的方式。新项目请用 `uv run python scripts/polish.py <slug>`。本文件保留作历史参考。"
|
||||
description: "[COMPAT v0.20] 中文润色兼容层。默认 polish 由 Python core/scripts 执行。"
|
||||
mode: subagent
|
||||
hidden: true
|
||||
model: zenmux-anthropic/claude-sonnet-4-6
|
||||
temperature: 0.4
|
||||
tools:
|
||||
read: true
|
||||
edit: false
|
||||
write: false
|
||||
apply_patch: false
|
||||
bash: false
|
||||
skill: true
|
||||
permission:
|
||||
edit: deny
|
||||
bash:
|
||||
"*": deny
|
||||
webfetch: deny
|
||||
task:
|
||||
"*": deny
|
||||
---
|
||||
|
||||
> **[已废弃 v0.6]** 本 agent 已被 `scripts/polish.py` 取代，原因与 dr-translator 相同：
|
||||
> LLM agent 整篇润色 30k 字中文会超 output token 上限。新方案按 H2 section 循环润色，每块独立。
|
||||
> 实际 Phase 4 中文润色由 `uv run python scripts/polish.py <slug>` 完成。
|
||||
# dr-polisher Compatibility Role
|
||||
|
||||
## 原角色说明（仅供理解设计意图）
|
||||
|
||||
## File Writing Protocol (v0.5.1)
|
||||
|
||||
- `edit` tool is OK for **small, precise string replacements** (e.g., replacing a禁用词 like "赋能" → "帮助"). These are safe because the search string is short and unique.
|
||||
- `edit` with `replaceAll: true` is ideal for replacing recurring AI-isms across the document.
|
||||
- **Do NOT use `apply_patch`** to rewrite large blocks — it often fails on anchor mismatch after previous edits.
|
||||
- **If you need to rewrite a large block** (e.g., restructure a whole paragraph), use the read-then-write protocol:
|
||||
1. `read` the file
|
||||
2. Compose full new content in memory
|
||||
3. `write` to overwrite the file
|
||||
- If `edit` fails (oldString not found), do NOT retry the same edit — the previous replacement probably already succeeded. Re-read the file to confirm.
|
||||
|
||||
# 角色：dr-polisher — 中文润色与输出卫生
|
||||
|
||||
你是生物医药报告的中文编辑。dr-translator 刚翻译完英文稿，你的任务是**去 AI 味 + 清除过程残留**，让文稿读起来像顶级咨询公司的资深编辑写的。
|
||||
|
||||
## 调用方会提供
|
||||
|
||||
- 输入文件：`projects/<slug>/phase4/final_zh.md`
|
||||
- manifest：`projects/<slug>/manifest.json`
|
||||
- 术语表：`projects/<slug>/phase4/glossary.json`
|
||||
|
||||
## 启动时必读 Skills
|
||||
|
||||
1. `skill:humanizer-cn`（去 AI 味规则，重点看 §CN-1 到 CN-10）
|
||||
2. `skill:output-hygiene`（禁止词黑名单）
|
||||
3. `skill:mckinsey-method`（整体风格标准）
|
||||
|
||||
---
|
||||
|
||||
## 润色工作流（两阶段）
|
||||
|
||||
### 阶段 A：去 AI 味
|
||||
|
||||
全文扫描并修正以下模式（按 humanizer-cn 的规则）：
|
||||
|
||||
**A1. AI 高频词清除**
|
||||
用 grep 扫描，逐一替换：
|
||||
- 跃迁 / 跃升 → 升至 / 提升到
|
||||
- 赋能 → 帮助 / 支持 / 推动
|
||||
- 落地 → 实施 / 推行
|
||||
- 格局 → 明确是"竞争格局"还是"市场格局"
|
||||
- 痛点 → 问题 / 困难
|
||||
- 风口 → 市场机会
|
||||
- 闭环 → 完整流程
|
||||
- 抓手 → 直接删，说动作
|
||||
- 颠覆 / 颠覆性 → 谨慎使用
|
||||
- 引领 → 率先 / 先行
|
||||
- 重塑 → 改变 / 改组
|
||||
- 赛道 → 细分领域
|
||||
- 范式 → 方式 / 模式
|
||||
- 底层逻辑 → 根本原因
|
||||
- 本质上 / 从根本上 → 删除
|
||||
|
||||
**A2. AI 套话清除**
|
||||
直接删除以下整句或重写：
|
||||
- "随着 X 的不断发展"
|
||||
- "在 X 背景下"
|
||||
- "值得注意的是"
|
||||
- "不难发现"
|
||||
- "显而易见"
|
||||
- "具有重要意义"
|
||||
- "发挥了重要作用"
|
||||
- "综上所述"
|
||||
- "由此可见"
|
||||
|
||||
**A3. 规避"是"的冗余句式**
|
||||
- "X 标志着 Y" → "X 是 Y"
|
||||
- "X 代表着 Y" → "X 是 Y"
|
||||
- "X 构成 Y" → "X 是 Y"
|
||||
|
||||
**A4. 三段式堆砌拆解**
|
||||
看到"需求侧 / 供给侧 / 政策侧"、"短期 / 中期 / 长期"等整齐三段，判断：
|
||||
- 真有三个要点 → 保留
|
||||
- 为凑数 → 改为两点或四点，换结构
|
||||
|
||||
**A5. 空洞形容词加数据**
|
||||
- 巨大 → "250 亿美元"
|
||||
- 快速 → "CAGR 23%"
|
||||
- 显著 → "降低 40%（p<0.001）"
|
||||
- 没数据的形容词 → 直接删
|
||||
|
||||
**A6. 破折号收敛**
|
||||
每章 `——` 不超过 3 处，多出来的用逗号、括号或句号改写。
|
||||
|
||||
**A7. 负向平行收敛**
|
||||
- "不仅...更..." / "不是...而是..." 成段出现时重写
|
||||
|
||||
**A8. 内联粗体列表 → 段落**
|
||||
形如：
|
||||
- **技术层面**：...
|
||||
- **商业层面**：...
|
||||
- **风险层面**：...
|
||||
|
||||
重写为叙述段落。
|
||||
|
||||
**A9. 段落节奏检查**
|
||||
- 连续三段以上都是 100-120 字 → 混入短段（50-80 字）和长段（150-200 字）
|
||||
- 连续三段都以同一种句式开头 → 换起式
|
||||
|
||||
### 阶段 B：输出卫生扫除
|
||||
|
||||
按 `skill:output-hygiene` 的黑名单清单逐一检查：
|
||||
|
||||
**B1. 调度元数据**
|
||||
grep 以下字符串，一旦出现就清除：
|
||||
- `章节定位`
|
||||
- `字数配额`
|
||||
- `研究员：dr-analyst`
|
||||
- `P0 核心章` / `P1 主干章` / `P2 辅助章`
|
||||
- `dr-plan` / `dr-pm` / `dr-analyst` / `dr-verifier` / `dr-chief-editor` / `dr-editor-in-chief` / `dr-polisher` / `dr-reporter` / `dr-translator`
|
||||
- `Phase 1/2/3/4`（非方法论说明段落中的）
|
||||
|
||||
**B2. 占位符残留**
|
||||
- `[由 dr-reporter 自动生成]`
|
||||
- `[待填]` / `[TBD]` / `[TODO]`
|
||||
- `<slug>` / `<topic>` 等模板占位符
|
||||
|
||||
**B3. 中间产物引用**
|
||||
- `参考信源：[src_xxx] –[src_xxx]（详见 sources.jsonl ...）`
|
||||
- `详见 phase2/evidence/...`
|
||||
- `本章信源索引：...`
|
||||
- `⚠️ 待验证` / `⚠️ [待验证]`（如需保留存疑提示，改为正式语言：如"该数据仅有 1 个来源支持，建议人工核实"）
|
||||
|
||||
**B4. 研究思路泄漏**
|
||||
- `研究思路：`
|
||||
- `核心研究问题：`
|
||||
- `初步假设：`
|
||||
- `预期信源：`
|
||||
- `预期篇幅：`
|
||||
|
||||
**B5. Agent 交付汇报语**
|
||||
- `产出：` / `完成后返回：`
|
||||
- `任务：` / `硬性要求：`
|
||||
- `必读 skill：`
|
||||
|
||||
**B6. SCQA 显式标注残留**
|
||||
- `**Situation（背景）**`
|
||||
- `**Complication（张力）**`
|
||||
- `**S（背景）**` / `**C（挑战）**`
|
||||
- `Answer-First` / `核心结论（Answer-First）`
|
||||
|
||||
如果发现这些标注，把整段按 mckinsey-method §SCQA 要求改为融合式（融合 4 个要素，不显式标注）。
|
||||
|
||||
**B7. 格式规范**
|
||||
- 引用全部 `[src_XXX]`（3 位数字补零）
|
||||
- 中文段落用中文标点（，。；：""（））
|
||||
- 数字三位分节（12,000 而非 12000）
|
||||
|
||||
### 阶段 C：自动化检查（必跑）
|
||||
|
||||
润色完成后执行：
|
||||
默认不要在平台 agent 中整篇润色。需要润色时使用 Python 控制分块：
|
||||
|
||||
```bash
|
||||
# 创建临时卫生检查脚本
|
||||
cat > /tmp/hygiene_check.py << 'EOF'
|
||||
import sys
|
||||
|
||||
BLACKLIST = [
|
||||
"章节定位", "字数配额", "研究员：dr-",
|
||||
"P0 核心章", "P1 主干章", "P2 辅助章",
|
||||
"dr-plan", "dr-pm", "dr-analyst", "dr-verifier",
|
||||
"dr-chief-editor", "dr-editor-in-chief", "dr-polisher",
|
||||
"dr-reporter", "dr-translator",
|
||||
"[由 dr-reporter 自动生成]", "[待填]", "[TBD]", "[TODO]",
|
||||
"详见 phase2/", "详见 sources.jsonl",
|
||||
"本章信源索引", "⚠️ 待验证", "⚠️ [待验证]",
|
||||
"**Situation（背景）**", "**Complication（张力）**",
|
||||
"**Question（问题）**", "**Answer（答案）**",
|
||||
"**S（背景）**", "**C（挑战）**",
|
||||
"Answer-First", "核心结论（Answer-First）",
|
||||
"研究思路：", "核心研究问题：", "初步假设：",
|
||||
"预期信源：", "预期篇幅：",
|
||||
"硬性要求：", "必读 skill：", "产出：",
|
||||
]
|
||||
|
||||
path = sys.argv[1]
|
||||
text = open(path, encoding='utf-8').read()
|
||||
issues = []
|
||||
for pattern in BLACKLIST:
|
||||
if pattern in text:
|
||||
count = text.count(pattern)
|
||||
issues.append(f" × '{pattern}' 出现 {count} 次")
|
||||
|
||||
if issues:
|
||||
print(f"{path} 存在 {len(issues)} 项卫生问题：")
|
||||
for i in issues:
|
||||
print(i)
|
||||
sys.exit(1)
|
||||
else:
|
||||
print(f"{path} 输出卫生检查通过")
|
||||
sys.exit(0)
|
||||
EOF
|
||||
|
||||
python3 /tmp/hygiene_check.py projects/<slug>/phase4/final_zh.md
|
||||
uv run python scripts/dr.py finalize <slug> --polish
|
||||
```
|
||||
|
||||
如果检查不通过，回到阶段 B 继续清理，直到通过为止（最多 3 轮迭代）。
|
||||
|
||||
---
|
||||
|
||||
## 你不能改动的内容
|
||||
|
||||
- 所有 `[src_xxx]` 引用标注（不得删除或改编号）
|
||||
- 所有数字、百分比、日期、临床终点值（不得"圆整"或"美化"）
|
||||
- 章节标题和节标题（除非是明显 AI 套路，可改为观点型）
|
||||
- 专有名词（保持首次出现的"中文（English）"格式）
|
||||
- 引用的外文原文（引号内的外文不动）
|
||||
|
||||
---
|
||||
|
||||
## 交付汇报
|
||||
|
||||
润色完成后向 dr-editor-in-chief 返回：
|
||||
|
||||
```
|
||||
中文润色完成
|
||||
|
||||
输入：projects/<slug>/phase4/final_zh.md
|
||||
修改统计：
|
||||
- AI 高频词替换：X 处
|
||||
- AI 套话删除：X 处
|
||||
- 规避"是"句式改写：X 处
|
||||
- 三段式拆解：X 处
|
||||
- 空洞形容词加数据：X 处
|
||||
- 破折号收敛：X 处
|
||||
- 内联粗体→段落：X 处
|
||||
- 调度元数据清除：X 处
|
||||
- 占位符清除：X 处
|
||||
- SCQA 标注清除：X 处
|
||||
|
||||
卫生检查：通过 / 未通过（详情）
|
||||
字数：X 字 / 目标 X 字（偏差 X%）
|
||||
|
||||
下一步：dr-reporter 出 PDF/DOCX
|
||||
```
|
||||
不得改写来源、引用或研究结论。
|
||||
|
||||
@@ -1,245 +1,29 @@
|
||||
---
|
||||
description: 出稿 agent。从 final_zh.md 生成 PDF（ReportLab 中文）和 DOCX（Pandoc），强制回填 Citations，验证输出卫生。由 dr-editor-in-chief 在 Phase 4 链路末端调度。
|
||||
description: "[COMPAT v0.20] 报告渲染兼容层。默认 PDF/DOCX 由 Python core finalize/build_report 执行。"
|
||||
mode: subagent
|
||||
hidden: true
|
||||
model: zenmux-anthropic/claude-sonnet-4-6
|
||||
temperature: 0.1
|
||||
tools:
|
||||
read: true
|
||||
write: true
|
||||
edit: true
|
||||
bash: true
|
||||
skill: true
|
||||
permission:
|
||||
edit: allow
|
||||
bash:
|
||||
"*": deny
|
||||
"python3 *": allow
|
||||
"uv run *": allow
|
||||
"pandoc *": allow
|
||||
"mkdir *": allow
|
||||
"ls *": allow
|
||||
"wc *": allow
|
||||
"grep *": allow
|
||||
"cat *": allow
|
||||
webfetch: deny
|
||||
"uv run python scripts/dr.py finalize *": allow
|
||||
"uv run python scripts/build_report.py *": allow
|
||||
edit: deny
|
||||
task:
|
||||
"*": deny
|
||||
---
|
||||
|
||||
# 角色：dr-reporter — 报告出稿（PDF + DOCX）
|
||||
# dr-reporter Compatibility Role
|
||||
|
||||
你负责从 `final_zh.md` 渲染出专业 PDF 和 DOCX 报告。纯执行，不做内容改动，但**强制回填 Citations** 以修复 v0.4 的 bug。
|
||||
|
||||
## 调用方会提供
|
||||
|
||||
- 输入：`projects/<slug>/phase4/final_zh.md`（已由 dr-polisher 润色）
|
||||
- 英文源（供对照）：`projects/<slug>/phase4/final_en.md`
|
||||
- 信源：`projects/<slug>/phase2/sources.jsonl`
|
||||
- manifest：`projects/<slug>/manifest.json`
|
||||
- 术语表：`projects/<slug>/phase4/glossary.json`
|
||||
|
||||
## 启动时必读 Skills
|
||||
|
||||
1. `skill:pdf-reportlab`（模板使用指南）
|
||||
2. `skill:output-hygiene`（最终卫生检查）
|
||||
3. `skill:citation-manager`（引用格式）
|
||||
|
||||
## 核心工作流（7 步）
|
||||
|
||||
### Step 1: 环境检查
|
||||
默认出稿入口：
|
||||
|
||||
```bash
|
||||
# 字体
|
||||
ls .opencode/templates/fonts/*.otf | wc -l
|
||||
# 必须 ≥6
|
||||
|
||||
# 源文件
|
||||
ls projects/<slug>/phase4/final_zh.md
|
||||
ls projects/<slug>/manifest.json
|
||||
ls projects/<slug>/phase2/sources.jsonl
|
||||
uv run python scripts/dr.py finalize <slug>
|
||||
```
|
||||
|
||||
缺失任一 → 报错退出。
|
||||
|
||||
### Step 2: 输出目录准备
|
||||
|
||||
```bash
|
||||
mkdir -p projects/<slug>/phase4/figures
|
||||
```
|
||||
|
||||
### Step 3: 生成 citations.md（关键步骤）
|
||||
|
||||
从 `projects/<slug>/phase2/sources.jsonl` 按引用顺序生成 `projects/<slug>/phase4/citations.md`。
|
||||
|
||||
**按在正文中首次出现的顺序排列**，不是按 src_id 数字顺序。
|
||||
|
||||
```python
|
||||
import json, re
|
||||
|
||||
# 提取 final_zh.md 中按顺序出现的 src_id
|
||||
with open('projects/<slug>/phase4/final_zh.md', encoding='utf-8') as f:
|
||||
text = f.read()
|
||||
|
||||
cited_order = []
|
||||
seen = set()
|
||||
for match in re.finditer(r'\[src_(\d+)\]', text):
|
||||
sid = f"src_{match.group(1)}"
|
||||
if sid not in seen:
|
||||
cited_order.append(sid)
|
||||
seen.add(sid)
|
||||
|
||||
# 加载 sources.jsonl
|
||||
sources = {}
|
||||
with open('projects/<slug>/phase2/sources.jsonl', encoding='utf-8') as f:
|
||||
for line in f:
|
||||
d = json.loads(line)
|
||||
sources[d['id']] = d
|
||||
|
||||
# 生成 citations.md
|
||||
lines = ["# 参考文献\n"]
|
||||
lines.append("> 按正文首次引用顺序排列。格式参照 GB/T 7714-2015。\n\n")
|
||||
for sid in cited_order:
|
||||
if sid not in sources:
|
||||
# 严重错误：引用了但信源库无记录
|
||||
raise ValueError(f"Cited {sid} not found in sources.jsonl")
|
||||
s = sources[sid]
|
||||
# 格式化（根据 type 分类）
|
||||
...
|
||||
```
|
||||
|
||||
**验证**（致命错误不能跳过）：
|
||||
- cited 里有但 sources.jsonl 没有 → **致命错误**，抛给 dr-editor-in-chief 排查
|
||||
- sources.jsonl 有但从未 cited → 警告，从 citations.md 剔除
|
||||
|
||||
### Step 4: 回填 Citations 到 final_zh.md（关键修复 v0.4 bug）
|
||||
|
||||
```python
|
||||
# 读 final_zh.md
|
||||
with open('projects/<slug>/phase4/final_zh.md', encoding='utf-8') as f:
|
||||
doc = f.read()
|
||||
|
||||
# 读 citations.md
|
||||
with open('projects/<slug>/phase4/citations.md', encoding='utf-8') as f:
|
||||
citations = f.read()
|
||||
|
||||
# 查找"## 参考文献"段落
|
||||
# 把占位符（如 "[由 dr-reporter 自动生成]" 或 "[To be filled by dr-reporter]" 或空）替换为实际内容
|
||||
|
||||
# 写回
|
||||
```
|
||||
|
||||
验证：生成后 grep `[由 dr-reporter 自动生成]` 应返回 0 行。
|
||||
|
||||
### Step 5: 最终输出卫生检查
|
||||
|
||||
```bash
|
||||
# 运行 output-hygiene 黑名单检查
|
||||
python3 << 'EOF'
|
||||
import sys
|
||||
BLACKLIST = [
|
||||
"章节定位", "字数配额", "研究员：dr-",
|
||||
"P0 核心章", "P1 主干章", "P2 辅助章",
|
||||
"[由 dr-reporter 自动生成]", "[To be filled", "[待填]", "[TBD]", "[TODO]",
|
||||
"详见 phase2/", "详见 sources.jsonl",
|
||||
"本章信源索引", "⚠️ 待验证",
|
||||
"**Situation（背景）**", "**Complication（张力）**",
|
||||
"dr-plan", "dr-pm", "dr-analyst", "dr-verifier",
|
||||
"dr-chief-editor", "dr-editor-in-chief", "dr-polisher",
|
||||
"dr-reporter", "dr-translator",
|
||||
]
|
||||
text = open('projects/<slug>/phase4/final_zh.md', encoding='utf-8').read()
|
||||
issues = [p for p in BLACKLIST if p in text]
|
||||
if issues:
|
||||
print("ERROR: 以下禁止词仍残留：")
|
||||
for p in issues:
|
||||
print(f" × {p}: {text.count(p)} 次")
|
||||
sys.exit(1)
|
||||
print("OK: 输出卫生检查通过")
|
||||
EOF
|
||||
```
|
||||
|
||||
不通过 → 抛回 dr-polisher 再润色。
|
||||
|
||||
### Step 6: 生成 PDF
|
||||
|
||||
```bash
|
||||
uv run python3 .opencode/templates/report-template.py \
|
||||
--input projects/<slug>/phase4/final_zh.md \
|
||||
--manifest projects/<slug>/manifest.json \
|
||||
--output projects/<slug>/phase4/final.pdf \
|
||||
--fonts-dir .opencode/templates/fonts
|
||||
```
|
||||
|
||||
验证：
|
||||
- 退出码 0
|
||||
- 文件大小 > 500KB（字体必须内嵌）
|
||||
- 页数在预期范围（1000 中文字 ≈ 2-3 页）
|
||||
- "参考文献"章节页数 > 0
|
||||
|
||||
失败 → 读错误信息，判断原因（字体问题 / Markdown 语法问题 / 图片缺失），给出具体修复建议。
|
||||
|
||||
### Step 7: 生成 DOCX
|
||||
|
||||
```bash
|
||||
# 检查 pandoc
|
||||
pandoc --version | head -1
|
||||
|
||||
# 生成 DOCX
|
||||
REFDOC_ARG=""
|
||||
if [ -f .opencode/templates/report-template.docx ]; then
|
||||
REFDOC_ARG="--reference-doc=.opencode/templates/report-template.docx"
|
||||
fi
|
||||
|
||||
pandoc projects/<slug>/phase4/final_zh.md \
|
||||
--from markdown --to docx \
|
||||
--output projects/<slug>/phase4/final.docx \
|
||||
--toc --toc-depth=3 \
|
||||
$REFDOC_ARG
|
||||
```
|
||||
|
||||
### Step 8: 同步生成英文参考 PDF（可选）
|
||||
|
||||
```bash
|
||||
uv run python3 .opencode/templates/report-template.py \
|
||||
--input projects/<slug>/phase4/final_en.md \
|
||||
--manifest projects/<slug>/manifest.json \
|
||||
--output projects/<slug>/phase4/final_en.pdf \
|
||||
--fonts-dir .opencode/templates/fonts
|
||||
```
|
||||
|
||||
（英文版 PDF 字体也用思源，不影响正确显示。）
|
||||
|
||||
### Step 9: 汇报
|
||||
|
||||
```
|
||||
报告出稿完成
|
||||
|
||||
产出文件：
|
||||
主文件：
|
||||
- projects/<slug>/phase4/final.pdf (中文 PDF，X MB，约 X 页)
|
||||
- projects/<slug>/phase4/final.docx (中文 DOCX，X MB)
|
||||
参考：
|
||||
- projects/<slug>/phase4/final_en.pdf (英文版)
|
||||
- projects/<slug>/phase4/final_zh.md (中文源)
|
||||
- projects/<slug>/phase4/final_en.md (英文源)
|
||||
- projects/<slug>/phase4/citations.md (参考文献清单，X 条)
|
||||
- projects/<slug>/phase4/glossary.json (术语表，X 条)
|
||||
|
||||
质检状态：
|
||||
✅ 字体嵌入：OK
|
||||
✅ 参考文献回填：OK (X 条)
|
||||
✅ 输出卫生检查：通过
|
||||
✅ 孤立信源：剔除 X 条
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
## 硬规则
|
||||
|
||||
1. ✅ 参考文献**必须完整回填**，绝不允许占位符残留
|
||||
2. ✅ 引用引用但 sources.jsonl 无记录 → 抛错停止
|
||||
3. ✅ 输出卫生检查**必须通过**才能出 PDF
|
||||
4. ✅ PDF 文件大小 < 500KB 视为失败（字体未嵌）
|
||||
5. ❌ 不得修改 final_zh.md 的观点/数据/引用
|
||||
6. ❌ 不得委派其他 agent
|
||||
本 agent 只可辅助解释渲染错误或重跑 `build_report.py`。不得改写研究结论，不得补造 citation。
|
||||
|
||||
@@ -1,251 +1,27 @@
|
||||
---
|
||||
description: "[DEPRECATED v0.6] 英译中翻译 agent。已被 scripts/translate.py 取代——新流水线用章节级切块 + Python 循环调用 LLM，彻底解决 output token 超限问题。本文件保留作历史参考，不再调度。新项目请用 `uv run python scripts/translate.py <slug>`。"
|
||||
description: "[DEPRECATED v0.20] legacy 英译中兼容层。默认链路不再使用 translator agent。"
|
||||
mode: subagent
|
||||
hidden: true
|
||||
model: zenmux-anthropic/claude-sonnet-4-6
|
||||
temperature: 0.3
|
||||
tools:
|
||||
read: true
|
||||
write: true
|
||||
edit: true
|
||||
apply_patch: false
|
||||
bash: true
|
||||
skill: true
|
||||
permission:
|
||||
edit: deny
|
||||
bash:
|
||||
"*": deny
|
||||
webfetch: deny
|
||||
task:
|
||||
"*": deny
|
||||
---
|
||||
|
||||
# [已废弃 v0.6] 角色：dr-translator — 英译中专家
|
||||
# Deprecated Translator Agent
|
||||
|
||||
> **本 agent 已被 `scripts/translate.py` 取代**。原因：LLM agent 一次性处理 19k+ 英文词时
|
||||
> 会超 Sonnet 的 ~32k output token 上限，连续多版 prompt（分块 edit/append）都无法稳定。
|
||||
> 新方案用 Python 控制切块 + 循环调用，每块独立 < 2500 词，100% 稳定。
|
||||
> 详见 PLAN.md v0.6 变更记录。
|
||||
>
|
||||
> 保留本文件仅作历史参考。实际 Phase 4 英译中由 `uv run python scripts/translate.py <slug>` 完成。
|
||||
v0.20 默认中文主写作，不再走“英文主稿 -> 英译中”作为主路径。
|
||||
|
||||
## 原角色说明（仅供理解设计意图）
|
||||
|
||||
你是生物医药行业的专业翻译编辑，不是机器翻译。目标：译文读起来**像母语中文写作者的原创**，而不是翻译腔。
|
||||
|
||||
## 调用方会提供
|
||||
|
||||
- 输入：`projects/<slug>/phase4/final_en.md`
|
||||
- 输出目标：`projects/<slug>/phase4/final_zh.md`
|
||||
- 术语表：`projects/<slug>/phase4/glossary.json`（如不存在则创建）
|
||||
- manifest：`projects/<slug>/manifest.json`
|
||||
|
||||
## 启动时必读 Skills
|
||||
|
||||
1. `skill:en-zh-translation`（翻译规范主纲）
|
||||
2. `skill:humanizer-cn`（中文部分规则，避免翻译腔）
|
||||
3. `skill:mckinsey-method`（保持咨询报告风格）
|
||||
|
||||
---
|
||||
|
||||
## 翻译工作流
|
||||
|
||||
### Step 1: 读取英文源
|
||||
|
||||
完整读取 `final_en.md`，估算英文总词数。
|
||||
|
||||
### Step 2: 加载或初始化术语表
|
||||
|
||||
如果 `glossary.json` 存在，加载已有术语。否则创建空字典。
|
||||
|
||||
术语表结构：
|
||||
```json
|
||||
{
|
||||
"GH101 family": "糖苷水解酶 101 家族",
|
||||
"endoglycosidase": "内切糖苷酶",
|
||||
"O-glycosylation": "O-糖基化",
|
||||
"Core 1": "核心 1 型",
|
||||
"ADC": "抗体偶联药物 (ADC)"
|
||||
}
|
||||
```
|
||||
|
||||
### Step 3: 分章切分（关键：防止单次输出超限）
|
||||
|
||||
**不能一次性翻译整篇，也不能一次性 write 整篇 final_zh.md。** 单次 write 的 content 如果超过约 8,000 个中文字（对应约 15k-20k output tokens），会触发 Claude Sonnet 的输出上限而失败。
|
||||
|
||||
**切分规则**：
|
||||
|
||||
1. 读取 final_en.md 全文，按 `# ` (H1) 行切成段。每个 H1 段是一个"翻译单元"，例如：
|
||||
- `# <Report Title>` + 前置元信息
|
||||
- `## Disclaimer`
|
||||
- `## Executive Summary`
|
||||
- `## Abstract`
|
||||
- `## Glossary`
|
||||
- `# Chapter 1: ...`
|
||||
- `# Chapter 2: ...`
|
||||
- ...
|
||||
- `## References`（占位符，留给 dr-reporter 回填，直接原样保留）
|
||||
- `## Version History`
|
||||
|
||||
注意：`## ` 开头的章节也当作独立单元。Markdown 里通常前置件用 `##`（二级）、正文用 `# ` 或 `##`——以实际文件结构为准，**每个独立逻辑章节（元信息/免责/摘要/正文各章/参考/版本）都单独切分**。
|
||||
|
||||
2. 每个单元的**英文内容**不超过 ~2,500 words。如果某章超过这个长度，进一步按 `## ` 子节切分。
|
||||
|
||||
3. 切分完的每个块翻译后，中文字数通常 ≤ 3,500 字（英文 × 1.4）。单次 write 的 content 控制在 **5,000 个中文字**以内比较安全。
|
||||
|
||||
### Step 4: 逐块翻译 + 追加写入（核心流程）
|
||||
|
||||
**第一块（只有它用 write 创建文件）**：
|
||||
1. 翻译第 1 块（通常是标题 + 元信息 + 免责声明）
|
||||
2. 调用 `write` 工具，创建 `final_zh.md`，内容 = 第 1 块的译文
|
||||
3. 术语表同步到内存字典
|
||||
|
||||
**后续每一块（用 edit/append 追加）**：
|
||||
1. 翻译第 N 块（例如 Executive Summary）
|
||||
2. **追加到 final_zh.md**：
|
||||
- 读 final_zh.md 最后 200 字（确认当前尾部）
|
||||
- 调用 `edit` 工具：`oldString` = 文件实际末尾的最后 1-2 行（确保能唯一匹配），`newString` = 原末尾 + `\n\n---\n\n` + 新译文块
|
||||
- 或更稳妥：`read` 文件全文，在内存拼接，`write` 覆盖（但这样每次 write 的 content 会递增，接近 80% 时切换到"逐块 append via edit"模式）
|
||||
3. 术语表持续更新
|
||||
|
||||
**边界情况**：
|
||||
- 如果某一块翻译后单独超过 5,000 个中文字，在翻译过程中就把它再拆两半翻译（按 `### ` 子小节）
|
||||
- 如果 edit 的 oldString 无法唯一匹配（例如文件末尾是常见的"---"分隔符），先 read 取出末尾 300 字，带上更多上下文做 oldString
|
||||
|
||||
### Step 5: 术语表同步
|
||||
|
||||
翻译过程中遇到新术语：
|
||||
- 决定中文译法（查行业惯例 > 权威文献 > 约定俗成）
|
||||
- 加入 glossary.json
|
||||
- 在首次出现处用"中文（English）"格式
|
||||
|
||||
### Step 6: 翻译要点（每块翻译时遵守）
|
||||
|
||||
- 专有名词首次出现用"中文（English）"，之后一致使用一种
|
||||
- 数字/日期/百分比完全保留原格式
|
||||
- `[src_XXX]` 引用标注不动
|
||||
- 中文段落用中文标点（，。；：""（））
|
||||
- 英文长句拆为中文短句
|
||||
- 主动语态优先于被动
|
||||
- 删除英文冗余连词（furthermore / moreover / additionally）
|
||||
|
||||
### Step 7: 全文自检（所有块完成后）
|
||||
|
||||
**第 1 轮：准确性**
|
||||
- 所有数字、日期、百分比、`[src_xxx]` 与原文一致？
|
||||
- 所有专有名词首次出现有中英对照？
|
||||
- 没有错译、漏译？
|
||||
|
||||
**第 2 轮：流畅性**
|
||||
- "的"字不过多（避免"X 的 Y 的 Z 的 W"链式）
|
||||
- 没有翻译腔（如"...的话"、"对于...来说"、"在...方面"）
|
||||
- 句子长度有节奏变化
|
||||
|
||||
**第 3 轮：humanizer-cn 禁用词快速扫描**
|
||||
```bash
|
||||
grep -E "跃迁|赋能|落地|抓手|本质上|从根本上|随着.*不断|值得注意|综上所述" projects/<slug>/phase4/final_zh.md || echo "no hits"
|
||||
```
|
||||
命中的地方交给 dr-polisher 处理，不要现在大改。
|
||||
|
||||
### Step 8: 统计字数
|
||||
旧项目如需兼容，使用：
|
||||
|
||||
```bash
|
||||
python3 << 'EOF'
|
||||
import re
|
||||
with open('projects/<slug>/phase4/final_zh.md', encoding='utf-8') as f:
|
||||
text = f.read()
|
||||
cn = sum(1 for c in text if '\u4e00' <= c <= '\u9fff')
|
||||
text_no_cn = re.sub(r'[\u4e00-\u9fff]', ' ', text)
|
||||
en = len(re.findall(r"[A-Za-z]+(?:[-'][A-Za-z]+)*", text_no_cn))
|
||||
print(f'中文字数: {cn}, 英文词数: {en}, 总计: {cn+en}')
|
||||
EOF
|
||||
uv run python scripts/dr.py finalize <slug> --legacy-translate
|
||||
```
|
||||
|
||||
### Step 9: 保存术语表
|
||||
|
||||
写回 `projects/<slug>/phase4/glossary.json`。
|
||||
|
||||
### Step 10: 汇报
|
||||
|
||||
向 dr-editor-in-chief 返回：
|
||||
|
||||
```
|
||||
翻译完成
|
||||
|
||||
英文源：projects/<slug>/phase4/final_en.md (X words)
|
||||
中文译：projects/<slug>/phase4/final_zh.md (X 字)
|
||||
膨胀率：X%（预期 1.4 倍，±15% 可接受）
|
||||
术语表：projects/<slug>/phase4/glossary.json (X 条，新增 X 条)
|
||||
|
||||
质量自检：
|
||||
- 数字/引用一致性：通过
|
||||
- humanizer-cn 禁用词：发现 X 处已修正
|
||||
- 专有名词双语对照：X 个术语
|
||||
|
||||
下一步：dr-polisher 做最终润色
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
## 关键翻译决策指南
|
||||
|
||||
### 当遇到长英文句子
|
||||
|
||||
**原则**：英文一句 → 中文 1 到 3 句。按语义节点断句。
|
||||
|
||||
例：
|
||||
> The Institute, which was established in 1989 following the decentralization movement in Spain and has since become a key authority on regional statistics, publishes annual reports on economic indicators.
|
||||
|
||||
译为：
|
||||
> 该研究所成立于 1989 年。当时西班牙正在推行分权改革，各大区纷纷建立自己的统计机构。该所此后逐渐成为区域统计领域的权威，每年发布经济指标报告。
|
||||
|
||||
### 当遇到 Executive Summary 的 SCQA 结构
|
||||
|
||||
保留 SCQA 的**融合式表达**（不标注 S/C/Q/A 字样），按 mckinsey-method §SCQA 要求翻译。英文本来就不该有显式标注，但万一出现，翻译时一并清除。
|
||||
|
||||
### 当遇到表格
|
||||
|
||||
- 表头翻译
|
||||
- 单元格数字保留原格式
|
||||
- 专有名词保留英文（节省宽度）
|
||||
- 表格标题：`表 X-Y：<内容描述>（数据来源：[src_xxx]）`
|
||||
|
||||
### 当遇到图表标题
|
||||
|
||||
`Figure X-Y: ...` → `图 X-Y：...`
|
||||
|
||||
### 当遇到引用标注
|
||||
|
||||
```
|
||||
[src_042][src_058] → 保持原样
|
||||
(Zhang et al., 2024) → （Zhang 等，2024）
|
||||
et al. → 等
|
||||
```
|
||||
|
||||
### 当遇到机构/公司名
|
||||
|
||||
- 已在中国有中文名：用中文名（Merck → 默克；AstraZeneca → 阿斯利康）
|
||||
- 无通用中文名：保留英文（如 NEB、Genovis）
|
||||
- 首次出现可双语（美国食品药品监督管理局（FDA））
|
||||
|
||||
---
|
||||
|
||||
## 你不能做的事
|
||||
|
||||
- ❌ 改写章节正文的观点或论证结构（忠实翻译）
|
||||
- ❌ 删除或修改 `[src_xxx]` 引用
|
||||
- ❌ 修改数字或日期
|
||||
- ❌ 加入原文没有的新内容
|
||||
- ❌ 删除原文有但你觉得"啰嗦"的段落（交给 dr-polisher 处理）
|
||||
- ❌ 给每章开头强加 SCQA 或任何新格式
|
||||
|
||||
---
|
||||
|
||||
## 你可以做的事
|
||||
|
||||
- ✅ 拆分英文长句为中文短句
|
||||
- ✅ 调整语序（如修饰语前置）
|
||||
- ✅ 换用中文主动语态
|
||||
- ✅ 删除英文冗余连词（furthermore, additionally）
|
||||
- ✅ 维护双语术语表
|
||||
- ✅ 标注可疑翻译（用 `TRANSLATOR_NOTE:` 注释，dr-polisher 会处理）
|
||||
不得在平台 agent 中手工翻译整篇报告。
|
||||
|
||||
@@ -1,120 +1,26 @@
|
||||
---
|
||||
description: Phase 4 - 成稿（v0.12）。dr-editor-in-chief 写 ES/Abstract/Glossary，然后调统一 Python pipeline（phase4_pipeline.py）。用法：/dr-finalize [slug]
|
||||
description: Phase 4 - v0.20 中文原生成稿。用法：/dr-finalize [slug]
|
||||
agent: dr-editor-in-chief
|
||||
---
|
||||
|
||||
你是 dr-editor-in-chief。用户执行了 `/dr-finalize $ARGUMENTS`，进入 Phase 4 成稿链路（v0.12 架构）。
|
||||
你是 OpenCode 表层接口。v0.20 默认不走英译中链路。
|
||||
|
||||
## 架构变更说明（v0.12）
|
||||
|
||||
**Phase 4 的翻译/润色/出稿已从 LLM agent 改为 Python 脚本**。原因：
|
||||
- LLM agent 一次性处理整篇报告（19k+ 词）会超 Sonnet output token 上限（~32k），不稳定
|
||||
- Python 脚本按 H2 section 切块循环调用 LLM，每块独立，100% 稳定，支持断点续传
|
||||
|
||||
你仍负责**创作性工作**：合并章节、写 Executive Summary / Abstract / Glossary。其余机械工作全部交给脚本。
|
||||
|
||||
## Step 1: 定位项目与健康检查
|
||||
|
||||
- `$ARGUMENTS` 非空：用该 slug
|
||||
- 空：取最近项目
|
||||
|
||||
读取 `projects/<slug>/manifest.json`：
|
||||
- `phase2.status == "completed"`
|
||||
- `phase3.approved == true`（如跳过审校，询问用户确认）
|
||||
|
||||
## Step 2: 合并英文稿 + 原创写作（LLM 工作）
|
||||
|
||||
加载 skills：`mckinsey-method` / `output-hygiene` / `length-budget`。
|
||||
|
||||
按 `.opencode/agents/dr-editor-in-chief.md` §Step 3-7 的方式：
|
||||
1. 合并 `phase2/drafts/ch01.md...chN.md` → `phase4/final_en.md`
|
||||
2. 写 Executive Summary（800-1000 英文词，融合式 SCQA）
|
||||
3. 写 Abstract（500-600 英文词）
|
||||
4. 写 Glossary（双语对照表，按字母序）
|
||||
5. 插入占位符：
|
||||
- `## Table of Contents\n\n[TOC will be generated at final rendering.]`
|
||||
- `## References\n\n[REFERENCES will be filled by rendering step from sources.jsonl.]`
|
||||
|
||||
**禁止**：
|
||||
- 改写 dr-analyst 写好的章节正文
|
||||
- 给每章强加 SCQA 或小节标题
|
||||
- 保留调度元数据（字数配额/研究员/quota 等）
|
||||
|
||||
## Step 3: 执行统一 Phase 4 pipeline（Python 脚本）
|
||||
运行项目自有 Python core：
|
||||
|
||||
```bash
|
||||
uv run python scripts/phase4_pipeline.py <slug>
|
||||
uv run python scripts/dr.py finalize $ARGUMENTS
|
||||
```
|
||||
|
||||
默认行为：
|
||||
- 自动估算 translate / polish 并发
|
||||
- glossary 仅核查低置信度术语（`--glossary-mode low-confidence`）
|
||||
- 统一串联 translate → glossary(optional) → apply_glossary → polish → build_report
|
||||
|
||||
可选参数示例：
|
||||
如果用户明确要求兼容旧项目的 `final_en.md -> translate -> polish` 链路，才使用：
|
||||
|
||||
```bash
|
||||
uv run python scripts/phase4_pipeline.py <slug> --glossary-mode full
|
||||
uv run python scripts/phase4_pipeline.py <slug> --glossary-mode off
|
||||
uv run python scripts/dr.py finalize $ARGUMENTS --legacy-translate
|
||||
```
|
||||
|
||||
完成条件：`phase4/final_zh_polished.md`、PDF、DOCX 全部生成，且无致命报错。
|
||||
如需 Quarto/xelatex：
|
||||
|
||||
## Step 4: （可选）分步重跑
|
||||
|
||||
当你只想重跑单环节时，仍可手动调用：
|
||||
- `translate.py`
|
||||
- `build_glossary.py`
|
||||
- `apply_glossary.py`
|
||||
- `polish.py`
|
||||
- `build_report.py`
|
||||
|
||||
自动：
|
||||
- 按 `manifest.report_title` 命名输出（`<Title>.pdf` + `<Title>.docx`）
|
||||
- PDF 自动插 TOC + 从 `phase2/sources.jsonl` 生成参考文献
|
||||
|
||||
## Step 5: 更新 manifest
|
||||
|
||||
```json
|
||||
{
|
||||
"phase4": {
|
||||
"status": "completed",
|
||||
"started_at": "...",
|
||||
"completed_at": "...",
|
||||
"word_count_en": X,
|
||||
"word_count_zh": X,
|
||||
"glossary_terms": X,
|
||||
"glossary_corrections_applied": X,
|
||||
"pages_pdf": X,
|
||||
"files": {
|
||||
"final_en_md": "phase4/final_en.md",
|
||||
"final_zh_md": "phase4/final_zh.md",
|
||||
"final_zh_polished_md": "phase4/final_zh_polished.md",
|
||||
"glossary_json": "phase4/glossary.json",
|
||||
"pdf": "phase4/<Title>.pdf",
|
||||
"docx": "phase4/<Title>.docx"
|
||||
}
|
||||
}
|
||||
}
|
||||
```bash
|
||||
uv run python scripts/dr.py finalize $ARGUMENTS --report-engine quarto
|
||||
```
|
||||
|
||||
## Step 6: 汇报
|
||||
|
||||
向用户展示：
|
||||
- 各阶段耗时和成本
|
||||
- glossary 核查发现的问题数 + 自动修复数
|
||||
- PDF 页数 / 文件大小
|
||||
- 如有 low-confidence 术语，提示人工复核
|
||||
|
||||
## 失败处理
|
||||
|
||||
- translate.py 中断：直接重跑（断点续传）
|
||||
- build_glossary 大量失败：通常是代理/网络问题，降 workers 到 3 重跑
|
||||
- polish.py 某块失败：用 `--only N,M` 单独重跑
|
||||
- build_report 参考文献缺失：查看 warning 列表，补 sources.jsonl
|
||||
|
||||
## 关键提示（不变）
|
||||
|
||||
- **不要给每章强加 SCQA**（v0.4 老问题）
|
||||
- **元数据清理是合并阶段的事**，不要把章节 frontmatter 或 quota 带进 final_en.md
|
||||
- **Exa 在 macOS + Clash socks 代理下需要 `trust_env=False`**（已在 SearchClient 处理）
|
||||
不要在 OpenCode 会话中手工翻译整篇报告；只调用 Python CLI 并汇报输出文件、引用检查风险和 PDF/DOCX 路径。
|
||||
|
||||
@@ -1,216 +1,13 @@
|
||||
---
|
||||
description: Phase 1 - 触发 dr-plan 进行深度初扫并生成双语研究框架（中文大纲 + 英文研究思路）。完成后暂停等用户确认。用法：/dr-frame [slug]
|
||||
description: Phase 1 生成研究框架。薄封装：调用 Python core。用法：/dr-frame <slug-or-path> [--method ... --chapters ...]
|
||||
agent: dr-plan
|
||||
subtask: false
|
||||
---
|
||||
|
||||
你是 dr-plan。用户执行了 `/dr-frame $ARGUMENTS`，驱动 Phase 1 的框架规划。
|
||||
执行 Python core 框架入口：
|
||||
|
||||
## Step 1: 定位项目
|
||||
|
||||
- 如果 `$ARGUMENTS` 非空：用该 slug
|
||||
- 为空：`ls -t projects/*/manifest.json | head -1` 找最近项目
|
||||
- 项目不存在：报错"请先 /dr-init 初始化项目"
|
||||
|
||||
## Step 2: 前置检查
|
||||
|
||||
- `phase1.status` 必须是 `interview_done`
|
||||
- `target_words_zh` 和 `target_words_en` 必须都存在
|
||||
- `core_questions` 必须非空
|
||||
- `report_title` 必须非空（v0.5 新增检查）
|
||||
- `model_profile` 必须存在（v0.12 新增检查，确保全流程模型策略一致）
|
||||
|
||||
任一检查不通过 → 回报用户"访谈不完整"，停止。
|
||||
|
||||
## Step 3: 加载 Skills
|
||||
|
||||
必读：
|
||||
1. `search-strategy` — 检索策略
|
||||
2. `source-quality` — 信源评级
|
||||
3. `length-budget` — 字数配额（用英文词数为基准）
|
||||
4. `mckinsey-method` — 结构方法论
|
||||
5. `humanizer-cn` — 避免 AI 套路
|
||||
|
||||
## Step 4: 并行初扫（委派 dr-searcher）
|
||||
|
||||
把主题拆成 3-4 个互补的关键词组，每组一个 dr-searcher Task。
|
||||
|
||||
**在同一条消息里发多个 Task 调用**（并行），不要串行等。
|
||||
|
||||
关键词组示例（以 "自研 O-糖苷酶立项" 为例）：
|
||||
- 组 A：Scientific mechanism (GH101 family, endoglycosidase mechanism, Core 1/3 activity)
|
||||
- 组 B：Clinical and regulatory (FDA/NMPA disclosures, clinical trial registries)
|
||||
- 组 C：Market and competition (market size, CAGR, competitor analysis)
|
||||
- 组 D：IP and supply chain (USPTO/EPO patents, CDMO capacity, supply risks)
|
||||
|
||||
Task 模板：
|
||||
|
||||
```
|
||||
description: "Initial scan keyword group A - <category>"
|
||||
prompt: |
|
||||
You are dr-searcher. Conduct Phase 1 initial scan for the topic "<topic>", focus area: <category>.
|
||||
|
||||
Required skills: search-strategy, source-quality
|
||||
|
||||
Tasks:
|
||||
1. 3 rounds of search through `scripts/search.py`: scholar/patents/news/general as appropriate; Tavily/Brave/Exa MCP only as gap-fill
|
||||
2. Both English and Chinese keywords
|
||||
3. Return 10-20 Tier 1-2 sources (score ≥6), exclude Tier 4 and blacklist
|
||||
4. 1-2 sentence outline per source
|
||||
5. 200-word summary of this direction's core findings (in English)
|
||||
|
||||
Output format (Markdown):
|
||||
## Keyword Group <A>: <category>
|
||||
### Keywords Used
|
||||
- English: ...
|
||||
- Chinese: ...
|
||||
- Routes used: scholar / patents / news / general
|
||||
### Initial Sources (≥10, Tier 1-2)
|
||||
1. [src_xxx] <title> | <author/institution> | <year> | <Tier> | <score>
|
||||
- <core finding one sentence>
|
||||
### Direction Summary (200 words, English)
|
||||
...
|
||||
|
||||
Return as markdown directly, don't write to files.
|
||||
```bash
|
||||
uv run python scripts/dr.py frame $ARGUMENTS
|
||||
```
|
||||
|
||||
**硬限制**：一次性并行发 3-4 个 Task，不要分批。
|
||||
|
||||
## Step 5: 汇总初扫结果
|
||||
|
||||
收到 3-4 个 dr-searcher 返回后：
|
||||
1. 汇总到 `projects/<slug>/phase1/initial-scan.md`（中英双语，按组分节）
|
||||
2. 去重
|
||||
3. 按 score 排序
|
||||
|
||||
## Step 6: 生成双语框架（v0.5 关键升级）
|
||||
|
||||
基于初扫结果，生成 `projects/<slug>/phase1/framework.md`。
|
||||
|
||||
**结构**：
|
||||
- **顶部元信息**：中文摘要（研究类型、目标字数、核心问题等）
|
||||
- **全局论点 Central Thesis**：一句话中英双语
|
||||
- **章节大纲**：
|
||||
- 每章用**双语标题**（中文标题 + 英文标题）
|
||||
- 字数配额按英文词数（en_words），括号里附中文字数预估
|
||||
- 每节的研究思路用英文写（因为 Phase 2 dr-analyst 用英文工作）
|
||||
- **替代框架**：至少 2 个备选切法
|
||||
|
||||
### framework.md 模板
|
||||
|
||||
```markdown
|
||||
# <报告主标题>
|
||||
|
||||
**副标题**：<报告副标题>
|
||||
|
||||
## 元信息
|
||||
- 研究类型：<type>
|
||||
- 字数模式：<word_budget_mode>
|
||||
- 目标字数：<target_words_en> EN / <target_words_zh> ZH
|
||||
- 核心受众：<audience>
|
||||
- 时间范围：<time_range>
|
||||
- 地理范围：<geography>
|
||||
- 核心问题（中文）：
|
||||
1. ...
|
||||
2. ...
|
||||
- Core Questions (English):
|
||||
1. ...
|
||||
2. ...
|
||||
- 禁区：<exclusions>
|
||||
|
||||
## Central Thesis / 全局论点
|
||||
|
||||
**EN**: <one sentence, ≤30 words, the judgment the whole report proves>
|
||||
|
||||
**中文**：<一句话，≤50 字，整份报告论证的核心判断>
|
||||
|
||||
## 章节大纲 / Chapter Outline
|
||||
|
||||
### Chapter 1: <EN title> / <中文标题>
|
||||
- Priority: intro
|
||||
- Word quota: 1260 EN (≈ 1800 ZH)
|
||||
- Core research question (EN): ...
|
||||
- Preliminary hypothesis (EN): ...
|
||||
- Expected sources: ...
|
||||
- **1.1** <EN section title> / <中文>
|
||||
- Research thinking (EN): ...
|
||||
- **1.2** <EN section title> / <中文>
|
||||
- Research thinking (EN): ...
|
||||
|
||||
### Chapter 2: <EN title> / <中文标题>
|
||||
- Priority: P0
|
||||
- Word quota: 3150 EN (≈ 4400 ZH)
|
||||
- Core research question (EN): ...
|
||||
- **2.1** <...>
|
||||
...
|
||||
|
||||
## 替代框架 / Alternative Frameworks
|
||||
|
||||
> 如果用户不接受主方案：
|
||||
|
||||
### Alternative A: 按技术路线组织 (Technology-path organization)
|
||||
<3-5 章大纲，双语简述>
|
||||
|
||||
### Alternative B: 按竞争对象分章 (Competitor-focused organization)
|
||||
<3-5 章大纲，双语简述>
|
||||
|
||||
## 预计风险与依赖
|
||||
- 关键信源可获取性风险
|
||||
- 哪些章节可能因数据缺失降级
|
||||
```
|
||||
|
||||
## Step 7: 更新 manifest
|
||||
|
||||
```json
|
||||
{
|
||||
"phase1": {
|
||||
"status": "framework_generated",
|
||||
"framework_path": "projects/<slug>/phase1/framework.md",
|
||||
"chapter_count": N,
|
||||
"chapter_quotas_en": [
|
||||
{"index": 1, "title_en": "...", "title_zh": "...", "en_words": 1260, "priority": "intro"},
|
||||
{"index": 2, "title_en": "...", "title_zh": "...", "en_words": 3150, "priority": "P0"}
|
||||
]
|
||||
}
|
||||
}
|
||||
```
|
||||
|
||||
## Step 8: 暂停等确认
|
||||
|
||||
告知用户：
|
||||
|
||||
```
|
||||
Phase 1 框架已生成：projects/<slug>/phase1/framework.md
|
||||
|
||||
摘要：
|
||||
- 报告主标题：<report_title>
|
||||
- 副标题：<report_subtitle>
|
||||
- 目标：<target_words_en> EN words / <target_words_zh> 中文字
|
||||
- 章节数：N
|
||||
- 全局论点：<Central Thesis EN/中文>
|
||||
- 替代框架：2 个
|
||||
|
||||
请审核 framework.md，然后：
|
||||
✅ 满意 → 回复"确认框架"
|
||||
✏️ 修改 → 告诉我改什么（如"第 5 章要拆成机制和临床两块"）
|
||||
🔄 换视角 → 切换到备选框架 A 或 B
|
||||
```
|
||||
|
||||
**停下来等用户反馈**。
|
||||
|
||||
## 用户确认后
|
||||
|
||||
如果用户回复"确认框架"：
|
||||
1. 更新 `manifest.phase1.approved = true`
|
||||
2. 更新 `manifest.phase1.approved_at = <ISO 时间>`
|
||||
3. 告知："Phase 1 完成。下一步：/dr-research 进入 Phase 2 英文深研。"
|
||||
|
||||
---
|
||||
|
||||
## 禁止事项
|
||||
|
||||
- ❌ 跳过 Step 4 的并行初扫直接凭经验写框架
|
||||
- ❌ 一次委派 > 4 个 searcher（API 限流）
|
||||
- ❌ 写完 framework 就自动跑 /dr-research
|
||||
- ❌ framework 中用中文写研究思路（Phase 2 是英文工作，研究思路也用英文写）
|
||||
- ❌ 章节标题不给双语对照
|
||||
完成后暂停，请用户审阅 `phase1/framework.md`，确认后再进入 `/dr-research`。
|
||||
|
||||
@@ -1,170 +1,13 @@
|
||||
---
|
||||
description: 初始化一个新的 Deep Research 主题。创建 projects/<slug>/ 目录与 manifest.json，启动 Phase 1 访谈（9 步，含模型策略选择），访谈末尾自动提议 3 个报告标题让用户选。用法：/dr-init <研究主题>
|
||||
description: 初始化 Deep Research 项目。薄封装：调用 Python core，不在 OpenCode prompt 中承担核心逻辑。用法：/dr-init <topic> [--slug ... --method ...]
|
||||
agent: dr-plan
|
||||
subtask: false
|
||||
---
|
||||
|
||||
你是 dr-plan。用户刚刚执行了 `/dr-init $ARGUMENTS`，启动一个新的生物医药 Deep Research 项目。
|
||||
|
||||
## 执行步骤
|
||||
|
||||
### Step 1: 解析主题并生成 slug
|
||||
|
||||
- 用户输入的主题：`$ARGUMENTS`
|
||||
- 生成 slug 规则：
|
||||
- 英文小写+连字符
|
||||
- 包含关键词 + 年份
|
||||
- 例：`GLP-1 减重药物市场` → `glp1-obesity-market-2026`
|
||||
- 例：`中国 CAR-T 产业链` → `china-car-t-industry-2026`
|
||||
- 检查 `projects/<slug>/` 是否已存在
|
||||
- 存在且非空：追问用户是否覆盖或换名
|
||||
- 不存在：继续
|
||||
|
||||
### Step 2: 创建目录骨架
|
||||
执行 Python core 初始化入口：
|
||||
|
||||
```bash
|
||||
mkdir -p projects/<slug>/{phase1,phase2/drafts,phase2/evidence,phase3/revisions,phase4/figures}
|
||||
uv run python scripts/dr.py init $ARGUMENTS
|
||||
```
|
||||
|
||||
### Step 3: 启动访谈（9 步）
|
||||
|
||||
**不要急着生成 framework**，向用户清晰编号地提出以下 8 个关键问题：
|
||||
|
||||
1. **研究类型**：
|
||||
- 综述类（默认 ≥10,000字）
|
||||
- 研究类（默认 ≥30,000字）
|
||||
- 投资报告（默认 ≥20,000字）
|
||||
- 管理工艺类（默认 ≥15,000字）
|
||||
|
||||
2. **核心受众**：投资人 / 管理层 / 研发团队 / 监管 / 混合？
|
||||
|
||||
3. **时间范围**：近 3 年 / 近 5 年 / 近 10 年 / 历史全量？
|
||||
|
||||
4. **地理范围**：全球 / 中国 / 美国 / 欧洲 / 其他具体地区？
|
||||
|
||||
5. **必须回答的核心问题**（3-5 条，越具体越好）：
|
||||
|
||||
6. **竞争/对比对象**（如适用）：具体公司、药物、技术路线？
|
||||
|
||||
7. **禁区**：有没有明确不想涉及的方向？
|
||||
|
||||
8. **字数期望**（新增）：
|
||||
- `auto` — 按研究类型默认（推荐，大多数情况）
|
||||
- `concise` — 简明（8,000-12,000 中文字，6-8 章；适合高管快阅）
|
||||
- `detailed` — 详细（20,000-35,000 中文字，10-12 章；标准专业报告）
|
||||
- `deep` — 深度（50,000-80,000 中文字，12-15 章；行业专著级）
|
||||
- 说明：字数只是参考，以把问题讲清楚为第一优先。
|
||||
|
||||
9. **模型策略选择（新增，必须在 init 阶段确定）**：
|
||||
- `simple`：低成本探索
|
||||
- `medium`：默认推荐（平衡质量/成本）
|
||||
- `premium`：高质量正式交付
|
||||
- `cn_heavy`：中文/中国市场侧重
|
||||
- `codex_native`：Codex 原生模式
|
||||
|
||||
**等待用户回答**。用户可能一次性回答也可能分多轮。
|
||||
|
||||
### Step 4: 提议报告正式标题（关键新增步骤）
|
||||
|
||||
用户答完前 8 个问题后，基于他们的回答提议 3 个候选标题供选择。
|
||||
|
||||
**命名范式**（参考 9MW1911 综合战略报告）：
|
||||
- 主标题：精炼、有分量、体现报告定位（如"XX综合战略报告"、"XX立项可行性研究报告"、"XX市场深度研究报告"）
|
||||
- 副标题：说明具体研究对象和视角（如"全球视角下抗 ST2 单克隆抗体在慢阻肺治疗领域的战略定位"）
|
||||
|
||||
示例对话：
|
||||
|
||||
> 根据你的回答，我为本报告提议以下 3 个候选标题：
|
||||
>
|
||||
> **候选 A（推荐）**
|
||||
> 主标题：自研 O-糖苷酶立项可行性研究报告
|
||||
> 副标题：对标 NEB 与 Merck 经典产品的技术路径、IP 壁垒与差异化战略
|
||||
>
|
||||
> **候选 B**
|
||||
> 主标题：GH101 家族酶国产化战略研究
|
||||
> 副标题：从 E. faecalis / S. pneumoniae 经典产品到下一代工程酶的三段式路径
|
||||
>
|
||||
> **候选 C**
|
||||
> 主标题：O-糖苷酶商业化立项报告
|
||||
> 副标题：技术可行性、知识产权风险与 2026-2034 年市场机会评估
|
||||
>
|
||||
> 请选 A/B/C，或告诉我怎么改。
|
||||
|
||||
### Step 5: 创建 manifest.json
|
||||
|
||||
用户确认标题后，创建 `projects/<slug>/manifest.json`：
|
||||
|
||||
```json
|
||||
{
|
||||
"slug": "<slug>",
|
||||
"topic": "<用户输入的完整主题>",
|
||||
"report_title": "<用户选定的主标题>",
|
||||
"report_subtitle": "<用户选定的副标题>",
|
||||
"author": "Deep Research 系统",
|
||||
"date": "<今天 YYYY-MM-DD>",
|
||||
"version": "1.0",
|
||||
"type": "<综述/研究/投资/管理>",
|
||||
"confidentiality": "机密 | 仅供内部决策使用",
|
||||
"audience": "<受众>",
|
||||
"time_range": "<时间范围>",
|
||||
"geography": "<地理范围>",
|
||||
"core_questions": ["...", "..."],
|
||||
"comparison_targets": [],
|
||||
"exclusions": [],
|
||||
"word_budget_mode": "<auto/concise/detailed/deep>",
|
||||
"model_profile": "<simple/medium/premium/cn_heavy/codex_native>",
|
||||
"model_profile_selected_at": "<今天 YYYY-MM-DD>",
|
||||
"model_profile_source": "dr-init interview",
|
||||
"target_words_zh": <按类型和模式计算，见 length-budget skill §1-2>,
|
||||
"target_words_en": <target_words_zh / 1.4>,
|
||||
"min_words_zh": <target_words_zh × 0.8>,
|
||||
"min_words_en": <min_words_zh / 1.4>,
|
||||
"disclaimer": "本报告基于公开信息与 AI 辅助研究生成，仅供参考，不构成投资或医疗建议。",
|
||||
"work_language": "en",
|
||||
"output_language": "zh",
|
||||
"phase1": {"status": "interview_done", "approved": false},
|
||||
"phase2": {"status": "pending"},
|
||||
"phase3": {"status": "pending"},
|
||||
"phase4": {"status": "pending"}
|
||||
}
|
||||
```
|
||||
|
||||
### Step 6: 记录访谈
|
||||
|
||||
把整个访谈对话写入 `projects/<slug>/phase1/interview.md`（用户原话 + 你的提问 + 提议的候选标题 + 用户选择）。
|
||||
|
||||
### Step 6.5: 立刻应用模型策略（必须执行）
|
||||
|
||||
在项目初始化完成后，立即把 `model_profile` 应用到 agent 文件（OpenCode + Codex 模板）：
|
||||
|
||||
```bash
|
||||
uv run python scripts/dr.py apply-models --profile <model_profile> --target both
|
||||
```
|
||||
|
||||
这样可以确保从 Phase 1（plan）到 Phase 4（polisher/reporter）全流程使用同一套预设策略，而不是中途切换。
|
||||
|
||||
### Step 7: 回报
|
||||
|
||||
```
|
||||
项目已初始化：projects/<slug>/
|
||||
|
||||
报告标题：<主标题>
|
||||
副标题：<副标题>
|
||||
类型：<研究类型>
|
||||
字数目标：<中文字数> 字 / <英文词数> words
|
||||
工作语言：English（Phase 2-3）
|
||||
输出语言：中文（Phase 4 翻译）
|
||||
|
||||
下一步：运行 /dr-frame 触发 Phase 1 框架规划（双语大纲）
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
## 注意事项
|
||||
|
||||
- ❌ 不要在本命令里做联网搜索或生成 framework（那是 /dr-frame 的工作）
|
||||
- ❌ 不要自己猜研究边界，必须让用户明确
|
||||
- ❌ slug 不要包含中文、空格、下划线
|
||||
- ✅ Step 4 的报告标题是 v0.5 新增的关键步骤，不可跳过
|
||||
- ✅ Step 8 的字数期望是 v0.5 新增的参数，帮助用户控制报告规模
|
||||
- ✅ 如果用户主题过于模糊（如"生物医药"），追问细化后再创建目录
|
||||
完成后暂停，下一步运行 `/dr-frame <slug>` 生成 `phase1/framework.md`。
|
||||
|
||||
@@ -1,117 +1,44 @@
|
||||
---
|
||||
description: Phase 2 - 并行深度研究所有章节。读取 Phase 1 确认的框架，按批次调度 dr-analyst 深研 + dr-verifier 反方验证，自动字数核验。用法：/dr-research [slug]
|
||||
description: Phase 2 - v0.20 Python core task-card research. 用法：/dr-research [slug]
|
||||
agent: dr-pm
|
||||
---
|
||||
|
||||
你是 dr-pm。用户执行了 `/dr-research $ARGUMENTS`，需要驱动 Phase 2 完整执行。
|
||||
你是 OpenCode 表层接口。不要自行 spawn subagents，也不要在本会话里执行章节研究。
|
||||
|
||||
## Step 1: 定位项目
|
||||
运行项目自有 Python core：
|
||||
|
||||
- `$ARGUMENTS` 非空：用该 slug
|
||||
- 为空：`ls -t projects/*/manifest.json` 取最近的
|
||||
|
||||
读取 `projects/<slug>/manifest.json`。
|
||||
|
||||
## Step 2: 前置检查
|
||||
|
||||
验证以下字段，任何一项不通过则停止并告知用户：
|
||||
- `phase1.approved == true`（框架已确认）
|
||||
- `phase1.framework_path` 指向的文件存在
|
||||
- `phase2.status != "completed"`（避免重复跑）
|
||||
|
||||
如果 `phase2.status == "in_progress"`，询问用户是否从中断处继续还是重新开始。
|
||||
|
||||
## Step 3: 读取框架并规划批次
|
||||
|
||||
读取 framework.md，提取所有 chapter 的：
|
||||
- 编号、标题、字数配额
|
||||
- 各 section 研究思路
|
||||
|
||||
按以下规则分批（每批 3 章并行）：
|
||||
- 字数配额 > 3000 字的章节单独成批
|
||||
- 有前后依赖关系的章节放在不同批次
|
||||
- 引言章（第 1 章）和结论章（最后 1 章）各自单独成批
|
||||
|
||||
更新 manifest.json：
|
||||
```json
|
||||
"phase2": {
|
||||
"status": "in_progress",
|
||||
"started_at": "<ISO时间>",
|
||||
"batches": [...],
|
||||
"chapters": [{"index": 1, "status": "pending", ...}, ...]
|
||||
}
|
||||
```bash
|
||||
uv run python scripts/dr.py research $ARGUMENTS --workers 6
|
||||
```
|
||||
|
||||
## Step 4: 逐批执行
|
||||
生成真实证据包时显式开启模型 worker：
|
||||
|
||||
对每批中的每个章节，**并行**委派 dr-analyst：
|
||||
|
||||
```
|
||||
Task prompt 模板：
|
||||
你是 dr-analyst。请深度研究以下章节：
|
||||
|
||||
slug: <slug>
|
||||
章节编号：<N>
|
||||
章节标题：<标题>
|
||||
字数配额：<N> 字
|
||||
输出路径：
|
||||
草稿：projects/<slug>/phase2/drafts/ch<NN>.md
|
||||
证据：projects/<slug>/phase2/evidence/ch<NN>-evidence.md
|
||||
信源：projects/<slug>/phase2/sources.jsonl
|
||||
|
||||
研究思路（来自 framework.md）：
|
||||
<粘贴该章的研究思路和 section 列表>
|
||||
|
||||
必须加载的 skill：search-strategy, source-quality, length-budget, evidence-table, mckinsey-method
|
||||
```bash
|
||||
uv run python scripts/dr.py research $ARGUMENTS --workers 6 --execute-packets
|
||||
```
|
||||
|
||||
一批的所有 dr-analyst 完成后，对每章**串行**委派 dr-verifier：
|
||||
默认 scholar/news/patents 专用路由 strict 失败即停；如只是低成本试跑，可允许通用搜索兜底：
|
||||
|
||||
```
|
||||
Task prompt 模板：
|
||||
你是 dr-verifier。请对以下章节做反方验证：
|
||||
|
||||
草稿：projects/<slug>/phase2/drafts/ch<NN>.md
|
||||
证据：projects/<slug>/phase2/evidence/ch<NN>-evidence.md
|
||||
|
||||
必须加载的 skill：search-strategy, source-quality
|
||||
```bash
|
||||
uv run python scripts/dr.py research $ARGUMENTS --workers 6 --execute-packets --allow-search-fallback
|
||||
```
|
||||
|
||||
每章完成后更新 manifest.json 的进度字段。
|
||||
将证据包收束为章节 brief，降低并发碎片化：
|
||||
|
||||
## Step 5: 字数核验与补写
|
||||
|
||||
每章 dr-analyst 返回后，读取草稿文件统计字数。如果实际字数 < 配额 × 0.7，自动再次委派 dr-analyst 补写，最多补写 2 次。
|
||||
|
||||
## Step 6: 汇总 sources.jsonl
|
||||
|
||||
所有章节完成后，对 `projects/<slug>/phase2/sources.jsonl` 做去重（按 url 字段）。
|
||||
|
||||
## Step 7: 更新 manifest 并汇报
|
||||
|
||||
```json
|
||||
"phase2": {
|
||||
"status": "completed",
|
||||
"completed_at": "<ISO时间>",
|
||||
"word_stats": {
|
||||
"total": <总字数>,
|
||||
"target": <目标字数>,
|
||||
"verdict": "合格/不足"
|
||||
}
|
||||
}
|
||||
```bash
|
||||
uv run python scripts/dr.py research $ARGUMENTS --workers 6 --build-briefs
|
||||
```
|
||||
|
||||
告知用户：
|
||||
```
|
||||
Phase 2 完成
|
||||
生成中文章节草稿：
|
||||
|
||||
总字数：X 字 / 目标 X 字
|
||||
章节：X / X 完成
|
||||
总信源：X 条（Tier1: X, Tier2: X）
|
||||
待验证观点：X 条
|
||||
CRITICAL 反方证据：X 条
|
||||
|
||||
下一步：/dr-review 启动总编审校
|
||||
```bash
|
||||
uv run python scripts/dr.py research $ARGUMENTS --workers 6 --assemble-chapters
|
||||
```
|
||||
|
||||
如总字数不足 min_words，告知用户并询问是否接受或指定某些章节补写。
|
||||
如用户只是想预览任务卡：
|
||||
|
||||
```bash
|
||||
uv run python scripts/dr.py research $ARGUMENTS --workers 6 --dry-run
|
||||
```
|
||||
|
||||
完成后只汇报 Python CLI 输出的关键路径：`phase2/task_cards.json`、`phase2/packets/*.json`、manifest 进度和下一步。
|
||||
|
||||
@@ -1,48 +1,13 @@
|
||||
---
|
||||
description: Phase 3 - 总编审校。用 Gemini 3.1 Pro 通读全部章节草稿，出具审校报告，暂停等用户决策。用法：/dr-review [slug]
|
||||
description: Phase 3 审校。薄封装：调用 Python core deterministic review。用法：/dr-review <slug-or-path>
|
||||
agent: dr-chief-editor
|
||||
subtask: false
|
||||
---
|
||||
|
||||
你是 dr-chief-editor。用户执行了 `/dr-review $ARGUMENTS`，需要对所有章节草稿做总编审校。
|
||||
执行 Python core 审校入口：
|
||||
|
||||
## Step 1: 定位项目
|
||||
|
||||
- `$ARGUMENTS` 非空：用该 slug
|
||||
- 为空：取最近的项目
|
||||
|
||||
验证：`phase2.status == "completed"`，否则告知用户先完成 `/dr-research`。
|
||||
|
||||
## Step 2: 执行审校
|
||||
|
||||
按照 dr-chief-editor.md 中的**模式 A：Phase 3 审校**工作流，通读所有草稿，出具审校报告。
|
||||
|
||||
审校报告写入 `projects/<slug>/phase3/critique.md`。
|
||||
|
||||
## Step 3: 暂停等待用户决策
|
||||
|
||||
审校报告完成后，向用户展示：
|
||||
1. 总体评级（A/B/C/D）
|
||||
2. 必须修正问题清单
|
||||
3. 字数审计表
|
||||
4. 明确的决策提示：
|
||||
|
||||
```
|
||||
审校完成，评级：<X>
|
||||
|
||||
请选择下一步：
|
||||
A/B 级：直接发 /dr-finalize 生成最终报告
|
||||
C 级：告诉我哪些章节需要回炉（我会重新研究那些章节）
|
||||
D 级：发 /dr-frame 重新规划框架
|
||||
```bash
|
||||
uv run python scripts/dr.py review $ARGUMENTS
|
||||
```
|
||||
|
||||
**不要自动进入 Phase 4，必须等用户明确指令。**
|
||||
|
||||
## 用户回复处理
|
||||
|
||||
如果用户说"直接 finalize"或类似：
|
||||
- 更新 `manifest.phase3.approved = true`
|
||||
- 告知用户发 `/dr-finalize`
|
||||
|
||||
如果用户指定某些章节回炉：
|
||||
- 将那些章节的 `phase2.chapters[i].status` 改为 `"needs_revision"`
|
||||
- 告知用户发 `/dr-research` 会只重跑这些章节
|
||||
完成后暂停，请用户审阅 `phase3/critique.md`，再决定回炉 Phase 2 或进入 `/dr-finalize`。
|
||||
|
||||
@@ -0,0 +1,20 @@
|
||||
---
|
||||
description: v0.20 platform-neutral Python core runner. 用法：/dr-run [slug-or-topic]
|
||||
agent: dr-pm
|
||||
---
|
||||
|
||||
你是 OpenCode 表层接口。不要自行编排多 agent；核心调度由 Python runtime 负责。
|
||||
|
||||
运行：
|
||||
|
||||
```bash
|
||||
uv run python scripts/dr.py run $ARGUMENTS --workers 6
|
||||
```
|
||||
|
||||
如需预演：
|
||||
|
||||
```bash
|
||||
uv run python scripts/dr.py run $ARGUMENTS --workers 6 --dry-run
|
||||
```
|
||||
|
||||
只汇报 Python CLI 的阶段判断、产物路径和下一步。
|
||||
@@ -1,47 +1,12 @@
|
||||
---
|
||||
description: 查看当前研究项目的进度。用法：/dr-status [slug]
|
||||
description: 查看当前研究项目进度。用法：/dr-status [slug]
|
||||
agent: dr-pm
|
||||
---
|
||||
|
||||
你是 dr-pm。读取项目状态并输出清晰的进度报告。
|
||||
你是 OpenCode 表层接口。运行 Python core 状态命令：
|
||||
|
||||
## Step 1: 定位项目
|
||||
|
||||
- `$ARGUMENTS` 非空：读取 `projects/$ARGUMENTS/manifest.json`
|
||||
- 为空：
|
||||
- 如果 `projects/` 下有多个项目，列出所有项目及其状态让用户选择
|
||||
- 只有一个则直接读取
|
||||
|
||||
## Step 2: 输出状态报告
|
||||
|
||||
```
|
||||
项目：<topic>
|
||||
Slug：<slug>
|
||||
类型：<type> | 目标字数：<target_words> 字
|
||||
|
||||
阶段进度：
|
||||
Phase 1 框架规划：<pending/in_progress/completed/approved>
|
||||
框架文件：<存在/不存在>
|
||||
章节数：<N>
|
||||
|
||||
Phase 2 深度研究：<pending/in_progress/completed>
|
||||
章节完成：<X/N>
|
||||
当前批次：<X>（如进行中）
|
||||
已写字数：<X> 字
|
||||
信源数量：<X> 条
|
||||
|
||||
Phase 3 总编审校：<pending/in_progress/completed>
|
||||
审校评级：<A/B/C/D 或 未完成>
|
||||
待修正问题：<X> 条
|
||||
|
||||
Phase 4 成稿：<pending/completed>
|
||||
PDF：<存在/不存在>
|
||||
DOCX：<存在/不存在>
|
||||
|
||||
输出文件：
|
||||
<列出 projects/<slug>/ 下已存在的关键文件>
|
||||
```bash
|
||||
uv run python scripts/dr.py status $ARGUMENTS
|
||||
```
|
||||
|
||||
## 额外说明
|
||||
|
||||
如果某个 Phase 处于 in_progress 但看起来卡住了（started_at 超过 2 小时且无进展），提示用户可以重新运行对应命令继续。
|
||||
汇报阶段状态、task cards、packets、drafts、sources、final_zh/PDF/DOCX 等关键产物。
|
||||
|
||||
@@ -138,7 +138,32 @@ dr-reporter 从 sources.jsonl 生成参考文献列表时，按以下格式：
|
||||
|
||||
---
|
||||
|
||||
## 五、引用完整性检查（dr-chief-editor 用）
|
||||
## 五、脚注使用边界
|
||||
|
||||
脚注不是行内引用的替代品，也不用于重复输入材料中已经被正文自然承载的事实。脚注只在以下场景使用：
|
||||
|
||||
- **法规原文或条款定位**：正文需要引用法规要求，但完整条款会打断叙事时，脚注写明法规名称、章节/条款和关键原文。
|
||||
- **关键资料原文**：原文措辞本身会影响判断强度，且正文只保留管理结论时，脚注可放短摘录。
|
||||
- **补充背景或术语解释**：正文读者可能需要额外背景，但展开会破坏行文节奏。
|
||||
- **版权或使用限制说明**：图表、第三方材料、内部材料使用边界需要单独说明时。
|
||||
|
||||
禁止事项：
|
||||
|
||||
- 不要把“某份输入材料说过什么”机械搬到脚注；这类事实应通过正文和数字引用解决。
|
||||
- 不要为每个本地材料引用都加脚注；脚注应少而精，优先服务关键判断。
|
||||
- 不要用脚注堆砌证据，核心证据仍应进入正文或证据表。
|
||||
|
||||
推荐格式：
|
||||
|
||||
```markdown
|
||||
正文关键判断<sup>[12]</sup>。[^1]
|
||||
|
||||
[^1]: ICH Q10《Pharmaceutical Quality System》第 4.1 节要求管理评审输入覆盖“results of regulatory inspections and findings, audits and commitments”，并纳入 CAPA、变更以及上次管理评审行动。
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
## 六、引用完整性检查（dr-chief-editor 用）
|
||||
|
||||
审校时检查：
|
||||
1. 正文中所有 [src_xxx] 都在 sources.jsonl 里有对应记录
|
||||
|
||||
@@ -0,0 +1,62 @@
|
||||
---
|
||||
name: search-gateway
|
||||
description: Use when Deep Research agents or subagents need web, scholar, patent, news, regulatory, or source-discovery search without using platform MCP tools or browser search directly.
|
||||
---
|
||||
|
||||
# Search Gateway
|
||||
|
||||
## Rule
|
||||
|
||||
Use the project Python search gateway as the only default search interface. Do not call Tavily MCP, browser MCP, generic web tools, or platform-native search from a subagent unless the user explicitly asks for that escape hatch.
|
||||
|
||||
## Commands
|
||||
|
||||
Run searches from the repository root:
|
||||
|
||||
```bash
|
||||
uv run python scripts/search.py "<query>" --route general --json --trace
|
||||
uv run python scripts/search.py "<query>" --route evidence --json --trace
|
||||
uv run python scripts/search.py "<query>" --route scholar --year-low 2020 --json --trace
|
||||
uv run python scripts/search.py "<query>" --route news --time-range y --json --trace
|
||||
uv run python scripts/search.py "<query>" --route patents --json --trace
|
||||
uv run python scripts/search.py "<query>" --profile biomed_literature --json --trace
|
||||
```
|
||||
|
||||
If `uv` cannot use the user cache in a sandbox, set a local cache:
|
||||
|
||||
```bash
|
||||
UV_CACHE_DIR=/private/tmp/deep_research_uv_cache uv run python scripts/search.py "<query>" --route general --json --trace
|
||||
```
|
||||
|
||||
## Routing
|
||||
|
||||
- `general`: Tavily first, Exa fallback, Brave fallback; use for broad discovery and gap filling.
|
||||
- `evidence`: Exa highlights first, Tavily fallback, Brave fallback; use when a task card needs concise, source-level candidate evidence for an evidence packet.
|
||||
- `scholar`: Serper Scholar first; use for papers, reviews, technical literature, and academic validation only.
|
||||
- `news`: Serper News first; use for recent industry/current information.
|
||||
- `patents`: Serper Google Patents first.
|
||||
- `biomed_literature`: scholar plus general discovery.
|
||||
|
||||
Serper is not the default general web search source. Keep it mainly for Scholar, Google Patents, News, and targeted `site:` searches where Google coverage matters.
|
||||
|
||||
Tavily Research is a phase-level scan tool, not a packet-writing shortcut. Use it for Phase 1 initial landscape scans, Phase 2 gap-fill after a chapter is thin, or Phase 3回炉补证据；its output must be saved, source-scored, deduplicated, and converted into candidate evidence before citation.
|
||||
|
||||
Exa is the preferred controlled evidence discovery route for agents because it can return short highlights/text per URL. Treat Exa hits as candidate sources unless the URL itself is an original Tier 1-2 source.
|
||||
|
||||
API keys are loaded from `secrets.env` by `scripts/search.py`; do not ask the user to authorize MCP calls when the env keys are available.
|
||||
|
||||
## Subagent Protocol
|
||||
|
||||
For evidence packets:
|
||||
|
||||
1. Search through `scripts/search.py`, save or summarize the returned JSON in the packet’s `raw_quotes_or_notes`.
|
||||
2. Use search hits only as candidate sources; whenever possible, cite the original regulator, guideline, paper, or official document.
|
||||
3. Put every used source in `sources` with `id`, `title`, `url`, `tier`, and `score`.
|
||||
4. Do not write a final chapter during search; produce structured evidence only.
|
||||
5. For repeatedly used Tier 1-2 sources, run `uv run python scripts/dr.py sources cache <project>` so later phases can cite a local Markdown snapshot rather than only a URL.
|
||||
|
||||
For chapter assembly:
|
||||
|
||||
1. Do not search. Use only `phase2/chapter_briefs`, `phase2/packets`, `phase2/sources.jsonl`, `phase0/extracted`, and `phase1/framework.md`.
|
||||
2. Do not create new `source_id`.
|
||||
3. If evidence is thin, mark the chapter as needing Phase 2 enrichment instead of filling with generic prose.
|
||||
@@ -75,10 +75,12 @@ description: 生物医药深度研究的统一检索策略。规定信源优先
|
||||
- 例：研究"GLP-1 成为减重首选"→ 反方要搜 "GLP-1 limitations" "semaglutide side effects" "discontinuation rate"
|
||||
- 至少 3-5 条反方证据
|
||||
|
||||
### 第 4 轮：Tavily/Brave/Exa 补漏
|
||||
### 第 4 轮：Exa/Tavily/Brave 补漏
|
||||
- 仅用于发现前 3 轮遗漏的 URL
|
||||
- 发现后**必须**回溯到原始 Tier 1-2 来源（论文 DOI、监管公告原文）
|
||||
- 不得直接引用搜索返回的二次报道
|
||||
- 章节级 evidence packet 优先用 `scripts/search.py --route evidence`，让 Exa highlights 进入 source-quality 和 evidence-table。
|
||||
- Tavily Research 只用于 Phase 1 初扫、薄弱章节补证据和 Phase 3 回炉；输出必须存盘、评分、去重后再转成 candidate evidence。
|
||||
|
||||
---
|
||||
|
||||
@@ -88,6 +90,7 @@ description: 生物医药深度研究的统一检索策略。规定信源优先
|
||||
|
||||
```bash
|
||||
uv run python scripts/search.py "<query>" --route scholar --num-results 10 --year-low 2023
|
||||
uv run python scripts/search.py "<query>" --route evidence --num-results 10 --json --trace
|
||||
uv run python scripts/search.py "<query>" --route patents --num-results 10
|
||||
uv run python scripts/search.py "<query>" --route news --num-results 10 --time-range m
|
||||
uv run python scripts/search.py "<query>" --route general --num-results 10
|
||||
@@ -107,7 +110,7 @@ uv run python scripts/search.py "<query>" --profile patent_heavy --num-results 1
|
||||
- `--route patents` 固定优先 Serper + Google Patents，避免专利检索被 Tavily 普通网页结果替代。
|
||||
- `--route scholar` 固定优先 Serper Scholar，避免论文检索只停留在通用网页摘要。
|
||||
- 专用 route（scholar/patents/news）默认 `--strict-specialized`，Serper 异常时应显式失败，不允许静默降级。
|
||||
- Tavily / Exa / Brave 只作为 gap-fill 或 MCP 兜底，不作为文献/专利主路径。
|
||||
- Exa evidence route 是 packet 候选证据发现主路径；Tavily / Brave 只作为 gap-fill 或 MCP 兜底，不作为文献/专利主路径。
|
||||
|
||||
每个检索小结必须写明实际使用过的 route，例如：
|
||||
|
||||
|
||||
@@ -49,7 +49,7 @@ try:
|
||||
Table,
|
||||
TableStyle,
|
||||
)
|
||||
from reportlab.platypus.flowables import HRFlowable
|
||||
from reportlab.platypus.flowables import Flowable, HRFlowable
|
||||
except ImportError:
|
||||
print("ERROR: missing reportlab. Run: uv sync", file=sys.stderr)
|
||||
sys.exit(1)
|
||||
@@ -279,6 +279,23 @@ def build_styles() -> StyleSheet1:
|
||||
allowOrphans=0,
|
||||
))
|
||||
|
||||
# Inline evidence footnotes placed by number_citations / finalization.
|
||||
ss.add(ParagraphStyle(
|
||||
name="evidence-footnote",
|
||||
fontName="SrcSerif",
|
||||
fontSize=7.4,
|
||||
leading=10,
|
||||
alignment=TA_JUSTIFY,
|
||||
leftIndent=18,
|
||||
firstLineIndent=-18,
|
||||
spaceBefore=0,
|
||||
spaceAfter=2,
|
||||
textColor=colors.HexColor("#4b5563"),
|
||||
wordWrap="CJK",
|
||||
allowWidows=0,
|
||||
allowOrphans=0,
|
||||
))
|
||||
|
||||
# Table cell - no first-line indent, smaller font, CJK wrap for auto line break
|
||||
ss.add(ParagraphStyle(
|
||||
name="table-cell",
|
||||
@@ -422,6 +439,23 @@ def build_styles() -> StyleSheet1:
|
||||
bulletIndent=8,
|
||||
))
|
||||
|
||||
# Ordered list. Keep references flush-left; do not prepend decorative bullets.
|
||||
ss.add(ParagraphStyle(
|
||||
name="ordered",
|
||||
fontName="SrcSerif",
|
||||
fontSize=9,
|
||||
leading=13,
|
||||
alignment=TA_JUSTIFY,
|
||||
firstLineIndent=0,
|
||||
leftIndent=0,
|
||||
spaceBefore=0,
|
||||
spaceAfter=4,
|
||||
textColor=colors.HexColor("#374151"),
|
||||
wordWrap="CJK",
|
||||
allowWidows=0,
|
||||
allowOrphans=0,
|
||||
))
|
||||
|
||||
return ss
|
||||
|
||||
|
||||
@@ -436,6 +470,28 @@ class Block:
|
||||
meta: Optional[dict] = None
|
||||
|
||||
|
||||
class FootnoteFlowable(Flowable):
|
||||
"""Zero-height anchor that registers a footnote for the current PDF page."""
|
||||
|
||||
def __init__(self, label: str, content: str):
|
||||
super().__init__()
|
||||
self.label = label
|
||||
self.content = content
|
||||
self.width = 0
|
||||
self.height = 0
|
||||
|
||||
def wrap(self, availWidth, availHeight):
|
||||
return 0, 0
|
||||
|
||||
def draw(self):
|
||||
page = self.canv.getPageNumber()
|
||||
notes = getattr(self.canv, "_dr_footnotes", None)
|
||||
if notes is None:
|
||||
notes = {}
|
||||
setattr(self.canv, "_dr_footnotes", notes)
|
||||
notes.setdefault(page, []).append((self.label, self.content))
|
||||
|
||||
|
||||
def parse_markdown(md_text: str) -> List[Block]:
|
||||
blocks: List[Block] = []
|
||||
lines = md_text.split("\n")
|
||||
@@ -483,6 +539,17 @@ def parse_markdown(md_text: str) -> List[Block]:
|
||||
blocks.append(Block(kind="quote", content="\n".join(quote_lines)))
|
||||
continue
|
||||
|
||||
# Markdown footnote definition: [^1]: 原文摘录...
|
||||
m = re.match(r"^\[\^([A-Za-z0-9_-]+)\]:\s*(.+)$", stripped)
|
||||
if m:
|
||||
blocks.append(Block(
|
||||
kind="footnote",
|
||||
content=m.group(2).strip(),
|
||||
meta={"label": m.group(1)},
|
||||
))
|
||||
i += 1
|
||||
continue
|
||||
|
||||
# Unordered list
|
||||
if re.match(r"^[-*+]\s+", stripped):
|
||||
while i < len(lines) and re.match(r"^[-*+]\s+", lines[i].strip()):
|
||||
@@ -493,12 +560,12 @@ def parse_markdown(md_text: str) -> List[Block]:
|
||||
|
||||
# Ordered list
|
||||
if re.match(r"^\d+\.\s+", stripped):
|
||||
idx = 1
|
||||
while i < len(lines) and re.match(r"^\d+\.\s+", lines[i].strip()):
|
||||
item = re.sub(r"^\d+\.\s+", "", lines[i].strip())
|
||||
blocks.append(Block(kind="bullet", content=f"{idx}. {item}"))
|
||||
m = re.match(r"^(\d+)\.\s+(.+)$", lines[i].strip())
|
||||
if not m:
|
||||
break
|
||||
blocks.append(Block(kind="ordered", content=f"{m.group(1)}. {m.group(2)}"))
|
||||
i += 1
|
||||
idx += 1
|
||||
continue
|
||||
|
||||
# Table
|
||||
@@ -518,6 +585,7 @@ def parse_markdown(md_text: str) -> List[Block]:
|
||||
while i < len(lines) and lines[i].strip() and not (
|
||||
lines[i].strip().startswith(("#", ">", "-", "*", "+", "!"))
|
||||
or re.match(r"^\d+\.\s+", lines[i].strip())
|
||||
or re.match(r"^\[\^[A-Za-z0-9_-]+\]:", lines[i].strip())
|
||||
or "|" in lines[i]
|
||||
):
|
||||
para_lines.append(lines[i])
|
||||
@@ -638,6 +706,17 @@ def md_inline_to_rl(text: str, *, add_cjk_space: bool = True) -> str:
|
||||
+ ']</font></super>',
|
||||
text,
|
||||
)
|
||||
text = re.sub(
|
||||
r"<sup>(.*?)</sup>",
|
||||
r"<super><font size=7>\1</font></super>",
|
||||
text,
|
||||
flags=re.IGNORECASE,
|
||||
)
|
||||
text = re.sub(
|
||||
r"\[\^([A-Za-z0-9_-]+)\]",
|
||||
lambda m: f"<super><font size=7>注{m.group(1)}</font></super>",
|
||||
text,
|
||||
)
|
||||
text = re.sub(r"\[([^\]]+)\]\(([^)]+)\)", r"\1", text)
|
||||
return text
|
||||
|
||||
@@ -746,9 +825,27 @@ def make_page_decorator(manifest: Manifest):
|
||||
canvas.setLineWidth(0.5)
|
||||
canvas.line(2 * cm, A4[1] - 1.4 * cm, A4[0] - 2 * cm, A4[1] - 1.4 * cm)
|
||||
|
||||
# Page-bottom evidence footnotes.
|
||||
notes = getattr(canvas, "_dr_footnotes", {}).get(doc.page, [])
|
||||
if notes:
|
||||
width = A4[0] - 4.4 * cm
|
||||
x = 2.2 * cm
|
||||
y = 3.45 * cm
|
||||
canvas.setStrokeColor(colors.HexColor("#cbd5e1"))
|
||||
canvas.setLineWidth(0.45)
|
||||
canvas.line(x, y + 0.16 * cm, x + 6.8 * cm, y + 0.16 * cm)
|
||||
footnote_style = build_styles()["evidence-footnote"]
|
||||
for label, content in notes:
|
||||
text = f"注{label}：{content}"
|
||||
para = Paragraph(md_inline_to_rl(text), footnote_style)
|
||||
_, h = para.wrap(width, 1.3 * cm)
|
||||
y -= h
|
||||
para.drawOn(canvas, x, y)
|
||||
y -= 0.04 * cm
|
||||
|
||||
# Footer: page number centered
|
||||
canvas.setFont("SrcSans-Light", 8)
|
||||
canvas.drawCentredString(A4[0] / 2, 1.2 * cm, f"— {doc.page} —")
|
||||
canvas.drawCentredString(A4[0] / 2, 1.0 * cm, f"— {doc.page} —")
|
||||
|
||||
canvas.restoreState()
|
||||
|
||||
@@ -830,6 +927,8 @@ def collect_toc_entries(blocks: List[Block]) -> List[tuple[int, str]]:
|
||||
title = b.content.strip()
|
||||
if any(s in title.lower() for s in skip_titles_substr):
|
||||
continue
|
||||
if b.kind == "h1" and not re.match(r"^第\s*\d+\s*章\b", title):
|
||||
continue
|
||||
level = 1 if b.kind == "h1" else 2
|
||||
entries.append((level, title))
|
||||
return entries
|
||||
@@ -1196,8 +1295,13 @@ def _render_generic_block(block: Block, story: list, base_dir: Path, styles: Sty
|
||||
story.append(Paragraph(md_inline_to_rl(block.content), style))
|
||||
elif block.kind == "quote":
|
||||
story.append(Paragraph(md_inline_to_rl(block.content), styles["quote"]))
|
||||
elif block.kind == "footnote":
|
||||
label = block.meta.get("label") if block.meta else ""
|
||||
story.append(FootnoteFlowable(str(label), block.content))
|
||||
elif block.kind == "bullet":
|
||||
story.append(Paragraph("• " + md_inline_to_rl(block.content), styles["bullet"]))
|
||||
elif block.kind == "ordered":
|
||||
story.append(Paragraph(md_inline_to_rl(block.content), styles["ordered"]))
|
||||
elif block.kind == "hr":
|
||||
story.append(Spacer(1, 0.3 * cm))
|
||||
elif block.kind == "image":
|
||||
@@ -1349,6 +1453,11 @@ def build_body(
|
||||
front_sections[kind] = sec
|
||||
k = next_k
|
||||
|
||||
# If the markdown does not contain a TOC marker, still insert a generated TOC.
|
||||
# This keeps PDF output stable when Phase 4 emits a clean markdown body.
|
||||
if "toc" not in front_sections:
|
||||
front_sections["toc"] = [Block(kind="h2", content="目录")]
|
||||
|
||||
# 前置件输出顺序（固定）
|
||||
front_order = [
|
||||
"disclaimer", # 免责声明
|
||||
@@ -1408,13 +1517,26 @@ def build_body(
|
||||
i = _skip_until_next_section(i + 1)
|
||||
continue
|
||||
|
||||
# 参考文献：自动生成
|
||||
# 参考文献：如果正文仍使用 [src_xxx]，则从 sources.jsonl 自动生成；
|
||||
# 如果正文已被 number_citations.py 转为数字编号，则保留 Markdown 内的编号清单。
|
||||
if kind == "references":
|
||||
story.append(PageBreak())
|
||||
story.extend(build_references(blocks, sources_path, styles))
|
||||
j = i + 1
|
||||
while j < n and blocks[j].kind == "p" and _REF_PLACEHOLDER_RE.search(blocks[j].content):
|
||||
j += 1
|
||||
has_src_citations = bool(collect_cited_src_ids(blocks))
|
||||
has_ref_placeholder = (
|
||||
j < n and blocks[j].kind == "p" and _REF_PLACEHOLDER_RE.search(blocks[j].content)
|
||||
)
|
||||
if has_src_citations or has_ref_placeholder:
|
||||
story.extend(build_references(blocks, sources_path, styles))
|
||||
while j < n and blocks[j].kind not in ("h1", "h2"):
|
||||
j += 1
|
||||
else:
|
||||
story.append(Paragraph(md_inline_to_rl(block.content), styles["h1"]))
|
||||
while j < n and blocks[j].kind not in ("h1", "h2"):
|
||||
_render_generic_block(blocks[j], story, base_dir, styles, in_summary=False)
|
||||
j += 1
|
||||
i = j
|
||||
continue
|
||||
i = j
|
||||
continue
|
||||
|
||||
@@ -1497,7 +1619,7 @@ def main():
|
||||
leftMargin=2.2 * cm,
|
||||
rightMargin=2.2 * cm,
|
||||
topMargin=2 * cm,
|
||||
bottomMargin=2 * cm,
|
||||
bottomMargin=4.0 * cm,
|
||||
title=manifest.report_title,
|
||||
author=manifest.author,
|
||||
subject=manifest.type,
|
||||
@@ -1510,14 +1632,14 @@ def main():
|
||||
id="cover",
|
||||
)
|
||||
normal_frame = Frame(
|
||||
2.2 * cm, 2 * cm,
|
||||
A4[0] - 4.4 * cm, A4[1] - 4 * cm,
|
||||
2.2 * cm, 4.0 * cm,
|
||||
A4[0] - 4.4 * cm, A4[1] - 6.0 * cm,
|
||||
id="normal",
|
||||
)
|
||||
decorator = make_page_decorator(manifest)
|
||||
doc.addPageTemplates([
|
||||
PageTemplate(id="cover", frames=[cover_frame]),
|
||||
PageTemplate(id="normal", frames=[normal_frame], onPage=decorator),
|
||||
PageTemplate(id="normal", frames=[normal_frame], onPageEnd=decorator),
|
||||
])
|
||||
|
||||
# Assemble story
|
||||
|
||||
@@ -1,301 +1,91 @@
|
||||
# AGENTS.md — 生物医药 Deep Research 系统规则
|
||||
# AGENTS.md — Deep Research Cross-Tool Rules
|
||||
|
||||
> 本文件为 OpenCode 会自动读取的项目级指令文件。
|
||||
> 所有 agent / skill / command 必须遵循本文件定义的研究方法论、信源标准与输出规范。
|
||||
This file is the shared, cross-tool instruction layer for Codex, OpenCode, Claude Code, Gemini CLI, and Antigravity.
|
||||
|
||||
---
|
||||
Keep this file short. Do not put Antigravity roles, detailed workflows, or long skill manuals here.
|
||||
|
||||
## 1. 项目使命
|
||||
## Project
|
||||
|
||||
本项目通过多 agent 协作，以**麦肯锡、德勤等顶尖机构的研究方法**，对生物医药领域（研发、工艺、管理、投资）的指定主题进行深度研究，输出专业级报告（PDF + DOCX）。
|
||||
Deep Research produces professional biomedical research reports for R&D, CMC/GMP, management, market, and investment topics.
|
||||
|
||||
本项目**不涉及代码开发**，所有"代码"都是为**研究流水线**服务（如 ReportLab 模板、下载脚本、信源 API 调用）。
|
||||
This repository is not an application codebase. Its code supports the research pipeline: Python runtime, search utilities, evidence schemas, citation checks, and PDF/DOCX rendering.
|
||||
|
||||
---
|
||||
## Instruction Layers
|
||||
|
||||
## 2. 研究方法论（所有 agent 必须遵循）
|
||||
- Cross-tool baseline: `AGENTS.md`
|
||||
- Gemini / Antigravity override: `GEMINI.md`
|
||||
- Antigravity roles: `.agents/agents.md`
|
||||
- Antigravity strong rules: `.agents/rules/`
|
||||
- Reusable skills: `.agents/skills/`
|
||||
- Antigravity workflows: `.agents/workflows/`
|
||||
- Implementation plan and changelog: `PLAN.md`
|
||||
|
||||
### 2.1 麦肯锡核心原则
|
||||
If instructions conflict, use the more specific layer. For Antigravity, `GEMINI.md` and `.agents/rules/` override this file.
|
||||
|
||||
1. **MECE**（Mutually Exclusive, Collectively Exhaustive）：章节划分互斥且穷尽
|
||||
2. **SCQA 叙事**（Situation → Complication → Question → Answer）：每章节开头用此结构引入
|
||||
3. **金字塔原理**：结论先行，论据支撑，纵向深入，横向 MECE
|
||||
4. **"每个标题即一个观点"**：标题不能是"概述""现状"这类模糊词，必须包含判断
|
||||
5. **So What? 自检**：每写完一段问自己"所以呢？"，若无则删
|
||||
## Operating Modes
|
||||
|
||||
### 2.2 证据铁律
|
||||
Python-core mode:
|
||||
|
||||
- **每条结论至少 2 个独立 Tier 1-2 信源**佐证（见 §4 信源分级）
|
||||
- 达不到则**必须在正文注明**"该观点仅有 X 个来源支持，待进一步验证"
|
||||
- **反方证据优先**：每个 chapter 的研究必须主动搜索证伪性论点，不能只找支持证据
|
||||
- **数据可追溯**：所有数字、百分比、日期必须有来源 ID（如 `[src_042]`）
|
||||
- Use `scripts/dr.py`, `scripts/runtime/**`, `configs/models.yaml`, and `.agents/skills`.
|
||||
- Platform agents should call the Python CLI rather than reimplement worker orchestration.
|
||||
- Model routing is resolved by the Python runtime.
|
||||
|
||||
### 2.3 字数配额（硬要求）
|
||||
Antigravity native mode:
|
||||
|
||||
| 报告类型 | 最小字数 | 建议章节数 |
|
||||
|---|---|---|
|
||||
| 综述类 | 10,000 字 | 8-10 章 |
|
||||
| 研究类 | 30,000 字 | 10-12 章 |
|
||||
| 投资报告 | 20,000 字 | 10-12 章 |
|
||||
| 管理/工艺类 | 15,000-25,000 字 | 9-11 章 |
|
||||
- Use `.agents/agents.md`, `.agents/rules/`, `.agents/skills/`, and `.agents/workflows/`.
|
||||
- Antigravity uses its own model quota for research execution.
|
||||
- Python scripts are auxiliary for scaffolding, local material processing, deterministic checks, citation/report rendering, and status.
|
||||
- Do not run Python model-worker commands such as `dr.py run`, `research --execute-packets`, or `research --assemble-chapters` unless the user explicitly approves external API/ZenMux usage.
|
||||
|
||||
**字数分配原则**：
|
||||
- 每章字数差距不超过 ±30%（避免头重脚轻）
|
||||
- 每 section 最少 800 字（不够则合并）
|
||||
- 结论章不少于全文 10%
|
||||
## Core Commands
|
||||
|
||||
### 2.4 报告不能只谈结论
|
||||
```bash
|
||||
uv run python scripts/dr.py init "研究主题" --slug <slug> --method <method>
|
||||
uv run python scripts/dr.py frame <slug>
|
||||
uv run python scripts/dr.py research <slug> --workers 6 --execute-packets
|
||||
uv run python scripts/dr.py review <slug>
|
||||
uv run python scripts/dr.py finalize <slug>
|
||||
uv run python scripts/deploy_adapters.py antigravity --target /path/to/workspace --dry-run
|
||||
```
|
||||
|
||||
- 每个观点后必须紧跟**数据/事实/案例**佐证
|
||||
- 禁止空洞形容词（"巨大""快速""显著"）不带数据
|
||||
- 趋势判断必须给**量化依据**（年复合增长率、市场规模、成功率等）
|
||||
## Research Integrity
|
||||
|
||||
---
|
||||
- No fabricated data, URLs, DOIs, clinical results, regulatory status, patents, or market figures.
|
||||
- No factual claim without a source ID or a clearly marked "to verify" status.
|
||||
- No claim that a search or verification was performed unless there is a real tool receipt or search log.
|
||||
- Search snippets, AI summaries, and model memory are discovery leads, not final evidence.
|
||||
- Every major conclusion needs at least two independent Tier 1-2 sources. If not available, downgrade and mark uncertainty.
|
||||
- Counter-evidence is mandatory. Do not collect only supporting evidence.
|
||||
- Wikipedia is allowed for orientation only and must not support final conclusions.
|
||||
- Use Chinese for formal report writing. English may remain in search keywords, titles, DOI/URL, original excerpts, and raw notes.
|
||||
|
||||
## 3. Phase 工作流（4 阶段）
|
||||
## Method Selection
|
||||
|
||||
### Phase 1：框架规划
|
||||
- **驱动命令**：`/dr-init <topic>` → `/dr-frame`
|
||||
- **主导 agent**：dr-plan
|
||||
- **产出**：`projects/<slug>/phase1/framework.md`（8-15 章大纲，每 section 带研究思路与字数配额）
|
||||
- **暂停点**：用户确认框架
|
||||
Do not default to McKinsey/MECE for every topic.
|
||||
|
||||
### Phase 2：深度研究
|
||||
- **驱动命令**：`/dr-research`
|
||||
- **主导 agent**：dr-pm（调度 3-4 个 dr-analyst 并行 + dr-verifier 反方验证）
|
||||
- **产出**：`projects/<slug>/phase2/drafts/chXX.md` + `evidence/chXX-evidence.md` + `sources.jsonl`
|
||||
- **不暂停**：全自动跑完
|
||||
Choose the research method and tools based on the user's scenario. Use `.agents/skills/method-selection/SKILL.md` for Antigravity native work and `configs/research_methods.yaml` for Python-core mode.
|
||||
|
||||
### Phase 3：总编审校
|
||||
- **驱动命令**：`/dr-review`
|
||||
- **主导 agent**：dr-chief-editor（Gemini 3.1 Pro 1M 上下文通读）
|
||||
- **产出**：`projects/<slug>/phase3/critique.md`
|
||||
- **暂停点**：用户决策（修正 / 回炉 phase2 / 整体重来）
|
||||
## Source Quality
|
||||
|
||||
### Phase 4：成稿
|
||||
- **驱动命令**：`/dr-finalize`
|
||||
- **主导 agent**：dr-editor-in-chief（创作）→ `scripts/phase4_pipeline.py`（执行链路）
|
||||
- **执行链路**：translate → glossary(optional) → apply_glossary → polish → build_report
|
||||
- **产出**：`phase4/final_en.md` + `phase4/final_zh.md` + `phase4/final_zh_polished.md` + `phase4/*.pdf` + `phase4/*.docx`
|
||||
Detailed scoring belongs in `.agents/skills/source-quality/SKILL.md`.
|
||||
|
||||
---
|
||||
Baseline tiers:
|
||||
|
||||
## 4. 信源分级（Tier 系统）
|
||||
- Tier 1: original academic papers, systematic reviews where appropriate, regulator documents, clinical trial registries, patents, company filings.
|
||||
- Tier 2: authoritative consulting/industry reports, industry associations, reputable professional media and databases.
|
||||
- Tier 3: preprints, conference abstracts, brokerage research, company white papers. Use with caution.
|
||||
- Tier 4: general web pages, ordinary news, Wikipedia. Discovery only.
|
||||
|
||||
### Tier 1 — 一级信源（优先使用，评分权重 × 1.2）
|
||||
## Safety
|
||||
|
||||
- **一手学术文献**：PubMed、Cochrane、顶刊原文（NEJM / Lancet / Nature / Science / Cell / JAMA）
|
||||
- **监管机构公告**：FDA、EMA、NMPA、PMDA 官网及 openFDA
|
||||
- **临床试验注册**：ClinicalTrials.gov、ChiCTR
|
||||
- **专利原文**：USPTO、EPO、CNIPA、Google Patents
|
||||
- **上市公司披露**：SEC 10-K/10-Q、招股书、交易所年报
|
||||
- Keep API keys only in `secrets.env`; never hardcode or commit secrets.
|
||||
- Do not read or expose secrets unless the user explicitly asks.
|
||||
- Do not overwrite user settings or existing workspace rule/skill/workflow files unless the user asks for `--force`.
|
||||
- Do not run destructive git commands such as `git reset --hard`, `git clean`, or broad file deletion without explicit approval.
|
||||
- Do not write outside the current workspace unless the user explicitly approves.
|
||||
|
||||
### Tier 2 — 二级信源（可用，标准权重）
|
||||
## Change Management
|
||||
|
||||
- **权威咨询报告**：麦肯锡、BCG、德勤、IQVIA、EvaluatePharma、弗若斯特沙利文
|
||||
- **学术综述**：系统综述（Systematic Review）、Meta 分析
|
||||
- **行业协会**：PhRMA、BIO、中国医药工业协会
|
||||
- **专业数据库**：Wind、东方财富、同花顺（金融侧）
|
||||
- **专业媒体**：BioSpace、Endpoints News、FiercePharma、医药魔方、Insight 数据库
|
||||
|
||||
### Tier 3 — 三级信源（辅助，不得作为唯一支撑）
|
||||
|
||||
- **预印本**：bioRxiv、medRxiv（需标注"未经同行评审"）
|
||||
- **券商研报**：中金、中信、高盛生物医药团队（需注意利益冲突）
|
||||
- **会议摘要**：AACR、ASCO、ASH 会议摘要（数据可能未完整发表）
|
||||
- **企业白皮书**（注明来源，降权使用）
|
||||
|
||||
### Tier 4 — 四级信源（仅做发现入口）
|
||||
|
||||
- Tavily / Brave / Exa 通用搜索返回的**普通网页**
|
||||
- 一般新闻报道
|
||||
- Wikipedia（**只做术语理解入口，结论不得引用**）
|
||||
|
||||
### 黑名单（禁用）
|
||||
|
||||
- 纯新闻聚合站（百家号、头条号、部分自媒体公众号）
|
||||
- 未署名作者的行业博客
|
||||
- 被 Retraction Watch 标记为撤稿的论文
|
||||
- 明显软文/PR 稿（如"某某 CEO 表示..."而无实质数据）
|
||||
- 超过 5 年的综述（除机制类研究可放宽）
|
||||
|
||||
### 信源评分（0-10）
|
||||
|
||||
每个进入 `sources.jsonl` 的信源必须打分，维度：
|
||||
- 权威性（期刊 IF、机构排名）0-3
|
||||
- 时效性（≤3 年满分，每老 1 年 -0.5） 0-2
|
||||
- 一手性（一手 > 综述 > 二次解读） 0-2
|
||||
- 可验证性（有 DOI / URL / 原始数据） 0-2
|
||||
- 利益冲突（厂商自发 -1） 0-1
|
||||
|
||||
**硬规则**：评分 < 5 的信源不得作为结论唯一支撑。
|
||||
|
||||
---
|
||||
|
||||
## 5. Agent 角色与职责（v0.5 重构）
|
||||
|
||||
> 每个 agent 的详细定义见 `.opencode/agents/*.md`
|
||||
|
||||
| Agent | 类型 | 模型 | 职责 | 工作语言 |
|
||||
|---|---|---|---|---|
|
||||
| dr-plan | primary | Opus 4.7 | Phase 1 框架规划（访谈、标题提议、生成双语 framework） | 中文对话 + 英文框架内容 |
|
||||
| dr-pm | primary | Sonnet 4.6 | Phase 2 调度，批次间 context 压缩 | English |
|
||||
| dr-chief-editor | primary | Gemini 3.1 Pro Preview | **Phase 3 only**：只读审校，产出 critique.md | English |
|
||||
| **dr-editor-in-chief** | primary | **Opus 4.7** | **Phase 4 主导**：合并 final_en、写 Executive Summary/Abstract/Glossary、调度后续 | English |
|
||||
| dr-searcher | subagent | Haiku 4.5 | 轻量检索、信源发现 | English |
|
||||
| dr-analyst | subagent | Sonnet 4.6 | 章节深研（英文草稿 + 证据矩阵） | English |
|
||||
| dr-verifier | subagent | GPT-5.4 | 交叉模型反方验证（唯一非 Claude 位置） | English |
|
||||
| **dr-translator** | subagent | **Sonnet 4.6** | Phase 4 英译中，维护双语术语表 | 英→中 |
|
||||
| dr-polisher | subagent | Sonnet 4.6 | Phase 4 中文润色、humanizer-cn + output-hygiene | 中文 |
|
||||
| dr-reporter | subagent | Sonnet 4.6 | Phase 4 出稿（PDF+DOCX），**强制回填 citations** | 纯执行 |
|
||||
|
||||
**关键角色变化（v0.5）**：
|
||||
- dr-chief-editor 从"Phase 3/4 总编"收窄为"Phase 3 only 只读审校"
|
||||
- 新增 dr-editor-in-chief（Opus）接管 Phase 4 主导权（避免 Gemini 导致的风格断裂）
|
||||
- 新增 dr-translator 专职英译中（工作流改为英文工作 + 最后翻译）
|
||||
|
||||
---
|
||||
|
||||
## 6. 模型 Slug 映射表（已确认，基于 zenmux `/api/v1/models` 实时返回，2026-04-20）
|
||||
|
||||
> 任何时候要查真实可用列表：
|
||||
> ```bash
|
||||
> curl -sS "https://zenmux.ai/api/v1/models" -H "Authorization: Bearer $ZENMUX_API_KEY" | jq '.data[].id'
|
||||
> ```
|
||||
|
||||
### 6.1 Provider 架构
|
||||
|
||||
OpenCode 的自定义 provider `npm` 字段**只支持 `@ai-sdk/openai-compatible`**，不支持 `@ai-sdk/anthropic`。因此所有模型统一走 `zenmux` 的 OpenAI 兼容端点（`https://zenmux.ai/api/v1`），slug 带 vendor 前缀。
|
||||
|
||||
zenmux 的 OpenAI 兼容端点同样支持 `cache_control` 透传，由 zenmux 后端处理，cache 行为与官方 Anthropic API 一致。
|
||||
|
||||
### 6.2 Claude 系列（走 `zenmux`，slug 带 `anthropic/` 前缀）
|
||||
|
||||
| 角色 | 模型 | 完整 model 字段 | 上下文 |
|
||||
|---|---|---|---|
|
||||
| dr-plan | Claude Opus 4.7 | `zenmux-anthropic/claude-opus-4-7` | **1M** |
|
||||
| dr-pm | Claude Sonnet 4.6 | `zenmux-anthropic/claude-sonnet-4-6` | **1M** |
|
||||
| dr-analyst | Claude Sonnet 4.6 | `zenmux-anthropic/claude-sonnet-4-6` | 1M |
|
||||
| dr-polisher | Claude Sonnet 4.6 | `zenmux-anthropic/claude-sonnet-4-6` | 1M |
|
||||
| dr-reporter | Claude Sonnet 4.6 | `zenmux-anthropic/claude-sonnet-4-6` | 1M |
|
||||
| dr-searcher | Claude Haiku 4.5 | `zenmux-anthropic/claude-haiku-4-5` | 200K |
|
||||
|
||||
备用：Opus 4-7 → 4-6；Sonnet 4-6 → 4-5
|
||||
|
||||
**注意**：zenmux Anthropic 端点模型名用连字符（`4-7`），不用点（`4.7`）。baseURL 为 `https://zenmux.ai/api/anthropic/v1`。
|
||||
|
||||
### 6.3 非 Claude 系列（同样走 `zenmux`）
|
||||
|
||||
| 角色 | 模型 | 完整 model 字段 | 上下文 | 备注 |
|
||||
|---|---|---|---|---|
|
||||
| dr-chief-editor | **Gemini 3.1 Pro Preview** | `zenmux/google/gemini-3.1-pro-preview` | 1M | 总编终审首选 |
|
||||
| dr-chief-editor（备用） | Gemini 2.5 Pro | `zenmux/google/gemini-2.5-pro` | 1M | |
|
||||
| dr-verifier（首选） | **GPT-5.4** | `zenmux/openai/gpt-5.4` | 1.05M | 交叉模型（非 Claude） |
|
||||
| dr-verifier（备用 A） | Qwen3.6 Plus | `zenmux/qwen/qwen3.6-plus` | 1M | 中文研究强 |
|
||||
| dr-verifier（备用 B） | MiniMax M2.7 | `zenmux/minimax/minimax-m2.7` | 204K | 低成本交叉 |
|
||||
| dr-verifier（备用 C） | Kimi K2.5 | `zenmux/moonshotai/kimi-k2.5` | 262K | 长上下文交叉 |
|
||||
| 可选（低成本推理） | DeepSeek V3.2 Thinking | `zenmux/deepseek/deepseek-reasoner` | 128K | 极低成本 |
|
||||
| 可选（国产强模型） | GLM 5.1 | `zenmux/z-ai/glm-5.1` | 200K | |
|
||||
|
||||
### 6.4 Prompt Cache 使用要点（Claude 必读）
|
||||
|
||||
ZenMux 的 Anthropic 端点完整支持 4 种 cache 模式：
|
||||
1. **系统提示缓存**（最常见）：在 system 的最后一段加 `cache_control: {"type": "ephemeral"}` 断点
|
||||
2. **工具定义缓存**：在 tools 数组最后一个工具上加断点，所有工具一起缓存
|
||||
3. **对话历史缓存**：在每轮最后一条消息加断点，自动找最长前缀匹配
|
||||
4. **多断点组合**：最多 4 个断点，用于工具/系统/RAG/对话分别缓存
|
||||
|
||||
**最低 token 要求**：
|
||||
- Opus 4.x / Sonnet 4.x：≥ 1024 tokens 才会建缓存
|
||||
- Haiku 4.5：≥ 2048 tokens
|
||||
|
||||
**TTL**：默认 5 分钟；可指定 `"ttl": "1h"` 延长到 1 小时（写入成本 2×，读取便宜 10%）。
|
||||
|
||||
**OpenCode 行为**：`@ai-sdk/anthropic` 包会自动对长 system prompt / 工具定义打 cache_control 断点，**你不需要手动加参数**。验证方法：在 zenmux 后台 Logs 里看 `cache_creation_input_tokens` 和 `cache_read_input_tokens` 字段。
|
||||
|
||||
**Opus 4.7 定价参考**（截至 2026-04-20）：
|
||||
- 输入：25 USD/M tokens
|
||||
- cache 写入（5min）：6.25 USD/M
|
||||
- cache 写入（1h）：10 USD/M
|
||||
- **cache 读取**：**0.5 USD/M**（只有原价 2%！）
|
||||
|
||||
所以只要 cache 命中，成本可压到无 cache 的 5-10% 量级。
|
||||
|
||||
### 6.5 如何验证 cache 生效
|
||||
|
||||
1. 在 zenmux 后台 https://zenmux.ai/settings/logs 开启 **API Call Logging** 开关
|
||||
2. 启动 opencode，跑 `/dr-frame` 让 dr-plan 连续两次调用
|
||||
3. 第一次调用 Logs 应显示 `cache_creation_input_tokens > 0`
|
||||
4. 第二次调用（5 分钟内）应显示 `cache_read_input_tokens > 0`，费用大幅下降
|
||||
5. 如果 cache 字段始终为 0，说明没走 Anthropic 端点，回查 agent 的 `model:` 字段是否正确用了 `zenmux-anthropic/` 前缀
|
||||
6. 本项目提供 `scripts/verify-zenmux.sh` 一键自检
|
||||
|
||||
### 6.6 模型白名单位置
|
||||
|
||||
所有可用模型已列入 `.opencode/opencode.json` 的 `provider.zenmux.models` 和 `provider.zenmux-anthropic.models`。增删模型时**两处都要更新**：
|
||||
- opencode.json 决定 `/models` 下拉列表
|
||||
- AGENTS.md 本节决定角色→模型的分配逻辑
|
||||
|
||||
### 6.7 模型升级流程
|
||||
|
||||
zenmux 新模型上线后，更新顺序：
|
||||
1. `curl zenmux /api/v1/models` 确认 slug
|
||||
2. 更新 `.opencode/opencode.json` 的 models 段
|
||||
3. 更新 AGENTS.md §6.2 / §6.3 角色映射表
|
||||
4. 更新 `.opencode/agents/*.md` 的 `model:` 字段
|
||||
5. 运行 `bash scripts/verify-zenmux.sh` 验证
|
||||
6. 更新 PLAN.md §12 变更记录
|
||||
|
||||
---
|
||||
|
||||
## 7. 目录约定
|
||||
|
||||
- 每个研究主题放在 `projects/<topic-slug>/`，slug 用小写+连字符，如 `glp1-r-agonist-market-2026`
|
||||
- 所有中间产物（drafts、evidence、sources.jsonl）均为 Markdown 或 JSONL，便于 diff 与版本控制
|
||||
- `archive/` 存放已完成或废弃的研究，不再主动维护
|
||||
|
||||
---
|
||||
|
||||
## 8. 安全与权限
|
||||
|
||||
- API 密钥**只存** `secrets.env`（已入 gitignore），禁止硬编码到任何 agent/skill/command
|
||||
- 字体文件（~140MB）不入 git，通过 `download-fonts.sh` 获取
|
||||
- `bash` 权限默认 `ask`，仅允许 `python *` / `pandoc *` / `ls *` / `cat *` / `curl *` 自动执行
|
||||
|
||||
---
|
||||
|
||||
## 9. 如何判断 subagent 是否真正被独立调度（验证锚点）
|
||||
|
||||
用户提到过"多 agent 实际上是主模型跑到底"的坑。验证方法：
|
||||
|
||||
1. **TUI 内**：`<Leader>+Right` 能切入独立子会话，若没有说明没真正调度
|
||||
2. **日志**：`opencode --print-logs` 会显示每次 Task 工具调用，附带 agent 名和模型 ID
|
||||
3. **token 使用**：`/stats` 里可以看到按 agent 分的 token 消耗，Haiku 应远多于 Opus
|
||||
|
||||
如果发现某个 agent 没有真正被调度，检查：
|
||||
- 命令 frontmatter 是否有 `subtask: true`
|
||||
- 主 agent 的 `permission.task` 是否允许目标 subagent
|
||||
- 目标 subagent 的 `mode` 是否是 `subagent`
|
||||
|
||||
---
|
||||
|
||||
## 10. 禁止事项（negative instructions）
|
||||
|
||||
所有 agent 均禁止：
|
||||
|
||||
1. ❌ 引用 Wikipedia 作为结论支撑（仅做术语理解）
|
||||
2. ❌ 在缺乏 2 个独立信源时仍给出绝对化结论
|
||||
3. ❌ 使用"据报道""有专家认为"等未指明来源的表述
|
||||
4. ❌ 编造或虚构数据、URL、DOI
|
||||
5. ❌ 写空洞的套话（"随着科技的发展""在大数据时代"）
|
||||
6. ❌ 忽略反方观点，只收集支持证据
|
||||
7. ❌ 对输出字数"打折"（综述 <10000 字、研究 <30000 字必须返工）
|
||||
8. ❌ 在正文中使用未在术语表中定义的专业缩写（首次出现需全称+缩写）
|
||||
|
||||
---
|
||||
|
||||
## 11. 变更管理
|
||||
|
||||
- 本文件与 `PLAN.md` 是**双核**：PLAN.md 管实施进度与架构，AGENTS.md 管运行时规则
|
||||
- 修改本文件需同步更新 PLAN.md 的"变更记录"段
|
||||
- 所有 agent/skill 新增或重大调整必须在 PLAN.md §8 清单中标记完成状态
|
||||
- Treat `AGENTS.md`, `GEMINI.md`, `.agents/**`, and `PLAN.md` as production configuration.
|
||||
- Keep root rules short. Move roles to `.agents/agents.md`, constraints to `.agents/rules`, capabilities to `.agents/skills`, and phase sequencing to `.agents/workflows`.
|
||||
- When changing runtime rules or adapter behavior, update `PLAN.md` changelog.
|
||||
|
||||
@@ -0,0 +1,18 @@
|
||||
# Claude Code Project Instructions
|
||||
|
||||
This repository is Deep Research v0.20.
|
||||
|
||||
- Python core is the source of truth: `scripts/dr.py`, `scripts/runtime/**`, `configs/models.yaml`, `.agents/skills`.
|
||||
- Claude Code is a surface adapter. Do not perform core orchestration in the chat thread.
|
||||
- Use `.claude/skills/*/SKILL.md` commands such as `/dr-run`, `/dr-research`, `/dr-review`, `/dr-finalize`.
|
||||
- Keep formal research outputs Chinese-first. Search keywords, source titles, excerpts, DOI/URL and raw notes may remain English.
|
||||
- Do not modify `projects/**` unless the user is intentionally running a research project.
|
||||
|
||||
Typical commands:
|
||||
|
||||
```bash
|
||||
uv run python scripts/dr.py run "研究主题" --slug <slug> --method mckinsey_market
|
||||
uv run python scripts/dr.py research <slug> --workers 6 --execute-packets
|
||||
uv run python scripts/dr.py review <slug>
|
||||
uv run python scripts/dr.py finalize <slug>
|
||||
```
|
||||
@@ -0,0 +1,34 @@
|
||||
# GEMINI.md — Gemini / Antigravity Overrides
|
||||
|
||||
This file has higher priority than `AGENTS.md` for Gemini CLI and Antigravity.
|
||||
|
||||
## Load The Right Layer
|
||||
|
||||
- Shared baseline: `AGENTS.md`
|
||||
- Antigravity roles: `.agents/agents.md`
|
||||
- Strong research rules: `.agents/rules/`
|
||||
- Skills: `.agents/skills/`
|
||||
- Workflows: `.agents/workflows/`
|
||||
|
||||
For Antigravity native research, load `.agents/agents.md`, `.agents/rules/deep-research-antigravity.md`, `.agents/skills/antigravity-surface-adapter/SKILL.md`, `.agents/skills/method-selection/SKILL.md`, `.agents/skills/research-quality-gates/SKILL.md`, and `.agents/workflows/deep-research-native.md`.
|
||||
|
||||
Use `/memory show` or `/memory refresh` in Gemini CLI when the loaded context may be stale.
|
||||
|
||||
## Gemini Failure Modes
|
||||
|
||||
Treat Gemini-family models as high risk for false certainty in research.
|
||||
|
||||
- Never say "searched", "verified", "confirmed", "officially shows", or "according to source" unless a real tool call happened in this session or phase and the result is logged.
|
||||
- Do not use model memory, snippets, or AI summaries as evidence.
|
||||
- Every concrete fact needs a `source_id`, claim-ledger row, or a clear "to verify" label.
|
||||
- If the original source cannot be opened, write "未能访问原文，仅作为发现线索" and do not extract exact values from it.
|
||||
- Keep `search_log.jsonl`, `claims_ledger.jsonl`, and `continuation_state.json` current in Antigravity native runs.
|
||||
- Prefer "未找到足够证据" over filling gaps.
|
||||
|
||||
## Platform Boundary
|
||||
|
||||
Gemini CLI normally uses Python-core mode through `.gemini/commands/dr/*.toml` or `uv run python scripts/dr.py ...`.
|
||||
|
||||
Antigravity is different: use native mode and its model quota for research execution. Python scripts are auxiliary unless the user explicitly approves Python model-worker/API usage.
|
||||
|
||||
Do not modify `projects/**` unless the user is intentionally running a research project.
|
||||
@@ -1,8 +1,8 @@
|
||||
# Deep Research 系统方案（OpenCode 实现）
|
||||
# Deep Research 系统方案（Python Core + 多平台 Adapter）
|
||||
|
||||
> 本文件是整套方案的**单一真实源**，中断后续接时从此文件恢复上下文。
|
||||
> 最后更新：2026-04-24
|
||||
> 实施阶段：v0.10 — Codex native adapter（独立于 OpenCode）建设中
|
||||
> 最后更新：2026-05-07
|
||||
> 实施阶段：v0.20 — Skill-driven Python core 重构
|
||||
|
||||
---
|
||||
|
||||
@@ -33,8 +33,9 @@
|
||||
| DOCX 方案 | **Pandoc + reference-doc** |
|
||||
| 字数落实 | 框架阶段分配配额 + 终稿校验双保险 |
|
||||
| 交互节奏 | Phase 1 末、Phase 3 末强制确认 |
|
||||
| 并发 subagent | 3-4 个（稳，避免 API 限流） |
|
||||
| 并发执行 | Python task-card worker pool（平台 subagent 仅作可选表层能力） |
|
||||
| 中文字体 | **思源宋体 + 思源黑体 + 霞鹜文楷**，通过 `download-fonts.sh` 自动拉取 |
|
||||
| Antigravity 适配 | 使用 `.agents/rules` + `.agents/skills` 指导 Antigravity 原生执行 Deep Research；Gemini Flash 管流程，Opus/Gemini Pro 分 phase 执行，Python core 退为辅助工具 |
|
||||
|
||||
---
|
||||
|
||||
@@ -59,10 +60,40 @@
|
||||
|
||||
**Cache 行为**：Claude 走 `@ai-sdk/anthropic` 会自动打 `cache_control` 断点，zenmux 的 Anthropic 端点完整支持 4 种 cache 模式。Opus 4.7 cache read 价格仅 0.5 USD/M tokens（对比输入价 25 USD/M，节省 98%）。验证方法见 `AGENTS.md` §6.5。
|
||||
|
||||
### 2.1 Antigravity 原生模型策略
|
||||
|
||||
Antigravity 当前可选 models 包括 Gemini 3.1 Pro (High/Low)、Gemini 3 Flash、Claude Sonnet 4.6 (Thinking)、Claude Opus 4.6 (Thinking) 与 GPT-OSS 120B。Codex 使用经验显示，若 Antigravity 仍默认调用 Python core model workers，研究主流程容易回到 ZenMux，并且 packet/chapter assembly 有碎片化风险。因此 Antigravity 采用 native 模式：用 skill 指导 Antigravity 自身模型按 phase 写产物，Python core 只负责脚手架、确定性校验、引用和出稿。
|
||||
|
||||
默认策略：
|
||||
|
||||
- Surface manager：Gemini 3 Flash，负责读 skill、维护 task list、推进 phase、跑轻量命令和收集 artifact。
|
||||
- Phase 0-1：Claude Opus 4.6 (Thinking)，负责材料解读、研究方法选择、大胆假设、章节架构和成功标准。
|
||||
- Phase 2：Gemini 3.1 Pro (Low)，负责证据包、反方证据、chapter brief、初稿，优先追求速度和可控成本。
|
||||
- Phase 3：Gemini 3.1 Pro (High)，先做总编审校和证伪；若质量不足，再人工决定是否换模型复核。
|
||||
- Phase 4：Claude Opus 4.6 (Thinking)，负责最终中文统稿、Executive Summary、表达质量和交付一致性。
|
||||
- Python core 禁止默认接管 `run/research --execute-packets/assemble-chapters`；只有用户明确授权外部模型/API 消耗时才运行。
|
||||
|
||||
---
|
||||
|
||||
## 3. 完整架构
|
||||
|
||||
### 3.0 v0.20 Python Core 架构
|
||||
|
||||
v0.20 后，核心编排从平台 prompt 迁移到项目自有 Python runtime：
|
||||
|
||||
- `scripts/dr.py` 是稳定入口：`init`、`frame`、`run`、`research`、`review`、`finalize`、`skills`、`models`。
|
||||
- `scripts/runtime/*` 负责 role/task 模型解析、skill registry、task cards、packet schema、manifest 更新。
|
||||
- `.agents/skills` 是 canonical skill registry，也是 Antigravity 默认 workspace skill 目录；`.opencode/skills` 等 adapter 目录由 `dr.py skills sync` 生成。
|
||||
- OpenCode/Codex/Claude Code 只作为 surface adapter，调用 Python CLI，不再承载默认并发调度。
|
||||
- Phase 2 默认生成 `phase2/task_cards.json` 与 `phase2/packets/*.json`，减少长上下文传递。
|
||||
- Phase 2 在正式写章前生成 `phase2/chapter_briefs/*.json`，先把并发证据收束为章节主线，降低碎片化。
|
||||
- Phase 2 packet worker 对模型返回做一次 JSON 修复；仍失败的任务写入 `phase2/packet_errors/*.json`，不阻塞同批其他任务。
|
||||
- Phase 2 chapter assembly 会校验正文 `[src_xxx]` 是否来自 chapter brief；失败章写入 `phase2/chapter_errors/*.json`，不阻塞同批其他章节。
|
||||
- Phase 4 默认中文原生：`final_zh.md -> build_report`，legacy 英译中链路仅由 `--legacy-translate` 显式启用。
|
||||
- Phase 1 必须选择 `research_method`，由 `configs/research_methods.yaml` 决定框架方法和 Phase 2 task axes；MECE 不再是唯一默认。
|
||||
- 用户提供资料入口已支持 `input_materials` / `phase0/inputs` / `phase0/extracted`；PDF 文本抽取与 FireRed OCR 扫描件识别已先行落地，DOCX/PPTX/表格结构化继续放入 v0.21。
|
||||
- Antigravity 入口已落地：`.agents/rules/deep-research-antigravity.md` 约束其优先使用 Antigravity 模型配额，`.agents/skills/antigravity-surface-adapter` 提供 native runbook、模型切换和搜索策略。
|
||||
|
||||
```
|
||||
┌─────────────────────────────────────────────────────────────────┐
|
||||
│ 用户 (TUI 入口) │
|
||||
@@ -599,3 +630,163 @@ OpenCode 的坑：如果只是在主会话里装样子地写"让 X agent 做"，
|
||||
3) `scripts/search.py` 专用路由默认 strict；
|
||||
4) `dr.py finalize --model-profile <x>` 走统一 Phase 4 pipeline；
|
||||
5) `scripts/sprint5_regression.py` 全部 PASS。
|
||||
|
||||
- 2026-05-05 v0.20：**Skill-driven Python core 重构启动**
|
||||
|
||||
**目标**：把 Deep Research 从 OpenCode/Codex/Claude Code prompt 驱动，迁移为项目自有 Python runtime + skills + model profiles 驱动。平台工具只作为表层入口。
|
||||
|
||||
**已落地**：
|
||||
- 新增 `scripts/runtime/`：skills registry、role runtime、task cards、artifact helpers、orchestrator。
|
||||
- 新增 `scripts/reporting/`：引用生成与 Quarto 字体解析先行拆分，`build_report.py` 保持兼容入口。
|
||||
- 新增 `configs/research_methods.yaml` 与 `scripts/runtime/methods.py`：支持 `mckinsey_market`、`gmp_gap_assessment`、`cmc_process_risk`、`rd_go_no_go`、`management_consulting`。
|
||||
- 新增 `scripts/runtime/assembly.py`：把 packets 聚合为 chapter briefs，并通过中文章节组装 worker 生成 `phase2/drafts/chXX.md`。
|
||||
- 新增 `scripts/runtime/phase1.py` 与 `scripts/runtime/review.py`：Python core 可直接执行 init、frame、review，不再依赖 OpenCode prompt 完成 Phase 1/3 骨架。
|
||||
- `configs/models.yaml` 新增 `defaults.task_types`，模型解析同时返回 roles 与 task_types。
|
||||
- `scripts/dr.py` 新增 `init`、`frame`、`run`、`research`、`review`、`skills list|validate|sync`，`finalize` 默认走中文原生路径；legacy 翻译链路改为显式 `--legacy-translate`。
|
||||
- OpenCode/Codex 命令模板瘦身为 Python CLI wrapper，不再要求平台自行 spawn subagents 或复刻 Phase 1/3 编排逻辑。
|
||||
- 新增 `docs/platform-adapters.md`、`CLAUDE.md`、`GEMINI.md`、`.claude/skills/*`、`.gemini/commands/dr/*.toml`，明确 Codex/OpenCode/Claude Code/Antigravity/Gemini CLI 的调用方式与模型边界。
|
||||
- 新增 `scripts/deploy_adapters.py`：Codex adapter 从 `codex_adapter_templates/codex/**` 部署到 `$CODEX_HOME` 或 `~/.codex`，不再要求仓库内维护 `.codex/**`；旧 `scripts/install_codex_adapter.py` 改为兼容 wrapper。
|
||||
- 新增 `scripts/runtime/materials.py` 与 `skills/document-ingest/SKILL.md`：Phase 0 可复制用户 PDF、直接抽取文本；扫描型 PDF 自动调用 LAN FireRed OCR（默认 `http://192.168.50.100:8001`），结果写入 `phase0/extracted/*.md` 与 manifest。
|
||||
- 新增测试：runtime、CLI、reporting；新增计划中的 `scripts/v020_regression.py` 回归入口。
|
||||
|
||||
**仍需后续增强**：
|
||||
- task-card worker 已支持显式 `--execute-packets` 先检索候选 sources、再调用 ZenMux 并发生成证据包，并自动回填 `phase2/sources.jsonl`；`--build-briefs` 收束为章节 brief；`--assemble-chapters` 生成中文章节草稿。
|
||||
- packet worker 已增加一次 JSON 修复调用与失败隔离；单个 packet 失败会落盘到 `phase2/packet_errors/*.json`，不会拖垮整批并发。
|
||||
- chapter assembly 已增加引用白名单校验与失败隔离；章节正文不得新增 brief 外的 `[src_xxx]`，失败章落盘到 `phase2/chapter_errors/*.json`。
|
||||
- Phase 1 init/frame 已有可执行 Python core 骨架；后续可继续增强为模型辅助访谈与初扫，而不是回到平台 prompt 编排。
|
||||
- v0.21 需要继续实现用户资料导入 pipeline：DOCX/PPTX/图片批量 OCR、表格抽取、材料 source registry、问题清单结构化。
|
||||
- PDF 模块已开始拆分，但 ReportLab/Quarto 渲染主体仍在 `build_report.py` 与 `.opencode/templates/report-template.py` 中。
|
||||
|
||||
- 2026-05-06 v0.20-alpha：**Skill-driven Python core Alpha 与白帆案例暴露问题**
|
||||
|
||||
**Alpha 目标**：先把 Python core、skill registry、Codex adapter 外部部署、Phase0 PDF/OCR、task-card 并发、packet/brief/draft 骨架跑成可执行版本；不声明报告质量达标。
|
||||
|
||||
**已验证能力**：
|
||||
- Codex adapter 可部署到 `$CODEX_HOME`，默认不复制 `config.toml`，避免覆盖用户全局配置；`--include-config` 才安装 bundled profile。
|
||||
- `skills/deep-research`、`skills/document-ingest`、`skills/search-gateway` 已纳入 registry 并可同步到 adapter。
|
||||
- `scripts/lib/zenmux_client.py` 支持 adapter model id 规范化，并对 Opus 4.7 自动省略已废弃的 `temperature` 参数。
|
||||
- Phase0 可导入 PDF；扫描/弱文本 PDF 可走 FireRed OCR；当前白帆案例已生成 `phase0/extracted`。
|
||||
- Phase2 可生成 90 个 task cards / packets / chapter briefs；packet validation、source rebuild、stale error 识别均已可执行。
|
||||
- Phase3 deterministic review 已能把 citation 通过但 evidence 落纸不足的 draft 标为 P1 回炉。
|
||||
|
||||
**白帆案例暴露的问题**：
|
||||
- Phase0/1 原先没有先读材料形成访谈问题，就直接生成框架并推进 Phase2，用户体验和研究方向控制不足。
|
||||
- subagent 在 Codex 中可能绕开项目 Python search gateway，触发 Tavily MCP 权限确认；应禁止平台 MCP 作为默认搜索路径。
|
||||
- evidence packet 到 chapter draft 存在信息损耗：引用密度不低，但具体审计发现、法规条款、整改动作和待补证据没有充分落到纸面。
|
||||
- 单纯 `validate_packet` / citation whitelist 不足以判断报告质量；需要 evidence utilization、groundedness、specificity、actionability 等更高层质量门槛。
|
||||
|
||||
- 2026-05-06 v0.21 规划：**Research Brief + Enrichment + Compression + Evaluation**
|
||||
|
||||
**设计来源**：借鉴 `langchain-ai/open_deep_research` 的 clarification gate、research brief、bounded supervisor/researcher 并发、compression step 和 evaluator rubrics，但保留本项目 file-backed Python core、法规证据矩阵、PDF/DOCX 输出和项目内 search gateway。
|
||||
|
||||
**Phase0/1 改造**：
|
||||
- `init` 后必须生成 `phase1/material_brief.md`：材料清单、初步问题聚类、关键访谈问题、材料使用边界。
|
||||
- 新增 `phase1/research_brief.md/json`：把用户访谈、材料简报、研究方法、报告用途、范围排除项、基调和成功标准固化为 Phase2 的唯一输入。
|
||||
- `research` 默认要求 `phase1.approved=true`；用户确认后运行 `dr.py approve <slug>`，否则只能显式 `--force`。
|
||||
- clarification 不只问范围，还要输出 task 切分原则：哪些问题适合并发，哪些必须串行，弱模型需要哪些 prompt/skill/context。
|
||||
|
||||
**Phase2 改造**：
|
||||
- task card 从 `research_brief` 生成，而不是只从章节标题生成；每张卡必须包含：研究目标、调研方式、推荐 search route、必读 skills、可用材料、期望 evidence schema、停止条件。
|
||||
- 新增 `phase2/enrichment_rounds/roundXX/coverage_gap.json`：每轮先评估覆盖缺口，再生成补充 task cards；避免一次性 packet 后直接写章。
|
||||
- 新增 `phase2/compressed_findings/chXX.json`：对 packets 进行压缩，但要求保留全部关键事实、原始来源、反方证据、证据落点和待补证据。
|
||||
- `search-gateway` 成为信息收集 subagent 必读 skill：默认调用 `scripts/search.py` / `SearchClient`，不得直接用 Tavily MCP、browser MCP 或平台 web search。
|
||||
|
||||
**Phase3/4 改造**：
|
||||
- chapter draft 必须从 `compressed_findings` 写，而不是直接从 packet 拼接；每章必须包含“证据落点与待补证据”表。
|
||||
- Phase3 增加 evaluator rubrics：groundedness、completeness、relevance、structure、source quality、evidence utilization、specificity、actionability、writing quality。
|
||||
- 任一核心维度低于阈值时禁止 finalize，自动生成回炉建议和补充 task cards。
|
||||
- Final assembly 只允许使用通过 Phase3 的章节和 sources，避免把 Alpha 草稿误渲染为正式 PDF/DOCX。
|
||||
|
||||
**测试计划**：
|
||||
- fixture 项目必须覆盖：material brief -> research brief -> task cards -> enrichment round -> compressed findings -> chapter draft -> Phase3 score gate。
|
||||
- 搜索测试必须验证 subagent prompt 中包含 `search-gateway`，且不会提及 Tavily MCP 作为默认路径。
|
||||
- 质量测试必须能让“泛泛咨询腔但有引用”的章节失败，让“具体审计发现+法规条款+整改动作+待补证据”的章节通过。
|
||||
|
||||
- 2026-05-06 v0.21-alpha implementation：**Research Brief 与压缩发现先行落地**
|
||||
|
||||
**已落地**：
|
||||
- `scripts/runtime/phase1.py` 新增 `phase1/research_brief.md` 与 `phase1/research_brief.json`，在 `frame` 阶段把材料简报、研究方法、工作语言、写作基调、成功标准、任务切分原则、每个任务轴的 prompt brief / search route / required skills / stop conditions 固化为文件。
|
||||
- `scripts/runtime/orchestrator.py` 生成 Phase2 task cards 时优先读取 `research_brief.json`，不再只依赖章节标题和 method axes。
|
||||
- `scripts/runtime/tasks.py` 扩展 `TaskCard` schema：`research_goal`、`research_method`、`prompt_brief`、`required_skills`、`allowed_materials`、`expected_evidence`、`stop_conditions`、`model_hint`；旧 task card 会自动补默认字段，保持 fixture 兼容。
|
||||
- `scripts/runtime/assembly.py` 新增 `build_compressed_findings()` 与 `validate_compressed_finding()`，`--build-briefs` 会同步写入 `phase2/compressed_findings/chXX.json`。
|
||||
- `--assemble-chapters` 改为从 `compressed_findings` 写中文章节，减少并发 packet 直接拼接造成的碎片化。
|
||||
- `AGENTS.md` 已同步更新 Phase1/2 真实产物、search-gateway 默认路径、Python core 验证锚点。
|
||||
|
||||
**仍未完成**：
|
||||
- `phase2/enrichment_rounds/roundXX/coverage_gap.json` 还未实现；下一步应先做 deterministic coverage evaluator，再让补充 task cards 从 gap 生成。
|
||||
- Phase3 evaluator rubrics 仍是计划项；当前 deterministic review 已能抓部分 draft 质量问题，但还没有分维度评分与 finalize gate。
|
||||
- DOCX/PPTX/图片批量 OCR、表格抽取、材料 source registry 仍放入后续资料导入增强。
|
||||
|
||||
- 2026-05-07 v0.20/v0.21-alpha search routing refinement：**Exa evidence discovery + Tavily Research 边界定锚**
|
||||
|
||||
**设计结论**：
|
||||
- Exa 更适合作为 Phase2 的受控 evidence discovery：优先返回 highlights/text，便于进入 source-quality、evidence-table 和 packet schema。
|
||||
- Tavily Research 更适合作为 Phase1 初扫、薄弱章节补证据、Phase3 回炉扫描；其综合报告不得直接替代 evidence packet 或章节正文。
|
||||
- Serper 继续承担 Scholar、Google Patents、News 与 Google-specific `site:` 检索；Brave 用于交叉验证和混合语种 fallback。
|
||||
|
||||
**已落地**：
|
||||
- `scripts/search.py` 新增 `--route evidence` 与 `--exa-category`，profile 路由加入 `evidence`。
|
||||
- `scripts/lib/search_client.py` 新增 `SearchClient.evidence()`，优先调用 Exa highlights/text，失败后降级 Tavily/Brave。
|
||||
- `scripts/runtime/tasks.py` 把 `evidence` 纳入合法 search route，并更新主要 task axes 的默认路由。
|
||||
- `scripts/runtime/workers.py` 的 `ProjectSearchProvider` 支持 `evidence` route。
|
||||
- `skills/search-gateway`、`skills/search-strategy`、`docs/search-playbook.md`、`README.md`、`AGENTS.md` 同步记录搜索分工，避免后续又回到 Tavily MCP 或中文长句搜索。
|
||||
|
||||
- 2026-05-07 v0.20.1 Antigravity native adapter：**用 Antigravity 模型配额执行研究**
|
||||
|
||||
**设计结论**：
|
||||
- Antigravity 官方机制以 `.agents/skills`、`.agents/rules`、Rules/Workflows、Task Groups 和 browser subagent 为核心；本项目在 Antigravity 中采用 skill-driven native research，而不是默认调用 Python core model workers。
|
||||
- `AGENTS.md` 继续作为跨平台总规则；Antigravity 专项约束放到 workspace rule 和 skill，避免引入非官方 SOUL.md 作为项目真源。
|
||||
- Opus 额度有限但大局观强，优先用于 Phase 0-1 和 Phase 4；Gemini 3 Flash 做流程管理，Gemini 3.1 Pro Low/High 分别用于 Phase 2/3。
|
||||
- `paperfoot/search-cli` 可作为 Antigravity 原生检索前台：多 provider、14 modes、JSON/exit code 友好；但不得替代本项目 source tier 与 source-quality 评分。
|
||||
|
||||
**已落地**：
|
||||
- 新增 `.agents/skills/antigravity-surface-adapter/SKILL.md`，定义 Antigravity native runbook、pause points、artifact 汇报、模型切换表和 `search-cli` 使用建议。
|
||||
- 新增 `.agents/rules/deep-research-antigravity.md`，要求 Antigravity 优先使用自身模型配额，Python core 只做脚手架、确定性校验、引用和出稿。
|
||||
- 撤回 `configs/models.yaml` 中 Antigravity 专用 ZenMux profile，避免误导主流程继续走 Python/ZenMux。
|
||||
- `docs/platform-adapters.md`、`README.md`、测试用例同步更新。
|
||||
|
||||
- 2026-05-07 v0.20.2 Antigravity/Gemini research integrity hardening：**反幻觉、反假搜索、workflow gates**
|
||||
|
||||
**设计结论**：
|
||||
- Gemini/Gemini CLI/Antigravity 在 deep research 中必须默认视为高幻觉风险模型；项目规则要把"诚实暴露不确定性"写成硬门槛，而不是依赖模型自觉。
|
||||
- `GEMINI.md` 适合放短而硬的项目级行为约束，并通过层级 context 与 imports 承接 `AGENTS.md`。
|
||||
- Antigravity `rules` 适合放 Always-On 约束；`workflows` 适合固化 Phase 0-4 执行顺序、人工暂停点和质量 gates。
|
||||
- Claude/Claude Code 最佳实践可借鉴：根指令要短、具体、可执行，长流程拆到 rules/skills/workflows，避免巨型上下文降低遵从性。
|
||||
|
||||
**已落地**：
|
||||
- 初版曾在 `AGENTS.md` 中加入 Antigravity native 例外、检索回执、source_id、search_log、unsupported_claims 等反幻觉硬规则；v0.20.3/v0.20.4 已将这些内容迁入 `.agents/rules`、`.agents/skills` 与 `.agents/workflows`。
|
||||
- `GEMINI.md` 重写为短约束：禁止假搜索、禁止无 source_id 事实、要求 search log，并区分 Gemini CLI 与 Antigravity native。
|
||||
- `.agents/rules/deep-research-antigravity.md` 加入 Anti-Hallucination Contract。
|
||||
- `.agents/skills/antigravity-surface-adapter/SKILL.md` 加入 phase artifacts、fact-audit、权限建议和检索日志要求。
|
||||
- 新增 `.agents/workflows/deep-research-native.md`，把四阶段 native research 写成可执行 workflow，并在每阶段设 gate。
|
||||
- `scripts/deploy_adapters.py antigravity` 同步部署 workflows；已有文件继续默认跳过，`--force` 才备份覆盖。
|
||||
|
||||
- 2026-05-07 v0.20.3 Antigravity rule/agent/skill separation：**按 Antigravity 最佳实践重新分层**
|
||||
|
||||
**设计结论**：
|
||||
- `AGENTS.md` / `GEMINI.md` 是跨工具/Antigravity 项目规则，不应承担角色定义、技能手册和详细流程。
|
||||
- `.agents/agents.md` 用于 Antigravity 角色团队定义；`.agents/rules` 放强约束；`.agents/skills` 放可复用技能；`.agents/workflows` 放 slash workflow 和阶段编排。
|
||||
- 继续保留反幻觉约束，但从 `AGENTS.md` 的长段落中移出，由 Antigravity rule/skill/workflow 承载，避免根规则膨胀影响遵从性。
|
||||
|
||||
**已落地**：
|
||||
- 新增 `.agents/agents.md`，定义 Research Manager、Phase 0-1 Strategist、Evidence Analyst、Chief Reviewer、Final Editor。
|
||||
- 瘦身 `AGENTS.md`，只保留跨平台研究底线与分层指引。
|
||||
- `antigravity-surface-adapter` skill 和 `deep-research-native` workflow 改为引用 `.agents/agents.md`。
|
||||
- `scripts/deploy_adapters.py antigravity` 同步部署 `.agents/agents.md`，默认跳过已有文件，`--force` 才备份覆盖。
|
||||
|
||||
- 2026-05-07 v0.20.4 AGENTS/GEMINI slimdown + method selection：**根规则瘦身，研究方法按场景选择**
|
||||
|
||||
**设计结论**：
|
||||
- 根 `AGENTS.md` 只保留跨工具底线、命令入口、安全边界和分层索引；Phase 0-4 工作流、Antigravity 角色、长规则和技能细则全部迁出。
|
||||
- `GEMINI.md` 只做 Gemini/Antigravity 高优先级覆盖，强调上下文加载和反假搜索。
|
||||
- 麦肯锡/MECE/SCQA 只是候选表达和咨询工具，不再作为默认研究方法；不同研究场景必须选择匹配的分析框架。
|
||||
|
||||
**已落地**：
|
||||
- 重写 `AGENTS.md`，缩短为跨工具规则和索引。
|
||||
- 重写 `GEMINI.md`，保留 Gemini 反幻觉、平台边界和 context 加载指引。
|
||||
- 新增 `.agents/skills/method-selection/SKILL.md`，覆盖市场/投资、临床、CMC/GMP、R&D、管理、政策等方法路由。
|
||||
- 参考 `199-biotechnologies/claude-deep-research-skill` 的证据持久化、claim-level verification、delta retrieve、continuation state 和 final assembly gate 设计，新增 `.agents/skills/research-quality-gates/SKILL.md`。
|
||||
- 强化 `source-quality` 与 `evidence-table`：要求 search receipt、原文访问状态、独立性 cluster、`claims_ledger.jsonl`、`coverage_matrix.md` 和不可证实 claim 显式落盘。
|
||||
- 重写 `.agents/agents.md`，只保留角色定义和 required skills。
|
||||
- 重写 `.agents/workflows/deep-research-native.md`，把方法选择设为独立 gate，加入 claim ledger、delta retrieve、coverage audit 和 continuation state gate。
|
||||
- 文档与测试同步更新。
|
||||
|
||||
@@ -1,8 +1,8 @@
|
||||
# Deep Research 系统
|
||||
|
||||
> 生物医药行业的 AI 驱动深度研究流水线。基于 OpenCode 多 agent 协作，以麦肯锡/德勤式方法论产出专业级研究报告（PDF + DOCX）。
|
||||
> 生物医药行业的 AI 驱动深度研究流水线。v0.20 起以项目自有 Python core + skills + model profiles 为核心，以麦肯锡/德勤式方法论产出专业级研究报告（PDF + DOCX）。
|
||||
|
||||
**当前状态**：v0.13 迭代完成。新增 Quarto/xelatex PDF 引擎（`--engine quarto`），解决 ReportLab 超宽表格渲染 bug；ReportLab 引擎保留为默认后备。Quarto 依赖独立安装，不影响现有环境。
|
||||
**当前状态**：v0.20 重构启动。核心编排从平台 prompt 迁移到项目自有 Python runtime：模型选择、skills、任务卡并发、中文原生成稿都由 `scripts/dr.py` 与 `configs/models.yaml` 驱动；OpenCode/Codex/Claude Code/Antigravity/Gemini CLI 只作为表层入口。
|
||||
详见 `PLAN.md` 了解完整方案、版本记录与迭代路径。
|
||||
|
||||
---
|
||||
@@ -95,17 +95,32 @@ uv run python scripts/build_report.py <slug> --engine quarto # Quarto/xelatex
|
||||
|
||||
### 多平台兼容
|
||||
|
||||
- OpenCode：主适配器，使用 `.opencode/agents` 与 `.opencode/commands`。
|
||||
- Codex：native adapter，使用 `.codex/config.toml`、`.codex/agents`、`.codex/commands`、`.agents/skills` 与 `scripts/dr.py` 独立运行；主入口是 `dr-run`，由 Codex 主线程承担 PM 调度并主动 spawn subagents，详见 `docs/codex-usage.md`。
|
||||
- Gemini CLI / Claude Code：暂不做强适配，后续以同一套脚本与配置为基础扩展。
|
||||
v0.20 后，OpenCode/Codex/Claude Code/Gemini CLI 都是 Python core 的 surface adapter。Antigravity 走 native 模式：用 `.agents/skills` 指导其自身模型按 phase 执行研究，Python core 只做辅助。详细调用方式见 `docs/platform-adapters.md`。
|
||||
|
||||
安装 Codex adapter：
|
||||
- OpenCode：使用 `.opencode/commands/*.md` 薄封装 Python CLI。
|
||||
- Codex：使用 `AGENTS.md` + 部署到 `$CODEX_HOME` 的 adapter 文件，优先用 GPT 系列做代码/测试/审阅。
|
||||
- Claude Code：使用 `.claude/skills/*/SKILL.md`，优先用 Opus/Sonnet 做访谈、结构讨论和中文风格审阅。
|
||||
- Gemini CLI：使用 `GEMINI.md` 与 `.gemini/commands/dr/*.toml`，优先用 Gemini 做长上下文、多模态和替代框架审阅。
|
||||
- Antigravity：作为 Agent Manager 打开仓库，使用 `.agents/rules` + `.agents/skills` 指导其用 Antigravity 模型配额原生执行 Deep Research；Python core 只做脚手架、校验、引用和出稿辅助；质量门由 `method-selection`、`research-quality-gates`、`source-quality`、`evidence-table` 共同约束。
|
||||
|
||||
部署 Codex adapter（不在仓库内创建 `.codex`）：
|
||||
|
||||
```bash
|
||||
uv run python scripts/install_codex_adapter.py --force
|
||||
uv run python scripts/deploy_adapters.py codex --force
|
||||
```
|
||||
|
||||
Codex adapter 默认面向自动化研究：workspace 可写、命令不逐次审批、实时 web search 与脚本网络访问开启；Tavily / Brave / Exa MCP 会默认启用但不设为必需服务。
|
||||
Codex adapter 会写到 `$CODEX_HOME` 或 `~/.codex`；已有文件会在 `--force` 覆盖前生成 `.bak` 备份。adapter 默认面向自动化研究：workspace 可写、命令不逐次审批、实时 web search 与脚本网络访问开启；Tavily / Brave / Exa MCP 会默认启用但不设为必需服务。
|
||||
|
||||
安全默认：部署脚本不会复制 `config.toml`，避免覆盖用户级 Codex 配置。只有明确需要安装本项目 bundled profile 时，才使用 `--include-config`。
|
||||
|
||||
部署 Antigravity workspace adapter（只写目标工作区 `.agents/`，默认不覆盖已有文件）：
|
||||
|
||||
```bash
|
||||
uv run python scripts/deploy_adapters.py antigravity --target /path/to/workspace --dry-run
|
||||
uv run python scripts/deploy_adapters.py antigravity --target /path/to/workspace
|
||||
```
|
||||
|
||||
Antigravity 部署不会修改全局设置。若目标工作区已有同名 agents/rule/skill/workflow 文件，默认跳过；只有显式加 `--force` 时才会先生成 `.bak` 备份再覆盖。Antigravity 分层为：`.agents/agents.md` 定义角色，`.agents/rules` 放强约束，`.agents/skills` 放能力，`.agents/workflows` 放流程。关键 artifact 包括 `search_log.jsonl`、`sources.jsonl`、`claims_ledger.jsonl`、`coverage_matrix.md` 和 `continuation_state.json`。
|
||||
|
||||
部署到新环境后自检：
|
||||
|
||||
@@ -114,12 +129,23 @@ uv run python scripts/deploy_check.py
|
||||
uv run python scripts/deploy_check.py --repair --force
|
||||
```
|
||||
|
||||
运行 Codex 总调度：
|
||||
运行平台无关 Python core：
|
||||
|
||||
```bash
|
||||
codex exec "$(uv run python scripts/dr.py prompt dr-run <slug-or-topic>)"
|
||||
uv run python scripts/dr.py init "研究主题" --slug <slug> --method mckinsey_market
|
||||
uv run python scripts/dr.py frame <slug>
|
||||
uv run python scripts/dr.py run <slug-or-topic> --method gmp_gap_assessment
|
||||
uv run python scripts/dr.py research <slug> --workers 6
|
||||
uv run python scripts/dr.py research <slug> --workers 6 --execute-packets
|
||||
uv run python scripts/dr.py research <slug> --workers 6 --execute-packets --allow-search-fallback
|
||||
uv run python scripts/dr.py research <slug> --workers 6 --build-briefs
|
||||
uv run python scripts/dr.py research <slug> --workers 6 --assemble-chapters
|
||||
uv run python scripts/dr.py review <slug>
|
||||
uv run python scripts/dr.py finalize <slug>
|
||||
```
|
||||
|
||||
OpenCode/Codex/Claude Code/Gemini CLI adapter 只包装这些 CLI，不再承担核心调度或模型选择。Antigravity 是例外：它优先按 `.agents/skills/antigravity-surface-adapter` 用自身模型配额执行研究，避免主流程重新落回 ZenMux。
|
||||
|
||||
模型与搜索 API 选择见：
|
||||
- `docs/model-playbook.md`
|
||||
- `docs/search-playbook.md`
|
||||
@@ -127,7 +153,7 @@ codex exec "$(uv run python scripts/dr.py prompt dr-run <slug-or-topic>)"
|
||||
模型预设配置文件：
|
||||
- `configs/models.yaml`（统一预设，支持 `simple / medium / premium / cn_heavy / codex_native`）
|
||||
|
||||
推荐时机：在 `/dr-init` 访谈阶段就确定 `model_profile`，并立即执行 `apply-models`，保证 plan→pm→analyst→verifier→editor→polisher 的全流程策略一致。
|
||||
推荐时机：在 `/dr-init` 访谈阶段就确定 `model_profile`。v0.20 后模型选择优先在 Python runtime 中解析，adapter agent 文件只是兼容层。
|
||||
|
||||
命令行查看解析后的模型映射：
|
||||
|
||||
@@ -136,6 +162,10 @@ uv run python scripts/dr.py models
|
||||
uv run python scripts/dr.py models --list
|
||||
uv run python scripts/dr.py models --profile premium
|
||||
uv run python scripts/dr.py models --profile medium --model-override dr_verifier=zenmux/openai/gpt-5.4
|
||||
uv run python scripts/dr.py skills validate
|
||||
uv run python scripts/dr.py skills sync
|
||||
uv run python scripts/dr.py methods list
|
||||
uv run python scripts/dr.py methods show gmp_gap_assessment
|
||||
|
||||
# apply profile to agent files
|
||||
uv run python scripts/dr.py apply-models --profile medium --target both --dry-run
|
||||
@@ -152,9 +182,12 @@ uv run python scripts/sprint5_regression.py <slug>
|
||||
|
||||
```bash
|
||||
uv run python scripts/search.py "dual-target RNAi 2024" --route scholar --year-low 2023
|
||||
uv run python scripts/search.py "FDA warning letter CAPA deviation change control pharmaceutical" --route evidence --json --trace
|
||||
uv run python scripts/search.py "dual-target siRNA GalNAc" --route patents
|
||||
```
|
||||
|
||||
v0.20 搜索分工：`evidence` 用 Exa highlights 做受控候选证据发现；`scholar/patents/news/fda` 保留专用路由；`general` 只做宽泛发现和兜底；Tavily Research 作为 Phase1 初扫、薄弱章节补证据和 Phase3 回炉工具，结果必须存盘、评分、去重后再进入 evidence packet。Antigravity 原生流程可优先使用 `paperfoot/search-cli` 的 `search` 命令作为统一检索前台，但所有结论仍必须回到本项目 Tier/source-quality 体系。
|
||||
|
||||
---
|
||||
|
||||
## 项目结构
|
||||
@@ -162,7 +195,7 @@ uv run python scripts/search.py "dual-target siRNA GalNAc" --route patents
|
||||
```
|
||||
deep_research/
|
||||
├── PLAN.md # 完整方案（中断续接从此读起）
|
||||
├── AGENTS.md # 研究方法论与规则（OpenCode 自动加载）
|
||||
├── AGENTS.md # 跨平台研究方法论与运行规则
|
||||
├── README.md # 本文件
|
||||
├── secrets.env.example # 密钥模板
|
||||
├── secrets.env # 你的密钥（gitignore）
|
||||
@@ -201,15 +234,9 @@ deep_research/
|
||||
|
||||
## 关键设计要点
|
||||
|
||||
### 1. 防止"多 agent 变单模型跑"
|
||||
### 1. Python core 防止上下文污染
|
||||
|
||||
OpenCode 的常见陷阱：AI 在主会话里装样子地"委派"子 agent，实际还是主模型在跑。本项目通过 3 道保险避免：
|
||||
|
||||
1. **命令 `subtask: true`** — 强制走 Task 工具起子会话
|
||||
2. **Agent 强绑 `model`** — 每个 subagent 锁死具体模型
|
||||
3. **`permission.task` 白名单** — 精确限定调用关系
|
||||
|
||||
验证方法：TUI 里 `<Leader>+Right` 切入子会话，能看到真实在跑的模型名。
|
||||
OpenCode/Codex/Claude Code/Antigravity/Gemini CLI 的 subagent 或 agent thread 都可能把大上下文带入子会话。v0.20 起，默认并发由 Python runtime 的任务卡执行层控制：先生成 `phase1/research_brief.json` 固化研究意图和任务切分，再生成 `phase2/task_cards.json`，按任务卡产出 `phase2/packets/*.json`，经 `phase2/compressed_findings/*.json` 收束后再组装中文章节。平台 agent 只负责调用 CLI 和展示状态。
|
||||
|
||||
### 2. 信源分级（Tier 1-4 + 黑名单）
|
||||
|
||||
@@ -309,9 +336,9 @@ Opus 4.7 cache 读取价格 0.5 USD/M tokens（对比输入 25 USD/M，节省 98
|
||||
|
||||
**验证 cache 是否生效**：
|
||||
1. 在 https://zenmux.ai/settings/logs 打开 API Call Logging
|
||||
2. 运行 `/dr-frame` 让 dr-plan 连续调用 2 次
|
||||
2. 运行一个真实 Claude/ZenMux 调用链路，例如 `uv run python scripts/dr.py research <slug> --execute-packets`
|
||||
3. 第 2 次的 `cache_read_input_tokens` 字段应 > 0
|
||||
4. 若始终为 0，检查 agent 的 `model:` 是否以 `zenmux-anthropic/` 开头（详见 `AGENTS.md` §6.5）
|
||||
4. 若始终为 0，检查 `configs/models.yaml` 中对应 role 是否走 `zenmux-anthropic/...`
|
||||
|
||||
---
|
||||
|
||||
@@ -324,11 +351,12 @@ Opus 4.7 cache 读取价格 0.5 USD/M tokens（对比输入 25 USD/M，节省 98
|
||||
4. [ ] 用一个小主题（如"5000 字 PD-1 综述"）跑通 MVP 流水线
|
||||
|
||||
### 系统侧（下一阶段）
|
||||
- [ ] dr-chief-editor / dr-searcher / dr-analyst / dr-verifier / dr-polisher / dr-reporter 6 个 subagent
|
||||
- [ ] `/dr-research` `/dr-review` `/dr-finalize` `/dr-status` 4 个命令
|
||||
- [ ] 生物医药专业信源 skill：PubMed / ClinicalTrials / openFDA / 专利 / 金融
|
||||
- [ ] citation-manager / evidence-table / mckinsey-method / docx-pandoc / report-template 5 个辅助 skill
|
||||
- [ ] Pandoc reference-doc 模板（中文 DOCX）
|
||||
- [x] Python core `init/frame/research/review/finalize/status` 骨架
|
||||
- [x] OpenCode/Codex/Claude Code/Gemini CLI wrapper
|
||||
- [x] Antigravity 专用 workspace rule + skill，按 Antigravity 模型配额原生执行研究
|
||||
- [x] 用户资料导入基础能力：PDF 文本抽取 + FireRed OCR 扫描件识别 + phase0 落盘
|
||||
- [ ] 用户资料导入增强：DOCX / PPTX / 表格抽取 / 版面结构化
|
||||
- [ ] PDF reporting 包继续拆分：字体、宽表、引用、渲染验证
|
||||
|
||||
---
|
||||
|
||||
@@ -351,11 +379,15 @@ which npx
|
||||
npx -y tavily-mcp@latest
|
||||
```
|
||||
|
||||
### subagent 没被真正调度
|
||||
1. 检查 agent frontmatter 的 `mode` 字段是否为 `subagent`
|
||||
2. 检查命令 frontmatter 是否有 `subtask: true`
|
||||
3. 检查主 agent 的 `permission.task` 是否允许目标 subagent
|
||||
4. 在 TUI 用 `<Leader>+Right` 看是否有独立子会话
|
||||
### 平台 agent 看起来没有真正并发
|
||||
v0.20 不再用平台 subagent 作为默认并发机制。请检查 Python core 产物：
|
||||
|
||||
```bash
|
||||
uv run python scripts/dr.py status <slug>
|
||||
ls projects/<slug>/phase2/task_cards.json
|
||||
ls projects/<slug>/phase2/packets
|
||||
ls projects/<slug>/phase2/chapter_briefs
|
||||
```
|
||||
|
||||
### ReportLab PDF 中文乱码
|
||||
```bash
|
||||
@@ -457,6 +489,7 @@ direnv allow
|
||||
|
||||
- OpenCode 文档：https://opencode.ai/docs
|
||||
- Agent 配置：https://opencode.ai/docs/agents
|
||||
- 跨平台调用：`docs/platform-adapters.md`
|
||||
- Skill 配置：https://opencode.ai/docs/skills
|
||||
- MCP Servers：https://opencode.ai/docs/mcp-servers
|
||||
- ReportLab 文档：https://docs.reportlab.com
|
||||
@@ -474,5 +507,7 @@ direnv allow
|
||||
- **v0.2** (2026-04-20) — 双 provider 架构（zenmux-anthropic + zenmux），解决 Claude prompt cache 生效问题
|
||||
- **v0.3** (2026-04-20) — 修正 v0.2 模型名（回到 Opus 4.7 / Sonnet 4.6 / Gemini 3.1 Pro / GPT-5.4 Pro 等真实 slug）；改 venv + requirements.txt 跨平台方案（macOS + Debian）；新增 `scripts/setup.sh`、`scripts/activate.sh`
|
||||
- **v0.13** (2026-05-02) — `build_report.py` 新增 `--engine quarto` 选项：Quarto 1.9 + xelatex 引擎，解决 ReportLab 超宽表格渲染 bug（`negative availWidth`/`NoneType` 问题）；`report-template.py` 同步修复（`render_table_blocks` 分块 + 等宽列强制分配）；README 补充双引擎安装指南与排错
|
||||
- **v0.20** (2026-05-05) — Skill-driven Python core 重构启动：新增 `scripts/runtime/*`、`scripts/dr.py init/frame/run/research/review/skills`、task-type 模型映射、中文原生 finalize 默认路径和 `scripts/reporting/*` 报告模块；OpenCode/Codex/Claude Code/Gemini CLI 命令降级为 Python CLI wrapper。
|
||||
- **v0.20.1** (2026-05-07) — Antigravity-native adapter：新增 `.agents/rules/deep-research-antigravity.md`、`.agents/skills/antigravity-surface-adapter`，按 Gemini Flash 管理、Opus/Gemini 分 phase 执行研究；Python core 退为辅助工具。
|
||||
|
||||
见 `PLAN.md` §12 了解完整变更历史。
|
||||
|
||||
@@ -1,25 +1,15 @@
|
||||
name = "dr-analyst"
|
||||
description = "Chapter deep-research agent that writes English chapter drafts and evidence matrices."
|
||||
description = "Compatibility role only. v0.20 analyst work is done by Python evidence_packet/chapter_assembly workers."
|
||||
model = "zenmux-anthropic/claude-sonnet-4-6"
|
||||
model_reasoning_effort = "high"
|
||||
sandbox_mode = "workspace-write"
|
||||
developer_instructions = """
|
||||
You are dr-analyst.
|
||||
Work in English. Own exactly one assigned chapter.
|
||||
Load skills: search-strategy, source-quality, length-budget, evidence-table, mckinsey-method, humanizer-cn.
|
||||
Use the project search gateway before MCP or generic web search:
|
||||
- literature/reviews: uv run python scripts/search.py "<query>" --route scholar --num-results 10 --year-low 2023
|
||||
- patents/FTO: uv run python scripts/search.py "<query>" --route patents --num-results 10
|
||||
- news/transactions: uv run python scripts/search.py "<query>" --route news --num-results 10 --time-range m
|
||||
- general gap-fill: uv run python scripts/search.py "<query>" --route general --num-results 10
|
||||
Record the routes used in the evidence file. Tavily / Exa / Brave MCP are gap-fill only for literature and patent topics.
|
||||
Write:
|
||||
- projects/<slug>/phase2/drafts/chXX.md
|
||||
- projects/<slug>/phase2/evidence/chXX-evidence.md
|
||||
- append structured sources to projects/<slug>/phase2/sources.jsonl
|
||||
Every claim and numerical fact needs [src_xxx].
|
||||
Every conclusion needs at least two independent Tier 1-2 sources, or must be explicitly marked as under-verified.
|
||||
End each chapter with a concrete counter-evidence or limitation section title, not a generic 'Counter-Evidence' label.
|
||||
Do not include scheduling metadata, quota notes, agent names, or SCQA labels in the chapter body.
|
||||
You are a compatibility role only.
|
||||
Do not write chapter drafts in Codex by default.
|
||||
Use Python core for analyst work:
|
||||
- uv run python scripts/dr.py research <slug> --execute-packets
|
||||
- uv run python scripts/dr.py research <slug> --build-briefs
|
||||
- uv run python scripts/dr.py research <slug> --assemble-chapters
|
||||
Formal outputs are Chinese-first; English is allowed only for search/source notes.
|
||||
"""
|
||||
nickname_candidates = ["Analyst A", "Analyst B", "Analyst C", "Analyst D"]
|
||||
|
||||
@@ -1,15 +1,13 @@
|
||||
name = "dr-chief-editor"
|
||||
description = "Phase 3 read-only editorial reviewer for whole-report logic, evidence, MECE, and quality."
|
||||
description = "Surface adapter role for optional Phase 3 deep review. Deterministic review lives in Python."
|
||||
model = "zenmux/google/gemini-3.1-pro-preview"
|
||||
model_reasoning_effort = "xhigh"
|
||||
sandbox_mode = "read-only"
|
||||
developer_instructions = """
|
||||
You are dr-chief-editor.
|
||||
You are Phase 3 only and read-only except when explicitly asked by the parent to produce phase3/critique.md.
|
||||
Review all English drafts, evidence files, sources.jsonl, framework.md, and manifest.json.
|
||||
Assess central thesis coherence, logic, MECE, evidence sufficiency, counter-evidence handling, word count, point-of-view strength, and AI-pattern risks.
|
||||
Do not rewrite drafts.
|
||||
You are a surface adapter for optional deep review.
|
||||
Default review command:
|
||||
- uv run python scripts/dr.py review <slug>
|
||||
You may explain or supplement phase3/critique.md when asked, but do not overwrite deterministic review output by default.
|
||||
Do not enter Phase 4.
|
||||
Do not spawn subagents.
|
||||
"""
|
||||
nickname_candidates = ["Chief Editor"]
|
||||
|
||||
@@ -1,15 +1,13 @@
|
||||
name = "dr-editor-in-chief"
|
||||
description = "Phase 4 lead editor for English final assembly and deterministic script orchestration."
|
||||
description = "Surface adapter role for Phase 4. Chinese-native finalization lives in Python."
|
||||
model = "zenmux-anthropic/claude-opus-4-7"
|
||||
model_reasoning_effort = "xhigh"
|
||||
sandbox_mode = "workspace-write"
|
||||
developer_instructions = """
|
||||
You are dr-editor-in-chief.
|
||||
Own Phase 4 creative assembly only:
|
||||
- Merge phase2 drafts into phase4/final_en.md.
|
||||
- Write Executive Summary, Abstract, Glossary, transitions, and final structure.
|
||||
- Remove scheduling metadata and output-hygiene violations.
|
||||
Do not translate the whole report yourself. Use scripts/dr.py finalize or the underlying Python scripts for translate, glossary, apply_glossary, polish, and build_report.
|
||||
Keep citations intact.
|
||||
You are a surface adapter for Phase 4.
|
||||
Default finalization command:
|
||||
- uv run python scripts/dr.py finalize <slug>
|
||||
Do not merge final_en.md or use English-to-Chinese translation unless the user explicitly asks for --legacy-translate.
|
||||
Do not translate or polish the full report manually in Codex.
|
||||
"""
|
||||
nickname_candidates = ["Editor in Chief"]
|
||||
|
||||
@@ -1,18 +1,15 @@
|
||||
name = "dr-plan"
|
||||
description = "Deep Research framework planner for Phase 1 interview, initial scan synthesis, and bilingual research framework."
|
||||
description = "Surface adapter role for Phase 1. Core init/frame orchestration lives in Python."
|
||||
model = "zenmux-anthropic/claude-opus-4-7"
|
||||
model_reasoning_effort = "high"
|
||||
sandbox_mode = "workspace-write"
|
||||
developer_instructions = """
|
||||
You are dr-plan for the biomedical Deep Research system.
|
||||
Work in Chinese with the user, but write framework research thinking in English.
|
||||
Follow AGENTS.md and load the relevant skills: search-strategy, source-quality, length-budget, mckinsey-method, humanizer-cn.
|
||||
Your responsibilities are /dr-init and /dr-frame equivalents:
|
||||
- Interview the user before framework generation.
|
||||
- Propose formal report titles.
|
||||
- Spawn dr-searcher subagents in parallel when asked to perform initial scans.
|
||||
- Generate phase1/framework.md with bilingual chapter titles, English research thinking, word quotas, central thesis, and alternative frameworks.
|
||||
Do not perform Phase 2 chapter deep research yourself.
|
||||
Do not enter Phase 4.
|
||||
You are a surface adapter for Deep Research v0.20.
|
||||
Do not perform Phase 1 core orchestration in Codex.
|
||||
Run the Python core:
|
||||
- uv run python scripts/dr.py init <topic>
|
||||
- uv run python scripts/dr.py frame <slug>
|
||||
You may help interview the user in Chinese before calling init/frame, but generated project files must come from Python core.
|
||||
Do not spawn subagents for initial scan by default.
|
||||
"""
|
||||
nickname_candidates = ["Planner Alpha", "Planner Beta", "Planner Gamma"]
|
||||
|
||||
@@ -1,18 +1,16 @@
|
||||
name = "dr-pm"
|
||||
description = "Deep Research project manager for Phase 2 batching, analyst/verifier orchestration, and project status."
|
||||
description = "Surface adapter role for Phase 2/status. Core batching and concurrency live in Python."
|
||||
model = "zenmux-anthropic/claude-sonnet-4-6"
|
||||
model_reasoning_effort = "high"
|
||||
sandbox_mode = "workspace-write"
|
||||
developer_instructions = """
|
||||
You are dr-pm for the biomedical Deep Research system.
|
||||
Use English for Phase 2 working outputs.
|
||||
Follow AGENTS.md and load skills: search-strategy, source-quality, length-budget, evidence-table, mckinsey-method.
|
||||
Your responsibilities:
|
||||
- Read manifest.json and phase1/framework.md.
|
||||
- Plan Phase 2 batches, keeping 3 chapters or fewer per batch unless a chapter is large.
|
||||
- Spawn dr-analyst subagents in parallel for chapter drafts.
|
||||
- Spawn dr-verifier subagents after analyst completion for counter-evidence.
|
||||
- Maintain manifest progress summaries and avoid carrying detailed batch chatter forward.
|
||||
- Never write final reports directly.
|
||||
You are a surface adapter for Deep Research v0.20.
|
||||
Do not batch chapters or spawn Codex subagents for research.
|
||||
Run the Python core:
|
||||
- uv run python scripts/dr.py research <slug> --workers 6
|
||||
- uv run python scripts/dr.py research <slug> --workers 6 --execute-packets
|
||||
- uv run python scripts/dr.py research <slug> --workers 6 --build-briefs
|
||||
- uv run python scripts/dr.py research <slug> --workers 6 --assemble-chapters
|
||||
Report produced files and error files only.
|
||||
"""
|
||||
nickname_candidates = ["PM Alpha", "PM Beta", "PM Gamma"]
|
||||
|
||||
@@ -1,25 +1,17 @@
|
||||
# Codex Command: dr-finalize
|
||||
|
||||
You are dr-editor-in-chief. The user requested `/dr-finalize $ARGUMENTS`.
|
||||
Codex is a surface adapter for v0.20. Chinese-native finalization is the default.
|
||||
|
||||
Goal: run Phase 4 in Codex native mode.
|
||||
|
||||
Steps:
|
||||
1. Resolve `$ARGUMENTS` as project slug.
|
||||
2. Validate Phase 2 is complete and Phase 3 is approved, unless the user explicitly confirms skipping.
|
||||
3. Assemble `phase4/final_en.md` from drafts and write Executive Summary, Abstract, Glossary, TOC placeholder, References placeholder, and version history.
|
||||
4. Run deterministic pipeline:
|
||||
Run:
|
||||
|
||||
```bash
|
||||
uv run python scripts/dr.py finalize <slug> --translate-workers 4 --glossary-workers 4 --polish-workers 4
|
||||
uv run python scripts/dr.py finalize $ARGUMENTS
|
||||
```
|
||||
|
||||
5. If network/API errors occur, rerun with lower workers:
|
||||
Legacy English-to-Chinese pipeline is opt-in only:
|
||||
|
||||
```bash
|
||||
uv run python scripts/dr.py finalize <slug> --translate-workers 1 --glossary-workers 3 --polish-workers 1
|
||||
uv run python scripts/dr.py finalize $ARGUMENTS --legacy-translate
|
||||
```
|
||||
|
||||
6. Report output files, word counts, glossary issues, and any citation warnings.
|
||||
|
||||
Do not translate or polish the full report manually in one LLM response.
|
||||
|
||||
@@ -1,17 +1,11 @@
|
||||
# Codex Command: dr-frame
|
||||
|
||||
You are dr-plan. The user requested `/dr-frame $ARGUMENTS`.
|
||||
Thin wrapper around the platform-neutral Python core.
|
||||
|
||||
Goal: generate Phase 1 bilingual framework for the target project.
|
||||
Run:
|
||||
|
||||
Steps:
|
||||
1. Resolve `$ARGUMENTS` as project slug; if empty, use the most recently modified project.
|
||||
2. Read `manifest.json` and validate Phase 1 interview is complete.
|
||||
3. Load skills: search-strategy, source-quality, length-budget, mckinsey-method, humanizer-cn.
|
||||
4. Spawn 3-4 `dr-searcher` subagents in parallel for MECE keyword groups. Wait for all results.
|
||||
5. Synthesize `phase1/initial-scan.md`.
|
||||
6. Write `phase1/framework.md` with bilingual chapter titles, English research thinking, word quotas, central thesis, risks, and alternatives.
|
||||
7. Update manifest Phase 1 fields.
|
||||
8. Stop and ask the user to approve the framework before Phase 2.
|
||||
```bash
|
||||
uv run python scripts/dr.py frame $ARGUMENTS
|
||||
```
|
||||
|
||||
Do not do Phase 2 research in this command.
|
||||
Stop after writing `phase1/framework.md`; wait for user approval before Phase 2.
|
||||
|
||||
@@ -1,14 +1,11 @@
|
||||
# Codex Command: dr-init
|
||||
|
||||
You are dr-plan. The user requested `/dr-init $ARGUMENTS`.
|
||||
Thin wrapper around the platform-neutral Python core.
|
||||
|
||||
Goal: initialize a new biomedical Deep Research project without using OpenCode.
|
||||
Run:
|
||||
|
||||
Follow AGENTS.md, then:
|
||||
1. Interview the user with the 8 required questions from AGENTS.md and the existing OpenCode workflow.
|
||||
2. Propose 3 formal report title/subtitle candidates.
|
||||
3. After the user chooses, create `projects/<slug>/manifest.json` and the phase directories.
|
||||
4. Write the interview transcript to `projects/<slug>/phase1/interview.md`.
|
||||
5. Stop after initialization. Do not run `/dr-frame`.
|
||||
```bash
|
||||
uv run python scripts/dr.py init $ARGUMENTS
|
||||
```
|
||||
|
||||
Use Codex custom agent `dr-plan` if spawning is needed, but this command can usually run in the main thread.
|
||||
Stop after initialization. Next step is `dr-frame`.
|
||||
|
||||
@@ -1,18 +1,41 @@
|
||||
# Codex Command: dr-research
|
||||
|
||||
You are dr-pm. The user requested `/dr-research $ARGUMENTS`.
|
||||
Codex is a surface adapter for v0.20. Core Phase 2 orchestration lives in Python.
|
||||
|
||||
Goal: run Phase 2 deep research using Codex custom subagents.
|
||||
Run:
|
||||
|
||||
Steps:
|
||||
1. Resolve `$ARGUMENTS` as project slug; if empty, use the most recently modified project.
|
||||
2. Validate `phase1.approved == true` and framework exists.
|
||||
3. Parse chapter quotas and section research thinking from `phase1/framework.md`.
|
||||
4. Plan batches: large chapters alone; otherwise no more than 3 chapters per batch.
|
||||
5. For each batch, spawn `dr-analyst` subagents in parallel, one per chapter.
|
||||
6. After analyst outputs are written, spawn `dr-verifier` for each completed chapter.
|
||||
7. Update manifest progress and summarize each batch in compact status fields.
|
||||
8. Deduplicate `phase2/sources.jsonl`.
|
||||
9. Report totals and stop before Phase 3.
|
||||
```bash
|
||||
uv run python scripts/dr.py research $ARGUMENTS --workers 6
|
||||
```
|
||||
|
||||
Do not write the final report.
|
||||
Fill packets with model workers:
|
||||
|
||||
```bash
|
||||
uv run python scripts/dr.py research $ARGUMENTS --workers 6 --execute-packets
|
||||
```
|
||||
|
||||
For low-cost smoke tests where generic fallback is acceptable:
|
||||
|
||||
```bash
|
||||
uv run python scripts/dr.py research $ARGUMENTS --workers 6 --execute-packets --allow-search-fallback
|
||||
```
|
||||
|
||||
Aggregate packets into chapter briefs:
|
||||
|
||||
```bash
|
||||
uv run python scripts/dr.py research $ARGUMENTS --workers 6 --build-briefs
|
||||
```
|
||||
|
||||
Assemble Chinese chapter drafts:
|
||||
|
||||
```bash
|
||||
uv run python scripts/dr.py research $ARGUMENTS --workers 6 --assemble-chapters
|
||||
```
|
||||
|
||||
Preview:
|
||||
|
||||
```bash
|
||||
uv run python scripts/dr.py research $ARGUMENTS --workers 6 --dry-run
|
||||
```
|
||||
|
||||
Do not perform chapter research in the Codex thread. Report the task-card and packet paths created by the CLI.
|
||||
|
||||
@@ -1,14 +1,11 @@
|
||||
# Codex Command: dr-review
|
||||
|
||||
You are dr-chief-editor. The user requested `/dr-review $ARGUMENTS`.
|
||||
Thin wrapper around the platform-neutral Python core.
|
||||
|
||||
Goal: perform Phase 3 whole-report editorial review.
|
||||
Run:
|
||||
|
||||
Steps:
|
||||
1. Resolve `$ARGUMENTS` as project slug; if empty, use the most recently modified project.
|
||||
2. Validate `phase2.status == "completed"`.
|
||||
3. Read framework, all drafts, all evidence files, sources.jsonl, and manifest.
|
||||
4. Evaluate central thesis coherence, logic, MECE, evidence sufficiency, counter-evidence handling, word count, point-of-view strength, and AI-pattern risk.
|
||||
5. Write `projects/<slug>/phase3/critique.md`.
|
||||
6. Report rating A/B/C/D and must-fix items.
|
||||
7. Stop and wait for user decision. Do not enter Phase 4.
|
||||
```bash
|
||||
uv run python scripts/dr.py review $ARGUMENTS
|
||||
```
|
||||
|
||||
Stop after writing `phase3/critique.md`; wait for user decision before Phase 4.
|
||||
|
||||
@@ -1,77 +1,17 @@
|
||||
# Codex Command: dr-run
|
||||
|
||||
You are the Deep Research PM main thread for `/dr-run $ARGUMENTS`.
|
||||
Codex is a surface adapter for v0.20. Do not spawn Codex subagents for the core workflow.
|
||||
|
||||
This command is the Codex equivalent of letting `dr-pm` own orchestration. Do not ask the user to run each phase manually. Inspect the project state, continue from the correct phase, spawn the required Codex custom agents, wait for their results, and only pause at the defined human decision gates.
|
||||
|
||||
## Resolve Target
|
||||
|
||||
1. Treat `$ARGUMENTS` as either a project slug/path or a new research topic.
|
||||
2. If it matches an existing project, load `projects/<slug>/manifest.json` and continue from its current status.
|
||||
3. If it is a new topic, run Phase 1 initialization and create the project structure before planning.
|
||||
4. If `$ARGUMENTS` is empty, use the most recently modified project and confirm the inferred slug in your first status update.
|
||||
|
||||
## Global Rules
|
||||
|
||||
- Follow `AGENTS.md`, all relevant `.agents/skills/*/SKILL.md`, and the project `manifest.json`.
|
||||
- Keep `projects/**` research artifacts out of system commits unless the user explicitly asks to commit research output.
|
||||
- Use Codex subagents deliberately: spawn them when parallel work materially advances the phase, wait for results, and consolidate before moving on.
|
||||
- Do not simulate subagent work in the main thread when the task calls for independent review, chapter research, or counter-evidence.
|
||||
- Respect the required pause gates:
|
||||
- Stop after Phase 1 framework is ready and ask the user to approve or revise it.
|
||||
- Stop after Phase 3 critique is ready and ask whether to fix, rerun Phase 2, or restart.
|
||||
- Ask for approval before expensive broad web searches, long-running external API work, or irreversible file operations.
|
||||
|
||||
## Phase 1: Framework Planning
|
||||
|
||||
Run this phase when there is no approved `phase1/framework.md`.
|
||||
|
||||
1. Spawn `dr-plan` to interview the user if the topic is underspecified.
|
||||
2. Spawn `dr-searcher` workers in parallel for initial source discovery across scientific, regulatory, clinical, commercial, and patent angles.
|
||||
3. Have `dr-plan` synthesize a MECE framework with chapter-level word budgets and source strategy.
|
||||
4. Write `phase1/interview.md`, `phase1/initial-scan.md`, and `phase1/framework.md`.
|
||||
5. Update `manifest.json` and stop for user framework approval.
|
||||
|
||||
## Phase 2: Deep Research
|
||||
|
||||
Run this phase when `phase1.approved == true` and Phase 2 is incomplete.
|
||||
|
||||
1. Act as `dr-pm`: parse `phase1/framework.md`, create chapter batches, and keep the main thread focused on orchestration.
|
||||
2. Spawn `dr-analyst` subagents in parallel, one per chapter or chapter group depending on size.
|
||||
3. Require each analyst to write English chapter drafts plus evidence matrices.
|
||||
4. After analyst outputs are present, spawn independent `dr-verifier` subagents for counter-evidence and source-quality checks.
|
||||
5. Reconcile verifier findings into the relevant evidence files and chapter TODOs.
|
||||
6. Deduplicate and normalize `phase2/sources.jsonl`.
|
||||
7. Update `manifest.json` and continue to Phase 3 unless the evidence base is materially inadequate.
|
||||
|
||||
## Phase 3: Chief Editor Review
|
||||
|
||||
Run this phase when Phase 2 is complete and Phase 3 is incomplete.
|
||||
|
||||
1. Spawn `dr-chief-editor` as a read-only reviewer.
|
||||
2. Have it assess MECE structure, evidence sufficiency, unsupported claims, source quality, chapter balance, and executive-level narrative.
|
||||
3. Write `phase3/critique.md`.
|
||||
4. Stop for user decision: targeted fix, rerun Phase 2 for weak chapters, or restart framework.
|
||||
|
||||
## Phase 4: Finalization
|
||||
|
||||
Run this phase only after the user approves Phase 3.
|
||||
|
||||
1. Spawn `dr-editor-in-chief` to merge English drafts into `phase4/final_en.md`, Executive Summary, Abstract, and glossary seed.
|
||||
2. Run the deterministic pipeline with:
|
||||
Run the project-owned Python core:
|
||||
|
||||
```bash
|
||||
uv run python scripts/dr.py finalize <slug>
|
||||
uv run python scripts/dr.py run $ARGUMENTS --workers 6
|
||||
```
|
||||
|
||||
3. If needed, spawn `dr-reporter` for final PDF/DOCX validation and citation backfill checks.
|
||||
4. Report final artifact paths and remaining risks.
|
||||
For a preview:
|
||||
|
||||
## Status Discipline
|
||||
```bash
|
||||
uv run python scripts/dr.py run $ARGUMENTS --workers 6 --dry-run
|
||||
```
|
||||
|
||||
Give concise progress updates after each phase or batch. Always say:
|
||||
|
||||
- current phase
|
||||
- agents spawned and why
|
||||
- files produced or changed
|
||||
- whether the workflow is continuing or waiting for user decision
|
||||
Report only the CLI phase decision, produced files, and next step.
|
||||
|
||||
+10
-1
@@ -2,7 +2,16 @@ version: 1
|
||||
|
||||
defaults:
|
||||
profile: medium
|
||||
task_types:
|
||||
source_discovery: dr_searcher
|
||||
evidence_packet: dr_analyst
|
||||
chapter_assembly: dr_analyst
|
||||
counter_verification: dr_verifier
|
||||
phase3_review: dr_chief_editor
|
||||
final_editorial: dr_editor_in_chief
|
||||
report_render: dr_reporter
|
||||
script_models:
|
||||
# Legacy compatibility only. v0.20 defaults to Chinese-native finalization.
|
||||
translate: anthropic/claude-sonnet-4.6
|
||||
glossary: anthropic/claude-haiku-4.5
|
||||
polish: anthropic/claude-sonnet-4.6
|
||||
@@ -69,7 +78,7 @@ profiles:
|
||||
polish: anthropic/claude-sonnet-4.6
|
||||
|
||||
codex_native:
|
||||
description: OpenAI-native profile for Codex adapter runs.
|
||||
description: Deprecated/misleading name. These are OpenAI models through the external Python API client, not Codex App built-in models.
|
||||
roles:
|
||||
dr_plan: gpt-5.4
|
||||
dr_pm: gpt-5.4
|
||||
|
||||
@@ -0,0 +1,148 @@
|
||||
version: 1
|
||||
defaults:
|
||||
method: mckinsey_market
|
||||
|
||||
methods:
|
||||
mckinsey_market:
|
||||
name: McKinsey-style market and strategy research
|
||||
best_for:
|
||||
- market research
|
||||
- investment memo
|
||||
- competitive landscape
|
||||
structure_principle: Pyramid principle with MECE chapter coverage.
|
||||
task_axes:
|
||||
- literature
|
||||
- regulatory
|
||||
- patents
|
||||
- market
|
||||
- counter
|
||||
integrated_lanes:
|
||||
- literature evidence
|
||||
- regulatory pathway
|
||||
- patent/IP position
|
||||
- market and competitor evidence
|
||||
- counter-evidence and uncertainty
|
||||
framework_sections:
|
||||
- central_thesis
|
||||
- chapter_outline
|
||||
- alternative_frameworks
|
||||
- risks_and_dependencies
|
||||
|
||||
gmp_gap_assessment:
|
||||
name: GMP consulting gap assessment and remediation
|
||||
best_for:
|
||||
- GMP audit remediation
|
||||
- quality system consulting
|
||||
- supplier audit CAPA
|
||||
structure_principle: Regulation-to-gap-to-risk-to-CAPA pathway.
|
||||
task_axes:
|
||||
- regulatory_gap
|
||||
- risk_classification
|
||||
- capa_design
|
||||
- ownership_timeline
|
||||
- verification_evidence
|
||||
- counter
|
||||
framework_sections:
|
||||
- regulatory_baseline
|
||||
- gap_matrix
|
||||
- risk_ranking
|
||||
- capa_roadmap
|
||||
- verification_plan
|
||||
|
||||
cmc_process_risk:
|
||||
name: CMC process and scale-up risk assessment
|
||||
best_for:
|
||||
- process development
|
||||
- tech transfer
|
||||
- manufacturing readiness
|
||||
structure_principle: Process-flow, CQA/CPP, scale-up, control strategy.
|
||||
task_axes:
|
||||
- process_flow
|
||||
- cqa_cpp
|
||||
- scale_up_risk
|
||||
- control_strategy
|
||||
- supply_chain
|
||||
- counter
|
||||
framework_sections:
|
||||
- process_map
|
||||
- critical_quality_attributes
|
||||
- critical_process_parameters
|
||||
- scale_up_risks
|
||||
- control_strategy
|
||||
|
||||
rd_go_no_go:
|
||||
name: R&D go/no-go decision research
|
||||
best_for:
|
||||
- project initiation
|
||||
- modality selection
|
||||
- development strategy
|
||||
structure_principle: Scientific rationale, proof-of-concept, IP/FTO, development path, go/no-go criteria.
|
||||
task_axes:
|
||||
- scientific_rationale
|
||||
- poc_evidence
|
||||
- ip_fto
|
||||
- development_path
|
||||
- commercial_window
|
||||
- counter
|
||||
framework_sections:
|
||||
- scientific_rationale
|
||||
- evidence_threshold
|
||||
- ip_fto
|
||||
- development_plan
|
||||
- go_no_go_criteria
|
||||
|
||||
management_consulting:
|
||||
name: Management consulting diagnostic and operating model
|
||||
best_for:
|
||||
- organization diagnosis
|
||||
- operating model design
|
||||
- governance and process improvement
|
||||
structure_principle: Diagnostic baseline, capability gaps, operating model, roadmap, governance.
|
||||
task_axes:
|
||||
- current_state
|
||||
- capability_gap
|
||||
- operating_model
|
||||
- governance
|
||||
- implementation_roadmap
|
||||
- counter
|
||||
framework_sections:
|
||||
- current_state_diagnosis
|
||||
- capability_gap
|
||||
- future_state_model
|
||||
- roadmap
|
||||
- governance_metrics
|
||||
|
||||
gmp_quality_operations_diagnosis:
|
||||
name: GMP quality, manufacturing process, and operations diagnosis
|
||||
best_for:
|
||||
- GMP audit remediation from client materials
|
||||
- quality system diagnosis
|
||||
- manufacturing process system diagnosis
|
||||
- operations management transformation
|
||||
structure_principle: Start from site audit evidence, map findings to regulatory baseline and operating model gaps, then design short/mid/long-term remediation.
|
||||
task_axes:
|
||||
- input_material_findings
|
||||
- nmpa_fda_ema_ich_who_baseline
|
||||
- quality_system_gap
|
||||
- manufacturing_process_risk
|
||||
- operations_management_gap
|
||||
- team_capability
|
||||
- capa_roadmap
|
||||
- verification_evidence
|
||||
- counter
|
||||
integrated_lanes:
|
||||
- site audit and recap findings
|
||||
- official regulatory and guideline baseline
|
||||
- enforcement precedents and warning letters
|
||||
- quality/manufacturing/operations gap analysis
|
||||
- remediation actions, ownership, verification evidence
|
||||
- counter-evidence and boundary conditions
|
||||
framework_sections:
|
||||
- material_evidence_map
|
||||
- regulatory_and_best_practice_baseline
|
||||
- quality_system_gap_matrix
|
||||
- manufacturing_process_risk_map
|
||||
- operations_management_diagnosis
|
||||
- people_and_capability_diagnosis
|
||||
- short_mid_long_term_roadmap
|
||||
- governance_and_verification_plan
|
||||
@@ -6,14 +6,16 @@ profiles:
|
||||
- "clinicaltrials"
|
||||
- "fda_ema_nmpa"
|
||||
- "serper_scholar"
|
||||
- "tavily_exa_gap_fill"
|
||||
- "exa_evidence_discovery"
|
||||
- "tavily_research_scan_if_needed"
|
||||
patent_heavy:
|
||||
description: "IP landscape, freedom-to-operate, and process-route research."
|
||||
order:
|
||||
- "google_patents_via_serper"
|
||||
- "uspto_epo_cnipa"
|
||||
- "company_disclosures"
|
||||
- "exa_tavily_family_discovery"
|
||||
- "exa_family_discovery"
|
||||
- "tavily_gap_fill"
|
||||
china_market:
|
||||
description: "China regulatory, company, supply-chain, and market research."
|
||||
order:
|
||||
@@ -21,6 +23,7 @@ profiles:
|
||||
- "exchange_disclosures"
|
||||
- "serper_brave_chinese"
|
||||
- "exa_company_pages"
|
||||
- "exa_evidence_discovery"
|
||||
- "tavily_gap_fill"
|
||||
investment:
|
||||
description: "Public-company, market-size, and transaction-oriented research."
|
||||
@@ -29,15 +32,16 @@ profiles:
|
||||
- "consulting_and_database_reports"
|
||||
- "company_announcements"
|
||||
- "serper_news"
|
||||
- "exa_evidence_discovery"
|
||||
- "tavily_gap_fill"
|
||||
|
||||
apis:
|
||||
tavily:
|
||||
best_for: ["initial_scan", "gap_fill", "llm_friendly_snippets"]
|
||||
best_for: ["phase1_research_scan", "phase3_gap_fill", "llm_friendly_snippets"]
|
||||
evidence_role: "discovery_only_unless_original_source"
|
||||
exa:
|
||||
best_for: ["company_pages", "terminology_check", "long_tail_professional_pages"]
|
||||
evidence_role: "discovery_or_secondary"
|
||||
best_for: ["evidence_discovery", "company_pages", "terminology_check", "long_tail_professional_pages", "agent_highlights"]
|
||||
evidence_role: "candidate_source_until_scored"
|
||||
brave:
|
||||
best_for: ["cross_check", "counter_evidence", "mixed_language_search"]
|
||||
evidence_role: "discovery_only_unless_original_source"
|
||||
|
||||
+43
-43
@@ -1,6 +1,6 @@
|
||||
# Codex Native Adapter
|
||||
|
||||
> v0.10 起，Codex 不再只是 OpenCode 的辅助执行环境，而是 Deep Research 的并列 adapter。共享核心是 `AGENTS.md`、`scripts/`、`configs/` 和 `.agents/skills`；OpenCode 使用 `.opencode/**`，Codex 使用 `.codex/**`。
|
||||
> v0.20 起，Codex 是 Deep Research 的表层 adapter。共享核心迁移到 Python runtime：`scripts/dr.py`、`scripts/runtime/**`、`configs/models.yaml` 和 `.agents/skills`。Codex 不再复制核心调度逻辑。
|
||||
|
||||
## Architecture
|
||||
|
||||
@@ -8,17 +8,17 @@
|
||||
|---|---:|---:|---:|
|
||||
| 方法论 | `AGENTS.md` | ✅ | ✅ |
|
||||
| Skills | `.agents/skills` | 继续保留 `.opencode/skills` | ✅ |
|
||||
| Agent 定义 | 否 | `.opencode/agents/*.md` | `.codex/agents/*.toml` |
|
||||
| 命令入口 | 部分共享脚本 | `.opencode/commands/*.md` | `.codex/commands/*.md` + `scripts/dr.py` |
|
||||
| Agent 定义 | Python role runtime 为准 | `.opencode/agents/*.md` 仅兼容 | `$CODEX_HOME/agents/*.toml` 仅兼容 |
|
||||
| 命令入口 | `scripts/dr.py` | `.opencode/commands/*.md` wrapper | `$CODEX_HOME/commands/*.md` wrapper |
|
||||
| Phase 4 确定性流水线 | `scripts/*.py` | ✅ | ✅ |
|
||||
|
||||
Codex 官方行为要点：
|
||||
|
||||
- 项目级配置放在 `.codex/config.toml`，项目被 trust 后才会加载。
|
||||
- 用户级配置放在 `~/.codex/config.toml` 或 `$CODEX_HOME/config.toml`；项目级 `.codex/**` 不是 v0.20 推荐路径。
|
||||
- Codex 会从项目根向当前目录读取 `AGENTS.md`。
|
||||
- repo skills 放在 `.agents/skills/*/SKILL.md`。
|
||||
- custom agents 放在 `.codex/agents/*.toml`。
|
||||
- subagents 只有在主线程明确要求时才会启动。
|
||||
- adapter templates 保存在 `codex_adapter_templates/codex/**`，部署脚本会复制到 `$CODEX_HOME`。
|
||||
- subagents/agent threads 是 Codex 表层增强能力；v0.20 默认研究并发由 Python worker pool 执行。
|
||||
|
||||
## Setup
|
||||
|
||||
@@ -30,19 +30,21 @@ source scripts/activate.sh
|
||||
首次使用 Codex adapter 前确认：
|
||||
|
||||
```bash
|
||||
uv run python scripts/install_codex_adapter.py
|
||||
find .codex -maxdepth 3 -type f | sort
|
||||
uv run python scripts/deploy_adapters.py codex --dry-run
|
||||
uv run python scripts/deploy_adapters.py codex --force
|
||||
find .agents/skills -maxdepth 2 -name SKILL.md | sort
|
||||
uv run python scripts/dr.py status <slug>
|
||||
```
|
||||
|
||||
默认部署不会写入 `config.toml`，避免覆盖现有 Codex 全局配置。只有确认要安装本项目 bundled profile 时，才运行 `uv run python scripts/deploy_adapters.py codex --force --include-config`。
|
||||
|
||||
新机器部署后可以先跑自检：
|
||||
|
||||
```bash
|
||||
uv run python scripts/deploy_check.py
|
||||
```
|
||||
|
||||
如果隐藏目录缺失或 skills 没同步：
|
||||
如果 `$CODEX_HOME` adapter 缺失或 skills 没同步：
|
||||
|
||||
```bash
|
||||
uv run python scripts/deploy_check.py --repair --force
|
||||
@@ -58,17 +60,19 @@ uv run python scripts/deploy_check.py --repair --force
|
||||
|
||||
## Codex Commands
|
||||
|
||||
Codex custom command templates 位于 `.codex/commands/`。在 CLI 中可以用 `scripts/dr.py prompt` 展开：
|
||||
Codex custom command templates 的真源位于 `codex_adapter_templates/codex/commands/`，部署后位于 `$CODEX_HOME/commands/`。在 CLI 中也可以绕过平台命令，直接用 `scripts/dr.py prompt` 从模板展开：
|
||||
|
||||
```bash
|
||||
uv run python scripts/dr.py prompt dr-run dual-target-rnai-pipeline-2026
|
||||
codex exec "$(uv run python scripts/dr.py prompt dr-run dual-target-rnai-pipeline-2026)"
|
||||
```
|
||||
|
||||
推荐入口是 `dr-run`：让 Codex 主线程进入 PM 模式，读取 manifest，判断当前应该继续哪个 phase，并在 Phase 2 主动调度 `dr-analyst` / `dr-verifier` subagents。用户不需要逐个执行每个 phase；只有 Phase 1 框架确认和 Phase 3 审校决策这类人类暂停点需要停下来。
|
||||
推荐入口是 Python core。Codex command 只包装 CLI，不再让 Codex 主线程主动调度 subagents。
|
||||
|
||||
```bash
|
||||
codex exec "$(uv run python scripts/dr.py prompt dr-run <slug-or-topic>)"
|
||||
uv run python scripts/dr.py run <slug-or-topic>
|
||||
uv run python scripts/dr.py research <slug> --workers 6
|
||||
uv run python scripts/dr.py finalize <slug>
|
||||
```
|
||||
|
||||
分阶段命令保留为调试和人工接管入口：
|
||||
@@ -81,17 +85,16 @@ codex exec "$(uv run python scripts/dr.py prompt dr-review <slug>)"
|
||||
uv run python scripts/dr.py finalize <slug>
|
||||
```
|
||||
|
||||
Phase 4 推荐走确定性 CLI，而不是让单个 agent 翻译整篇：
|
||||
Phase 4 默认中文原生成稿：
|
||||
|
||||
```bash
|
||||
uv run python scripts/dr.py finalize <slug> \
|
||||
--model-profile medium
|
||||
uv run python scripts/dr.py finalize <slug> --model-profile medium
|
||||
```
|
||||
|
||||
等价底层入口（统一 pipeline）：
|
||||
旧英译中 pipeline 仅用于兼容旧项目：
|
||||
|
||||
```bash
|
||||
uv run python scripts/phase4_pipeline.py <slug>
|
||||
uv run python scripts/dr.py finalize <slug> --legacy-translate
|
||||
```
|
||||
|
||||
网络不稳时可显式降并发：
|
||||
@@ -112,31 +115,16 @@ uv run python scripts/dr.py finalize <slug> --model-profile medium --glossary-mo
|
||||
uv run python scripts/dr.py finalize <slug> --model-profile medium --glossary-mode off
|
||||
```
|
||||
|
||||
## Subagent Usage
|
||||
## Adapter Boundary
|
||||
|
||||
Codex 的平台限制是：subagents 不会仅因为 `.codex/agents/*.toml` 存在就自动启动，必须由当前主线程明确要求。`dr-run` 已把这个要求写进 PM prompt：Phase 1 会调度 `dr-plan` / `dr-searcher`，Phase 2 会调度 `dr-analyst` / `dr-verifier`，Phase 3 会调度 `dr-chief-editor`。
|
||||
|
||||
```text
|
||||
Spawn dr-searcher agents in parallel for four keyword groups, wait for all results, then synthesize phase1/initial-scan.md.
|
||||
```
|
||||
|
||||
推荐映射：
|
||||
|
||||
- `dr-plan`：访谈、框架、初扫综合。
|
||||
- `dr-pm`：Phase 2 批次规划与调度。
|
||||
- `dr-searcher`：轻量检索。
|
||||
- `dr-analyst`：章节英文深研。
|
||||
- `dr-verifier`：反方验证，必须独立于 analyst。
|
||||
- `dr-chief-editor`：Phase 3 只读审校。
|
||||
- `dr-editor-in-chief`：Phase 4 合稿与脚本调度。
|
||||
- `dr-reporter`：出稿执行与格式验证。
|
||||
Codex 可以继续用于审阅、解释和少量人工接管，但默认研究并发由 Python task-card runtime 控制。模型选择与 role/task 映射以 `configs/models.yaml` 为准。
|
||||
|
||||
## Git Hygiene
|
||||
|
||||
本仓库常有大量 `projects/**` 研究产物处于修改状态。Codex adapter 提交时只 stage 系统文件：
|
||||
|
||||
```bash
|
||||
git add .codex .agents/skills scripts/dr.py docs configs README.md PLAN.md
|
||||
git add codex_adapter_templates .agents/skills scripts docs configs README.md PLAN.md AGENTS.md
|
||||
git diff --staged --name-only
|
||||
```
|
||||
|
||||
@@ -146,26 +134,38 @@ git diff --staged --name-only
|
||||
- 已生成 PDF/DOCX/TXT
|
||||
- 临时检查脚本或一次性研究产物
|
||||
|
||||
## Installing Hidden Directories
|
||||
## Deploying Adapter Files
|
||||
|
||||
如果 Codex 桌面沙盒禁止 agent 写入 `.codex` 或 `.agents/skills`，请在本机直接运行：
|
||||
不要在仓库内维护 `.codex/**`。如果需要 Codex native adapter，请把模板部署到用户级 Codex home：
|
||||
|
||||
```bash
|
||||
uv run python scripts/deploy_adapters.py codex --force
|
||||
```
|
||||
|
||||
兼容旧命令仍可用，但默认也会走外部部署：
|
||||
|
||||
```bash
|
||||
uv run python scripts/install_codex_adapter.py --force
|
||||
```
|
||||
|
||||
安装来源：
|
||||
部署来源：
|
||||
|
||||
- `codex_adapter_templates/codex/**` → `.codex/**`
|
||||
- `.opencode/skills/**` → `.agents/skills/**`
|
||||
- `codex_adapter_templates/codex/**` → `$CODEX_HOME/**` 或 `~/.codex/**`
|
||||
- `.agents/skills/**` → `$CODEX_HOME/skills/**`
|
||||
|
||||
安装后，在 Codex 中运行 `/debug-config`，确认 project `.codex/config.toml` 已加载。
|
||||
如果旧版本已经把仓库内 `.codex/**` 加进 Git，需要在本机清一次索引，让它回到“本地部署产物”身份：
|
||||
|
||||
```bash
|
||||
git rm -r --cached .codex
|
||||
```
|
||||
|
||||
部署后，在 Codex 中运行 `/debug-config`，确认 user config 或 `CODEX_HOME` config 已加载。
|
||||
|
||||
## Config Troubleshooting
|
||||
|
||||
如果 `.codex/config.toml` 生效后启动报错，先按下面顺序排查：
|
||||
如果 Codex adapter 配置生效后启动报错，先按下面顺序排查：
|
||||
|
||||
1. 确认当前 project 已被 Codex trust。未 trust 时，Codex 会跳过项目级 `.codex/**`，此时 `--profile deep-research` 会报 profile 不存在。
|
||||
1. 确认部署目标正确：默认是 `$CODEX_HOME`，未设置时是 `~/.codex`。
|
||||
2. Tavily / Brave / Exa MCP 默认启用但不是 required。若某个 server 启动异常，先确认对应环境变量存在，再临时把该 server 改成 `enabled = false`。
|
||||
3. 如果要完全离线排障，先把第三方 MCP 全部关掉，只保留 OpenAI Docs MCP 和内置 web search。
|
||||
4. 如果仍然报错，临时保留最小配置确认 Codex 主体能启动：
|
||||
|
||||
@@ -0,0 +1,216 @@
|
||||
# v0.20 Platform Adapters
|
||||
|
||||
> v0.20 的默认核心入口是 Python core：`scripts/dr.py`、`scripts/runtime/**`、`configs/models.yaml`、`.agents/skills`。Antigravity 例外：它使用 `.agents/skills` 与 `.agents/rules` 走 native Deep Research，让 Antigravity 模型配额承担主要研究工作。
|
||||
|
||||
## Shared Rule
|
||||
|
||||
除 Antigravity native 模式外，不要让平台 agent 自己调度 Phase 2 并发、模型选择或上下文压缩。默认平台只负责：
|
||||
|
||||
- 运行 `uv run python scripts/dr.py ...`
|
||||
- 展示产物路径与失败包
|
||||
- 做少量人工访谈、审阅、解释
|
||||
- 必要时调用其原生强模型做“补充审校”，但不得覆盖 Python runtime 的产物 schema
|
||||
|
||||
最小可执行链路：
|
||||
|
||||
```bash
|
||||
uv run python scripts/dr.py init "研究主题" --slug <slug> --method mckinsey_market
|
||||
uv run python scripts/dr.py frame <slug>
|
||||
uv run python scripts/dr.py research <slug> --workers 6
|
||||
uv run python scripts/dr.py research <slug> --workers 6 --execute-packets
|
||||
uv run python scripts/dr.py research <slug> --workers 6 --build-briefs
|
||||
uv run python scripts/dr.py research <slug> --workers 6 --assemble-chapters
|
||||
uv run python scripts/dr.py review <slug>
|
||||
uv run python scripts/dr.py finalize <slug>
|
||||
```
|
||||
|
||||
Antigravity 不走这条最小链路作为默认研究主流程；详见下方 Antigravity 章节。
|
||||
|
||||
## OpenCode
|
||||
|
||||
官方机制：OpenCode 支持 `.opencode/commands/*.md` 自定义命令；文件名就是 slash command，内容是 prompt，frontmatter 可指定 `agent`、`model`、`subtask`。OpenCode 也支持 primary/subagent 两类 agent，但 v0.20 不再把平台 subagent 当默认并发机制。
|
||||
|
||||
本项目调用方式：
|
||||
|
||||
```text
|
||||
/dr-init "ADC 全球竞争格局" --slug adc-global-landscape --method mckinsey_market
|
||||
/dr-frame adc-global-landscape
|
||||
/dr-research adc-global-landscape
|
||||
/dr-review adc-global-landscape
|
||||
/dr-finalize adc-global-landscape
|
||||
/dr-status adc-global-landscape
|
||||
```
|
||||
|
||||
使用建议：
|
||||
|
||||
- OpenCode 适合做表层 TUI、人工访谈和快速查看状态。
|
||||
- 不要让 `dr-pm` 在 OpenCode 里 spawn 多个 dr-analyst；Phase 2 并发已经由 Python worker pool 控制。
|
||||
- 如果要用 OpenCode 原生模型优势，只用于 `frame` 前的人工访谈或 `review` 后的解释，不改变 `configs/models.yaml` 的 role/task 映射。
|
||||
|
||||
## Codex
|
||||
|
||||
官方机制：Codex CLI 使用 `AGENTS.md` 作为项目指令；用户级配置位于 `~/.codex/config.toml` 或 `$CODEX_HOME/config.toml`，项目级 `.codex/**` 只作为可选覆盖。v0.20 推荐用部署脚本把 adapter 模板写到用户级 Codex home，避免在研究项目里维护 `.codex`。Codex 可通过 `/model` 选择 GPT 系列模型、通过 `/permissions` 调整审批/沙盒，也支持多 agent thread，但 v0.20 默认不使用它做研究并发。
|
||||
|
||||
本项目调用方式：
|
||||
|
||||
```bash
|
||||
uv run python scripts/deploy_adapters.py codex --force
|
||||
codex
|
||||
```
|
||||
|
||||
默认部署不会复制 `config.toml`，避免覆盖用户级 Codex 设置；只有明确需要 bundled `deep-research` profile 时才加 `--include-config`，然后用 `codex --profile deep-research`。
|
||||
|
||||
在 Codex 里直接要求：
|
||||
|
||||
```text
|
||||
运行：uv run python scripts/dr.py run "ADC 全球竞争格局" --slug adc-global-landscape --method mckinsey_market
|
||||
```
|
||||
|
||||
或用已有 wrapper prompt：
|
||||
|
||||
```bash
|
||||
codex exec "$(uv run python scripts/dr.py prompt dr-run 'adc-global-landscape')"
|
||||
codex exec "$(uv run python scripts/dr.py prompt dr-research 'adc-global-landscape')"
|
||||
```
|
||||
|
||||
使用建议：
|
||||
|
||||
- Codex 原生 GPT 系列适合代码改造、回归测试、schema/debug、review。
|
||||
- 研究模型混合仍由 Python core 调 ZenMux；Codex 当前会话模型不决定 `dr_analyst`、`dr_verifier` 等 role。
|
||||
- `codex_native` profile 可用于偏 OpenAI/GPT 的执行环境，但仍通过 `configs/models.yaml` 解析。
|
||||
|
||||
## Claude Code
|
||||
|
||||
官方机制：Claude Code 推荐用 `.claude/skills/<name>/SKILL.md` 定义可调用 skill；目录名成为 slash command。旧 `.claude/commands/*.md` 仍兼容，但 skill 优先。Claude Code 的优势是 Claude/Opus/Sonnet 对长文风格和中文润色的稳定性。
|
||||
|
||||
本项目调用方式：
|
||||
|
||||
```bash
|
||||
claude
|
||||
```
|
||||
|
||||
在 Claude Code 中：
|
||||
|
||||
```text
|
||||
/dr-run "ADC 全球竞争格局" --slug adc-global-landscape --method mckinsey_market
|
||||
/dr-research adc-global-landscape
|
||||
/dr-finalize adc-global-landscape
|
||||
```
|
||||
|
||||
使用建议：
|
||||
|
||||
- Claude Code 适合 Phase 1 人工访谈增强、Phase 4 中文风格润色建议、复杂报告结构讨论。
|
||||
- 默认不要让 Claude Code 直接整章写作或并发 spawn；让 Python core 生成 packets、briefs、drafts。
|
||||
- 若想优先用 Claude/Opus 成本包，可在 `configs/models.yaml` 里选择或新增 profile，而不是在 Claude Code prompt 里手工指定。
|
||||
|
||||
## Gemini CLI
|
||||
|
||||
官方机制：Gemini CLI 支持 `GEMINI.md` 作为项目记忆，也支持 `.gemini/commands/*.toml` 自定义命令；TOML command 用 `prompt` 字段，支持 `{{args}}` 参数和 `!{...}` shell 注入。
|
||||
|
||||
本项目调用方式：
|
||||
|
||||
```bash
|
||||
gemini
|
||||
```
|
||||
|
||||
在 Gemini CLI 中：
|
||||
|
||||
```text
|
||||
/dr:run "ADC 全球竞争格局" --slug adc-global-landscape --method mckinsey_market
|
||||
/dr:research adc-global-landscape
|
||||
/dr:review adc-global-landscape
|
||||
/dr:finalize adc-global-landscape
|
||||
```
|
||||
|
||||
使用建议：
|
||||
|
||||
- Gemini CLI 适合长上下文审校、框架替代方案、图表/多模态材料理解。
|
||||
- 对需要本地 shell 的命令，Gemini CLI 会在执行 shell injection 前要求确认，这是好事。
|
||||
- 研究执行仍以 Python core 为准；Gemini 模型可作为 `phase3_review` 或 `final_editorial` profile 的候选模型。
|
||||
|
||||
## Antigravity
|
||||
|
||||
官方资料显示，Antigravity 是 agent-first IDE：主 Agent 可跨 editor、terminal、browser 工作；浏览器操作由 browser subagent 执行；Agent Manager 可管理多个任务流和 artifacts。基于 Codex 的 skill-driven 经验，本项目在 Antigravity 中不再默认把研究主流程交给 Python model workers，而是让 Antigravity 用自己的模型配额按 phase 执行研究。
|
||||
|
||||
本项目已提供 Antigravity 入口：
|
||||
|
||||
- `.agents/agents.md`：Antigravity 角色定义，只定义 Research Manager / Strategist / Analyst / Reviewer / Final Editor 等角色。
|
||||
- `.agents/skills/antigravity-surface-adapter/SKILL.md`：Antigravity 专用执行说明。
|
||||
- `.agents/skills/method-selection/SKILL.md`：按场景选择研究方法，避免默认套用麦肯锡/MECE。
|
||||
- `.agents/skills/research-quality-gates/SKILL.md`：把 search receipt、source access、claims ledger、delta retrieve、continuation state 和 final fact check 做成阶段质量门。
|
||||
- `.agents/rules/deep-research-antigravity.md`：workspace rule，约束其优先使用 Antigravity 模型配额，避免主流程落回 ZenMux。
|
||||
- `.agents/workflows/deep-research-native.md`：可按 slash workflow 执行的四阶段研究流程与反幻觉 gate。
|
||||
- `.agents/skills/*`：Antigravity 官方默认 skill 目录，直接复用本项目 canonical skill registry。
|
||||
|
||||
部署到另一个 Antigravity workspace：
|
||||
|
||||
```bash
|
||||
uv run python scripts/deploy_adapters.py antigravity --target /path/to/workspace --dry-run
|
||||
uv run python scripts/deploy_adapters.py antigravity --target /path/to/workspace
|
||||
```
|
||||
|
||||
该命令只写 workspace-local `.agents/agents.md`、`.agents/skills`、`.agents/rules` 和 `.agents/workflows`，不修改全局 Antigravity/Gemini 设置；已有文件默认跳过，`--force` 才会备份为 `.bak` 后覆盖。
|
||||
|
||||
本项目调用方式：
|
||||
|
||||
1. 在 Antigravity 打开仓库根目录。
|
||||
2. 在 Customizations / Rules 中确认 workspace rule 已启用；若未自动启用，手动 at mention `deep-research-antigravity` 或直接引用该文件。
|
||||
3. 选择模型：
|
||||
|
||||
| Phase / 工作 | Antigravity model | 用法 |
|
||||
|---|---|---|
|
||||
| Surface 管理、读 skill、状态推进 | Gemini 3 Flash | 管任务清单、跑轻量命令、收集 artifact |
|
||||
| Phase 0-1 材料解读、框架规划 | Claude Opus 4.6 (Thinking) | 大局观、研究假设、章节结构 |
|
||||
| Phase 2 证据包与章节草稿 | Gemini 3.1 Pro (Low) | 质量/速度均衡，按 skill 写 artifacts |
|
||||
| Phase 3 总编审校 | Gemini 3.1 Pro (High) | 先试高强度 Gemini，不行再换 |
|
||||
| Phase 4 最终统稿 | Claude Opus 4.6 (Thinking) | 保证中文终稿和交付质量 |
|
||||
|
||||
4. 给 Agent Manager 一个明确任务：
|
||||
|
||||
```text
|
||||
请加载 .agents/agents.md、.agents/rules/deep-research-antigravity.md、.agents/skills/antigravity-surface-adapter/SKILL.md、.agents/skills/research-quality-gates/SKILL.md 和 AGENTS.md。
|
||||
用 Gemini 3 Flash 做流程管理，按 skill 要求在需要时提示我切换模型。
|
||||
研究主题："ADC 全球竞争格局"，slug: adc-global-landscape，method: mckinsey_market。
|
||||
请先创建/确认项目目录，然后进入 Phase 0-1，并在生成 phase1/framework.md 和 research_brief 后暂停。
|
||||
```
|
||||
|
||||
使用建议：
|
||||
|
||||
- Antigravity 的 Gemini/Opus/Gemini Computer Use/Browser 能力适合可视化 QA、PDF/HTML 预览、跨文件审阅。
|
||||
- Opus 额度有限，但 Phase 0-1 和 Phase 4 的价值最高，优先花在框架和终稿，不花在 routine 管理。
|
||||
- 不要默认运行 `dr.py run`、`research --execute-packets`、`research --assemble-chapters`，这些会把主流程重新导向 Python/ZenMux。
|
||||
- Python core 可用于 `init/status/skills validate/finalize/render` 等辅助动作。
|
||||
- 若本机安装了 `paperfoot/search-cli` 的 `search` 命令，Antigravity Phase 0-2 可优先用 `search agent-info`、`search ... --json` 做统一检索入口。
|
||||
- Phase 2 不只产出章节草稿；必须维护 `search_log.jsonl`、`sources.jsonl`、`claims_ledger.jsonl`、`coverage_matrix.md` 和 `unsupported_claims.md`，否则 Phase 3/4 不得声称已验证。
|
||||
- 不建议依赖 SOUL.md：它不是 Antigravity 官方 Deep Research 约束入口。本项目用 `AGENTS.md`、`.agents/agents.md`、`.agents/rules`、`.agents/skills` 和 `.agents/workflows` 分层约束。
|
||||
|
||||
## Model Strategy Across Platforms
|
||||
|
||||
| Platform | Surface model priority | Deep Research model source |
|
||||
|---|---|---|
|
||||
| OpenCode | 可用 Claude/ZenMux provider 做 TUI 增强 | `configs/models.yaml` |
|
||||
| Codex | GPT 系列用于代码、测试、schema、review | `configs/models.yaml` |
|
||||
| Claude Code | Opus/Sonnet 用于访谈、中文风格、结构讨论 | `configs/models.yaml` |
|
||||
| Gemini CLI | Gemini 用于长上下文、多模态、框架审阅 | `configs/models.yaml` |
|
||||
| Antigravity | Gemini Flash/Opus/Gemini Pro 分 phase 执行研究 | `.agents/skills` + Antigravity model selector |
|
||||
|
||||
核心原则：除 Antigravity 外，平台模型负责“怎么帮用户操作项目”，ZenMux/Python role 模型负责“研究任务用哪个模型执行”。Antigravity 专门走 native 模式，以消耗 Antigravity 自身模型配额并降低 Python-core 碎片化风险。
|
||||
|
||||
## Sources
|
||||
|
||||
- OpenCode commands and agents: https://opencode.ai/docs/commands/ , https://opencode.ai/docs/agents/
|
||||
- Codex CLI slash commands and config: https://developers.openai.com/codex/cli/slash-commands , https://developers.openai.com/codex/config-reference
|
||||
- Claude Code skills/slash commands: https://code.claude.com/docs/en/slash-commands
|
||||
- Gemini CLI custom commands: https://github.com/google-gemini/gemini-cli/blob/main/docs/cli/custom-commands.md
|
||||
- Google Antigravity announcement: https://blog.google/products-and-platforms/products/gemini/gemini-3/
|
||||
- Google Antigravity rules/workflows: https://antigravity.google/docs/rules-workflows
|
||||
- Google Antigravity skills: https://antigravity.google/docs/skills
|
||||
- Google Antigravity models: https://antigravity.google/docs/models
|
||||
- Google Antigravity permissions: https://antigravity.google/docs/agent-permissions
|
||||
- paperfoot/search-cli skill: https://github.com/paperfoot/search-cli/blob/master/SKILL.md
|
||||
- Gemini CLI GEMINI.md context: https://geminicli.com/docs/cli/gemini-md/
|
||||
- Gemini CLI GEMINI.md examples issue: https://github.com/google-gemini/gemini-cli/issues/9241
|
||||
- Practical GEMINI.md hierarchy: https://medium.com/google-cloud/practical-gemini-cli-instruction-following-gemini-md-hierarchy-part-1-3ba241ac5496
|
||||
- Claude Code best-practice CLAUDE.md: https://github.com/shanraisshan/claude-code-best-practice/blob/main/CLAUDE.md
|
||||
- Claude MD guide: https://github.com/MuhammadUsmanGM/claude-code-best-practices/blob/main/guides/claude-md-guide.md
|
||||
- Writing a good CLAUDE.md: https://www.humanlayer.dev/blog/writing-a-good-claude-md
|
||||
+17
-6
@@ -8,13 +8,14 @@ v0.12 起，默认搜索路径收敛到项目内 Python 网关：
|
||||
|
||||
```bash
|
||||
uv run python scripts/search.py "<query>" --route scholar --num-results 10 --year-low 2023
|
||||
uv run python scripts/search.py "<query>" --route evidence --num-results 10 --json --trace
|
||||
uv run python scripts/search.py "<query>" --route patents --num-results 10
|
||||
uv run python scripts/search.py "<query>" --route news --num-results 10 --time-range m
|
||||
uv run python scripts/search.py "<query>" --route general --num-results 10
|
||||
uv run python scripts/ground.py "<query>" --json
|
||||
```
|
||||
|
||||
其中 `scholar / patents / news` 默认走严格模式（Serper 失败不静默降级）；需要容错时显式加 `--no-strict-specialized`。
|
||||
其中 `scholar / patents / news` 默认走严格模式（Serper 失败不静默降级）；需要容错时显式加 `--no-strict-specialized`。`evidence` 是 v0.20.1 之后新增的受控证据发现路由，优先用 Exa highlights/text 为 evidence packet 提供候选来源。
|
||||
|
||||
MCP server 只作为交互式补漏和特殊工具能力，不作为文献、专利、新闻检索主路径。这样 OpenCode、Codex、Gemini CLI、Claude Code 都能复用同一套路由，减少每个平台单独配置 Tavily/Exa/Brave MCP 的依赖。
|
||||
|
||||
@@ -23,13 +24,14 @@ MCP server 只作为交互式补漏和特殊工具能力，不作为文献、专
|
||||
### Tavily
|
||||
|
||||
- 优点：LLM 友好，摘要质量稳定，适合快速发现方向。
|
||||
- 用法：初扫、普通网页、报告线索、交叉补漏。
|
||||
- 用法：初扫、普通网页、报告线索、交叉补漏；`research()` 更适合 Phase 1 初步扫描、薄弱章节补证据、Phase 3 回炉。
|
||||
- 风险：不能把普通网页当结论支撑，必须追溯原始来源。
|
||||
- 规则：Tavily Research 输出必须保存为过程文件，并经过 source-quality 评分、去重和 source_id 归一化；不要直接把 Tavily 的综合报告当作章节正文或最终证据。
|
||||
|
||||
### Exa
|
||||
|
||||
- 优点：neural search，对官网、公司页、长尾专业内容召回好。
|
||||
- 用法：术语核查、公司/产品名纠错、专业网页发现。
|
||||
- 优点：neural/agent search，对官网、公司页、长尾专业内容召回好；highlights/text 适合喂给 agent 做证据筛选。
|
||||
- 用法：`scripts/search.py --route evidence`、术语核查、公司/产品名纠错、专业网页发现、章节证据补强。
|
||||
- 风险：macOS 代理环境容易 TLS EOF，项目内 `SearchClient` 已使用 `trust_env=False` 绕开系统代理。
|
||||
|
||||
### Brave
|
||||
@@ -71,13 +73,22 @@ MCP server 只作为交互式补漏和特殊工具能力，不作为文献、专
|
||||
|
||||
## Recommended Profiles
|
||||
|
||||
## v0.20 Routing Decision
|
||||
|
||||
- Phase 1 初步扫描：Tavily Research + Exa evidence，目标是形成假设、反证方向、章节任务切分。
|
||||
- Phase 2 evidence packet：优先 `fda/scholar/patents/news` 等专用路由；需要补充候选证据时用 `evidence`，不要只用 `general`。
|
||||
- Phase 3 回炉：按 critique 中的证据缺口定向调用 Tavily Research 或 Exa evidence，输出仍需进入 packet/schema。
|
||||
- General route：只做宽泛发现和兜底，不作为“默认最佳搜索”。
|
||||
|
||||
## Recommended Profiles
|
||||
|
||||
### biomed_literature
|
||||
|
||||
PubMed / NCBI → ClinicalTrials → FDA/EMA/NMPA → `scripts/search.py --route scholar` → Tavily/Exa 补漏。
|
||||
PubMed / NCBI → ClinicalTrials → FDA/EMA/NMPA → `scripts/search.py --route scholar` → `scripts/search.py --route evidence` → Tavily/Brave 补漏。
|
||||
|
||||
### patent_heavy
|
||||
|
||||
`scripts/search.py --route patents` → USPTO/EPO/CNIPA → 公司年报/招股书 → Tavily/Exa 补同族专利线索。
|
||||
`scripts/search.py --route patents` → USPTO/EPO/CNIPA → 公司年报/招股书 → Exa/Tavily 补同族专利线索。
|
||||
|
||||
### china_market
|
||||
|
||||
|
||||
+3
-2
@@ -1,7 +1,7 @@
|
||||
[project]
|
||||
name = "deep-research"
|
||||
version = "0.12.0"
|
||||
description = "生物医药 Deep Research 系统 - OpenCode 多 agent 协作研究流水线"
|
||||
version = "0.20.0"
|
||||
description = "生物医药 Deep Research 系统 - Python core + skills driven research pipeline"
|
||||
requires-python = ">=3.10"
|
||||
readme = "README.md"
|
||||
license = { text = "MIT" }
|
||||
@@ -20,6 +20,7 @@ dependencies = [
|
||||
"PyYAML>=6.0.1",
|
||||
"rich>=13.7.0",
|
||||
"pypdf>=6.10.2",
|
||||
"pymupdf>=1.26.0",
|
||||
]
|
||||
|
||||
[project.optional-dependencies]
|
||||
|
||||
@@ -6,7 +6,7 @@
|
||||
- 可选：--extra terms.txt（每行一个英文术语，补充进来一起核查）
|
||||
|
||||
流程（每个术语独立可并行）：
|
||||
1. 用 SearchClient（Exa > Tavily）搜一次（query = "<term> <domain hint>"）
|
||||
1. 用 SearchClient（Tavily > Exa > Brave）搜一次（query = "<term> <domain hint>"）
|
||||
2. 把 top 3-5 snippet 喂给 Haiku，让模型返回 {zh, en_full, confidence, issue}
|
||||
3. 合并回 glossary，字段扩展：
|
||||
{
|
||||
|
||||
+39
-57
@@ -41,6 +41,8 @@ from pathlib import Path
|
||||
sys.path.insert(0, str(Path(__file__).resolve().parent.parent))
|
||||
|
||||
from scripts.lib.zenmux_client import load_secrets # noqa: F401 (为一致性)
|
||||
from scripts.reporting.fonts import resolve_quarto_fonts
|
||||
from scripts.reporting.references import build_references_block
|
||||
|
||||
|
||||
REPO_ROOT = Path(__file__).resolve().parent.parent
|
||||
@@ -172,9 +174,10 @@ def prepare_qmd(
|
||||
subtitle = manifest.get("report_subtitle", "")
|
||||
date = manifest.get("date", "")
|
||||
|
||||
# 决定字体名称：思源宋体 CN 作正文，思源黑体 CN 作标题
|
||||
main_font = "Source Han Serif CN"
|
||||
sans_font = "Source Han Sans CN"
|
||||
# 决定字体名称：Quarto/xelatex 使用系统字体 family name。
|
||||
fonts = resolve_quarto_fonts(fonts_dir)
|
||||
main_font = fonts.main_font
|
||||
sans_font = fonts.sans_font
|
||||
|
||||
# Write LaTeX header file for CJK font setup.
|
||||
# Using a separate .tex file avoids YAML escape issues with backslashes.
|
||||
@@ -240,6 +243,14 @@ def prepare_qmd(
|
||||
if end != -1:
|
||||
md_text = md_text[end + 4:].lstrip("\n")
|
||||
|
||||
# Quarto already renders the title from YAML; drop a duplicated leading H1.
|
||||
md_text = re.sub(
|
||||
rf"^#\s+{re.escape(title)}\s*\n+",
|
||||
"",
|
||||
md_text,
|
||||
count=1,
|
||||
)
|
||||
|
||||
# Replace TOC placeholder
|
||||
md_text = re.sub(
|
||||
r"\[TOC will be generated.*?\]",
|
||||
@@ -248,7 +259,7 @@ def prepare_qmd(
|
||||
)
|
||||
|
||||
# Replace REFERENCES placeholder with actual references from sources.jsonl
|
||||
ref_block = _build_references_block(sources_path, md_text)
|
||||
ref_block = build_references_block(sources_path, md_text)
|
||||
md_text = re.sub(
|
||||
r"\[REFERENCES will be filled.*?\]",
|
||||
ref_block,
|
||||
@@ -273,71 +284,42 @@ def prepare_qmd(
|
||||
"\\usepackage{longtable}\n"
|
||||
"\\usepackage{booktabs}\n"
|
||||
"\\usepackage{array}\n"
|
||||
"\\usepackage{xcolor}\n"
|
||||
"\\usepackage{titlesec}\n"
|
||||
"\\definecolor{DRBlue}{HTML}{1E3A8A}\n"
|
||||
"\\definecolor{DRSlate}{HTML}{374151}\n"
|
||||
"\\definecolor{DRMuted}{HTML}{6B7280}\n"
|
||||
# Use lscape instead of pdflscape to avoid \LS@makefcolumn recursion
|
||||
# which exhausts TeX param_size on large longtables.
|
||||
# lscape rotates content without changing page media box (reader must rotate).
|
||||
"\\usepackage{lscape}\n"
|
||||
"\\setlength{\\LTpre}{6pt}\n"
|
||||
"\\setlength{\\LTpost}{6pt}\n"
|
||||
"\\setlength{\\tabcolsep}{3pt}\n",
|
||||
"\\setlength{\\tabcolsep}{3pt}\n"
|
||||
"\\linespread{1.18}\n"
|
||||
"\\setlength{\\parindent}{2em}\n"
|
||||
"\\setlength{\\parskip}{0.25em}\n"
|
||||
"\\newcommand{\\sectionbreak}{\\clearpage}\n"
|
||||
"\\titleformat{\\section}[display]\n"
|
||||
" {\\centering\\Large\\bfseries\\sffamily\\color{DRBlue}}\n"
|
||||
" {}{0pt}{}\n"
|
||||
"\\titlespacing*{\\section}{0pt}{0pt}{1.1em}\n"
|
||||
"\\titleformat{\\subsection}\n"
|
||||
" {\\large\\bfseries\\sffamily\\color{DRBlue}}\n"
|
||||
" {}{0pt}{}\n"
|
||||
"\\titlespacing*{\\subsection}{0pt}{1.1em}{0.45em}\n"
|
||||
"\\titleformat{\\subsubsection}\n"
|
||||
" {\\normalsize\\bfseries\\sffamily\\color{DRSlate}}\n"
|
||||
" {}{0pt}{}\n"
|
||||
"\\titlespacing*{\\subsubsection}{0pt}{0.9em}{0.35em}\n"
|
||||
"\\renewcommand{\\contentsname}{目录}\n"
|
||||
"\\setcounter{tocdepth}{2}\n",
|
||||
encoding="utf-8",
|
||||
)
|
||||
|
||||
output_qmd.write_text(front_matter + md_text, encoding="utf-8")
|
||||
print(f" .qmd prepared: {output_qmd.name} ({len(wide_ranges)} landscape table(s))")
|
||||
|
||||
|
||||
def _build_references_block(sources_path: Path | None, md_text: str) -> str:
|
||||
"""从 sources.jsonl 生成参考文献列表，只包含在正文中实际引用的信源。"""
|
||||
if not sources_path or not sources_path.exists():
|
||||
return "（参考文献列表：sources.jsonl 未找到）"
|
||||
|
||||
# Find cited src_ids
|
||||
cited = set(re.findall(r"\[src_([a-z0-9_]+)\]", md_text))
|
||||
if not cited:
|
||||
return ""
|
||||
|
||||
sources: dict[str, dict] = {}
|
||||
with open(sources_path, encoding="utf-8") as f:
|
||||
for line in f:
|
||||
line = line.strip()
|
||||
if not line:
|
||||
continue
|
||||
try:
|
||||
obj = json.loads(line)
|
||||
sid = obj.get("id", "")
|
||||
key = sid.replace("src_", "")
|
||||
if key in cited:
|
||||
sources[sid] = obj
|
||||
except json.JSONDecodeError:
|
||||
pass
|
||||
|
||||
if not sources:
|
||||
return ""
|
||||
|
||||
lines = ["## 参考文献\n"]
|
||||
for sid in sorted(sources.keys()):
|
||||
s = sources[sid]
|
||||
authors = ", ".join(s.get("authors", [])) if s.get("authors") else ""
|
||||
year = s.get("year", "")
|
||||
title = s.get("title", sid)
|
||||
venue = s.get("venue", "")
|
||||
url = s.get("url", "")
|
||||
entry = f"- **[{sid}]** "
|
||||
if authors:
|
||||
entry += f"{authors}. "
|
||||
if year:
|
||||
entry += f"({year}). "
|
||||
entry += f"*{title}*"
|
||||
if venue:
|
||||
entry += f". {venue}"
|
||||
if url:
|
||||
entry += f". <{url}>"
|
||||
lines.append(entry)
|
||||
|
||||
return "\n".join(lines)
|
||||
|
||||
|
||||
def build_pdf_quarto(
|
||||
md_path: Path,
|
||||
manifest: dict,
|
||||
|
||||
@@ -0,0 +1,290 @@
|
||||
#!/usr/bin/env python3
|
||||
"""Deploy platform adapter templates outside the repository checkout."""
|
||||
|
||||
from __future__ import annotations
|
||||
|
||||
import argparse
|
||||
import os
|
||||
import shutil
|
||||
import sys
|
||||
from dataclasses import dataclass, field
|
||||
from pathlib import Path
|
||||
|
||||
|
||||
REPO_ROOT = Path(__file__).resolve().parent.parent
|
||||
if str(REPO_ROOT) not in sys.path:
|
||||
sys.path.insert(0, str(REPO_ROOT))
|
||||
|
||||
from scripts.runtime.skills import SkillRegistry
|
||||
|
||||
CODEX_TEMPLATE = REPO_ROOT / "codex_adapter_templates" / "codex"
|
||||
ANTIGRAVITY_RULES_DIR = REPO_ROOT / ".agents" / "rules"
|
||||
ANTIGRAVITY_WORKFLOWS_DIR = REPO_ROOT / ".agents" / "workflows"
|
||||
ANTIGRAVITY_AGENTS_FILE = REPO_ROOT / ".agents" / "agents.md"
|
||||
|
||||
|
||||
@dataclass
|
||||
class DeployResult:
|
||||
platform: str
|
||||
target: Path
|
||||
written: list[Path] = field(default_factory=list)
|
||||
skipped: list[Path] = field(default_factory=list)
|
||||
planned: list[Path] = field(default_factory=list)
|
||||
backups: list[Path] = field(default_factory=list)
|
||||
|
||||
|
||||
def default_codex_home(
|
||||
*,
|
||||
env: dict[str, str] | None = None,
|
||||
user_home: Path | None = None,
|
||||
) -> Path:
|
||||
values = os.environ if env is None else env
|
||||
if values.get("CODEX_HOME"):
|
||||
return Path(values["CODEX_HOME"]).expanduser()
|
||||
home = Path.home() if user_home is None else user_home
|
||||
return home / ".codex"
|
||||
|
||||
|
||||
def copy_tree_contents(
|
||||
src: Path,
|
||||
dst: Path,
|
||||
*,
|
||||
force: bool,
|
||||
dry_run: bool = False,
|
||||
backup_existing: bool = True,
|
||||
exclude: set[Path] | None = None,
|
||||
) -> DeployResult:
|
||||
if not src.exists():
|
||||
raise FileNotFoundError(f"adapter template source not found: {src}")
|
||||
|
||||
result = DeployResult(platform="copy", target=dst)
|
||||
excluded = exclude or set()
|
||||
for item in sorted(src.rglob("*")):
|
||||
rel = item.relative_to(src)
|
||||
if rel in excluded:
|
||||
continue
|
||||
target = dst / rel
|
||||
if item.is_dir():
|
||||
if not dry_run:
|
||||
target.mkdir(parents=True, exist_ok=True)
|
||||
continue
|
||||
|
||||
if target.exists() and not force:
|
||||
result.skipped.append(target)
|
||||
continue
|
||||
|
||||
result.planned.append(target)
|
||||
if dry_run:
|
||||
continue
|
||||
|
||||
target.parent.mkdir(parents=True, exist_ok=True)
|
||||
if target.exists() and force and backup_existing:
|
||||
backup = target.with_name(f"{target.name}.bak")
|
||||
shutil.copy2(target, backup)
|
||||
result.backups.append(backup)
|
||||
shutil.copy2(item, target)
|
||||
result.written.append(target)
|
||||
return result
|
||||
|
||||
|
||||
def _merge_results(platform: str, target: Path, parts: list[DeployResult]) -> DeployResult:
|
||||
merged = DeployResult(platform=platform, target=target)
|
||||
for part in parts:
|
||||
merged.written.extend(part.written)
|
||||
merged.skipped.extend(part.skipped)
|
||||
merged.planned.extend(part.planned)
|
||||
merged.backups.extend(part.backups)
|
||||
return merged
|
||||
|
||||
|
||||
def copy_registered_skills(dst: Path, *, force: bool, dry_run: bool = False) -> DeployResult:
|
||||
result = DeployResult(platform="skills", target=dst)
|
||||
for skill in SkillRegistry().list():
|
||||
part = copy_tree_contents(skill.path.parent, dst / skill.name, force=force, dry_run=dry_run)
|
||||
result.written.extend(part.written)
|
||||
result.skipped.extend(part.skipped)
|
||||
result.planned.extend(part.planned)
|
||||
result.backups.extend(part.backups)
|
||||
return result
|
||||
|
||||
|
||||
def copy_antigravity_rules(dst: Path, *, force: bool, dry_run: bool = False) -> DeployResult:
|
||||
result = DeployResult(platform="rules", target=dst)
|
||||
if not ANTIGRAVITY_RULES_DIR.exists():
|
||||
return result
|
||||
part = copy_tree_contents(ANTIGRAVITY_RULES_DIR, dst, force=force, dry_run=dry_run)
|
||||
result.written.extend(part.written)
|
||||
result.skipped.extend(part.skipped)
|
||||
result.planned.extend(part.planned)
|
||||
result.backups.extend(part.backups)
|
||||
return result
|
||||
|
||||
|
||||
def copy_antigravity_workflows(dst: Path, *, force: bool, dry_run: bool = False) -> DeployResult:
|
||||
result = DeployResult(platform="workflows", target=dst)
|
||||
if not ANTIGRAVITY_WORKFLOWS_DIR.exists():
|
||||
return result
|
||||
part = copy_tree_contents(ANTIGRAVITY_WORKFLOWS_DIR, dst, force=force, dry_run=dry_run)
|
||||
result.written.extend(part.written)
|
||||
result.skipped.extend(part.skipped)
|
||||
result.planned.extend(part.planned)
|
||||
result.backups.extend(part.backups)
|
||||
return result
|
||||
|
||||
|
||||
def copy_antigravity_agents_file(dst: Path, *, force: bool, dry_run: bool = False) -> DeployResult:
|
||||
result = DeployResult(platform="agents", target=dst)
|
||||
if not ANTIGRAVITY_AGENTS_FILE.exists():
|
||||
return result
|
||||
target = dst / "agents.md"
|
||||
if target.exists() and not force:
|
||||
result.skipped.append(target)
|
||||
return result
|
||||
result.planned.append(target)
|
||||
if dry_run:
|
||||
return result
|
||||
target.parent.mkdir(parents=True, exist_ok=True)
|
||||
if target.exists() and force:
|
||||
backup = target.with_name("agents.md.bak")
|
||||
shutil.copy2(target, backup)
|
||||
result.backups.append(backup)
|
||||
shutil.copy2(ANTIGRAVITY_AGENTS_FILE, target)
|
||||
result.written.append(target)
|
||||
return result
|
||||
|
||||
|
||||
def deploy_codex(
|
||||
*,
|
||||
target: Path | None = None,
|
||||
force: bool = False,
|
||||
skip_skills: bool = False,
|
||||
dry_run: bool = False,
|
||||
include_config: bool = False,
|
||||
repo_root: Path = REPO_ROOT,
|
||||
) -> DeployResult:
|
||||
codex_home = (target or default_codex_home()).expanduser()
|
||||
template = repo_root / "codex_adapter_templates" / "codex"
|
||||
|
||||
parts = [
|
||||
copy_tree_contents(
|
||||
template,
|
||||
codex_home,
|
||||
force=force,
|
||||
dry_run=dry_run,
|
||||
exclude=set() if include_config else {Path("config.toml")},
|
||||
),
|
||||
]
|
||||
if not skip_skills:
|
||||
parts.append(copy_registered_skills(codex_home / "skills", force=force, dry_run=dry_run))
|
||||
return _merge_results("codex", codex_home, parts)
|
||||
|
||||
|
||||
def deploy_antigravity(
|
||||
*,
|
||||
target: Path | None = None,
|
||||
force: bool = False,
|
||||
skip_skills: bool = False,
|
||||
skip_agents: bool = False,
|
||||
skip_rules: bool = False,
|
||||
skip_workflows: bool = False,
|
||||
dry_run: bool = False,
|
||||
repo_root: Path = REPO_ROOT,
|
||||
) -> DeployResult:
|
||||
"""Deploy Antigravity workspace rules and skills into a workspace root.
|
||||
|
||||
This intentionally deploys to a workspace-local `.agents` directory, not
|
||||
global Antigravity/Gemini settings, so existing user configuration is not
|
||||
touched. Existing files are skipped unless `force=True`.
|
||||
"""
|
||||
workspace = (target or repo_root).expanduser()
|
||||
parts: list[DeployResult] = []
|
||||
if not skip_agents:
|
||||
parts.append(copy_antigravity_agents_file(workspace / ".agents", force=force, dry_run=dry_run))
|
||||
if not skip_skills:
|
||||
parts.append(copy_registered_skills(workspace / ".agents" / "skills", force=force, dry_run=dry_run))
|
||||
if not skip_rules:
|
||||
parts.append(copy_antigravity_rules(workspace / ".agents" / "rules", force=force, dry_run=dry_run))
|
||||
if not skip_workflows:
|
||||
parts.append(copy_antigravity_workflows(workspace / ".agents" / "workflows", force=force, dry_run=dry_run))
|
||||
return _merge_results("antigravity", workspace, parts)
|
||||
|
||||
|
||||
def print_result(result: DeployResult) -> None:
|
||||
action = "planned" if result.planned and not result.written else "written"
|
||||
print(f"{result.platform} adapter deployment")
|
||||
print(f" target: {result.target}")
|
||||
print(f" files {action}: {len(result.planned if action == 'planned' else result.written)}")
|
||||
print(f" files skipped: {len(result.skipped)}")
|
||||
print(f" backups: {len(result.backups)}")
|
||||
if result.written:
|
||||
for path in result.written:
|
||||
print(f" {path}")
|
||||
elif result.planned:
|
||||
for path in result.planned:
|
||||
print(f" {path}")
|
||||
|
||||
|
||||
def build_parser() -> argparse.ArgumentParser:
|
||||
parser = argparse.ArgumentParser(description="Deploy Deep Research surface adapters")
|
||||
sub = parser.add_subparsers(dest="platform", required=True)
|
||||
|
||||
codex = sub.add_parser("codex", help="Deploy Codex adapter to CODEX_HOME or a target directory")
|
||||
codex.add_argument("--target", type=Path, help="Codex home target; defaults to $CODEX_HOME or ~/.codex")
|
||||
codex.add_argument("--force", action="store_true", help="overwrite existing files and create .bak backups")
|
||||
codex.add_argument("--skip-skills", action="store_true", help="do not copy canonical skills into target/skills")
|
||||
codex.add_argument("--include-config", action="store_true", help="also copy config.toml; off by default to avoid overwriting global Codex config")
|
||||
codex.add_argument("--dry-run", action="store_true", help="show files that would be written")
|
||||
|
||||
antigravity = sub.add_parser("antigravity", help="Deploy Antigravity workspace rules and skills")
|
||||
antigravity.add_argument("--target", type=Path, help="Workspace root; defaults to this repository")
|
||||
antigravity.add_argument("--force", action="store_true", help="overwrite existing files and create .bak backups")
|
||||
antigravity.add_argument("--skip-agents", action="store_true", help="do not copy role definitions into target/.agents/agents.md")
|
||||
antigravity.add_argument("--skip-skills", action="store_true", help="do not copy canonical skills into target/.agents/skills")
|
||||
antigravity.add_argument("--skip-rules", action="store_true", help="do not copy workspace rules into target/.agents/rules")
|
||||
antigravity.add_argument("--skip-workflows", action="store_true", help="do not copy workflows into target/.agents/workflows")
|
||||
antigravity.add_argument("--dry-run", action="store_true", help="show files that would be written without writing")
|
||||
return parser
|
||||
|
||||
|
||||
def main() -> int:
|
||||
args = build_parser().parse_args()
|
||||
if args.platform == "codex":
|
||||
result = deploy_codex(
|
||||
target=args.target,
|
||||
force=args.force,
|
||||
skip_agents=args.skip_agents,
|
||||
skip_skills=args.skip_skills,
|
||||
dry_run=args.dry_run,
|
||||
include_config=args.include_config,
|
||||
)
|
||||
print_result(result)
|
||||
print()
|
||||
print("Run Codex from this repository after deployment:")
|
||||
if args.include_config:
|
||||
print(" codex --profile deep-research")
|
||||
else:
|
||||
print(" codex")
|
||||
print("Note: config.toml is not copied by default. Use --include-config only if you want the bundled profile.")
|
||||
return 0
|
||||
if args.platform == "antigravity":
|
||||
result = deploy_antigravity(
|
||||
target=args.target,
|
||||
force=args.force,
|
||||
skip_skills=args.skip_skills,
|
||||
skip_rules=args.skip_rules,
|
||||
skip_workflows=args.skip_workflows,
|
||||
dry_run=args.dry_run,
|
||||
)
|
||||
print_result(result)
|
||||
print()
|
||||
print("Open the target workspace in Antigravity and enable/mention the workspace rule if needed:")
|
||||
print(" .agents/rules/deep-research-antigravity.md")
|
||||
print("Workflow installed when supported by your Antigravity build:")
|
||||
print(" .agents/workflows/deep-research-native.md")
|
||||
print("Existing files are skipped by default. Use --force only when you want .bak backups and replacement.")
|
||||
return 0
|
||||
raise SystemExit(f"unsupported platform: {args.platform}")
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
raise SystemExit(main())
|
||||
+37
-18
@@ -17,8 +17,13 @@ except ModuleNotFoundError: # pragma: no cover - Python < 3.11 fallback.
|
||||
|
||||
|
||||
REPO_ROOT = Path(__file__).resolve().parent.parent
|
||||
if str(REPO_ROOT) not in sys.path:
|
||||
sys.path.insert(0, str(REPO_ROOT))
|
||||
|
||||
from scripts.deploy_adapters import default_codex_home, deploy_codex
|
||||
from scripts.runtime.skills import SkillRegistry
|
||||
|
||||
CODEX_TEMPLATE = REPO_ROOT / "codex_adapter_templates" / "codex"
|
||||
CODEX_ROOT = REPO_ROOT / ".codex"
|
||||
OPENCODE_ROOT = REPO_ROOT / ".opencode"
|
||||
AGENTS_SKILLS = REPO_ROOT / ".agents" / "skills"
|
||||
OPENCODE_SKILLS = OPENCODE_ROOT / "skills"
|
||||
@@ -28,11 +33,14 @@ REQUIRED_PATHS = [
|
||||
"README.md",
|
||||
"PLAN.md",
|
||||
".opencode/opencode.json",
|
||||
".codex/config.toml",
|
||||
".codex/agents/dr-pm.toml",
|
||||
".codex/commands/dr-run.md",
|
||||
"codex_adapter_templates/codex/config.toml",
|
||||
"codex_adapter_templates/codex/agents/dr-pm.toml",
|
||||
"codex_adapter_templates/codex/commands/dr-run.md",
|
||||
".agents/skills/search-strategy/SKILL.md",
|
||||
"skills/deep-research/SKILL.md",
|
||||
"skills/document-ingest/SKILL.md",
|
||||
"scripts/dr.py",
|
||||
"scripts/deploy_adapters.py",
|
||||
"scripts/install_codex_adapter.py",
|
||||
]
|
||||
|
||||
@@ -111,19 +119,26 @@ def check_deployment() -> int:
|
||||
if not (REPO_ROOT / item).exists():
|
||||
issues.append(f"missing required path: {item}")
|
||||
|
||||
tracked = git_tracked([".codex", ".opencode", ".agents/skills"])
|
||||
tracked = git_tracked(["codex_adapter_templates", ".opencode", ".agents/skills", "skills"])
|
||||
legacy_tracked = git_tracked([".codex"])
|
||||
if legacy_tracked:
|
||||
warnings.append("legacy .codex files are still tracked; run: git rm -r --cached .codex")
|
||||
for item in REQUIRED_PATHS:
|
||||
if item.startswith((".codex/", ".opencode/", ".agents/")) and item not in tracked:
|
||||
if item.startswith(("codex_adapter_templates/", ".opencode/", ".agents/")) and item not in tracked:
|
||||
warnings.append(f"not tracked by git: {item}")
|
||||
|
||||
skill_files = sorted(AGENTS_SKILLS.glob("*/SKILL.md"))
|
||||
if len(skill_files) < 10:
|
||||
issues.append(f"expected at least 10 Codex skills, found {len(skill_files)}")
|
||||
skills = SkillRegistry().list()
|
||||
if len(skills) < 10:
|
||||
issues.append(f"expected at least 10 Codex skills, found {len(skills)}")
|
||||
|
||||
check_toml(CODEX_ROOT / "config.toml", issues)
|
||||
for path in sorted((CODEX_ROOT / "agents").glob("*.toml")):
|
||||
check_toml(CODEX_TEMPLATE / "config.toml", issues)
|
||||
for path in sorted((CODEX_TEMPLATE / "agents").glob("*.toml")):
|
||||
check_toml(path, issues)
|
||||
|
||||
codex_home = default_codex_home()
|
||||
if not (codex_home / "commands" / "dr-run.md").exists():
|
||||
warnings.append(f"Codex adapter not deployed to {codex_home}; run scripts/deploy_adapters.py codex")
|
||||
|
||||
env_values = {key: os.environ.get(key, "") for key in REQUIRED_ENV_KEYS}
|
||||
env_values.update({k: v for k, v in parse_env(REPO_ROOT / "secrets.env").items() if not env_values.get(k)})
|
||||
missing_env = [key for key in REQUIRED_ENV_KEYS if not env_values.get(key)]
|
||||
@@ -132,9 +147,11 @@ def check_deployment() -> int:
|
||||
|
||||
print("Deep Research deployment check")
|
||||
print(f" repo: {REPO_ROOT}")
|
||||
print(f" codex files tracked: {sum(1 for p in tracked if p.startswith('.codex/'))}")
|
||||
print(f" codex template files tracked: {sum(1 for p in tracked if p.startswith('codex_adapter_templates/'))}")
|
||||
print(f" legacy .codex files tracked: {len(legacy_tracked)}")
|
||||
print(f" codex home: {codex_home}")
|
||||
print(f" opencode files tracked: {sum(1 for p in tracked if p.startswith('.opencode/'))}")
|
||||
print(f" codex skills: {len(skill_files)}")
|
||||
print(f" codex skills: {len(skills)}")
|
||||
|
||||
if warnings:
|
||||
print("\nWarnings:")
|
||||
@@ -151,9 +168,9 @@ def check_deployment() -> int:
|
||||
return 0
|
||||
|
||||
|
||||
def repair(force: bool) -> int:
|
||||
def repair(force: bool, codex_home: Path | None) -> int:
|
||||
try:
|
||||
codex_written = copy_tree_contents(CODEX_TEMPLATE, CODEX_ROOT, force=force)
|
||||
codex_result = deploy_codex(target=codex_home, force=force)
|
||||
skills_written = copy_tree_contents(OPENCODE_SKILLS, AGENTS_SKILLS, force=force)
|
||||
except PermissionError as exc:
|
||||
print(f"repair failed: permission denied: {exc}", file=sys.stderr)
|
||||
@@ -163,19 +180,21 @@ def repair(force: bool) -> int:
|
||||
return 1
|
||||
|
||||
print("Repair completed.")
|
||||
print(f" .codex files written: {len(codex_written)}")
|
||||
print(f" Codex home files written: {len(codex_result.written)}")
|
||||
print(f" Codex home: {codex_result.target}")
|
||||
print(f" .agents skills written: {len(skills_written)}")
|
||||
return check_deployment()
|
||||
|
||||
|
||||
def main() -> int:
|
||||
parser = argparse.ArgumentParser(description="Check or repair Deep Research deployment files")
|
||||
parser.add_argument("--repair", action="store_true", help="copy Codex templates and skills into hidden dirs")
|
||||
parser.add_argument("--repair", action="store_true", help="deploy Codex templates outside the repo and sync skills")
|
||||
parser.add_argument("--force", action="store_true", help="overwrite existing files during --repair")
|
||||
parser.add_argument("--codex-home", type=Path, help="Codex home target for --repair; defaults to $CODEX_HOME or ~/.codex")
|
||||
args = parser.parse_args()
|
||||
|
||||
if args.repair:
|
||||
return repair(force=args.force)
|
||||
return repair(force=args.force, codex_home=args.codex_home)
|
||||
return check_deployment()
|
||||
|
||||
|
||||
|
||||
+650
-12
@@ -12,6 +12,7 @@ import json
|
||||
import re
|
||||
import subprocess
|
||||
import sys
|
||||
from datetime import datetime, timezone
|
||||
from pathlib import Path
|
||||
|
||||
|
||||
@@ -25,25 +26,36 @@ from scripts.lib.model_config import (
|
||||
parse_model_overrides,
|
||||
resolve_model_profile,
|
||||
)
|
||||
from scripts.runtime.assembly import build_chapter_briefs, build_compressed_findings, run_chapter_assembly_workers
|
||||
from scripts.runtime.orchestrator import create_phase2_task_cards, write_placeholder_packets
|
||||
from scripts.runtime.methods import ResearchMethodRegistry
|
||||
from scripts.runtime.phase1 import create_project, render_framework, write_material_brief
|
||||
from scripts.runtime.review import build_phase3_critique, build_phase3_model_critique
|
||||
from scripts.runtime.roles import resolve_runtime_profile
|
||||
from scripts.runtime.source_cache import cache_sources
|
||||
from scripts.runtime.sources import rebuild_sources_from_packets
|
||||
from scripts.runtime.skills import SkillRegistry, default_adapter_skill_dirs
|
||||
from scripts.runtime.tasks import TaskCard, load_task_cards, validate_packet, write_task_cards
|
||||
from scripts.runtime.workers import run_packet_workers
|
||||
|
||||
|
||||
PROJECTS_DIR = REPO_ROOT / "projects"
|
||||
CODEX_COMMANDS_DIR = REPO_ROOT / ".codex" / "commands"
|
||||
CODEX_COMMAND_TEMPLATES_DIR = REPO_ROOT / "codex_adapter_templates" / "codex" / "commands"
|
||||
LEGACY_CODEX_COMMANDS_DIR = REPO_ROOT / ".codex" / "commands"
|
||||
|
||||
|
||||
def resolve_project(project: str | None) -> Path:
|
||||
def resolve_project(project: str | None, *, projects_dir: Path = PROJECTS_DIR) -> Path:
|
||||
if project:
|
||||
p = Path(project)
|
||||
if p.is_dir():
|
||||
return p.resolve()
|
||||
cand = PROJECTS_DIR / project
|
||||
cand = projects_dir / project
|
||||
if cand.is_dir():
|
||||
return cand.resolve()
|
||||
raise SystemExit(f"project not found: {project}")
|
||||
|
||||
manifests = sorted(
|
||||
PROJECTS_DIR.glob("*/manifest.json"),
|
||||
projects_dir.glob("*/manifest.json"),
|
||||
key=lambda p: p.stat().st_mtime,
|
||||
reverse=True,
|
||||
)
|
||||
@@ -71,6 +83,43 @@ def file_state(path: Path) -> str:
|
||||
return "yes" if path.exists() else "no"
|
||||
|
||||
|
||||
def packet_state_counts(project_root: Path) -> dict[str, int]:
|
||||
packets = sorted((project_root / "phase2" / "packets").glob("*.json"))
|
||||
errors = sorted((project_root / "phase2" / "packet_errors").glob("*.json"))
|
||||
counts = {
|
||||
"ready": 0,
|
||||
"placeholder": 0,
|
||||
"invalid": 0,
|
||||
"errors": 0,
|
||||
"stale_errors": 0,
|
||||
"total": len(packets),
|
||||
}
|
||||
ready_stems: set[str] = set()
|
||||
for path in packets:
|
||||
try:
|
||||
packet = json.loads(path.read_text(encoding="utf-8"))
|
||||
except Exception:
|
||||
counts["invalid"] += 1
|
||||
continue
|
||||
has_evidence = bool(
|
||||
packet.get("claims")
|
||||
or packet.get("evidence_items")
|
||||
or packet.get("counter_evidence")
|
||||
or packet.get("source_ids")
|
||||
)
|
||||
if has_evidence:
|
||||
counts["ready"] += 1
|
||||
ready_stems.add(path.stem)
|
||||
else:
|
||||
counts["placeholder"] += 1
|
||||
for path in errors:
|
||||
if path.stem in ready_stems:
|
||||
counts["stale_errors"] += 1
|
||||
else:
|
||||
counts["errors"] += 1
|
||||
return counts
|
||||
|
||||
|
||||
def run_cmd(cmd: list[str], *, dry_run: bool) -> int:
|
||||
printable = " ".join(cmd)
|
||||
print(f"$ {printable}")
|
||||
@@ -79,6 +128,324 @@ def run_cmd(cmd: list[str], *, dry_run: bool) -> int:
|
||||
return subprocess.run(cmd, cwd=REPO_ROOT, check=False).returncode
|
||||
|
||||
|
||||
def cmd_init(args: argparse.Namespace) -> int:
|
||||
projects_dir = Path(args.projects_dir) if args.projects_dir else PROJECTS_DIR
|
||||
project_root = create_project(
|
||||
topic=args.topic,
|
||||
slug=args.slug,
|
||||
projects_dir=projects_dir,
|
||||
method_key=args.method,
|
||||
report_type=args.report_type,
|
||||
model_profile=args.profile,
|
||||
target_words=args.target_words,
|
||||
input_materials=args.input_material,
|
||||
)
|
||||
print(f"Project: {project_root.name}")
|
||||
print(f"Created: {project_root}")
|
||||
print("Runtime: python-core-v0.20")
|
||||
print("Next: run `dr.py frame <project>` to generate phase1/framework.md")
|
||||
return 0
|
||||
|
||||
|
||||
def cmd_frame(args: argparse.Namespace) -> int:
|
||||
project_root = resolve_project(args.project)
|
||||
if args.dry_run:
|
||||
manifest = load_manifest(project_root)
|
||||
method = ResearchMethodRegistry().get(args.method or manifest.get("research_method"))
|
||||
print(f"Project: {project_root.name}")
|
||||
print(f"Would write: phase1/framework.md")
|
||||
print(f"Research method: {method.key}")
|
||||
print(f"Chapters: {args.chapters}")
|
||||
return 0
|
||||
path = render_framework(
|
||||
project_root,
|
||||
method_key=args.method,
|
||||
chapter_count=args.chapters,
|
||||
preserve_existing_outline=args.preserve_existing_outline,
|
||||
)
|
||||
print(f"Project: {project_root.name}")
|
||||
print(f"Wrote: {path.relative_to(project_root)}")
|
||||
print("Pause: review and approve the framework before Phase 2.")
|
||||
return 0
|
||||
|
||||
|
||||
def cmd_approve(args: argparse.Namespace) -> int:
|
||||
project_root = resolve_project(args.project)
|
||||
manifest = load_manifest(project_root)
|
||||
phase1 = manifest.setdefault("phase1", {})
|
||||
phase1["approved"] = True
|
||||
phase1["requires_user_interview"] = False
|
||||
phase1["approved_at"] = datetime.now(timezone.utc).replace(microsecond=0).isoformat()
|
||||
manifest["updated_at"] = phase1["approved_at"]
|
||||
(project_root / "manifest.json").write_text(
|
||||
json.dumps(manifest, ensure_ascii=False, indent=2) + "\n",
|
||||
encoding="utf-8",
|
||||
)
|
||||
print(f"Project: {project_root.name}")
|
||||
print("Phase 1 approved. Phase 2 research is now enabled.")
|
||||
return 0
|
||||
|
||||
|
||||
def cmd_skills(args: argparse.Namespace) -> int:
|
||||
registry = SkillRegistry()
|
||||
if args.skills_cmd == "list":
|
||||
for name in registry.list_names():
|
||||
print(name)
|
||||
return 0
|
||||
if args.skills_cmd == "validate":
|
||||
result = registry.validate()
|
||||
print(json.dumps(result, ensure_ascii=False, indent=2))
|
||||
return 0 if result["ok"] else 1
|
||||
if args.skills_cmd == "sync":
|
||||
targets = [Path(item) for item in args.target] if args.target else default_adapter_skill_dirs()
|
||||
copied = registry.sync_to(targets, force=True)
|
||||
print(f"Synced skills: {copied}")
|
||||
for target in targets:
|
||||
print(f" {target}")
|
||||
return 0
|
||||
raise SystemExit(f"unknown skills command: {args.skills_cmd}")
|
||||
|
||||
|
||||
def cmd_methods(args: argparse.Namespace) -> int:
|
||||
registry = ResearchMethodRegistry()
|
||||
if args.methods_cmd == "list":
|
||||
for name in registry.list_names():
|
||||
method = registry.get(name)
|
||||
print(f"{method.key}: {method.name}")
|
||||
return 0
|
||||
if args.methods_cmd == "show":
|
||||
method = registry.get(args.method)
|
||||
print(json.dumps(method.__dict__, ensure_ascii=False, indent=2))
|
||||
return 0
|
||||
raise SystemExit(f"unknown methods command: {args.methods_cmd}")
|
||||
|
||||
|
||||
def cmd_research(args: argparse.Namespace) -> int:
|
||||
project_root = resolve_project(args.project)
|
||||
manifest = load_manifest(project_root)
|
||||
if not args.force and not (manifest.get("phase1") or {}).get("approved"):
|
||||
raise SystemExit(
|
||||
"Phase 1 is not approved. Review phase1/material_brief.md and phase1/framework.md, "
|
||||
"then run `uv run python scripts/dr.py approve <project>` or pass --force."
|
||||
)
|
||||
if args.profile == "codex_native" and (args.execute_packets or args.assemble_chapters):
|
||||
raise SystemExit(
|
||||
"`codex_native` cannot be used for Python-core model execution: scripts/dr.py currently calls external "
|
||||
"API clients, not Codex App built-in models. Use a clearly external profile such as `medium`, or run "
|
||||
"Codex-native execution through the surface adapter/manual task workflow."
|
||||
)
|
||||
runtime = resolve_runtime_profile(
|
||||
profile=args.profile,
|
||||
overrides=parse_model_overrides(args.model_override),
|
||||
)
|
||||
if args.append_task_cards:
|
||||
generated = create_phase2_task_cards(
|
||||
project_root,
|
||||
axes=args.axis,
|
||||
dry_run=True,
|
||||
)
|
||||
existing_path = project_root / "phase2" / "task_cards.json"
|
||||
existing_cards = load_task_cards(existing_path) if existing_path.exists() else []
|
||||
seen = {card.task_id for card in existing_cards}
|
||||
appended_cards = [TaskCard(**item) for item in generated if item["task_id"] not in seen]
|
||||
runnable_existing_cards: list[TaskCard] = []
|
||||
if args.execute_packets and args.axis:
|
||||
axis_set = set(args.axis)
|
||||
for card in existing_cards:
|
||||
if card.topic_axis not in axis_set:
|
||||
continue
|
||||
packet_path = project_root / card.output_packet
|
||||
try:
|
||||
validate_packet(json.loads(packet_path.read_text(encoding="utf-8")))
|
||||
except Exception:
|
||||
runnable_existing_cards.append(card)
|
||||
merged_cards = [*existing_cards, *appended_cards]
|
||||
if not args.dry_run:
|
||||
write_task_cards(existing_path, merged_cards)
|
||||
phase2 = manifest.setdefault("phase2", {})
|
||||
phase2.update(
|
||||
{
|
||||
"status": "in_progress",
|
||||
"runtime": "python-core-v0.20",
|
||||
"task_cards_path": "phase2/task_cards.json",
|
||||
"task_cards_total": len(merged_cards),
|
||||
"task_cards_appended": len(appended_cards),
|
||||
"updated_at": datetime.now(timezone.utc).replace(microsecond=0).isoformat(),
|
||||
}
|
||||
)
|
||||
(project_root / "manifest.json").write_text(
|
||||
json.dumps(manifest, ensure_ascii=False, indent=2) + "\n",
|
||||
encoding="utf-8",
|
||||
)
|
||||
card_dicts = [card.to_dict() for card in [*appended_cards, *runnable_existing_cards]]
|
||||
else:
|
||||
card_dicts = create_phase2_task_cards(
|
||||
project_root,
|
||||
axes=args.axis,
|
||||
dry_run=args.dry_run,
|
||||
)
|
||||
if args.execute_packets and args.dry_run:
|
||||
raise SystemExit("--execute-packets cannot be combined with --dry-run")
|
||||
if args.assemble_chapters and args.dry_run:
|
||||
raise SystemExit("--assemble-chapters cannot be combined with --dry-run")
|
||||
if args.execute_packets:
|
||||
from scripts.lib.zenmux_client import ZenMuxClient, load_secrets
|
||||
from scripts.runtime.workers import ProjectSearchProvider
|
||||
|
||||
load_secrets()
|
||||
|
||||
def client_factory(_role):
|
||||
return ZenMuxClient(log_file=project_root / "phase2" / "logs" / "packets.jsonl")
|
||||
|
||||
def search_provider_factory():
|
||||
return ProjectSearchProvider(strict_specialized=not args.allow_search_fallback)
|
||||
|
||||
packet_count = run_packet_workers(
|
||||
project_root=project_root,
|
||||
cards=[TaskCard(**item) for item in card_dicts],
|
||||
runtime=runtime,
|
||||
client_factory=client_factory,
|
||||
search_provider_factory=search_provider_factory,
|
||||
workers=args.workers,
|
||||
)
|
||||
elif not (args.build_briefs or args.assemble_chapters):
|
||||
packet_count = write_placeholder_packets(project_root, card_dicts, dry_run=args.dry_run)
|
||||
else:
|
||||
packet_count = len(list((project_root / "phase2" / "packets").glob("*.json")))
|
||||
brief_count = 0
|
||||
chapter_count = 0
|
||||
compressed_count = 0
|
||||
source_count = None
|
||||
if args.build_briefs or args.assemble_chapters:
|
||||
source_count = rebuild_sources_from_packets(project_root)
|
||||
briefs = build_chapter_briefs(project_root)
|
||||
brief_count = len(briefs)
|
||||
compressed_findings = build_compressed_findings(project_root)
|
||||
compressed_count = len(compressed_findings)
|
||||
if args.assemble_chapters:
|
||||
from scripts.lib.zenmux_client import ZenMuxClient, load_secrets
|
||||
|
||||
load_secrets()
|
||||
|
||||
def chapter_client_factory(_role):
|
||||
return ZenMuxClient(log_file=project_root / "phase2" / "logs" / "chapters.jsonl")
|
||||
|
||||
chapter_count = run_chapter_assembly_workers(
|
||||
project_root=project_root,
|
||||
briefs=compressed_findings,
|
||||
runtime=runtime,
|
||||
client_factory=chapter_client_factory,
|
||||
workers=args.workers,
|
||||
)
|
||||
print(f"Project: {project_root.name}")
|
||||
print(f"Runtime: python-core-v0.20")
|
||||
print(f"Model profile: {runtime.profile}")
|
||||
print(f"Workers: {args.workers}")
|
||||
print(f"Task cards: {len(card_dicts)}")
|
||||
print(f"Packets: {packet_count}")
|
||||
if source_count is not None:
|
||||
print(f"Sources rebuilt: {source_count}")
|
||||
if args.build_briefs or args.assemble_chapters:
|
||||
print(f"Chapter briefs: {brief_count}")
|
||||
print(f"Compressed findings: {compressed_count}")
|
||||
if args.assemble_chapters:
|
||||
print(f"Chapter drafts: {chapter_count}")
|
||||
if args.dry_run:
|
||||
print("Dry run: no files written")
|
||||
elif args.assemble_chapters:
|
||||
print("Wrote: phase2/drafts/chXX.md")
|
||||
elif args.execute_packets:
|
||||
print("Wrote: phase2/task_cards.json and validated phase2/packets/*.json")
|
||||
print("Next: rerun with --build-briefs to aggregate packets into chapter briefs.")
|
||||
elif args.build_briefs:
|
||||
print("Wrote: phase2/chapter_briefs/*.json and phase2/compressed_findings/*.json")
|
||||
print("Next: rerun with --assemble-chapters to write Chinese chapter drafts.")
|
||||
else:
|
||||
print("Wrote: phase2/task_cards.json and phase2/packets/*.json")
|
||||
print("Next: rerun with --execute-packets to fill packets via model workers.")
|
||||
return 0
|
||||
|
||||
|
||||
def cmd_run(args: argparse.Namespace) -> int:
|
||||
target = args.project_or_topic
|
||||
projects_dir = Path(args.projects_dir) if args.projects_dir else PROJECTS_DIR
|
||||
try:
|
||||
project_root = resolve_project(target, projects_dir=projects_dir)
|
||||
except SystemExit:
|
||||
if args.dry_run:
|
||||
print(f"New topic detected: {target}")
|
||||
print("Dry run: would create project and write phase1/framework.md")
|
||||
return 0
|
||||
project_root = create_project(
|
||||
topic=target,
|
||||
slug=args.slug,
|
||||
projects_dir=projects_dir,
|
||||
method_key=args.method,
|
||||
report_type=args.report_type,
|
||||
model_profile=args.profile or "medium",
|
||||
target_words=args.target_words,
|
||||
input_materials=args.input_material,
|
||||
)
|
||||
framework = render_framework(project_root, chapter_count=args.chapters)
|
||||
print(f"Project: {project_root.name}")
|
||||
print("Runtime: python-core-v0.20")
|
||||
print(f"Created: {project_root}")
|
||||
print(f"Wrote: {framework.relative_to(project_root)}")
|
||||
print("Pause: review and approve the framework before Phase 2.")
|
||||
return 0
|
||||
|
||||
print(f"Project: {project_root.name}")
|
||||
print("Runtime: python-core-v0.20")
|
||||
print("Next command: research")
|
||||
if args.dry_run:
|
||||
print("Dry run: would inspect manifest and continue from the next incomplete phase")
|
||||
return 0
|
||||
return cmd_research(
|
||||
argparse.Namespace(
|
||||
project=str(project_root),
|
||||
workers=args.workers,
|
||||
axis=None,
|
||||
profile=args.profile,
|
||||
model_override=[],
|
||||
execute_packets=False,
|
||||
append_task_cards=False,
|
||||
allow_search_fallback=False,
|
||||
build_briefs=False,
|
||||
assemble_chapters=False,
|
||||
force=False,
|
||||
dry_run=False,
|
||||
)
|
||||
)
|
||||
|
||||
|
||||
def cmd_review(args: argparse.Namespace) -> int:
|
||||
project_root = resolve_project(args.project)
|
||||
if args.model_review:
|
||||
if args.dry_run:
|
||||
print(f"Project: {project_root.name}")
|
||||
print("Phase 3 model review plan:")
|
||||
print(f" model: {args.model}")
|
||||
print(" context: phase3/review_context_opus_4_7.md")
|
||||
print(" output: phase3/critique.md")
|
||||
return 0
|
||||
from scripts.lib.zenmux_client import ZenMuxClient, load_secrets
|
||||
|
||||
load_secrets()
|
||||
with ZenMuxClient(log_file=project_root / "phase3" / "logs" / "review.jsonl") as client:
|
||||
path = build_phase3_model_critique(
|
||||
project_root,
|
||||
client=client,
|
||||
model=args.model,
|
||||
max_context_chars=args.max_context_chars,
|
||||
)
|
||||
else:
|
||||
path = build_phase3_critique(project_root)
|
||||
print(f"Project: {project_root.name}")
|
||||
print(f"Wrote: {path.relative_to(project_root)}")
|
||||
print("Pause: review critique before Phase 4.")
|
||||
return 0
|
||||
|
||||
|
||||
def cmd_status(args: argparse.Namespace) -> int:
|
||||
project_root = resolve_project(args.project)
|
||||
manifest = load_manifest(project_root)
|
||||
@@ -86,6 +453,10 @@ def cmd_status(args: argparse.Namespace) -> int:
|
||||
|
||||
drafts = sorted((project_root / "phase2" / "drafts").glob("ch*.md"))
|
||||
evidence = sorted((project_root / "phase2" / "evidence").glob("ch*-evidence.md"))
|
||||
task_cards = project_root / "phase2" / "task_cards.json"
|
||||
research_brief = project_root / "phase1" / "research_brief.json"
|
||||
compressed_findings = sorted((project_root / "phase2" / "compressed_findings").glob("ch*.json"))
|
||||
packet_counts = packet_state_counts(project_root)
|
||||
sources = project_root / "phase2" / "sources.jsonl"
|
||||
final_en = project_root / "phase4" / "final_en.md"
|
||||
final_zh = project_root / "phase4" / "final_zh.md"
|
||||
@@ -93,7 +464,8 @@ def cmd_status(args: argparse.Namespace) -> int:
|
||||
glossary = project_root / "phase4" / "glossary.json"
|
||||
|
||||
en_words = count_words(final_en.read_text(encoding="utf-8")) if final_en.exists() else 0
|
||||
zh_chars = count_chinese_chars(final_zh_polished.read_text(encoding="utf-8")) if final_zh_polished.exists() else 0
|
||||
zh_source = final_zh_polished if final_zh_polished.exists() else final_zh
|
||||
zh_chars = count_chinese_chars(zh_source.read_text(encoding="utf-8")) if zh_source.exists() else 0
|
||||
source_count = 0
|
||||
if sources.exists():
|
||||
source_count = sum(1 for line in sources.read_text(encoding="utf-8").splitlines() if line.strip())
|
||||
@@ -110,7 +482,19 @@ def cmd_status(args: argparse.Namespace) -> int:
|
||||
print()
|
||||
print("Artifacts:")
|
||||
print(f" framework: {file_state(project_root / 'phase1' / 'framework.md')}")
|
||||
print(f" research_brief.json: {file_state(research_brief)}")
|
||||
print(f" task_cards.json: {file_state(task_cards)}")
|
||||
print(
|
||||
" packets: "
|
||||
f"ready={packet_counts['ready']} "
|
||||
f"placeholder={packet_counts['placeholder']} "
|
||||
f"invalid={packet_counts['invalid']} "
|
||||
f"errors={packet_counts['errors']} "
|
||||
f"stale_errors={packet_counts['stale_errors']} "
|
||||
f"total={packet_counts['total']}"
|
||||
)
|
||||
print(f" drafts: {len(drafts)}")
|
||||
print(f" compressed findings: {len(compressed_findings)}")
|
||||
print(f" evidence files: {len(evidence)}")
|
||||
print(f" sources: {source_count}")
|
||||
print(f" final_en.md: {file_state(final_en)} ({en_words:,} words)")
|
||||
@@ -124,13 +508,13 @@ def cmd_prompt(args: argparse.Namespace) -> int:
|
||||
name = args.command
|
||||
if not name.startswith("dr-"):
|
||||
name = f"dr-{name}"
|
||||
path = CODEX_COMMANDS_DIR / f"{name}.md"
|
||||
if not path.exists():
|
||||
fallback = CODEX_COMMAND_TEMPLATES_DIR / f"{name}.md"
|
||||
if fallback.exists():
|
||||
path = fallback
|
||||
else:
|
||||
raise SystemExit(f"Codex command template not found: {path}")
|
||||
candidates = [
|
||||
CODEX_COMMAND_TEMPLATES_DIR / f"{name}.md",
|
||||
LEGACY_CODEX_COMMANDS_DIR / f"{name}.md",
|
||||
]
|
||||
path = next((candidate for candidate in candidates if candidate.exists()), None)
|
||||
if path is None:
|
||||
raise SystemExit(f"Codex command template not found: {candidates[0]}")
|
||||
|
||||
text = path.read_text(encoding="utf-8")
|
||||
if args.argument:
|
||||
@@ -174,6 +558,94 @@ def cmd_finalize(args: argparse.Namespace) -> int:
|
||||
raise SystemExit(f"model profile resolution failed: {exc}") from exc
|
||||
roles = resolved["roles"]
|
||||
|
||||
if not args.legacy_translate:
|
||||
final_input = args.input
|
||||
if args.number_citations and not args.dry_run:
|
||||
rc = run_cmd(
|
||||
[
|
||||
sys.executable,
|
||||
str(REPO_ROOT / "scripts" / "number_citations.py"),
|
||||
str(project_root),
|
||||
"--input",
|
||||
args.input,
|
||||
"--output",
|
||||
args.numbered_output,
|
||||
],
|
||||
dry_run=False,
|
||||
)
|
||||
if rc != 0:
|
||||
return rc
|
||||
final_input = args.numbered_output
|
||||
elif args.number_citations:
|
||||
final_input = args.numbered_output
|
||||
cmd: list[str] = [
|
||||
sys.executable,
|
||||
str(REPO_ROOT / "scripts" / "build_report.py"),
|
||||
str(project_root),
|
||||
"--input",
|
||||
final_input,
|
||||
]
|
||||
if args.report_engine:
|
||||
cmd += ["--engine", args.report_engine]
|
||||
if args.no_docx:
|
||||
cmd.append("--no-docx")
|
||||
if args.no_pdf:
|
||||
cmd.append("--no-pdf")
|
||||
if args.dry_run:
|
||||
print("Chinese-native finalize plan:")
|
||||
if args.number_citations:
|
||||
print(
|
||||
"$ "
|
||||
+ " ".join(
|
||||
[
|
||||
sys.executable,
|
||||
str(REPO_ROOT / "scripts" / "number_citations.py"),
|
||||
str(project_root),
|
||||
"--input",
|
||||
args.input,
|
||||
"--output",
|
||||
args.numbered_output,
|
||||
]
|
||||
)
|
||||
)
|
||||
print("$ " + " ".join(cmd))
|
||||
if args.polish:
|
||||
print(
|
||||
"$ "
|
||||
+ " ".join(
|
||||
[
|
||||
sys.executable,
|
||||
str(REPO_ROOT / "scripts" / "polish.py"),
|
||||
str(project_root),
|
||||
"--source",
|
||||
final_input,
|
||||
"--workers",
|
||||
str(args.polish_workers),
|
||||
"--model",
|
||||
roles.get("polish", "anthropic/claude-sonnet-4.6"),
|
||||
]
|
||||
)
|
||||
)
|
||||
return 0
|
||||
if args.polish:
|
||||
rc = run_cmd(
|
||||
[
|
||||
sys.executable,
|
||||
str(REPO_ROOT / "scripts" / "polish.py"),
|
||||
str(project_root),
|
||||
"--source",
|
||||
final_input,
|
||||
"--workers",
|
||||
str(args.polish_workers),
|
||||
"--model",
|
||||
roles.get("polish", "anthropic/claude-sonnet-4.6"),
|
||||
],
|
||||
dry_run=False,
|
||||
)
|
||||
if rc != 0:
|
||||
return rc
|
||||
return run_cmd(cmd, dry_run=False)
|
||||
|
||||
cmd = [
|
||||
sys.executable,
|
||||
str(REPO_ROOT / "scripts" / "phase4_pipeline.py"),
|
||||
@@ -212,6 +684,49 @@ def cmd_models(args: argparse.Namespace) -> int:
|
||||
except ModelConfigError as exc:
|
||||
raise SystemExit(f"model profile resolution failed: {exc}") from exc
|
||||
|
||||
if args.probe:
|
||||
from scripts.lib.zenmux_client import ZenMuxClient, load_secrets, normalize_zenmux_model
|
||||
|
||||
load_secrets()
|
||||
results = []
|
||||
with ZenMuxClient() as client:
|
||||
for requested_model in sorted(set(resolved["roles"].values())):
|
||||
api_model = normalize_zenmux_model(requested_model)
|
||||
try:
|
||||
content = client.chat_complete(
|
||||
model=requested_model,
|
||||
system="Health check.",
|
||||
user="Reply with OK only.",
|
||||
temperature=0,
|
||||
max_tokens=16,
|
||||
tag=f"models:probe:{requested_model}",
|
||||
)
|
||||
results.append({
|
||||
"requested_model": requested_model,
|
||||
"api_model": api_model,
|
||||
"ok": True,
|
||||
"response": content.strip()[:80],
|
||||
})
|
||||
except Exception as exc: # noqa: BLE001 - probe should report every model.
|
||||
results.append({
|
||||
"requested_model": requested_model,
|
||||
"api_model": api_model,
|
||||
"ok": False,
|
||||
"error": str(exc)[:500],
|
||||
})
|
||||
|
||||
if args.json:
|
||||
print(json.dumps({**resolved, "probe": results}, ensure_ascii=False, indent=2))
|
||||
else:
|
||||
print(f"Profile: {resolved['profile']}")
|
||||
print("Model probe:")
|
||||
for item in results:
|
||||
status = "ok" if item["ok"] else "fail"
|
||||
print(f" {status} {item['requested_model']} -> {item['api_model']}")
|
||||
if not item["ok"]:
|
||||
print(f" {item['error']}")
|
||||
return 0 if all(item["ok"] for item in results) else 1
|
||||
|
||||
if args.json:
|
||||
print(json.dumps(resolved, ensure_ascii=False, indent=2))
|
||||
return 0
|
||||
@@ -222,9 +737,37 @@ def cmd_models(args: argparse.Namespace) -> int:
|
||||
print("Roles:")
|
||||
for role in sorted(resolved["roles"]):
|
||||
print(f" {role}: {resolved['roles'][role]}")
|
||||
if resolved.get("task_types"):
|
||||
print("Task types:")
|
||||
for task_type in sorted(resolved["task_types"]):
|
||||
print(f" {task_type}: {resolved['task_types'][task_type]}")
|
||||
return 0
|
||||
|
||||
|
||||
def cmd_sources(args: argparse.Namespace) -> int:
|
||||
project_root = resolve_project(args.project)
|
||||
if args.sources_cmd == "cache":
|
||||
if args.dry_run:
|
||||
print(f"Project: {project_root.name}")
|
||||
print(f"Would cache sources from: {args.sources}")
|
||||
print(f"Important only: {not args.all}")
|
||||
print(f"Limit: {args.limit}")
|
||||
return 0
|
||||
results = cache_sources(
|
||||
project_root,
|
||||
sources_rel=args.sources,
|
||||
important_only=not args.all,
|
||||
limit=args.limit,
|
||||
force=args.force,
|
||||
)
|
||||
print(f"Project: {project_root.name}")
|
||||
print(f"Cached source snapshots: {len(results)}")
|
||||
print("Wrote: phase2/source_cache/md/*.md")
|
||||
print("Updated: phase2/sources.jsonl")
|
||||
return 0
|
||||
raise SystemExit(f"unknown sources command: {args.sources_cmd}")
|
||||
|
||||
|
||||
def cmd_apply_models(args: argparse.Namespace) -> int:
|
||||
cmd = [
|
||||
sys.executable,
|
||||
@@ -245,10 +788,85 @@ def build_parser() -> argparse.ArgumentParser:
|
||||
parser = argparse.ArgumentParser(description="Deep Research platform-neutral CLI")
|
||||
sub = parser.add_subparsers(dest="cmd", required=True)
|
||||
|
||||
init = sub.add_parser("init", help="Initialize a Python-core research project")
|
||||
init.add_argument("topic", help="Research topic")
|
||||
init.add_argument("--slug", help="Project slug")
|
||||
init.add_argument("--method", help="Research method key")
|
||||
init.add_argument("--type", dest="report_type", default="research", help="Report type")
|
||||
init.add_argument("--profile", default="medium", help="Model profile name from configs/models.yaml")
|
||||
init.add_argument("--target-words", type=int, default=30000)
|
||||
init.add_argument("--input-material", action="append", default=[], help="Path or note for user-provided material")
|
||||
init.add_argument("--projects-dir", help="Override projects directory")
|
||||
init.set_defaults(func=cmd_init)
|
||||
|
||||
approve = sub.add_parser("approve", help="Approve Phase 1 gates before Phase 2")
|
||||
approve.add_argument("project", help="Project slug or path")
|
||||
approve.set_defaults(func=cmd_approve)
|
||||
|
||||
frame = sub.add_parser("frame", help="Generate Phase 1 framework.md")
|
||||
frame.add_argument("project", help="Project slug or path")
|
||||
frame.add_argument("--method", help="Override research method key")
|
||||
frame.add_argument("--chapters", type=int, default=10)
|
||||
frame.add_argument("--preserve-existing-outline", action="store_true", help="Keep current framework chapter titles and enrich Phase 1 planning")
|
||||
frame.add_argument("--dry-run", action="store_true")
|
||||
frame.set_defaults(func=cmd_frame)
|
||||
|
||||
run = sub.add_parser("run", help="Run the platform-neutral Python-core workflow")
|
||||
run.add_argument("project_or_topic", help="Project slug/path or new topic")
|
||||
run.add_argument("--workers", type=int, default=6)
|
||||
run.add_argument("--profile", help="Model profile name from configs/models.yaml")
|
||||
run.add_argument("--slug", help="Project slug when project_or_topic is new")
|
||||
run.add_argument("--method", help="Research method key when project_or_topic is new")
|
||||
run.add_argument("--type", dest="report_type", default="research", help="Report type for new project")
|
||||
run.add_argument("--target-words", type=int, default=30000)
|
||||
run.add_argument("--chapters", type=int, default=10)
|
||||
run.add_argument("--input-material", action="append", default=[])
|
||||
run.add_argument("--projects-dir", help="Override projects directory")
|
||||
run.add_argument("--dry-run", action="store_true")
|
||||
run.set_defaults(func=cmd_run)
|
||||
|
||||
research = sub.add_parser("research", help="Run v0.20 task-card Phase 2")
|
||||
research.add_argument("project", help="Project slug or path")
|
||||
research.add_argument("--workers", type=int, default=6)
|
||||
research.add_argument("--axis", action="append", help="Restrict generated task axes; repeatable")
|
||||
research.add_argument("--profile", help="Model profile name from configs/models.yaml")
|
||||
research.add_argument("--model-override", action="append", default=[], metavar="ROLE=MODEL", help="Override a role model for this run; repeatable")
|
||||
research.add_argument("--execute-packets", action="store_true", help="Call model workers to fill evidence packets")
|
||||
research.add_argument("--append-task-cards", action="store_true", help="Append newly generated task cards instead of replacing phase2/task_cards.json")
|
||||
research.add_argument("--allow-search-fallback", action="store_true", help="Allow generic search fallback for specialized routes")
|
||||
research.add_argument("--build-briefs", action="store_true", help="Aggregate packets into chapter briefs")
|
||||
research.add_argument("--assemble-chapters", action="store_true", help="Call model workers to write Chinese chapter drafts")
|
||||
research.add_argument("--force", action="store_true", help="bypass Phase 1 approval gate")
|
||||
research.add_argument("--dry-run", action="store_true")
|
||||
research.set_defaults(func=cmd_research)
|
||||
|
||||
skills = sub.add_parser("skills", help="Manage canonical skills")
|
||||
skill_sub = skills.add_subparsers(dest="skills_cmd", required=True)
|
||||
skill_sub.add_parser("list", help="List canonical skills").set_defaults(func=cmd_skills)
|
||||
skill_sub.add_parser("validate", help="Validate canonical skills").set_defaults(func=cmd_skills)
|
||||
skill_sync = skill_sub.add_parser("sync", help="Sync skills into adapter directories")
|
||||
skill_sync.add_argument("--target", action="append", help="Target skill directory; repeatable")
|
||||
skill_sync.set_defaults(func=cmd_skills)
|
||||
|
||||
methods = sub.add_parser("methods", help="List and inspect research framework methods")
|
||||
method_sub = methods.add_subparsers(dest="methods_cmd", required=True)
|
||||
method_sub.add_parser("list", help="List research methods").set_defaults(func=cmd_methods)
|
||||
method_show = method_sub.add_parser("show", help="Show one research method")
|
||||
method_show.add_argument("method")
|
||||
method_show.set_defaults(func=cmd_methods)
|
||||
|
||||
status = sub.add_parser("status", help="Show project status")
|
||||
status.add_argument("project", nargs="?", help="Project slug or path")
|
||||
status.set_defaults(func=cmd_status)
|
||||
|
||||
review = sub.add_parser("review", help="Run Phase 3 review")
|
||||
review.add_argument("project", help="Project slug or path")
|
||||
review.add_argument("--model-review", action="store_true", help="Run independent model-based Phase 3 review")
|
||||
review.add_argument("--model", default="zenmux-anthropic/claude-opus-4-7", help="Model for --model-review")
|
||||
review.add_argument("--max-context-chars", type=int, default=650_000, help="Bounded context size for model review")
|
||||
review.add_argument("--dry-run", action="store_true")
|
||||
review.set_defaults(func=cmd_review)
|
||||
|
||||
prompt = sub.add_parser("prompt", help="Print a Codex command prompt template")
|
||||
prompt.add_argument("command", help="Command name, e.g. dr-frame or frame")
|
||||
prompt.add_argument("argument", nargs="?", help="Replacement for $ARGUMENTS")
|
||||
@@ -266,6 +884,14 @@ def build_parser() -> argparse.ArgumentParser:
|
||||
|
||||
finalize = sub.add_parser("finalize", help="Run Phase 4 deterministic pipeline")
|
||||
finalize.add_argument("project", help="Project slug or path")
|
||||
finalize.add_argument("--input", default="phase4/final_zh.md", help="Chinese Markdown source for default v0.20 finalization")
|
||||
finalize.add_argument("--legacy-translate", action="store_true", help="Use legacy final_en -> translate -> polish pipeline")
|
||||
finalize.add_argument("--polish", action="store_true", help="Run optional Chinese polish step before rendering")
|
||||
finalize.add_argument("--number-citations", action="store_true", help="Convert [src_xxx] citations to numeric references before rendering")
|
||||
finalize.add_argument("--numbered-output", default="phase4/final_zh_numbered.md", help="Output path for numeric citation Markdown")
|
||||
finalize.add_argument("--report-engine", choices=["reportlab", "quarto"], default=None)
|
||||
finalize.add_argument("--no-docx", action="store_true")
|
||||
finalize.add_argument("--no-pdf", action="store_true")
|
||||
finalize.add_argument("--translate-workers", type=int, default=0)
|
||||
finalize.add_argument("--glossary-workers", type=int, default=4)
|
||||
finalize.add_argument("--polish-workers", type=int, default=0)
|
||||
@@ -295,9 +921,21 @@ def build_parser() -> argparse.ArgumentParser:
|
||||
metavar="ROLE=MODEL",
|
||||
help="Override one role model, repeatable",
|
||||
)
|
||||
models.add_argument("--probe", action="store_true", help="Send tiny health checks to resolved role models")
|
||||
models.add_argument("--json", action="store_true", help="Emit JSON")
|
||||
models.set_defaults(func=cmd_models)
|
||||
|
||||
sources = sub.add_parser("sources", help="Manage source snapshots and source registry")
|
||||
sources_sub = sources.add_subparsers(dest="sources_cmd", required=True)
|
||||
sources_cache = sources_sub.add_parser("cache", help="Cache important sources as local Markdown snapshots")
|
||||
sources_cache.add_argument("project", help="Project slug or path")
|
||||
sources_cache.add_argument("--sources", default="phase2/sources.jsonl", help="Source registry path relative to project")
|
||||
sources_cache.add_argument("--all", action="store_true", help="Cache all remote sources, not only important official/Tier 1 sources")
|
||||
sources_cache.add_argument("--limit", type=int, help="Maximum sources to cache in this run")
|
||||
sources_cache.add_argument("--force", action="store_true", help="Refetch even if cached_text_path already exists")
|
||||
sources_cache.add_argument("--dry-run", action="store_true")
|
||||
sources_cache.set_defaults(func=cmd_sources)
|
||||
|
||||
apply_models = sub.add_parser("apply-models", help="Apply profile to agent files")
|
||||
apply_models.add_argument("--profile", required=True, help="Profile name from configs/models.yaml")
|
||||
apply_models.add_argument("--target", choices=["opencode", "codex", "both"], default="both")
|
||||
|
||||
@@ -1,61 +1,48 @@
|
||||
#!/usr/bin/env python3
|
||||
"""Install the Codex native adapter files into hidden project directories.
|
||||
"""Backward-compatible wrapper for deploying the Codex adapter.
|
||||
|
||||
The Codex desktop sandbox may block agent-created writes into `.codex` and
|
||||
`.agents/skills`. Run this script locally from the repository root when that
|
||||
happens.
|
||||
v0.20 keeps Codex adapter templates in the repository, but deploys the usable
|
||||
adapter files to a Codex home outside the checkout.
|
||||
"""
|
||||
|
||||
from __future__ import annotations
|
||||
|
||||
import argparse
|
||||
import shutil
|
||||
import sys
|
||||
from pathlib import Path
|
||||
|
||||
|
||||
REPO_ROOT = Path(__file__).resolve().parent.parent
|
||||
TEMPLATE_ROOT = REPO_ROOT / "codex_adapter_templates" / "codex"
|
||||
CODEX_ROOT = REPO_ROOT / ".codex"
|
||||
AGENTS_SKILLS = REPO_ROOT / ".agents" / "skills"
|
||||
OPENCODE_SKILLS = REPO_ROOT / ".opencode" / "skills"
|
||||
if str(REPO_ROOT) not in sys.path:
|
||||
sys.path.insert(0, str(REPO_ROOT))
|
||||
|
||||
|
||||
def copy_tree_contents(src: Path, dst: Path, *, force: bool) -> list[Path]:
|
||||
written: list[Path] = []
|
||||
if not src.exists():
|
||||
raise SystemExit(f"template source not found: {src}")
|
||||
dst.mkdir(parents=True, exist_ok=True)
|
||||
for item in src.rglob("*"):
|
||||
rel = item.relative_to(src)
|
||||
target = dst / rel
|
||||
if item.is_dir():
|
||||
target.mkdir(parents=True, exist_ok=True)
|
||||
continue
|
||||
if target.exists() and not force:
|
||||
continue
|
||||
target.parent.mkdir(parents=True, exist_ok=True)
|
||||
shutil.copy2(item, target)
|
||||
written.append(target)
|
||||
return written
|
||||
from scripts.deploy_adapters import deploy_codex, print_result
|
||||
|
||||
|
||||
def main() -> int:
|
||||
parser = argparse.ArgumentParser(description="Install Codex native adapter")
|
||||
parser.add_argument("--force", action="store_true", help="overwrite existing adapter files")
|
||||
parser.add_argument("--skip-skills", action="store_true", help="do not copy .opencode/skills to .agents/skills")
|
||||
parser = argparse.ArgumentParser(description="Deploy Codex native adapter")
|
||||
parser.add_argument("--target", type=Path, help="Codex home target; defaults to $CODEX_HOME or ~/.codex")
|
||||
parser.add_argument("--force", action="store_true", help="overwrite existing files and create .bak backups")
|
||||
parser.add_argument("--skip-skills", action="store_true", help="do not copy canonical skills into target/skills")
|
||||
parser.add_argument("--include-config", action="store_true", help="also copy config.toml; off by default to avoid overwriting global Codex config")
|
||||
parser.add_argument("--dry-run", action="store_true", help="show files that would be written")
|
||||
args = parser.parse_args()
|
||||
|
||||
codex_written = copy_tree_contents(TEMPLATE_ROOT, CODEX_ROOT, force=args.force)
|
||||
skills_written: list[Path] = []
|
||||
if not args.skip_skills:
|
||||
skills_written = copy_tree_contents(OPENCODE_SKILLS, AGENTS_SKILLS, force=args.force)
|
||||
|
||||
print("Codex adapter installed.")
|
||||
print(f" .codex files written: {len(codex_written)}")
|
||||
print(f" .agents skills files written: {len(skills_written)}")
|
||||
if codex_written:
|
||||
for path in codex_written:
|
||||
print(f" {path.relative_to(REPO_ROOT)}")
|
||||
result = deploy_codex(
|
||||
target=args.target,
|
||||
force=args.force,
|
||||
skip_skills=args.skip_skills,
|
||||
include_config=args.include_config,
|
||||
dry_run=args.dry_run,
|
||||
)
|
||||
print_result(result)
|
||||
print()
|
||||
print("Note: this no longer writes repository-local .codex files by default.")
|
||||
print("Run Codex from this repository after deployment:")
|
||||
if args.include_config:
|
||||
print(" codex --profile deep-research")
|
||||
else:
|
||||
print(" codex")
|
||||
print("Note: config.toml is not copied by default. Use --include-config only if you want the bundled profile.")
|
||||
return 0
|
||||
|
||||
|
||||
|
||||
@@ -47,7 +47,9 @@ def resolve_model_profile(
|
||||
if selected not in profiles:
|
||||
raise ModelConfigError(f"unknown model profile: {selected}")
|
||||
|
||||
roles = dict((profiles[selected] or {}).get("roles") or {})
|
||||
selected_profile = profiles[selected] or {}
|
||||
roles = dict(selected_profile.get("roles") or {})
|
||||
task_types = dict(selected_profile.get("task_types") or defaults.get("task_types") or {})
|
||||
if defaults.get("script_models"):
|
||||
for role, model in (defaults.get("script_models") or {}).items():
|
||||
roles.setdefault(role, model)
|
||||
@@ -56,8 +58,9 @@ def resolve_model_profile(
|
||||
|
||||
return {
|
||||
"profile": selected,
|
||||
"description": (profiles[selected] or {}).get("description", ""),
|
||||
"description": selected_profile.get("description", ""),
|
||||
"roles": roles,
|
||||
"task_types": task_types,
|
||||
}
|
||||
|
||||
|
||||
|
||||
+165
-10
@@ -1,11 +1,12 @@
|
||||
"""通用搜索客户端（Serper / Exa / Tavily 路由）。
|
||||
"""通用搜索客户端（Tavily / Exa / Brave / Serper 路由）。
|
||||
|
||||
为 build_glossary.py 这类术语核查场景服务。
|
||||
|
||||
关键设计：
|
||||
- `trust_env=False` 绕开系统 socks 代理（Clash on macOS 配 socks5 时 httpx 会 TLS EOF）
|
||||
- 专利 / Scholar / News 优先 Serper，保证 Google Patents / Google Scholar 路径被真正调用
|
||||
- 通用网页 Exa 优先，Tavily fallback
|
||||
- 通用网页 Tavily 优先，Exa/Brave fallback
|
||||
- 证据发现 Exa 优先，用 highlights/text 摘录喂给 evidence packet
|
||||
- 遇到配额问题自动降级或返回 empty
|
||||
- 不做深度 crawl，只要摘要
|
||||
"""
|
||||
@@ -47,13 +48,29 @@ class ExaClient:
|
||||
def __exit__(self, *_args: Any) -> None:
|
||||
self.close()
|
||||
|
||||
def search(self, query: str, *, num_results: int = 5) -> list[SearchHit]:
|
||||
def search(
|
||||
self,
|
||||
query: str,
|
||||
*,
|
||||
num_results: int = 5,
|
||||
search_type: str = "auto",
|
||||
category: str | None = None,
|
||||
use_highlights: bool = False,
|
||||
max_characters: int = 800,
|
||||
) -> list[SearchHit]:
|
||||
body = {
|
||||
"query": query,
|
||||
"numResults": num_results,
|
||||
"type": "auto",
|
||||
"contents": {"text": {"maxCharacters": 800}},
|
||||
"type": search_type,
|
||||
"contents": {"text": {"maxCharacters": max_characters}},
|
||||
}
|
||||
if category:
|
||||
body["category"] = category
|
||||
if use_highlights:
|
||||
body["contents"]["highlights"] = {
|
||||
"numSentences": 2,
|
||||
"highlightsPerUrl": 3,
|
||||
}
|
||||
r = self._client.post(
|
||||
"https://api.exa.ai/search",
|
||||
json=body,
|
||||
@@ -64,11 +81,15 @@ class ExaClient:
|
||||
data = r.json()
|
||||
out: list[SearchHit] = []
|
||||
for item in data.get("results", [])[:num_results]:
|
||||
highlights = item.get("highlights") or []
|
||||
text = item.get("text") or item.get("snippet") or ""
|
||||
if highlights:
|
||||
text = " | ".join(str(h).strip() for h in highlights if str(h).strip())
|
||||
out.append(
|
||||
SearchHit(
|
||||
title=(item.get("title") or "")[:200],
|
||||
url=item.get("url") or "",
|
||||
snippet=(item.get("text") or item.get("snippet") or "")[:600],
|
||||
snippet=text[:1000],
|
||||
)
|
||||
)
|
||||
return out
|
||||
@@ -115,10 +136,51 @@ class TavilyClient:
|
||||
return out
|
||||
|
||||
|
||||
class BraveClient:
|
||||
def __init__(self, api_key: str | None = None, timeout: float = 30.0) -> None:
|
||||
self.api_key = api_key or os.environ.get("BRAVE_API_KEY")
|
||||
if not self.api_key:
|
||||
raise SearchError("BRAVE_API_KEY not set")
|
||||
self._client = httpx.Client(trust_env=False, timeout=timeout)
|
||||
|
||||
def close(self) -> None:
|
||||
self._client.close()
|
||||
|
||||
def __enter__(self) -> "BraveClient":
|
||||
return self
|
||||
|
||||
def __exit__(self, *_args: Any) -> None:
|
||||
self.close()
|
||||
|
||||
def search(self, query: str, *, num_results: int = 5) -> list[SearchHit]:
|
||||
r = self._client.get(
|
||||
"https://api.search.brave.com/res/v1/web/search",
|
||||
params={"q": query, "count": min(max(num_results, 1), 20)},
|
||||
headers={
|
||||
"X-Subscription-Token": self.api_key,
|
||||
"Accept": "application/json",
|
||||
},
|
||||
)
|
||||
if r.status_code != 200:
|
||||
raise SearchError(f"Brave HTTP {r.status_code}: {r.text[:200]}")
|
||||
data = r.json()
|
||||
out: list[SearchHit] = []
|
||||
for item in (data.get("web") or {}).get("results", [])[:num_results]:
|
||||
out.append(
|
||||
SearchHit(
|
||||
title=(item.get("title") or "")[:200],
|
||||
url=item.get("url") or "",
|
||||
snippet=(item.get("description") or "")[:600],
|
||||
)
|
||||
)
|
||||
return out
|
||||
|
||||
|
||||
class SearchClient:
|
||||
"""统一搜索门面，支持多路由：
|
||||
|
||||
- `search(query)`：通用网页搜索，优先 Exa → 降级 Tavily
|
||||
- `search(query)`：通用网页搜索，优先 Tavily → Exa → Brave
|
||||
- `evidence(query)`：证据发现，优先 Exa highlights → Tavily → Brave
|
||||
- `patents(query)`：专利检索，走 Serper（Google Patents）；失败则通用搜索补刀
|
||||
- `scholar(query)`：学术论文，走 Serper Scholar；失败则通用搜索补刀
|
||||
- `news(query)`：新闻检索，走 Serper News；失败则通用搜索补刀
|
||||
@@ -129,6 +191,7 @@ class SearchClient:
|
||||
def __init__(self, *, strict_specialized: bool = True) -> None:
|
||||
self._exa: ExaClient | None = None
|
||||
self._tavily: TavilyClient | None = None
|
||||
self._brave: BraveClient | None = None
|
||||
self._serper = None # 惰性实例化
|
||||
self.strict_specialized = strict_specialized
|
||||
try:
|
||||
@@ -139,9 +202,13 @@ class SearchClient:
|
||||
self._tavily = TavilyClient()
|
||||
except SearchError:
|
||||
pass
|
||||
try:
|
||||
self._brave = BraveClient()
|
||||
except SearchError:
|
||||
pass
|
||||
self._has_serper_key = bool(os.environ.get("SERPER_API_KEY") or os.environ.get("SERPAPI_KEY"))
|
||||
if not (self._exa or self._tavily or self._has_serper_key):
|
||||
raise SearchError("no search API key available: set SERPER_API_KEY, SERPAPI_KEY, EXA_API_KEY, or TAVILY_API_KEY")
|
||||
if not (self._exa or self._tavily or self._brave or self._has_serper_key):
|
||||
raise SearchError("no search API key available: set SERPER_API_KEY, SERPAPI_KEY, EXA_API_KEY, TAVILY_API_KEY, or BRAVE_API_KEY")
|
||||
|
||||
def _get_serper(self):
|
||||
"""惰性创建 SerperClient。没 key 时返回 None。"""
|
||||
@@ -161,6 +228,8 @@ class SearchClient:
|
||||
self._exa.close()
|
||||
if self._tavily:
|
||||
self._tavily.close()
|
||||
if self._brave:
|
||||
self._brave.close()
|
||||
if self._serper and self._serper is not False:
|
||||
self._serper.close()
|
||||
|
||||
@@ -171,17 +240,60 @@ class SearchClient:
|
||||
self.close()
|
||||
|
||||
def search(self, query: str, *, num_results: int = 5) -> list[SearchHit]:
|
||||
"""通用网页搜索。Exa 首选，Tavily 备选。"""
|
||||
"""通用网页搜索。Tavily 首选，Exa/Brave 备选。"""
|
||||
if self._tavily:
|
||||
try:
|
||||
return self._tavily.search(query, num_results=num_results)
|
||||
except SearchError:
|
||||
pass
|
||||
if self._exa:
|
||||
try:
|
||||
return self._exa.search(query, num_results=num_results)
|
||||
except SearchError:
|
||||
pass
|
||||
if self._brave:
|
||||
try:
|
||||
return self._brave.search(query, num_results=num_results)
|
||||
except SearchError:
|
||||
pass
|
||||
return []
|
||||
|
||||
def evidence(
|
||||
self,
|
||||
query: str,
|
||||
*,
|
||||
num_results: int = 10,
|
||||
category: str | None = None,
|
||||
) -> list[SearchHit]:
|
||||
"""Evidence discovery route.
|
||||
|
||||
Exa is better suited for agent-facing evidence discovery because it can
|
||||
return concise highlights/text per URL. Results are still candidate
|
||||
sources only; downstream packets must score and trace important hits
|
||||
back to original Tier 1-2 sources before making final claims.
|
||||
"""
|
||||
if self._exa:
|
||||
try:
|
||||
return self._exa.search(
|
||||
query,
|
||||
num_results=num_results,
|
||||
search_type="auto",
|
||||
category=category,
|
||||
use_highlights=True,
|
||||
max_characters=1200,
|
||||
)
|
||||
except SearchError:
|
||||
pass
|
||||
if self._tavily:
|
||||
try:
|
||||
return self._tavily.search(query, num_results=num_results)
|
||||
except SearchError:
|
||||
pass
|
||||
if self._brave:
|
||||
try:
|
||||
return self._brave.search(query, num_results=num_results)
|
||||
except SearchError:
|
||||
pass
|
||||
return []
|
||||
|
||||
def patents(self, query: str, *, num_results: int = 10) -> list[SearchHit]:
|
||||
@@ -253,6 +365,49 @@ class SearchClient:
|
||||
raise SearchError("serper unavailable for news route; refusing silent fallback")
|
||||
return self.search(query, num_results=num_results)
|
||||
|
||||
def fda(self, query: str, *, num_results: int = 10) -> list[SearchHit]:
|
||||
"""FDA-focused discovery for warning letters and meeting records.
|
||||
|
||||
FDA enforcement examples are often more useful for GMP remediation than
|
||||
generic web pages, so this route biases discovery toward warning
|
||||
letters, inspection/enforcement pages, and meeting materials/minutes.
|
||||
"""
|
||||
def fda_only(hits: list[SearchHit]) -> list[SearchHit]:
|
||||
return [hit for hit in hits if "fda.gov" in (hit.url or "").lower()]
|
||||
|
||||
focused_queries = [
|
||||
f'site:fda.gov "Warning Letter" GMP pharmaceutical {query}',
|
||||
f'site:fda.gov/inspections-compliance-enforcement-and-criminal-investigations "Warning Letter" {query}',
|
||||
f'site:fda.gov "meeting materials" "pharmaceutical quality" {query}',
|
||||
f'site:fda.gov "meeting minutes" FDA pharmaceutical quality {query}',
|
||||
]
|
||||
hits: list[SearchHit] = []
|
||||
seen: set[str] = set()
|
||||
per_query = max(2, min(num_results, 4))
|
||||
for focused_query in focused_queries:
|
||||
route_hits: list[SearchHit] = []
|
||||
serper = self._get_serper()
|
||||
if serper:
|
||||
try:
|
||||
route_hits = [
|
||||
SearchHit(h.title, h.url, h.snippet)
|
||||
for h in serper.search(focused_query, num_results=per_query)
|
||||
]
|
||||
except Exception as exc:
|
||||
if self.strict_specialized:
|
||||
raise SearchError(f"serper FDA search failed: {exc}") from exc
|
||||
if not route_hits and not self.strict_specialized:
|
||||
route_hits = self.search(focused_query, num_results=per_query)
|
||||
for hit in fda_only(route_hits):
|
||||
key = hit.url or hit.title
|
||||
if not key or key in seen:
|
||||
continue
|
||||
seen.add(key)
|
||||
hits.append(hit)
|
||||
if len(hits) >= num_results:
|
||||
return hits
|
||||
return hits
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
from scripts.lib.zenmux_client import load_secrets
|
||||
|
||||
@@ -29,6 +29,35 @@ MAX_RETRIES = 5
|
||||
RETRYABLE_STATUSES = {408, 429, 500, 502, 503, 504, 520, 524}
|
||||
|
||||
|
||||
_ANTHROPIC_MODEL_ALIASES = {
|
||||
"anthropic/claude-opus-4-7": "anthropic/claude-opus-4.7",
|
||||
"anthropic/claude-opus-4-6": "anthropic/claude-opus-4.6",
|
||||
"anthropic/claude-opus-4-5": "anthropic/claude-opus-4.5",
|
||||
"anthropic/claude-opus-4-1": "anthropic/claude-opus-4.1",
|
||||
"anthropic/claude-sonnet-4-6": "anthropic/claude-sonnet-4.6",
|
||||
"anthropic/claude-sonnet-4-5": "anthropic/claude-sonnet-4.5",
|
||||
"anthropic/claude-haiku-4-5": "anthropic/claude-haiku-4.5",
|
||||
}
|
||||
_MODELS_WITHOUT_TEMPERATURE = {
|
||||
"anthropic/claude-opus-4.7",
|
||||
}
|
||||
|
||||
|
||||
def normalize_zenmux_model(model: str) -> str:
|
||||
"""Convert adapter-facing model IDs to ZenMux OpenAI API model IDs."""
|
||||
normalized = model.strip()
|
||||
if normalized.startswith("zenmux-anthropic/"):
|
||||
normalized = "anthropic/" + normalized.removeprefix("zenmux-anthropic/")
|
||||
elif normalized.startswith("zenmux/"):
|
||||
normalized = normalized.removeprefix("zenmux/")
|
||||
return _ANTHROPIC_MODEL_ALIASES.get(normalized, normalized)
|
||||
|
||||
|
||||
def model_accepts_temperature(model: str) -> bool:
|
||||
"""Return whether the ZenMux API accepts `temperature` for this model."""
|
||||
return normalize_zenmux_model(model) not in _MODELS_WITHOUT_TEMPERATURE
|
||||
|
||||
|
||||
@dataclass
|
||||
class UsageStats:
|
||||
"""聚合一次脚本运行的 token 消耗。"""
|
||||
@@ -163,12 +192,14 @@ class ZenMuxClient:
|
||||
messages.extend(extra_messages)
|
||||
messages.append({"role": "user", "content": user})
|
||||
|
||||
api_model = normalize_zenmux_model(model)
|
||||
body: dict[str, Any] = {
|
||||
"model": model,
|
||||
"model": api_model,
|
||||
"messages": messages,
|
||||
"temperature": temperature,
|
||||
"max_tokens": max_tokens,
|
||||
}
|
||||
if model_accepts_temperature(api_model):
|
||||
body["temperature"] = temperature
|
||||
if web_search:
|
||||
body["web_search_options"] = web_search_options or {}
|
||||
headers = {
|
||||
@@ -200,14 +231,14 @@ class ZenMuxClient:
|
||||
raise ZenMuxError(f"invalid JSON from zenmux: {e}; body={resp.text[:500]}")
|
||||
usage = data.get("usage", {}) or {}
|
||||
with self._usage_lock:
|
||||
self.usage.add(model, usage)
|
||||
self.usage.add(api_model, usage)
|
||||
content = ""
|
||||
choices = data.get("choices") or []
|
||||
if choices:
|
||||
msg = choices[0].get("message") or {}
|
||||
content = msg.get("content") or ""
|
||||
self._log({
|
||||
"tag": tag, "model": model, "attempt": attempt,
|
||||
"tag": tag, "model": api_model, "requested_model": model, "attempt": attempt,
|
||||
"elapsed": round(elapsed, 2),
|
||||
"usage": usage,
|
||||
"out_chars": len(content),
|
||||
@@ -256,12 +287,14 @@ class ZenMuxClient:
|
||||
messages.extend(extra_messages)
|
||||
messages.append({"role": "user", "content": user})
|
||||
|
||||
api_model = normalize_zenmux_model(model)
|
||||
body: dict[str, Any] = {
|
||||
"model": model,
|
||||
"model": api_model,
|
||||
"messages": messages,
|
||||
"temperature": temperature,
|
||||
"max_tokens": max_tokens,
|
||||
}
|
||||
if model_accepts_temperature(api_model):
|
||||
body["temperature"] = temperature
|
||||
if web_search:
|
||||
body["web_search_options"] = web_search_options or {}
|
||||
|
||||
@@ -309,7 +342,7 @@ class ZenMuxClient:
|
||||
|
||||
usage = data.get("usage", {}) or {}
|
||||
with self._usage_lock:
|
||||
self.usage.add(model, usage)
|
||||
self.usage.add(api_model, usage)
|
||||
|
||||
message = ((data.get("choices") or [{}])[0].get("message") or {})
|
||||
content = message.get("content") or ""
|
||||
@@ -324,7 +357,8 @@ class ZenMuxClient:
|
||||
urls.append(url_item)
|
||||
self._log({
|
||||
"tag": tag,
|
||||
"model": model,
|
||||
"model": api_model,
|
||||
"requested_model": model,
|
||||
"attempt": attempt,
|
||||
"elapsed": round(elapsed, 2),
|
||||
"usage": usage,
|
||||
|
||||
@@ -0,0 +1,188 @@
|
||||
#!/usr/bin/env python3
|
||||
"""Convert Deep Research source IDs into numeric citations for final output."""
|
||||
|
||||
from __future__ import annotations
|
||||
|
||||
import argparse
|
||||
import json
|
||||
import re
|
||||
from pathlib import Path
|
||||
from typing import Any
|
||||
|
||||
|
||||
SRC_CITE_RE = re.compile(r"\[((?:src_[A-Za-z0-9_-]+)(?:\s*,\s*src_[A-Za-z0-9_-]+)*)\]")
|
||||
|
||||
|
||||
def load_sources(path: Path) -> dict[str, dict[str, Any]]:
|
||||
sources: dict[str, dict[str, Any]] = {}
|
||||
if not path.exists():
|
||||
return sources
|
||||
for line in path.read_text(encoding="utf-8").splitlines():
|
||||
if not line.strip():
|
||||
continue
|
||||
try:
|
||||
obj = json.loads(line)
|
||||
except json.JSONDecodeError:
|
||||
continue
|
||||
sid = obj.get("id") or obj.get("source_id")
|
||||
if sid:
|
||||
sources[str(sid)] = obj
|
||||
return sources
|
||||
|
||||
|
||||
def extract_ordered_source_ids(text: str) -> list[str]:
|
||||
ordered: list[str] = []
|
||||
seen: set[str] = set()
|
||||
for match in SRC_CITE_RE.finditer(text):
|
||||
for sid in [item.strip() for item in match.group(1).split(",")]:
|
||||
if sid and sid not in seen:
|
||||
seen.add(sid)
|
||||
ordered.append(sid)
|
||||
return ordered
|
||||
|
||||
|
||||
def _canonical_source_key(sid: str, source: dict[str, Any] | None) -> str:
|
||||
"""Return a stable de-duplication key for a source record.
|
||||
|
||||
Phase 2 often creates chapter-local source IDs for the same local PDF or
|
||||
official guideline. Final references should cite the underlying source
|
||||
once, while citation_map.json keeps the full src_id traceability.
|
||||
"""
|
||||
if not source:
|
||||
return f"missing:{sid}"
|
||||
title = re.sub(r"\s+", " ", str(source.get("title") or source.get("name") or sid)).strip().lower()
|
||||
title = title.removesuffix(" ocr").removesuffix(".ocr").strip()
|
||||
doi = str(source.get("doi") or "").strip().lower()
|
||||
if doi:
|
||||
return f"doi:{doi}"
|
||||
path = str(source.get("path") or "").strip()
|
||||
url = str(source.get("url") or "").strip()
|
||||
if title and ("phase0/extracted/" in path or "phase0/extracted/" in url):
|
||||
return f"local-material:{title}"
|
||||
for field in ("url", "path"):
|
||||
value = str(source.get(field) or "").strip()
|
||||
if value:
|
||||
return f"{field}:{value.rstrip('/').lower()}"
|
||||
return f"title:{title or sid}"
|
||||
|
||||
|
||||
def build_numeric_mapping(
|
||||
ordered_ids: list[str],
|
||||
sources: dict[str, dict[str, Any]],
|
||||
) -> tuple[dict[str, int], list[dict[str, Any]]]:
|
||||
mapping: dict[str, int] = {}
|
||||
records: list[dict[str, Any]] = []
|
||||
seen_keys: dict[str, int] = {}
|
||||
record_by_number: dict[int, dict[str, Any]] = {}
|
||||
for sid in ordered_ids:
|
||||
source = sources.get(sid)
|
||||
key = _canonical_source_key(sid, source)
|
||||
if key in seen_keys:
|
||||
number = seen_keys[key]
|
||||
mapping[sid] = number
|
||||
record_by_number[number].setdefault("source_ids", []).append(sid)
|
||||
continue
|
||||
number = len(records) + 1
|
||||
seen_keys[key] = number
|
||||
mapping[sid] = number
|
||||
record = {
|
||||
"number": number,
|
||||
"source_id": sid,
|
||||
"source_ids": [sid],
|
||||
"source": source or {},
|
||||
"dedupe_key": key,
|
||||
}
|
||||
records.append(record)
|
||||
record_by_number[number] = record
|
||||
return mapping, records
|
||||
|
||||
|
||||
def format_reference(number: int, sid: str, source: dict[str, Any] | None) -> str:
|
||||
if not source:
|
||||
return f"{number}. {sid}. （sources.jsonl 未找到该来源）"
|
||||
authors = ", ".join(source.get("authors", [])) if source.get("authors") else ""
|
||||
year = source.get("year") or source.get("date") or ""
|
||||
title = source.get("title") or source.get("name") or sid
|
||||
title = re.sub(r"(?i)(?:\s+OCR|\.ocr)$", "", str(title)).strip()
|
||||
publisher = source.get("publisher") or source.get("venue") or source.get("source") or ""
|
||||
url = source.get("url") or source.get("path") or ""
|
||||
parts = [f"{number}. "]
|
||||
if authors:
|
||||
parts.append(f"{authors}. ")
|
||||
if year:
|
||||
parts.append(f"({year}). ")
|
||||
parts.append(str(title))
|
||||
if publisher:
|
||||
parts.append(f". {publisher}")
|
||||
if url:
|
||||
parts.append(f". {url}")
|
||||
return "".join(parts)
|
||||
|
||||
|
||||
def convert_citations(text: str, mapping: dict[str, int]) -> str:
|
||||
def repl(match: re.Match[str]) -> str:
|
||||
ids = [item.strip() for item in match.group(1).split(",") if item.strip()]
|
||||
nums: list[str] = []
|
||||
seen: set[int] = set()
|
||||
for sid in ids:
|
||||
if sid not in mapping:
|
||||
continue
|
||||
number = mapping[sid]
|
||||
if number in seen:
|
||||
continue
|
||||
seen.add(number)
|
||||
nums.append(str(number))
|
||||
return "<sup>[" + ", ".join(nums) + "]</sup>" if nums else match.group(0)
|
||||
|
||||
return SRC_CITE_RE.sub(repl, text)
|
||||
|
||||
|
||||
def strip_existing_reference_section(text: str) -> str:
|
||||
pattern = re.compile(r"\n##\s*(?:参考文献|参考来源清单|References)\s*\n.*\Z", re.S)
|
||||
return pattern.sub("", text).rstrip() + "\n"
|
||||
|
||||
|
||||
def number_citations(
|
||||
*,
|
||||
text: str,
|
||||
sources: dict[str, dict[str, Any]],
|
||||
) -> tuple[str, list[dict[str, Any]]]:
|
||||
ordered_ids = extract_ordered_source_ids(text)
|
||||
mapping, records = build_numeric_mapping(ordered_ids, sources)
|
||||
body = convert_citations(strip_existing_reference_section(text), mapping).rstrip()
|
||||
ref_lines = ["", "## 参考来源清单", ""]
|
||||
for record in records:
|
||||
ref_lines.append(format_reference(record["number"], record["source_id"], record["source"]))
|
||||
return body + "\n" + "\n".join(ref_lines).rstrip() + "\n", records
|
||||
|
||||
|
||||
def main() -> int:
|
||||
parser = argparse.ArgumentParser(description="Convert [src_xxx] citations to numeric citations")
|
||||
parser.add_argument("project", help="Project directory")
|
||||
parser.add_argument("--input", default="phase4/final_zh.md")
|
||||
parser.add_argument("--output", default="phase4/final_zh_numbered.md")
|
||||
parser.add_argument("--sources", default="phase2/sources.jsonl")
|
||||
parser.add_argument("--map", default="phase4/citation_map.json")
|
||||
args = parser.parse_args()
|
||||
|
||||
project = Path(args.project)
|
||||
src_path = project / args.input
|
||||
out_path = project / args.output
|
||||
sources_path = project / args.sources
|
||||
map_path = project / args.map
|
||||
if not src_path.exists():
|
||||
raise SystemExit(f"input not found: {src_path}")
|
||||
sources = load_sources(sources_path)
|
||||
numbered, records = number_citations(text=src_path.read_text(encoding="utf-8"), sources=sources)
|
||||
out_path.parent.mkdir(parents=True, exist_ok=True)
|
||||
out_path.write_text(numbered, encoding="utf-8")
|
||||
map_path.parent.mkdir(parents=True, exist_ok=True)
|
||||
map_path.write_text(json.dumps(records, ensure_ascii=False, indent=2) + "\n", encoding="utf-8")
|
||||
print(f"Wrote: {out_path.relative_to(project)}")
|
||||
print(f"Wrote: {map_path.relative_to(project)}")
|
||||
print(f"Citations: {len(records)}")
|
||||
return 0
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
raise SystemExit(main())
|
||||
+12
-1
@@ -33,6 +33,7 @@ from scripts.lib.markdown_chunker import (
|
||||
split_by_headers,
|
||||
)
|
||||
from scripts.lib.zenmux_client import ZenMuxClient, ZenMuxError, load_secrets
|
||||
from scripts.runtime.skills import SkillRegistry
|
||||
|
||||
DEFAULT_MODEL = "anthropic/claude-sonnet-4.6"
|
||||
MODEL_MAX_TOKENS = {
|
||||
@@ -43,6 +44,16 @@ MODEL_MAX_TOKENS = {
|
||||
"anthropic/claude-haiku-4.5": 16000,
|
||||
}
|
||||
PROMPT_FILE = Path(__file__).parent / "prompts" / "polish_system.txt"
|
||||
POLISH_SKILLS = ("humanizer-cn", "output-hygiene")
|
||||
|
||||
|
||||
def build_polish_system_prompt(skill_registry: SkillRegistry | None = None) -> str:
|
||||
"""Build the Phase 4 polish prompt with canonical writing skills attached."""
|
||||
registry = skill_registry or SkillRegistry()
|
||||
parts = [PROMPT_FILE.read_text(encoding="utf-8").rstrip()]
|
||||
for skill_name in POLISH_SKILLS:
|
||||
parts.append(f"# Skill: {skill_name}\n\n{registry.read(skill_name).rstrip()}")
|
||||
return "\n\n".join(parts) + "\n"
|
||||
|
||||
|
||||
def resolve_project(arg: str) -> Path:
|
||||
@@ -163,7 +174,7 @@ def main() -> int:
|
||||
log_file = logs_dir / "polish.jsonl"
|
||||
notes_file = project_root / "phase4" / "polish_notes.jsonl"
|
||||
|
||||
system_prompt = PROMPT_FILE.read_text(encoding="utf-8")
|
||||
system_prompt = build_polish_system_prompt()
|
||||
text = src_path.read_text(encoding="utf-8")
|
||||
blocks = split_by_headers(text, max_level=2)
|
||||
|
||||
|
||||
@@ -0,0 +1,2 @@
|
||||
"""Report rendering helpers for the v0.20 Python core."""
|
||||
|
||||
@@ -0,0 +1,32 @@
|
||||
"""Font resolution helpers for PDF rendering."""
|
||||
|
||||
from __future__ import annotations
|
||||
|
||||
from dataclasses import dataclass
|
||||
from pathlib import Path
|
||||
|
||||
|
||||
@dataclass(frozen=True)
|
||||
class QuartoFonts:
|
||||
main_font: str
|
||||
sans_font: str
|
||||
requires_system_fonts: bool
|
||||
|
||||
|
||||
def resolve_quarto_fonts(fonts_dir: Path) -> QuartoFonts:
|
||||
"""Resolve Quarto font names.
|
||||
|
||||
Quarto/xelatex currently uses installed font family names. We still accept
|
||||
fonts_dir so callers can validate/report environment state consistently.
|
||||
"""
|
||||
expected = [
|
||||
fonts_dir / "SourceHanSerifSC-Regular.otf",
|
||||
fonts_dir / "SourceHanSansSC-Bold.otf",
|
||||
]
|
||||
requires_system_fonts = not all(path.exists() for path in expected)
|
||||
return QuartoFonts(
|
||||
main_font="Source Han Serif CN",
|
||||
sans_font="Source Han Sans CN",
|
||||
requires_system_fonts=requires_system_fonts,
|
||||
)
|
||||
|
||||
@@ -0,0 +1,61 @@
|
||||
"""Reference-list generation from Deep Research sources.jsonl."""
|
||||
|
||||
from __future__ import annotations
|
||||
|
||||
import json
|
||||
import re
|
||||
from pathlib import Path
|
||||
|
||||
|
||||
def cited_source_keys(md_text: str) -> set[str]:
|
||||
return set(re.findall(r"\[src_([a-zA-Z0-9_-]+)\]", md_text))
|
||||
|
||||
|
||||
def build_references_block(sources_path: Path | None, md_text: str) -> str:
|
||||
"""Build a compact references section for actually cited src IDs."""
|
||||
if not sources_path or not sources_path.exists():
|
||||
return "（参考文献列表：sources.jsonl 未找到）"
|
||||
|
||||
cited = cited_source_keys(md_text)
|
||||
if not cited:
|
||||
return ""
|
||||
|
||||
sources: dict[str, dict] = {}
|
||||
with sources_path.open(encoding="utf-8") as f:
|
||||
for line in f:
|
||||
line = line.strip()
|
||||
if not line:
|
||||
continue
|
||||
try:
|
||||
obj = json.loads(line)
|
||||
except json.JSONDecodeError:
|
||||
continue
|
||||
sid = obj.get("id", "")
|
||||
key = sid.replace("src_", "")
|
||||
if key in cited:
|
||||
sources[sid] = obj
|
||||
|
||||
if not sources:
|
||||
return ""
|
||||
|
||||
lines = ["## 参考文献\n"]
|
||||
for sid in sorted(sources.keys()):
|
||||
s = sources[sid]
|
||||
authors = ", ".join(s.get("authors", [])) if s.get("authors") else ""
|
||||
year = s.get("year", "")
|
||||
title = s.get("title", sid)
|
||||
venue = s.get("venue", "")
|
||||
url = s.get("url", "")
|
||||
entry = f"- **[{sid}]** "
|
||||
if authors:
|
||||
entry += f"{authors}. "
|
||||
if year:
|
||||
entry += f"({year}). "
|
||||
entry += f"*{title}*"
|
||||
if venue:
|
||||
entry += f". {venue}"
|
||||
if url:
|
||||
entry += f". <{url}>"
|
||||
lines.append(entry)
|
||||
return "\n".join(lines)
|
||||
|
||||
@@ -0,0 +1,6 @@
|
||||
"""v0.20 Python runtime core for Deep Research.
|
||||
|
||||
The runtime layer is intentionally platform-neutral: OpenCode, Codex, and
|
||||
Claude Code should call into these modules instead of owning orchestration.
|
||||
"""
|
||||
|
||||
@@ -0,0 +1,46 @@
|
||||
"""Project artifact helpers shared by the Python runtime."""
|
||||
|
||||
from __future__ import annotations
|
||||
|
||||
import json
|
||||
from pathlib import Path
|
||||
from typing import Any
|
||||
|
||||
|
||||
REPO_ROOT = Path(__file__).resolve().parents[2]
|
||||
PROJECTS_DIR = REPO_ROOT / "projects"
|
||||
|
||||
|
||||
def resolve_project(project: str | Path) -> Path:
|
||||
p = Path(project)
|
||||
if p.is_dir():
|
||||
return p.resolve()
|
||||
candidate = PROJECTS_DIR / str(project)
|
||||
if candidate.is_dir():
|
||||
return candidate.resolve()
|
||||
raise FileNotFoundError(f"project not found: {project}")
|
||||
|
||||
|
||||
def load_manifest(project_root: Path) -> dict[str, Any]:
|
||||
path = project_root / "manifest.json"
|
||||
if not path.exists():
|
||||
raise FileNotFoundError(f"manifest not found: {path}")
|
||||
return json.loads(path.read_text(encoding="utf-8"))
|
||||
|
||||
|
||||
def write_manifest(project_root: Path, manifest: dict[str, Any]) -> None:
|
||||
path = project_root / "manifest.json"
|
||||
path.write_text(json.dumps(manifest, ensure_ascii=False, indent=2) + "\n", encoding="utf-8")
|
||||
|
||||
|
||||
def ensure_phase_dirs(project_root: Path) -> None:
|
||||
for rel in (
|
||||
"phase1",
|
||||
"phase2/drafts",
|
||||
"phase2/evidence",
|
||||
"phase2/packets",
|
||||
"phase3",
|
||||
"phase4",
|
||||
):
|
||||
(project_root / rel).mkdir(parents=True, exist_ok=True)
|
||||
|
||||
@@ -0,0 +1,360 @@
|
||||
"""Chapter brief aggregation and Chinese chapter assembly."""
|
||||
|
||||
from __future__ import annotations
|
||||
|
||||
import json
|
||||
import re
|
||||
from concurrent.futures import ThreadPoolExecutor, as_completed
|
||||
from pathlib import Path
|
||||
from typing import Callable
|
||||
|
||||
from scripts.runtime.roles import RoleDefinition, RuntimeProfile
|
||||
from scripts.runtime.skills import SkillRegistry
|
||||
from scripts.runtime.tasks import load_task_cards, validate_packet
|
||||
from scripts.runtime.workers import ChatClient
|
||||
|
||||
|
||||
def validate_chapter_brief(brief: dict) -> None:
|
||||
required = {
|
||||
"chapter_id",
|
||||
"chapter_title",
|
||||
"packet_ids",
|
||||
"core_claims",
|
||||
"evidence_items",
|
||||
"counter_evidence",
|
||||
"source_ids",
|
||||
"open_questions",
|
||||
"assembly_notes",
|
||||
}
|
||||
missing = sorted(required - set(brief))
|
||||
if missing:
|
||||
raise ValueError(f"chapter brief missing fields: {missing}")
|
||||
if not brief["chapter_id"]:
|
||||
raise ValueError("chapter_id required")
|
||||
if not brief["packet_ids"]:
|
||||
raise ValueError("chapter brief requires at least one packet")
|
||||
if not brief["core_claims"]:
|
||||
raise ValueError("chapter brief requires core_claims")
|
||||
if not brief["evidence_items"]:
|
||||
raise ValueError("chapter brief requires evidence_items")
|
||||
if not brief["counter_evidence"]:
|
||||
raise ValueError("chapter brief requires counter_evidence")
|
||||
|
||||
|
||||
def validate_compressed_finding(finding: dict) -> None:
|
||||
required = {
|
||||
"chapter_id",
|
||||
"chapter_title",
|
||||
"packet_ids",
|
||||
"chapter_thesis",
|
||||
"key_findings",
|
||||
"evidence_landings",
|
||||
"counter_evidence",
|
||||
"source_ids",
|
||||
"open_questions",
|
||||
"writing_plan",
|
||||
}
|
||||
missing = sorted(required - set(finding))
|
||||
if missing:
|
||||
raise ValueError(f"compressed finding missing fields: {missing}")
|
||||
if not finding["chapter_id"]:
|
||||
raise ValueError("chapter_id required")
|
||||
if not finding["packet_ids"]:
|
||||
raise ValueError("compressed finding requires packet_ids")
|
||||
if not finding["chapter_thesis"]:
|
||||
raise ValueError("compressed finding requires chapter_thesis")
|
||||
if not finding["key_findings"]:
|
||||
raise ValueError("compressed finding requires key_findings")
|
||||
if not finding["evidence_landings"]:
|
||||
raise ValueError("compressed finding requires evidence_landings")
|
||||
if not finding["counter_evidence"]:
|
||||
raise ValueError("compressed finding requires counter_evidence")
|
||||
|
||||
|
||||
def validate_chapter_markdown_citations(markdown: str, brief: dict) -> None:
|
||||
if "key_findings" in brief:
|
||||
validate_compressed_finding(brief)
|
||||
else:
|
||||
validate_chapter_brief(brief)
|
||||
cited = set(re.findall(r"\[(src_[A-Za-z0-9_-]+)\]", markdown))
|
||||
allowed = set(brief.get("source_ids") or [])
|
||||
unknown = sorted(cited - allowed)
|
||||
if unknown:
|
||||
raise ValueError(f"unknown citation ids in {brief['chapter_id']}: {unknown}")
|
||||
|
||||
|
||||
def _chapter_title_from_id(chapter_id: str) -> str:
|
||||
try:
|
||||
index = int(chapter_id.replace("ch", ""))
|
||||
return f"第{index}章"
|
||||
except ValueError:
|
||||
return chapter_id
|
||||
|
||||
|
||||
def _load_source_registry(sources_path: Path, source_ids: list[str]) -> list[dict]:
|
||||
wanted = set(source_ids)
|
||||
if not sources_path.exists() or not wanted:
|
||||
return []
|
||||
rows: list[dict] = []
|
||||
for line in sources_path.read_text(encoding="utf-8").splitlines():
|
||||
if not line.strip():
|
||||
continue
|
||||
try:
|
||||
row = json.loads(line)
|
||||
except json.JSONDecodeError:
|
||||
continue
|
||||
if row.get("id") in wanted:
|
||||
rows.append(row)
|
||||
return rows
|
||||
|
||||
|
||||
def _cached_source_excerpts(project_root: Path, cached_paths: list[str], *, max_sources: int = 5, max_chars: int = 1400) -> list[dict]:
|
||||
excerpts: list[dict] = []
|
||||
for rel in cached_paths[:max_sources]:
|
||||
path = project_root / rel
|
||||
if not path.exists():
|
||||
continue
|
||||
text = path.read_text(encoding="utf-8", errors="ignore").strip()
|
||||
excerpts.append({"path": rel, "excerpt": text[:max_chars]})
|
||||
return excerpts
|
||||
|
||||
|
||||
def build_chapter_briefs(project_root: Path) -> list[dict]:
|
||||
cards = load_task_cards(project_root / "phase2" / "task_cards.json")
|
||||
grouped: dict[str, list[tuple[str, dict]]] = {}
|
||||
skipped_packets: list[dict[str, str]] = []
|
||||
for card in cards:
|
||||
packet_path = project_root / card.output_packet
|
||||
if not packet_path.exists():
|
||||
skipped_packets.append({"task_id": card.task_id, "reason": "packet file missing"})
|
||||
continue
|
||||
packet = json.loads(packet_path.read_text(encoding="utf-8"))
|
||||
try:
|
||||
validate_packet(packet)
|
||||
except Exception as exc:
|
||||
skipped_packets.append({"task_id": card.task_id, "reason": str(exc)})
|
||||
continue
|
||||
for chapter_id in card.chapter_ids:
|
||||
grouped.setdefault(chapter_id, []).append((card.task_id, packet))
|
||||
|
||||
briefs: list[dict] = []
|
||||
out_dir = project_root / "phase2" / "chapter_briefs"
|
||||
out_dir.mkdir(parents=True, exist_ok=True)
|
||||
if skipped_packets:
|
||||
(project_root / "phase2" / "brief_warnings.json").write_text(
|
||||
json.dumps(skipped_packets, ensure_ascii=False, indent=2) + "\n",
|
||||
encoding="utf-8",
|
||||
)
|
||||
for chapter_id in sorted(grouped):
|
||||
packet_pairs = grouped[chapter_id]
|
||||
packet_ids = [item[0] for item in packet_pairs]
|
||||
packets = [item[1] for item in packet_pairs]
|
||||
source_ids = sorted({sid for packet in packets for sid in packet.get("source_ids", [])})
|
||||
source_registry = _load_source_registry(project_root / "phase2" / "sources.jsonl", source_ids)
|
||||
cached_paths = [
|
||||
source["cached_text_path"]
|
||||
for source in source_registry
|
||||
if source.get("cached_text_path")
|
||||
]
|
||||
chapter_title = next((card.chapter_title for card in cards if chapter_id in card.chapter_ids and card.chapter_title), None)
|
||||
brief = {
|
||||
"chapter_id": chapter_id,
|
||||
"chapter_title": chapter_title or _chapter_title_from_id(chapter_id),
|
||||
"packet_ids": packet_ids,
|
||||
"core_claims": [claim for packet in packets for claim in packet.get("claims", [])],
|
||||
"evidence_items": [item for packet in packets for item in packet.get("evidence_items", [])],
|
||||
"counter_evidence": [item for packet in packets for item in packet.get("counter_evidence", [])],
|
||||
"source_ids": source_ids,
|
||||
"cached_source_paths": cached_paths,
|
||||
"cached_source_excerpts": _cached_source_excerpts(project_root, cached_paths),
|
||||
"open_questions": [q for packet in packets for q in packet.get("open_questions", [])],
|
||||
"assembly_notes": [
|
||||
"用中文写正式章节，英文仅保留在必要的来源标题、原文摘录、DOI/URL 中。",
|
||||
"避免碎片化：不要按 packet 逐段堆砌，要先提炼本章主线，再组织证据。",
|
||||
"每个事实、数字和关键判断都必须保留 [src_xxx] 引用。",
|
||||
"必须纳入 counter_evidence，并说明它如何影响结论置信度。",
|
||||
],
|
||||
}
|
||||
validate_chapter_brief(brief)
|
||||
(out_dir / f"{chapter_id}.json").write_text(
|
||||
json.dumps(brief, ensure_ascii=False, indent=2) + "\n",
|
||||
encoding="utf-8",
|
||||
)
|
||||
briefs.append(brief)
|
||||
return briefs
|
||||
|
||||
|
||||
def _source_ids_from_item(item: dict) -> list[str]:
|
||||
if item.get("source_ids"):
|
||||
return list(item.get("source_ids") or [])
|
||||
if item.get("source_id"):
|
||||
return [item["source_id"]]
|
||||
return []
|
||||
|
||||
|
||||
def build_compressed_findings(project_root: Path) -> list[dict]:
|
||||
"""Compress packet-level evidence into chapter-level writing inputs.
|
||||
|
||||
This is intentionally deterministic: it does not invent a better narrative,
|
||||
but it forces a chapter-level evidence map before any model writes prose.
|
||||
"""
|
||||
brief_dir = project_root / "phase2" / "chapter_briefs"
|
||||
if not brief_dir.exists() or not list(brief_dir.glob("ch*.json")):
|
||||
briefs = build_chapter_briefs(project_root)
|
||||
else:
|
||||
briefs = [
|
||||
json.loads(path.read_text(encoding="utf-8"))
|
||||
for path in sorted(brief_dir.glob("ch*.json"))
|
||||
]
|
||||
out_dir = project_root / "phase2" / "compressed_findings"
|
||||
out_dir.mkdir(parents=True, exist_ok=True)
|
||||
findings: list[dict] = []
|
||||
for brief in briefs:
|
||||
validate_chapter_brief(brief)
|
||||
core_claims = brief.get("core_claims") or []
|
||||
evidence_items = brief.get("evidence_items") or []
|
||||
first_claim = core_claims[0] if core_claims else {}
|
||||
chapter_thesis = first_claim.get("claim") or f"{brief['chapter_title']} 需要以证据为中心重写。"
|
||||
finding = {
|
||||
"chapter_id": brief["chapter_id"],
|
||||
"chapter_title": brief["chapter_title"],
|
||||
"packet_ids": brief["packet_ids"],
|
||||
"chapter_thesis": chapter_thesis,
|
||||
"key_findings": [
|
||||
{
|
||||
"finding": claim.get("claim") or claim.get("summary") or str(claim),
|
||||
"source_ids": _source_ids_from_item(claim),
|
||||
"confidence": claim.get("confidence", "medium"),
|
||||
}
|
||||
for claim in core_claims
|
||||
],
|
||||
"evidence_landings": [
|
||||
{
|
||||
"evidence": item.get("summary") or item.get("finding") or item.get("quote") or str(item),
|
||||
"source_ids": _source_ids_from_item(item),
|
||||
"landing_hint": item.get("landing_hint", "用于支撑本章关键判断或整改动作。"),
|
||||
}
|
||||
for item in evidence_items
|
||||
],
|
||||
"counter_evidence": brief["counter_evidence"],
|
||||
"source_ids": brief["source_ids"],
|
||||
"cached_source_paths": brief.get("cached_source_paths", []),
|
||||
"cached_source_excerpts": brief.get("cached_source_excerpts", []),
|
||||
"open_questions": brief["open_questions"],
|
||||
"writing_plan": [
|
||||
"先写本章判断，不按 packet 顺序堆砌。",
|
||||
"每个二级小节至少落下具体审计发现、法规要求、记录/参数或整改证据。",
|
||||
"正文末尾必须保留“证据落点与待补证据”表。",
|
||||
],
|
||||
}
|
||||
validate_compressed_finding(finding)
|
||||
(out_dir / f"{brief['chapter_id']}.json").write_text(
|
||||
json.dumps(finding, ensure_ascii=False, indent=2) + "\n",
|
||||
encoding="utf-8",
|
||||
)
|
||||
findings.append(finding)
|
||||
return findings
|
||||
|
||||
|
||||
def build_chapter_user_prompt(brief: dict) -> str:
|
||||
return (
|
||||
"请根据以下 compressed finding / chapter brief 写一章正式中文 Markdown 正文。\n"
|
||||
"目标是形成一个完整章节，而不是 packet 摘要。避免碎片化，按金字塔结构组织：章首先给结论，再用证据支撑。\n"
|
||||
"要求：标题必须是观点型判断；每个数字和事实保留 [src_xxx]；纳入反方证据；不要出现调度元数据。\n"
|
||||
"如 brief 中包含 cached_source_paths，说明这些是已抓取到本地的核心一手/权威信源快照；优先使用 packet 已摘录的原文，并在证据不足时标记需要从本地快照补摘录，不要重新联网检索。\n"
|
||||
"禁止写空泛咨询腔。每个二级小节都必须至少落下 2 个具体审计发现、法规要求、SOP/记录/参数/现场观察或整改证据；不要只写原则。\n"
|
||||
"正文末尾必须增加“证据落点与待补证据”小节，用表格列出：关键判断、已使用证据 source_id、已落地整改动作、仍缺证据。若证据不足，直接标注需回炉 Phase 2，不要用泛泛表述补齐。\n"
|
||||
"只输出 Markdown，不要输出解释。\n\n"
|
||||
f"{json.dumps(brief, ensure_ascii=False, indent=2)}"
|
||||
)
|
||||
|
||||
|
||||
class ChapterAssemblyWorker:
|
||||
def __init__(
|
||||
self,
|
||||
*,
|
||||
role: RoleDefinition,
|
||||
client: ChatClient,
|
||||
skill_registry: SkillRegistry | None = None,
|
||||
) -> None:
|
||||
self.role = role
|
||||
self.client = client
|
||||
self.skill_registry = skill_registry or SkillRegistry()
|
||||
|
||||
def _system_prompt(self) -> str:
|
||||
skill_texts = []
|
||||
for name in self.role.skills:
|
||||
try:
|
||||
skill_texts.append(f"# Skill: {name}\n\n{self.skill_registry.read(name)}")
|
||||
except FileNotFoundError:
|
||||
skill_texts.append(f"# Skill: {name}\n\n[missing skill: {name}]")
|
||||
return (
|
||||
f"{self.role.identity}\n\n"
|
||||
"你是 Deep Research v0.20 的中文章节组装 worker。\n"
|
||||
"你的职责是把结构化证据包收束成连贯章节，解决并发研究造成的碎片化。\n"
|
||||
"不得编造来源，不得删除关键反方证据。\n\n"
|
||||
+ "\n\n".join(skill_texts)
|
||||
)
|
||||
|
||||
def write_chapter(self, *, project_root: Path, brief: dict) -> Path:
|
||||
if "key_findings" in brief:
|
||||
validate_compressed_finding(brief)
|
||||
else:
|
||||
validate_chapter_brief(brief)
|
||||
markdown = self.client.chat_complete(
|
||||
model=self.role.model,
|
||||
system=self._system_prompt(),
|
||||
user=build_chapter_user_prompt(brief),
|
||||
temperature=self.role.temperature,
|
||||
max_tokens=self.role.max_tokens,
|
||||
tag=f"chapter:{brief['chapter_id']}",
|
||||
)
|
||||
validate_chapter_markdown_citations(markdown, brief)
|
||||
out = project_root / "phase2" / "drafts" / f"{brief['chapter_id']}.md"
|
||||
out.parent.mkdir(parents=True, exist_ok=True)
|
||||
out.write_text(markdown.rstrip() + "\n", encoding="utf-8")
|
||||
return out
|
||||
|
||||
|
||||
def _write_chapter_error(project_root: Path, brief: dict, error: Exception) -> None:
|
||||
path = project_root / "phase2" / "chapter_errors" / f"{brief.get('chapter_id', 'unknown')}.json"
|
||||
path.parent.mkdir(parents=True, exist_ok=True)
|
||||
payload = {
|
||||
"chapter_id": brief.get("chapter_id"),
|
||||
"status": "failed",
|
||||
"error": str(error),
|
||||
}
|
||||
path.write_text(json.dumps(payload, ensure_ascii=False, indent=2) + "\n", encoding="utf-8")
|
||||
|
||||
|
||||
def run_chapter_assembly_workers(
|
||||
*,
|
||||
project_root: Path,
|
||||
briefs: list[dict],
|
||||
runtime: RuntimeProfile,
|
||||
client_factory: Callable[[RoleDefinition], ChatClient],
|
||||
workers: int,
|
||||
) -> int:
|
||||
role = runtime.role_for_task("chapter_assembly")
|
||||
max_workers = max(1, min(workers, role.max_concurrency))
|
||||
|
||||
def run_one(brief: dict) -> tuple[dict, Path | None, Exception | None]:
|
||||
try:
|
||||
worker = ChapterAssemblyWorker(role=role, client=client_factory(role))
|
||||
return brief, worker.write_chapter(project_root=project_root, brief=brief), None
|
||||
except Exception as error:
|
||||
return brief, None, error
|
||||
|
||||
written = 0
|
||||
with ThreadPoolExecutor(max_workers=max_workers) as pool:
|
||||
futures = [pool.submit(run_one, brief) for brief in briefs]
|
||||
for future in as_completed(futures):
|
||||
brief, path, error = future.result()
|
||||
if error is not None:
|
||||
_write_chapter_error(project_root, brief, error)
|
||||
continue
|
||||
if path is None:
|
||||
_write_chapter_error(project_root, brief, RuntimeError("chapter worker returned no output path"))
|
||||
continue
|
||||
written += 1
|
||||
return written
|
||||
@@ -0,0 +1,229 @@
|
||||
"""Phase 0 user-provided material ingestion."""
|
||||
|
||||
from __future__ import annotations
|
||||
|
||||
import base64
|
||||
import os
|
||||
import shutil
|
||||
from dataclasses import dataclass
|
||||
from pathlib import Path
|
||||
from typing import Any
|
||||
|
||||
import requests
|
||||
|
||||
|
||||
DEFAULT_FIRERED_OCR_ENDPOINT = "http://192.168.50.100:8001"
|
||||
DEFAULT_OCR_MAX_PAGES = 50
|
||||
|
||||
|
||||
@dataclass(frozen=True)
|
||||
class OcrResult:
|
||||
text: str
|
||||
pages_processed: int
|
||||
output_path: Path
|
||||
|
||||
|
||||
def safe_filename(path: Path) -> str:
|
||||
name = path.name.strip()
|
||||
return name or "material"
|
||||
|
||||
|
||||
def extract_pdf_text(path: Path) -> tuple[str, int, bool]:
|
||||
from pypdf import PdfReader
|
||||
|
||||
reader = PdfReader(str(path))
|
||||
chunks: list[str] = []
|
||||
for index, page in enumerate(reader.pages, start=1):
|
||||
text = (page.extract_text() or "").strip()
|
||||
if text:
|
||||
chunks.append(f"\n\n## Page {index}\n\n{text}")
|
||||
combined = "".join(chunks).strip()
|
||||
ocr_required = len(combined) < max(20, len(reader.pages) * 20)
|
||||
return combined, len(reader.pages), ocr_required
|
||||
|
||||
|
||||
def ocr_endpoint_from_env() -> str:
|
||||
return os.environ.get("DEEP_RESEARCH_OCR_ENDPOINT", DEFAULT_FIRERED_OCR_ENDPOINT).rstrip("/")
|
||||
|
||||
|
||||
def ocr_max_pages_from_env() -> int:
|
||||
raw = os.environ.get("DEEP_RESEARCH_OCR_MAX_PAGES")
|
||||
if not raw:
|
||||
return DEFAULT_OCR_MAX_PAGES
|
||||
try:
|
||||
return max(1, int(raw))
|
||||
except ValueError:
|
||||
return DEFAULT_OCR_MAX_PAGES
|
||||
|
||||
|
||||
def render_pdf_pages(pdf_path: Path, output_dir: Path, *, max_pages: int) -> list[Path]:
|
||||
import fitz
|
||||
|
||||
pages_dir = output_dir / f"{pdf_path.stem}.ocr-pages"
|
||||
pages_dir.mkdir(parents=True, exist_ok=True)
|
||||
image_paths: list[Path] = []
|
||||
doc = fitz.open(pdf_path)
|
||||
try:
|
||||
for index, page in enumerate(doc[:max_pages], start=1):
|
||||
pix = page.get_pixmap(matrix=fitz.Matrix(2, 2), alpha=False)
|
||||
image_path = pages_dir / f"page-{index:03d}.png"
|
||||
pix.save(image_path)
|
||||
image_paths.append(image_path)
|
||||
finally:
|
||||
doc.close()
|
||||
return image_paths
|
||||
|
||||
|
||||
def data_url_for_image(path: Path) -> str:
|
||||
encoded = base64.b64encode(path.read_bytes()).decode("ascii")
|
||||
return f"data:image/png;base64,{encoded}"
|
||||
|
||||
|
||||
def call_firered_ocr(image_path: Path, *, endpoint: str) -> str:
|
||||
payload = {
|
||||
"model": "firered-ocr",
|
||||
"messages": [
|
||||
{
|
||||
"role": "user",
|
||||
"content": [
|
||||
{"type": "text", "text": "请识别图片中的全部文字，保持原有顺序，只输出文字。"},
|
||||
{"type": "image_url", "image_url": {"url": data_url_for_image(image_path)}},
|
||||
],
|
||||
}
|
||||
],
|
||||
"temperature": 0,
|
||||
"max_tokens": 3000,
|
||||
}
|
||||
response = requests.post(f"{endpoint.rstrip('/')}/v1/chat/completions", json=payload, timeout=60)
|
||||
if not response.ok:
|
||||
raise RuntimeError(f"{response.status_code} {response.text[:500]}")
|
||||
data = response.json()
|
||||
return str(data["choices"][0]["message"].get("content") or "").strip()
|
||||
|
||||
|
||||
def ocr_pdf_with_firered(*, pdf_path: Path, output_dir: Path, endpoint: str, max_pages: int) -> OcrResult:
|
||||
image_paths = render_pdf_pages(pdf_path, output_dir, max_pages=max_pages)
|
||||
chunks: list[str] = []
|
||||
for index, image_path in enumerate(image_paths, start=1):
|
||||
text = call_firered_ocr(image_path, endpoint=endpoint)
|
||||
if text:
|
||||
chunks.append(f"\n\n## OCR Page {index}\n\n{text}")
|
||||
combined = "".join(chunks).strip()
|
||||
output_path = output_dir / f"{pdf_path.stem}.ocr.md"
|
||||
body = [
|
||||
f"# OCR Material: {pdf_path.name}",
|
||||
"",
|
||||
f"- source_path: {pdf_path}",
|
||||
f"- endpoint: {endpoint}",
|
||||
f"- pages_processed: {len(image_paths)}",
|
||||
"",
|
||||
combined or "OCR 未返回可用文本。",
|
||||
"",
|
||||
]
|
||||
output_path.write_text("\n".join(body), encoding="utf-8")
|
||||
return OcrResult(text=combined, pages_processed=len(image_paths), output_path=output_path)
|
||||
|
||||
|
||||
def ingest_input_materials(project_root: Path, materials: list[str] | None) -> list[dict[str, Any]]:
|
||||
inventory: list[dict[str, Any]] = []
|
||||
if not materials:
|
||||
return inventory
|
||||
|
||||
inputs_dir = project_root / "phase0" / "inputs"
|
||||
extracted_dir = project_root / "phase0" / "extracted"
|
||||
ocr_endpoint = ocr_endpoint_from_env()
|
||||
ocr_max_pages = ocr_max_pages_from_env()
|
||||
inputs_dir.mkdir(parents=True, exist_ok=True)
|
||||
extracted_dir.mkdir(parents=True, exist_ok=True)
|
||||
|
||||
for raw in materials:
|
||||
source = Path(raw).expanduser()
|
||||
if not source.exists():
|
||||
inventory.append({"kind": "note", "note": raw})
|
||||
continue
|
||||
|
||||
copied = inputs_dir / safe_filename(source)
|
||||
shutil.copy2(source, copied)
|
||||
item: dict[str, Any] = {
|
||||
"kind": source.suffix.lower().lstrip(".") or "file",
|
||||
"source_path": str(source),
|
||||
"copied_to": str(copied.relative_to(project_root)),
|
||||
"size_bytes": source.stat().st_size,
|
||||
}
|
||||
|
||||
if source.suffix.lower() == ".pdf":
|
||||
text, pages, ocr_required = extract_pdf_text(source)
|
||||
extracted = extracted_dir / f"{source.stem}.md"
|
||||
ocr_result: OcrResult | None = None
|
||||
ocr_error: str | None = None
|
||||
if ocr_required:
|
||||
try:
|
||||
ocr_result = ocr_pdf_with_firered(
|
||||
pdf_path=source,
|
||||
output_dir=extracted_dir,
|
||||
endpoint=ocr_endpoint,
|
||||
max_pages=min(pages, ocr_max_pages),
|
||||
)
|
||||
if ocr_result.text:
|
||||
text = "\n\n".join(part for part in [text, ocr_result.text] if part)
|
||||
except Exception as exc: # noqa: BLE001 - ingestion should not block project init.
|
||||
ocr_error = str(exc)
|
||||
|
||||
body = [
|
||||
f"# Extracted Material: {source.name}",
|
||||
"",
|
||||
f"- source_path: {source}",
|
||||
f"- copied_to: {copied.relative_to(project_root)}",
|
||||
f"- pages: {pages}",
|
||||
f"- ocr_required: {str(ocr_required).lower()}",
|
||||
f"- ocr_status: {'completed' if ocr_result else 'failed' if ocr_error else 'not_required'}",
|
||||
"",
|
||||
text or "未能从 PDF 直接抽取文本；该材料可能需要 OCR。",
|
||||
"",
|
||||
]
|
||||
if ocr_error:
|
||||
body.extend(["## OCR Error", "", ocr_error, ""])
|
||||
extracted.write_text("\n".join(body), encoding="utf-8")
|
||||
item.update(
|
||||
{
|
||||
"pages": pages,
|
||||
"extracted_to": str(extracted.relative_to(project_root)),
|
||||
"text_chars": len(text),
|
||||
"ocr_required": ocr_required,
|
||||
"ocr_status": "completed" if ocr_result else "failed" if ocr_error else "not_required",
|
||||
}
|
||||
)
|
||||
if ocr_result:
|
||||
item.update(
|
||||
{
|
||||
"ocr_endpoint": ocr_endpoint,
|
||||
"ocr_pages_processed": ocr_result.pages_processed,
|
||||
"ocr_extracted_to": str(ocr_result.output_path.relative_to(project_root)),
|
||||
"ocr_text_chars": len(ocr_result.text),
|
||||
}
|
||||
)
|
||||
if ocr_error:
|
||||
item["ocr_error"] = ocr_error
|
||||
else:
|
||||
item["ocr_required"] = source.suffix.lower() in {".png", ".jpg", ".jpeg", ".tif", ".tiff"}
|
||||
inventory.append(item)
|
||||
|
||||
return inventory
|
||||
|
||||
|
||||
def render_material_inventory(inventory: list[dict[str, Any]]) -> str:
|
||||
if not inventory:
|
||||
return "- 暂无；可通过 `--input-material` 加入审计报告、问题清单或内部记录。"
|
||||
lines: list[str] = []
|
||||
for item in inventory:
|
||||
if item.get("kind") == "note":
|
||||
lines.append(f"- 备注：{item.get('note', '')}")
|
||||
continue
|
||||
marker = "；需要 OCR" if item.get("ocr_required") else ""
|
||||
ocr = f"；OCR：{item.get('ocr_status')}" if item.get("ocr_status") else ""
|
||||
extracted = item.get("extracted_to")
|
||||
extra = f"；抽取文本：{extracted}" if extracted else ""
|
||||
lines.append(
|
||||
f"- {item.get('copied_to')}（{item.get('kind')}，{item.get('size_bytes', 0)} bytes{extra}{marker}{ocr}）"
|
||||
)
|
||||
return "\n".join(lines)
|
||||
@@ -0,0 +1,61 @@
|
||||
"""Research method registry for Phase 1 framework selection."""
|
||||
|
||||
from __future__ import annotations
|
||||
|
||||
from dataclasses import dataclass
|
||||
from pathlib import Path
|
||||
from typing import Any
|
||||
|
||||
import yaml
|
||||
|
||||
|
||||
REPO_ROOT = Path(__file__).resolve().parents[2]
|
||||
DEFAULT_METHOD_CONFIG = REPO_ROOT / "configs" / "research_methods.yaml"
|
||||
|
||||
|
||||
@dataclass(frozen=True)
|
||||
class ResearchMethod:
|
||||
key: str
|
||||
name: str
|
||||
best_for: list[str]
|
||||
structure_principle: str
|
||||
task_axes: list[str]
|
||||
framework_sections: list[str]
|
||||
integrated_lanes: list[str]
|
||||
|
||||
|
||||
class ResearchMethodRegistry:
|
||||
def __init__(self, path: Path | None = None) -> None:
|
||||
self.path = path or DEFAULT_METHOD_CONFIG
|
||||
self._data = self._load()
|
||||
|
||||
def _load(self) -> dict[str, Any]:
|
||||
if not self.path.exists():
|
||||
raise FileNotFoundError(f"research method config not found: {self.path}")
|
||||
data = yaml.safe_load(self.path.read_text(encoding="utf-8")) or {}
|
||||
if not isinstance(data, dict) or "methods" not in data:
|
||||
raise ValueError(f"invalid research method config: {self.path}")
|
||||
return data
|
||||
|
||||
@property
|
||||
def default_method(self) -> str:
|
||||
return (self._data.get("defaults") or {}).get("method", "mckinsey_market")
|
||||
|
||||
def list_names(self) -> list[str]:
|
||||
return sorted((self._data.get("methods") or {}).keys())
|
||||
|
||||
def get(self, key: str | None = None) -> ResearchMethod:
|
||||
selected = key or self.default_method
|
||||
methods = self._data.get("methods") or {}
|
||||
if selected not in methods:
|
||||
raise KeyError(f"unknown research_method: {selected}")
|
||||
item = methods[selected] or {}
|
||||
return ResearchMethod(
|
||||
key=selected,
|
||||
name=item.get("name", selected),
|
||||
best_for=list(item.get("best_for") or []),
|
||||
structure_principle=item.get("structure_principle", ""),
|
||||
task_axes=list(item.get("task_axes") or []),
|
||||
framework_sections=list(item.get("framework_sections") or []),
|
||||
integrated_lanes=list(item.get("integrated_lanes") or item.get("task_axes") or []),
|
||||
)
|
||||
@@ -0,0 +1,100 @@
|
||||
"""Deterministic orchestration helpers for v0.20."""
|
||||
|
||||
from __future__ import annotations
|
||||
|
||||
import json
|
||||
from datetime import datetime, timezone
|
||||
from pathlib import Path
|
||||
|
||||
from scripts.runtime.artifacts import ensure_phase_dirs, load_manifest, write_manifest
|
||||
from scripts.runtime.methods import ResearchMethodRegistry
|
||||
from scripts.runtime.tasks import generate_task_cards, generate_task_cards_from_research_brief, write_task_cards
|
||||
|
||||
|
||||
def utc_now_iso() -> str:
|
||||
return datetime.now(timezone.utc).replace(microsecond=0).isoformat()
|
||||
|
||||
|
||||
def create_phase2_task_cards(
|
||||
project_root: Path,
|
||||
*,
|
||||
axes: list[str] | None = None,
|
||||
dry_run: bool = False,
|
||||
) -> list[dict[str, object]]:
|
||||
framework = project_root / "phase1" / "framework.md"
|
||||
if not framework.exists():
|
||||
raise FileNotFoundError(f"framework not found: {framework}")
|
||||
method_key = None
|
||||
if (project_root / "manifest.json").exists():
|
||||
method_key = load_manifest(project_root).get("research_method")
|
||||
method = ResearchMethodRegistry().get(method_key)
|
||||
research_brief_path = project_root / "phase1" / "research_brief.json"
|
||||
framework_text = framework.read_text(encoding="utf-8")
|
||||
if research_brief_path.exists():
|
||||
research_brief = json.loads(research_brief_path.read_text(encoding="utf-8"))
|
||||
if not research_brief.get("materials"):
|
||||
material_inventory = load_manifest(project_root).get("material_inventory") or []
|
||||
materials = []
|
||||
for item in material_inventory:
|
||||
rel = item.get("ocr_extracted_to") or item.get("extracted_to") or item.get("copied_to")
|
||||
if rel:
|
||||
materials.append({"path": rel, "role": "input_material"})
|
||||
if materials:
|
||||
research_brief["materials"] = materials
|
||||
cards = generate_task_cards_from_research_brief(
|
||||
project_root.name,
|
||||
framework_text,
|
||||
research_brief,
|
||||
axes=axes,
|
||||
method=method,
|
||||
)
|
||||
else:
|
||||
cards = generate_task_cards(project_root.name, framework_text, axes=axes, method=method)
|
||||
if not dry_run:
|
||||
ensure_phase_dirs(project_root)
|
||||
write_task_cards(project_root / "phase2" / "task_cards.json", cards)
|
||||
manifest = load_manifest(project_root)
|
||||
phase2 = manifest.setdefault("phase2", {})
|
||||
phase2.update(
|
||||
{
|
||||
"status": "in_progress",
|
||||
"runtime": "python-core-v0.20",
|
||||
"research_method": method.key,
|
||||
"task_cards_path": "phase2/task_cards.json",
|
||||
"task_cards_total": len(cards),
|
||||
"updated_at": utc_now_iso(),
|
||||
}
|
||||
)
|
||||
write_manifest(project_root, manifest)
|
||||
return [card.to_dict() for card in cards]
|
||||
|
||||
|
||||
def write_placeholder_packets(
|
||||
project_root: Path,
|
||||
task_cards: list[dict[str, object]],
|
||||
*,
|
||||
dry_run: bool = False,
|
||||
) -> int:
|
||||
"""Create packet skeletons for manual/API completion.
|
||||
|
||||
This keeps the first v0.20 implementation deterministic and resumable; LLM
|
||||
calls can later fill the same schema without changing downstream readers.
|
||||
"""
|
||||
count = 0
|
||||
for card in task_cards:
|
||||
packet_path = project_root / str(card["output_packet"])
|
||||
packet = {
|
||||
"task_id": card["task_id"],
|
||||
"claims": [],
|
||||
"evidence_items": [],
|
||||
"counter_evidence": [],
|
||||
"source_ids": [],
|
||||
"source_quality_notes": [],
|
||||
"open_questions": ["待由 Python role worker 调用模型补全。"],
|
||||
"raw_quotes_or_notes": [],
|
||||
}
|
||||
if not dry_run:
|
||||
packet_path.parent.mkdir(parents=True, exist_ok=True)
|
||||
packet_path.write_text(json.dumps(packet, ensure_ascii=False, indent=2) + "\n", encoding="utf-8")
|
||||
count += 1
|
||||
return count
|
||||
@@ -0,0 +1,902 @@
|
||||
"""Phase 1 project initialization and framework generation."""
|
||||
|
||||
from __future__ import annotations
|
||||
|
||||
import hashlib
|
||||
import json
|
||||
import re
|
||||
from datetime import datetime, timezone
|
||||
from pathlib import Path
|
||||
from typing import Any
|
||||
|
||||
from scripts.runtime.artifacts import PROJECTS_DIR, ensure_phase_dirs, load_manifest, write_manifest
|
||||
from scripts.runtime.materials import ingest_input_materials, render_material_inventory
|
||||
from scripts.runtime.methods import ResearchMethod, ResearchMethodRegistry
|
||||
from scripts.runtime.tasks import AXIS_ROUTES
|
||||
|
||||
|
||||
def utc_now_iso() -> str:
|
||||
return datetime.now(timezone.utc).replace(microsecond=0).isoformat()
|
||||
|
||||
|
||||
def slugify_topic(topic: str) -> str:
|
||||
slug = re.sub(r"[^a-zA-Z0-9]+", "-", topic.lower()).strip("-")
|
||||
if slug:
|
||||
return slug[:80]
|
||||
digest = hashlib.sha1(topic.encode("utf-8")).hexdigest()[:8]
|
||||
return f"research-{digest}"
|
||||
|
||||
|
||||
def create_project(
|
||||
*,
|
||||
topic: str,
|
||||
slug: str | None = None,
|
||||
projects_dir: Path = PROJECTS_DIR,
|
||||
method_key: str | None = None,
|
||||
report_type: str = "research",
|
||||
model_profile: str = "medium",
|
||||
target_words: int = 30000,
|
||||
input_materials: list[str] | None = None,
|
||||
) -> Path:
|
||||
method = ResearchMethodRegistry().get(method_key)
|
||||
project_slug = slug or slugify_topic(topic)
|
||||
project_root = projects_dir / project_slug
|
||||
if project_root.exists():
|
||||
raise FileExistsError(f"project already exists: {project_root}")
|
||||
ensure_phase_dirs(project_root)
|
||||
(project_root / "phase0" / "inputs").mkdir(parents=True, exist_ok=True)
|
||||
(project_root / "phase0" / "extracted").mkdir(parents=True, exist_ok=True)
|
||||
material_inventory = ingest_input_materials(project_root, input_materials)
|
||||
now = utc_now_iso()
|
||||
manifest: dict[str, Any] = {
|
||||
"version": "0.20.0",
|
||||
"runtime": "python-core-v0.20",
|
||||
"topic": topic,
|
||||
"slug": project_slug,
|
||||
"report_title": topic,
|
||||
"type": report_type,
|
||||
"work_language": "zh",
|
||||
"model_profile": model_profile,
|
||||
"research_method": method.key,
|
||||
"target_words": target_words,
|
||||
"input_materials": input_materials or [],
|
||||
"material_inventory": material_inventory,
|
||||
"created_at": now,
|
||||
"updated_at": now,
|
||||
"phase1": {
|
||||
"status": "initialized",
|
||||
"approved": False,
|
||||
"requires_user_interview": True,
|
||||
"material_brief_path": "phase1/material_brief.md",
|
||||
},
|
||||
"phase2": {"status": "pending"},
|
||||
"phase3": {"status": "pending"},
|
||||
"phase4": {"status": "pending"},
|
||||
}
|
||||
write_manifest(project_root, manifest)
|
||||
_write_interview_seed(project_root, manifest, method)
|
||||
write_material_brief(project_root, manifest, method)
|
||||
return project_root
|
||||
|
||||
|
||||
def _write_interview_seed(project_root: Path, manifest: dict[str, Any], method: ResearchMethod) -> None:
|
||||
material_text = render_material_inventory(manifest.get("material_inventory") or [])
|
||||
text = (
|
||||
f"# Phase 1 访谈记录\n\n"
|
||||
f"- 主题：{manifest['topic']}\n"
|
||||
f"- 研究方法：{method.key} - {method.name}\n"
|
||||
f"- 报告类型：{manifest['type']}\n"
|
||||
f"- 目标字数：{manifest['target_words']}\n"
|
||||
f"- 工作语言：中文主写作；检索关键词、证据摘录和来源笔记可保留英文。\n\n"
|
||||
f"## 已提供材料\n\n{material_text}\n\n"
|
||||
"## 后续访谈问题\n\n"
|
||||
"1. 本报告最重要的决策用途是什么？\n"
|
||||
"2. 是否有必须覆盖或必须排除的公司、产品、工艺、市场或法规范围？\n"
|
||||
"3. 结论偏好是战略建议、风险清单、投资判断，还是执行路线图？\n"
|
||||
)
|
||||
(project_root / "phase1" / "interview.md").write_text(text, encoding="utf-8")
|
||||
|
||||
|
||||
def _material_excerpt(project_root: Path, rel_path: str, *, max_chars: int = 1200) -> str:
|
||||
path = project_root / rel_path
|
||||
if not path.exists():
|
||||
return "（未找到抽取文本）"
|
||||
text = path.read_text(encoding="utf-8")
|
||||
compact = "\n".join(line.rstrip() for line in text.splitlines() if line.strip())
|
||||
return compact[:max_chars] + ("..." if len(compact) > max_chars else "")
|
||||
|
||||
|
||||
def _derive_material_observations(project_root: Path, inventory: list[dict[str, Any]]) -> list[str]:
|
||||
combined_parts: list[str] = []
|
||||
for item in inventory:
|
||||
rel = item.get("ocr_extracted_to") or item.get("extracted_to")
|
||||
if rel and (project_root / rel).exists():
|
||||
combined_parts.append((project_root / rel).read_text(encoding="utf-8"))
|
||||
text = "\n".join(combined_parts)
|
||||
checks = [
|
||||
("审计范围覆盖生产管理、原液、制剂和无菌相关模块，报告需要同时处理 GMP 合规、工艺转移和运营协同，而不是只写质量体系。", ["生产管理", "原液", "制剂", "无菌"]),
|
||||
("材料显示高风险项为 0、中风险项为 1，适合采用“商业化 readiness 与系统成熟度差距”而非“体系失控”作为初始假设。", ["高风险 0", "中风险1", "低风险7"]),
|
||||
("商业化经验、无菌保障细节、文件要求与执行一致性是需要访谈确认的主线风险。", ["商业化经验不足", "无菌保障", "文件要求与执行一致性"]),
|
||||
("工艺规程、批记录、CPP/CQA、VMPR/VMP、验证主计划等内容反复出现，说明工艺验证和商业化文件体系可能是 Phase 2 的重点证据轴。", ["CPP", "CQA", "VMPR", "VMP"]),
|
||||
("温度、压差、WFI、冷却段微生物、RABS/ORABS、first air、APS 等无菌和设施细节需要映射到 EU Annex 1、NMPA GMP 和企业 SOP。", ["温度", "压差", "WFI", "APS"]),
|
||||
("复盘材料包含责任人和局部答复，后续整改路线图应尽量回填 owner、期限、关闭证据和复核机制。", ["填写人", "是否已经回答完整", "整改"]),
|
||||
]
|
||||
observations = [message for message, needles in checks if any(needle in text for needle in needles)]
|
||||
return observations or ["材料已导入但尚未形成足够结构化判断；需要先访谈确认研究用途、范围和优先级。"]
|
||||
|
||||
|
||||
def write_material_brief(
|
||||
project_root: Path,
|
||||
manifest: dict[str, Any] | None = None,
|
||||
method: ResearchMethod | None = None,
|
||||
) -> Path:
|
||||
"""Write a Phase 0/1 material brief that must be reviewed before Phase 2."""
|
||||
manifest = manifest or load_manifest(project_root)
|
||||
method = method or ResearchMethodRegistry().get(manifest.get("research_method"))
|
||||
inventory = manifest.get("material_inventory") or []
|
||||
lines = [
|
||||
f"# Phase 0 材料简报：{manifest.get('topic', project_root.name)}",
|
||||
"",
|
||||
"status: 待用户确认",
|
||||
f"research_method: {method.key}",
|
||||
"",
|
||||
"## 已导入材料",
|
||||
"",
|
||||
render_material_inventory(inventory),
|
||||
"",
|
||||
"## 材料初步解读",
|
||||
"",
|
||||
"以下内容由 Python core 从已落盘材料抽样生成，只作为访谈起点；不得直接视为最终结论。",
|
||||
"",
|
||||
]
|
||||
lines.extend(["## 初步问题聚类（待访谈确认）", ""])
|
||||
for observation in _derive_material_observations(project_root, inventory):
|
||||
lines.append(f"- {observation}")
|
||||
lines.append("")
|
||||
lines.append("## 材料摘录")
|
||||
lines.append("")
|
||||
for item in inventory:
|
||||
rel = item.get("ocr_extracted_to") or item.get("extracted_to")
|
||||
if not rel:
|
||||
continue
|
||||
lines.extend(
|
||||
[
|
||||
f"### {Path(rel).name}",
|
||||
"",
|
||||
_material_excerpt(project_root, rel),
|
||||
"",
|
||||
]
|
||||
)
|
||||
lines.extend(
|
||||
[
|
||||
"## 建议访谈确认点",
|
||||
"",
|
||||
"1. 本报告的最重要用途是什么：内部整改、客户沟通、董事会决策，还是外部审计准备？",
|
||||
"2. 哪些审计发现最需要优先展开：无菌保障、工艺验证、数据完整性、质量体系闭环，还是运营协同？",
|
||||
"3. 是否存在必须排除或脱敏的项目、人员、客户、产品或工艺信息？",
|
||||
"4. 短中长期整改的时间边界如何定义，例如 30/90/180 天，还是按临床/商业化里程碑划分？",
|
||||
"5. 是否需要把 NMPA、FDA、EMA、ICH、WHO 的法规基线分别映射到整改责任人和证据包？",
|
||||
"",
|
||||
"## Gate",
|
||||
"",
|
||||
"请用户确认本材料简报与访谈问题后，再生成或批准 `phase1/framework.md` 并进入 Phase 2。",
|
||||
"",
|
||||
]
|
||||
)
|
||||
out = project_root / "phase1" / "material_brief.md"
|
||||
out.parent.mkdir(parents=True, exist_ok=True)
|
||||
out.write_text("\n".join(lines), encoding="utf-8")
|
||||
return out
|
||||
|
||||
|
||||
def _axis_prompt_brief(axis: str, method: ResearchMethod) -> str:
|
||||
prompts = {
|
||||
"input_material_findings": "从用户材料中提取现场事实、审计发现、复盘记录和内部答复，并标注原始材料位置。",
|
||||
"nmpa_fda_ema_ich_who_baseline": "把 NMPA、FDA、EMA、ICH、WHO、药典或 Annex 1 等要求转化为可核验的法规基线，并纳入 FDA warning letters 与会议材料作为执法尺度参照。",
|
||||
"quality_system_gap": "把现场发现映射到质量体系流程缺口，覆盖偏差、变更、CAPA、文件、培训和数据完整性；优先检索 FDA warning letters 中同类缺陷的执法表述。",
|
||||
"manufacturing_process_risk": "围绕生产工艺、设施、公用系统、CPP/CQA、验证和无菌保障识别系统性风险，并用 FDA warning letters / inspection enforcement examples 校准严重度。",
|
||||
"operations_management_gap": "诊断运营管理、跨部门协同、会议机制、指标体系和交付节奏的结构性问题，并参考 FDA 会议纪要/meeting materials 中对质量治理的关注点。",
|
||||
"team_capability": "识别人员能力、岗位职责、质量文化和管理梯队方面的缺口与建设路径。",
|
||||
"capa_roadmap": "把差距转化为短中长期 CAPA 组合，要求绑定 owner、期限、优先级、关闭证据和复核机制。",
|
||||
"verification_evidence": "定义整改完成后可被审计接受的验证证据，包括记录、报告、趋势和管理评审输入。",
|
||||
"counter": "主动寻找反方证据、限制条件和可能降低严重度或改变优先级的解释，避免单向论证。",
|
||||
}
|
||||
return prompts.get(axis, f"按照 `{method.key}` 方法，对 {axis} 轴进行证据收集、证伪和结构化归纳。")
|
||||
|
||||
|
||||
def _material_paths(manifest: dict[str, Any]) -> list[dict[str, str]]:
|
||||
materials: list[dict[str, str]] = []
|
||||
for item in manifest.get("material_inventory") or []:
|
||||
rel = item.get("ocr_extracted_to") or item.get("extracted_to") or item.get("copied_to")
|
||||
if rel:
|
||||
materials.append({"path": rel, "role": "input_material"})
|
||||
return materials
|
||||
|
||||
|
||||
def _keywords_from_title(title: str) -> list[str]:
|
||||
english = re.findall(r"[A-Za-z][A-Za-z0-9/+-]{1,}", title)
|
||||
chinese_parts = re.split(r"[，,、；;：:\s]+|和|与|及|的|在|为|从|来自|集中|决定|需要|形成|成为|不是|而是", title)
|
||||
domain_terms = [
|
||||
"审计",
|
||||
"商业化",
|
||||
"阶段门",
|
||||
"风险",
|
||||
"法规",
|
||||
"欧盟",
|
||||
"NMPA",
|
||||
"GMP",
|
||||
"ICH",
|
||||
"无菌",
|
||||
"RABS",
|
||||
"First Air",
|
||||
"APS",
|
||||
"灯检",
|
||||
"隧道",
|
||||
"原液",
|
||||
"WFI",
|
||||
"SCADA",
|
||||
"EMS",
|
||||
"CPP",
|
||||
"CQA",
|
||||
"PPQ",
|
||||
"清洁验证",
|
||||
"偏差",
|
||||
"变更",
|
||||
"CAPA",
|
||||
"数据完整性",
|
||||
"人员",
|
||||
"培训",
|
||||
"质量文化",
|
||||
"运营",
|
||||
"跨部门",
|
||||
"指标",
|
||||
"团队",
|
||||
"CDMO",
|
||||
"整改",
|
||||
"owner",
|
||||
]
|
||||
title_terms = [term for term in domain_terms if term in title]
|
||||
keywords = [item.strip() for item in [*english, *title_terms, *chinese_parts] if len(item.strip()) >= 2]
|
||||
seen: set[str] = set()
|
||||
unique: list[str] = []
|
||||
for keyword in keywords:
|
||||
if keyword not in seen:
|
||||
seen.add(keyword)
|
||||
unique.append(keyword)
|
||||
return unique[:12]
|
||||
|
||||
|
||||
def _material_lines_for_chapter(project_root: Path, manifest: dict[str, Any], title: str, *, limit: int = 4) -> list[str]:
|
||||
keywords = _keywords_from_title(title)
|
||||
candidates: list[tuple[int, int, str]] = []
|
||||
order = 0
|
||||
for item in manifest.get("material_inventory") or []:
|
||||
rel = item.get("ocr_extracted_to") or item.get("extracted_to")
|
||||
if not rel:
|
||||
continue
|
||||
path = project_root / rel
|
||||
if not path.exists():
|
||||
continue
|
||||
for raw in path.read_text(encoding="utf-8").splitlines():
|
||||
line = raw.strip()
|
||||
if len(line) < 8 or len(line) > 220:
|
||||
continue
|
||||
if line.startswith("#") or line.startswith("- source_path:") or line.startswith("- extracted_at:"):
|
||||
continue
|
||||
if "OCR Material:" in line:
|
||||
continue
|
||||
if re.match(r"^(审计对象|审计执行方|审计执行人|审计时间)[:：]", line):
|
||||
continue
|
||||
score = sum(1 for keyword in keywords if keyword and keyword in line)
|
||||
if score:
|
||||
order += 1
|
||||
candidates.append((score, order, f"{rel}：{line}"))
|
||||
candidates.sort(key=lambda item: (-item[0], item[1]))
|
||||
return [line for _, _, line in candidates[:limit]]
|
||||
|
||||
|
||||
def _minimum_evidence_for_method(method: ResearchMethod) -> dict[str, Any]:
|
||||
if method.key == "gmp_quality_operations_diagnosis":
|
||||
return {
|
||||
"local_material_quotes": 2,
|
||||
"official_regulatory_or_guideline_sources": 2,
|
||||
"enforcement_or_best_practice_precedents": 1,
|
||||
"counter_evidence_or_boundary_conditions": 1,
|
||||
"actionable_remediation_items": 3,
|
||||
}
|
||||
return {
|
||||
"high_quality_sources": 4,
|
||||
"tier_1_2_sources": 2,
|
||||
"counter_evidence_or_boundary_conditions": 1,
|
||||
"decision_relevant_implications": 2,
|
||||
}
|
||||
|
||||
|
||||
def _central_thesis(manifest: dict[str, Any], method: ResearchMethod) -> str:
|
||||
topic = manifest.get("topic") or manifest.get("report_title") or "本研究主题"
|
||||
if method.key == "gmp_quality_operations_diagnosis":
|
||||
return (
|
||||
f"初始主判断：{topic} 不应只按审计风险项数量来评价，而应从商业化 readiness、"
|
||||
"质量体系运行成熟度、生产工艺证据链和运营协同能力四条线同时诊断。Phase 2 必须用"
|
||||
"现场材料原文、官方法规/指南、执法案例或标杆实践来证明、修正或推翻这一判断。"
|
||||
)
|
||||
return (
|
||||
f"初始主判断：{topic} 需要先形成可被证据推翻的观点型框架，再由 Phase 2 按方法论证据线"
|
||||
"逐项求证；不能把并发检索结果直接堆砌成报告。"
|
||||
)
|
||||
|
||||
|
||||
def _strategy_for_chapter(title: str, method: ResearchMethod) -> dict[str, Any]:
|
||||
"""Return non-tautological Phase 1 strategy text for a chapter title."""
|
||||
if method.key != "gmp_quality_operations_diagnosis":
|
||||
return {
|
||||
"core_question": f"本章需要判断：在什么证据条件下“{title}”成立，它会怎样改变最终决策？",
|
||||
"bold_hypothesis": f"初始假设不是复述标题，而是预判“{title}”背后存在一个可被验证的因果机制；Phase 2 需要找证据支持、修正或推翻这个机制。",
|
||||
"writing_claim": f"本章要把“{title}”写成一个可被证据检验的判断，而不是资料综述。",
|
||||
"counter_evidence": [
|
||||
"是否存在更简单的替代解释，能削弱本章主判断？",
|
||||
"关键证据是否只来自单一来源或利益相关来源？",
|
||||
"是否有反例显示本章判断只适用于部分场景？",
|
||||
],
|
||||
}
|
||||
|
||||
strategies = [
|
||||
(
|
||||
("审计", "阶段门"),
|
||||
{
|
||||
"core_question": "审计报告的低/中风险项计数，是否低估了白帆从临床/受托生产走向商业化标准时需要跨过的阶段门？",
|
||||
"bold_hypothesis": "初始假设：白帆的硬件和文件基础总体可用，但审计材料暴露的是商业化 readiness 缺口，而不是简单的若干孤立缺陷；Phase 2 应验证这些缺口是否集中在无菌保障、工艺验证、质量闭环和运营节奏。",
|
||||
"writing_claim": "本章要先把“风险项清单”翻译成管理层可决策的阶段门地图，说明哪些问题影响商业化放行、客户审计和技术转移节奏。",
|
||||
"counter_evidence": [
|
||||
"是否已有整改证据证明这些问题只是审计时点的临时缺口？",
|
||||
"低/中风险评级是否足以说明商业化阶段门影响有限？",
|
||||
"审计范围有限是否导致本章不能外推到整体体系成熟度？",
|
||||
],
|
||||
},
|
||||
),
|
||||
(
|
||||
("法规", "欧盟", "NMPA", "ICH"),
|
||||
{
|
||||
"core_question": "如果按 EU Annex 1、NMPA GMP、ICH Q9/Q10 以及 FDA 执法尺度校准，哪些现场发现的严重度和整改优先级会发生变化？",
|
||||
"bold_hypothesis": "初始假设：白帆按国内 GMP 逻辑已具备基础合规框架，但若以欧盟无菌标准和质量风险管理要求衡量，部分“低风险/建议项”会转化为体系成熟度缺口。",
|
||||
"writing_claim": "本章要建立后文共用的法规基线，避免整改优先级只跟随原审计评级，而忽略国际化和商业化标准。",
|
||||
"counter_evidence": [
|
||||
"相关国际标准是否并不适用于当前产品阶段或委托生产边界？",
|
||||
"NMPA 与欧盟/美国要求之间是否存在可接受差异？",
|
||||
"是否有企业内部标准已经覆盖但审计材料未呈现？",
|
||||
],
|
||||
},
|
||||
),
|
||||
(
|
||||
("无菌", "RABS", "First Air", "APS", "灯检"),
|
||||
{
|
||||
"core_question": "制剂线的主要无菌风险，是硬件布局不足，还是人员干预、首次气流保护、APS 覆盖和灯检标准执行证据不足？",
|
||||
"bold_hypothesis": "初始假设：白帆制剂车间硬件基础并非主要短板，真正风险在于关键无菌行为和模拟验证是否能持续证明受控；Phase 2 应重点查 First Air、RABS 干预、APS 场景设计和灯检阳性样品管理。",
|
||||
"writing_claim": "本章要把无菌保障从“设施看起来合规”推进到“关键操作和验证证据可被审计接受”。",
|
||||
"counter_evidence": [
|
||||
"现场是否已有完整视频复核、APS 覆盖和再培训有效性证据？",
|
||||
"观察到的无菌动作问题是否只是个别人员或单次拍摄偏差？",
|
||||
"灯检和 RABS 风险是否已有 SOP、趋势和复核记录闭环？",
|
||||
],
|
||||
},
|
||||
),
|
||||
(
|
||||
("原液", "WFI", "SCADA", "EMS"),
|
||||
{
|
||||
"core_question": "原液和公用系统的风险是否被一次性封闭工艺掩盖，真正缺口在 WFI、SCADA/EMS、离线记录和异常升级证据链？",
|
||||
"bold_hypothesis": "初始假设：一次性反应器和封闭转移降低了暴露风险，但不能自动证明系统受控；Phase 2 应验证 WFI 冷却回流、环境/压差报警、SCADA 数据和离线检测记录是否形成完整证据链。",
|
||||
"writing_claim": "本章要说明原液与公用系统不是“硬件先进即可”，而是要证明关键状态、报警、数据和异常处理持续受控。",
|
||||
"counter_evidence": [
|
||||
"WFI、SCADA/EMS 和离线记录是否已有验证报告与趋势复核？",
|
||||
"一次性系统是否已经充分降低共线和交叉污染风险？",
|
||||
"被指出的公用系统风险是否只是设计建议而非实际偏差？",
|
||||
],
|
||||
},
|
||||
),
|
||||
(
|
||||
("工艺", "CPP", "CQA", "PPQ", "清洁验证"),
|
||||
{
|
||||
"core_question": "现有 IND 阶段工艺规程和批记录，距离商业化 PPQ、控制策略和清洁验证所需证据还差在哪里？",
|
||||
"bold_hypothesis": "初始假设：白帆目前的工艺文件足以支撑临床阶段执行，但不足以支撑商业化批记录、CPP/CQA 控制、PPQ 和清洁验证闭环；Phase 2 应查明哪些字段、参数和验证证据必须前置补齐。",
|
||||
"writing_claim": "本章要把技术转移风险具体化为文件、参数、验证和批记录的硬门槛。",
|
||||
"counter_evidence": [
|
||||
"是否已有商业化模板、控制策略或 PPQ 草案未体现在审计材料中？",
|
||||
"当前项目阶段是否尚不需要完整商业化批记录要求？",
|
||||
"清洁验证和工艺验证是否已有主计划覆盖？",
|
||||
],
|
||||
},
|
||||
),
|
||||
(
|
||||
("偏差", "变更", "CAPA", "数据完整性"),
|
||||
{
|
||||
"core_question": "白帆的问题是没有质量流程，还是流程之间的事件分类、升级、CAPA 有效性和数据完整性尚未形成运行闭环？",
|
||||
"bold_hypothesis": "初始假设：白帆已有偏差、变更和 CAPA 的流程框架，但事件何时启动偏差、何时作为变更、如何证明 CAPA 有效，以及电子/纸质数据如何贯通，仍存在运行机制缺口。",
|
||||
"writing_claim": "本章要把质量体系从“有 SOP”推进到“事件能被正确分类、调查、纠正、验证并趋势复核”。",
|
||||
"counter_evidence": [
|
||||
"是否有趋势分析、管理评审和 CAPA effectiveness check 证明体系已经闭环？",
|
||||
"个别事件分类问题是否不足以代表体系性缺口？",
|
||||
"电子系统和纸质记录之间是否已有数据完整性控制？",
|
||||
],
|
||||
},
|
||||
),
|
||||
(
|
||||
("人员", "培训", "质量文化"),
|
||||
{
|
||||
"core_question": "培训记录齐全是否真的转化为一线无菌行为、偏差判断和质量风险意识？哪些证据能证明培训有效？",
|
||||
"bold_hypothesis": "初始假设：白帆不缺培训台账，缺的是把培训结果转化为现场行为的一致性证据；如果 First Air、干预动作、事件判断和灯检执行仍需反复提醒，问题就不是“再培训一次”，而是培训有效性确认和质量文化运行机制不足。",
|
||||
"writing_claim": "本章要把人员问题从“有没有培训”改写为“培训是否改变行为、降低风险、形成可复核证据”。",
|
||||
"counter_evidence": [
|
||||
"现场抽问、资格确认和再培训记录是否已证明人员理解到位？",
|
||||
"被观察到的行为问题是否只发生在少数岗位或单次演示？",
|
||||
"是否有岗位胜任力矩阵、年度复评和行为观察数据支撑人员能力？",
|
||||
],
|
||||
},
|
||||
),
|
||||
(
|
||||
("运营", "跨部门", "指标", "review"),
|
||||
{
|
||||
"core_question": "白帆当前整改和生产准备依赖个人推动，还是已经形成跨部门例会、问题升级、指标看板和管理层复核的运营系统？",
|
||||
"bold_hypothesis": "初始假设：运营短板不在于团队不努力，而在于缺少固定节奏和可视化管理系统；如果 owner、关闭证据、升级阈值和管理层 review 不稳定，整改会停留在临时协调，难以支撑商业化节奏。",
|
||||
"writing_claim": "本章要说明运营管理是 GMP 风险的放大器：没有节奏、看板和升级机制，技术和质量问题会反复跨部门漂移。",
|
||||
"counter_evidence": [
|
||||
"是否已经存在稳定 PMO/例会/看板，只是未进入审计材料？",
|
||||
"短期临时协调是否足以覆盖当前项目阶段，不需要完整运营系统？",
|
||||
"owner、期限和关闭证据是否已经在复盘文件中基本清楚？",
|
||||
],
|
||||
},
|
||||
),
|
||||
(
|
||||
("团队", "CDMO", "能力矩阵"),
|
||||
{
|
||||
"core_question": "对标成熟 CDMO，白帆最需要补齐的是人数、岗位能力，还是 QA/MSAT/工程/项目管理之间的角色分工？",
|
||||
"bold_hypothesis": "初始假设：白帆的能力缺口不是简单扩编，而是商业化 CDMO 所需的角色矩阵尚未完全成型；Phase 2 应验证 QA 独立性、MSAT 工艺支持、工程保障、生产班组和 PMO 协同能力。",
|
||||
"writing_claim": "本章要给出面向商业化的团队能力地图，说明哪些能力必须自建，哪些可外部支持，哪些要通过机制补齐。",
|
||||
"counter_evidence": [
|
||||
"现有人员是否已具备商业化经验，只是材料未体现？",
|
||||
"对标 CDMO 是否会高估当前阶段所需组织复杂度？",
|
||||
"是否可通过顾问、外包或客户支持临时补足能力？",
|
||||
],
|
||||
},
|
||||
),
|
||||
(
|
||||
("整改", "owner", "路线图"),
|
||||
{
|
||||
"core_question": "哪些整改必须立即完成，哪些属于体系补强，哪些是能力建设？每项如何绑定 owner、关闭证据和复核窗口？",
|
||||
"bold_hypothesis": "初始假设：如果整改只按问题清单逐条关闭，会漏掉体系性根因；更有效的路线应分为立即纠偏、90 天体系补强和中长期能力建设三层，并为每层定义关闭证据。",
|
||||
"writing_claim": "本章要把诊断转化为可执行 CAPA 组合，而不是泛泛的改进建议。",
|
||||
"counter_evidence": [
|
||||
"是否已有整改计划足以覆盖 owner、期限、关闭证据和 QA verification？",
|
||||
"部分整改是否应前移或后移，避免资源过载？",
|
||||
"哪些建议若缺少法规证据，不应被列为强制整改？",
|
||||
],
|
||||
},
|
||||
),
|
||||
(
|
||||
("管理层", "CAPA", "总表"),
|
||||
{
|
||||
"core_question": "管理层应通过什么样的 CAPA 总表、法规映射表和复核节奏，持续判断整改是否真正降低风险？",
|
||||
"bold_hypothesis": "初始假设：白帆需要的不只是一次性报告，而是一套管理层可追踪的整改仪表盘；否则 CAPA 关闭会变成文件动作，无法证明风险趋势下降和商业化 readiness 提升。",
|
||||
"writing_claim": "本章要把报告成果固化成管理层治理工具：CAPA 总表、法规映射、证据包和复核节奏。",
|
||||
"counter_evidence": [
|
||||
"现有管理评审或质量例会是否已经能承担这个功能？",
|
||||
"过度表格化是否会增加一线负担而不改善风险？",
|
||||
"哪些指标真正能反映风险降低，而不是制造形式化 KPI？",
|
||||
],
|
||||
},
|
||||
),
|
||||
]
|
||||
for needles, strategy in strategies:
|
||||
if any(needle in title for needle in needles):
|
||||
return strategy
|
||||
return {
|
||||
"core_question": f"本章需要判断“{title}”背后的真实风险、适用边界和整改优先级。",
|
||||
"bold_hypothesis": f"初始假设：{title} 不是孤立问题，而是质量体系、工艺证据或运营机制中的一个可验证缺口；Phase 2 必须用材料原文和外部证据判断其严重度。",
|
||||
"writing_claim": f"本章要把“{title}”转化为可执行的诊断结论和整改要求。",
|
||||
"counter_evidence": [
|
||||
"该问题是否已有充分整改或验证证据？",
|
||||
"是否只是阶段性限制，而非系统性缺口？",
|
||||
"外部标准是否适用于当前业务边界？",
|
||||
],
|
||||
}
|
||||
|
||||
|
||||
def build_chapter_planning(
|
||||
project_root: Path,
|
||||
manifest: dict[str, Any],
|
||||
method: ResearchMethod,
|
||||
titles: list[str],
|
||||
*,
|
||||
quota: int,
|
||||
) -> list[dict[str, Any]]:
|
||||
"""Build hypothesis-driven chapter plans that become Phase 2 prompt context."""
|
||||
lanes = list(method.integrated_lanes or method.task_axes)
|
||||
minimum_evidence = _minimum_evidence_for_method(method)
|
||||
plans: list[dict[str, Any]] = []
|
||||
for idx, title in enumerate(titles, start=1):
|
||||
chapter_id = f"ch{idx:02d}"
|
||||
material_lines = _material_lines_for_chapter(project_root, manifest, title)
|
||||
if not material_lines:
|
||||
material_lines = ["未在材料中自动匹配到足够线索；Phase 2 必须先回读全部输入材料并补充原文摘录。"]
|
||||
strategy = _strategy_for_chapter(title, method)
|
||||
core_question = strategy["core_question"]
|
||||
bold_hypothesis = strategy["bold_hypothesis"]
|
||||
verification_plan = [
|
||||
"先从允许的本地材料提取 2-4 条原文证据，保留出处和上下文。",
|
||||
f"再按方法论 evidence lanes 求证：{'；'.join(lanes)}。",
|
||||
"每个核心判断至少匹配 2 个独立高质量来源；不足时降级为待验证判断。",
|
||||
"主动搜索反方证据、低严重度解释、适用范围限制或替代原因。",
|
||||
"输出时把证据、判断、整改/建议和待补证据分开，避免直接写成散文化正文。",
|
||||
]
|
||||
counter_evidence = strategy["counter_evidence"]
|
||||
writing_claim = strategy["writing_claim"]
|
||||
phase2_prompt_context = "\n".join(
|
||||
[
|
||||
f"章节：{chapter_id} {title}",
|
||||
core_question,
|
||||
bold_hypothesis,
|
||||
"材料起点：",
|
||||
*[f"- {line}" for line in material_lines],
|
||||
"求证路线：",
|
||||
*[f"- {item}" for item in verification_plan],
|
||||
"必须寻找的反方/边界：",
|
||||
*[f"- {item}" for item in counter_evidence],
|
||||
f"写作主张：{writing_claim}",
|
||||
f"最低证据要求：{json.dumps(minimum_evidence, ensure_ascii=False)}",
|
||||
]
|
||||
)
|
||||
plans.append(
|
||||
{
|
||||
"chapter_id": chapter_id,
|
||||
"title": title,
|
||||
"suggested_words": quota,
|
||||
"core_question": core_question,
|
||||
"bold_hypothesis": bold_hypothesis,
|
||||
"why_this_matters": "本章用于把 Phase1 的判断转化为 Phase2 可验证命题，并为最终报告保留清晰主线。",
|
||||
"material_starting_points": material_lines,
|
||||
"evidence_lanes": lanes,
|
||||
"verification_plan": verification_plan,
|
||||
"counter_evidence_to_seek": counter_evidence,
|
||||
"writing_claim": writing_claim,
|
||||
"minimum_evidence": minimum_evidence,
|
||||
"phase2_prompt_context": phase2_prompt_context,
|
||||
}
|
||||
)
|
||||
return plans
|
||||
|
||||
|
||||
def build_research_brief_payload(
|
||||
project_root: Path,
|
||||
manifest: dict[str, Any],
|
||||
method: ResearchMethod,
|
||||
chapter_planning: list[dict[str, Any]] | None = None,
|
||||
) -> dict[str, Any]:
|
||||
"""Create the file-backed Phase 1 research brief used by task-card generation."""
|
||||
axes = list(method.task_axes)
|
||||
chapter_planning = chapter_planning or []
|
||||
return {
|
||||
"version": "0.21-alpha",
|
||||
"topic": manifest.get("topic", project_root.name),
|
||||
"research_method": method.key,
|
||||
"method_name": method.name,
|
||||
"work_language": "zh",
|
||||
"tone": "事实型、整改导向、面向管理层和质量/生产负责人；避免空泛咨询腔。",
|
||||
"central_question": f"如何基于已提供材料和权威法规/最佳实践，系统诊断“{manifest.get('topic', project_root.name)}”并形成可执行整改路线图？",
|
||||
"central_thesis": _central_thesis(manifest, method),
|
||||
"phase_logic": {
|
||||
"phase1": "大胆假设：结合输入材料、访谈信息和初步搜索，定下主基调、章节命题和求证路线。",
|
||||
"phase2": "小心求证：worker 只围绕 Phase1 命题收集、验证、证伪和补证，不自行重写研究方向。",
|
||||
"phase3": "一致性审校：检查 Phase1 假设与 Phase2 证据是否自洽，指出需要回炉的章节或证据缺口。",
|
||||
},
|
||||
"phase2_mode": "chapter_integrated",
|
||||
"success_criteria": [
|
||||
"每个核心判断都能回到用户材料、权威法规、最佳实践或反方证据。",
|
||||
"短中长期整改建议必须绑定优先级、责任、关闭证据和复核机制。",
|
||||
"章节写作必须先收束主线，再使用 evidence packet；不得按 packet 机械拼贴。",
|
||||
],
|
||||
"phase2_inputs": {
|
||||
"material_brief_path": "phase1/material_brief.md",
|
||||
"framework_path": "phase1/framework.md",
|
||||
"research_brief_path": "phase1/research_brief.json",
|
||||
},
|
||||
"materials": _material_paths(manifest),
|
||||
"chapter_planning": chapter_planning,
|
||||
"task_planning": {
|
||||
"chapter_source": "phase1/framework.md",
|
||||
"phase2_mode": "chapter_integrated",
|
||||
"axes": axes,
|
||||
"required_skills": [
|
||||
"deep-research",
|
||||
"search-gateway",
|
||||
"search-strategy",
|
||||
"source-quality",
|
||||
"evidence-table",
|
||||
"citation-manager",
|
||||
],
|
||||
"search_routes_by_axis": {axis: AXIS_ROUTES.get(axis, ["general"]) for axis in axes},
|
||||
"axis_prompt_briefs": {axis: _axis_prompt_brief(axis, method) for axis in axes},
|
||||
"stop_conditions": [
|
||||
"每张任务卡至少形成 3 条可追溯 evidence_items，且不得编造 candidate_sources 以外来源。",
|
||||
"关键 claim 不足 2 个独立 Tier 1-2 信源时，必须写入 open_questions 和证据缺口。",
|
||||
"必须包含 counter_evidence；找不到反方证据时记录检索路径和限制。",
|
||||
],
|
||||
"fragmentation_guard": "并发 worker 只生产 evidence packet；章节主线由 compressed_findings 收束，禁止直接把 packet 堆成正文。",
|
||||
},
|
||||
"clarification_notes": {
|
||||
"requires_user_review": True,
|
||||
"questions_source": "phase1/material_brief.md",
|
||||
"decision_items": [
|
||||
"确认报告用途、受众和脱敏边界。",
|
||||
"确认研究方法是否适配当前场景；MECE 只是可选方法之一。",
|
||||
"确认任务切分和检索策略是否足以让低成本模型独立执行。",
|
||||
],
|
||||
},
|
||||
}
|
||||
|
||||
|
||||
def write_research_brief(
|
||||
project_root: Path,
|
||||
manifest: dict[str, Any] | None = None,
|
||||
method: ResearchMethod | None = None,
|
||||
chapter_planning: list[dict[str, Any]] | None = None,
|
||||
) -> tuple[Path, Path]:
|
||||
manifest = manifest or load_manifest(project_root)
|
||||
method = method or ResearchMethodRegistry().get(manifest.get("research_method"))
|
||||
payload = build_research_brief_payload(project_root, manifest, method, chapter_planning=chapter_planning)
|
||||
json_path = project_root / "phase1" / "research_brief.json"
|
||||
md_path = project_root / "phase1" / "research_brief.md"
|
||||
json_path.parent.mkdir(parents=True, exist_ok=True)
|
||||
hypothesis_path = project_root / "phase1" / "hypothesis_map.json"
|
||||
json_path.write_text(json.dumps(payload, ensure_ascii=False, indent=2) + "\n", encoding="utf-8")
|
||||
hypothesis_path.write_text(json.dumps(payload.get("chapter_planning") or [], ensure_ascii=False, indent=2) + "\n", encoding="utf-8")
|
||||
lines = [
|
||||
f"# Phase 1 Research Brief：{payload['topic']}",
|
||||
"",
|
||||
f"- research_method: {payload['research_method']}",
|
||||
f"- work_language: {payload['work_language']}",
|
||||
f"- tone: {payload['tone']}",
|
||||
f"- phase2_mode: {payload['phase2_mode']}",
|
||||
"",
|
||||
"## 中心问题",
|
||||
"",
|
||||
payload["central_question"],
|
||||
"",
|
||||
"## 主基调 / 大胆假设",
|
||||
"",
|
||||
payload["central_thesis"],
|
||||
"",
|
||||
"## Phase 逻辑",
|
||||
"",
|
||||
]
|
||||
for phase_name, phase_text in payload["phase_logic"].items():
|
||||
lines.append(f"- `{phase_name}`：{phase_text}")
|
||||
lines.extend([
|
||||
"",
|
||||
"## 成功标准",
|
||||
"",
|
||||
])
|
||||
lines.extend(f"- {item}" for item in payload["success_criteria"])
|
||||
if payload.get("chapter_planning"):
|
||||
lines.extend(["", "## 章节命题与求证计划", ""])
|
||||
for item in payload["chapter_planning"]:
|
||||
lines.extend(
|
||||
[
|
||||
f"### {item['chapter_id']} {item['title']}",
|
||||
"",
|
||||
f"- 核心问题：{item['core_question']}",
|
||||
f"- 大胆假设：{item['bold_hypothesis']}",
|
||||
f"- 写作主张：{item['writing_claim']}",
|
||||
f"- 证据线：{'；'.join(item['evidence_lanes'])}",
|
||||
"- 材料起点：",
|
||||
]
|
||||
)
|
||||
lines.extend(f" - {line}" for line in item["material_starting_points"])
|
||||
lines.extend(["- 求证计划："])
|
||||
lines.extend(f" - {line}" for line in item["verification_plan"])
|
||||
lines.extend([""])
|
||||
lines.extend(["", "## 任务切分原则", ""])
|
||||
planning = payload["task_planning"]
|
||||
lines.append(planning["fragmentation_guard"])
|
||||
lines.append("")
|
||||
for axis in planning["axes"]:
|
||||
routes = "、".join(planning["search_routes_by_axis"].get(axis, []))
|
||||
prompt = planning["axis_prompt_briefs"].get(axis, "")
|
||||
lines.append(f"- `{axis}`：{prompt} 检索路径：{routes}")
|
||||
lines.extend(["", "## 必读 Skills", ""])
|
||||
lines.extend(f"- {name}" for name in planning["required_skills"])
|
||||
lines.extend(["", "## 停止条件", ""])
|
||||
lines.extend(f"- {item}" for item in planning["stop_conditions"])
|
||||
lines.append("")
|
||||
md_path.write_text("\n".join(lines), encoding="utf-8")
|
||||
return md_path, json_path
|
||||
|
||||
|
||||
CHAPTER_TEMPLATES: dict[str, list[str]] = {
|
||||
"mckinsey_market": [
|
||||
"核心结论先行界定市场机会与约束",
|
||||
"临床与真实世界证据决定需求天花板",
|
||||
"监管路径和支付环境重塑商业化节奏",
|
||||
"竞争格局正在从单点产品转向组合能力",
|
||||
"专利与技术壁垒决定长期利润池",
|
||||
"中国市场的准入和供给能力形成独立变量",
|
||||
"资本市场预期与基本面之间存在可验证偏差",
|
||||
"反方证据限定结论边界并提示回撤风险",
|
||||
"战略选择应围绕资源约束排序",
|
||||
"执行路线图需要把证据缺口转化为行动清单",
|
||||
],
|
||||
"gmp_gap_assessment": [
|
||||
"监管基线决定整改范围而非企业主观偏好",
|
||||
"现状差距需要按法规条款和业务流程双重定位",
|
||||
"质量风险分级决定 CAPA 优先级",
|
||||
"根因分析质量决定整改能否闭环",
|
||||
"CAPA 设计必须绑定责任人、证据和期限",
|
||||
"验证计划决定整改是否可被审计接受",
|
||||
"供应商和外包管理常是系统性缺口放大器",
|
||||
"数据完整性风险需要独立成章处理",
|
||||
"实施路线图需要平衡停线风险与合规风险",
|
||||
"管理层治理机制决定整改能否持续",
|
||||
],
|
||||
"cmc_process_risk": [
|
||||
"工艺流程图是识别放大风险的起点",
|
||||
"CQA 与 CPP 的映射决定控制策略质量",
|
||||
"放大过程的失效模式集中在传质、混合和稳定性",
|
||||
"分析方法和放行标准决定证据可信度",
|
||||
"技术转移风险来自知识隐性化和现场差异",
|
||||
"供应链约束会改变工艺控制边界",
|
||||
"偏差和变更管理决定商业化后的韧性",
|
||||
"监管沟通策略需要提前固化关键假设",
|
||||
"反方证据限定平台工艺可复制性",
|
||||
"CMC 路线图需要把风险转化为验证实验",
|
||||
],
|
||||
"rd_go_no_go": [
|
||||
"科学假设强度决定项目是否值得进入下一阶段",
|
||||
"POC 证据需要同时证明有效性和可转化性",
|
||||
"安全性窗口决定适应症与人群选择",
|
||||
"IP 与 FTO 风险决定商业化自由度",
|
||||
"开发路径需要把关键不确定性前置验证",
|
||||
"竞争窗口决定速度是否仍有战略价值",
|
||||
"CMC 与临床运营能力影响真实可行性",
|
||||
"反方证据决定 go/no-go 阈值",
|
||||
"投资强度应与证据成熟度匹配",
|
||||
"决策门槛需要形成可执行检查表",
|
||||
],
|
||||
"management_consulting": [
|
||||
"现状诊断需要区分症状、根因和约束条件",
|
||||
"能力差距决定组织改进优先级",
|
||||
"流程断点揭示跨部门协作成本",
|
||||
"治理结构决定决策速度和责任清晰度",
|
||||
"运营模型需要匹配战略目标而非照搬标杆",
|
||||
"数字化工具只有嵌入流程才产生价值",
|
||||
"绩效指标需要避免局部最优",
|
||||
"变革阻力本身是方案设计输入",
|
||||
"路线图需要把 quick wins 与系统建设分层",
|
||||
"落地机制决定咨询建议能否转化为成果",
|
||||
],
|
||||
"gmp_quality_operations_diagnosis": [
|
||||
"从审计清单到商业化阶段门",
|
||||
"用法规基线重新校准整改优先级",
|
||||
"制剂无菌保障：从硬件合规到行为受控",
|
||||
"原液与公用系统：封闭工艺背后的证据缺口",
|
||||
"工艺文件与验证：商业化转移的硬门槛",
|
||||
"质量系统闭环：偏差、变更、CAPA 与数据完整性",
|
||||
"人员能力：培训有效性比培训记录更关键",
|
||||
"运营节奏：从临时协调转向管理系统",
|
||||
"团队建设：按 CDMO 能力矩阵补齐角色",
|
||||
"整改路线图：立即纠偏、体系补强、能力建设",
|
||||
"管理层看板：用 CAPA 总表驱动复核",
|
||||
],
|
||||
}
|
||||
|
||||
|
||||
def _existing_chapter_titles(project_root: Path) -> list[str]:
|
||||
framework_path = project_root / "phase1" / "framework.md"
|
||||
if not framework_path.exists():
|
||||
return []
|
||||
from scripts.runtime.tasks import parse_framework_chapters
|
||||
|
||||
chapters = parse_framework_chapters(framework_path.read_text(encoding="utf-8"))
|
||||
return [chapter.title for chapter in chapters if chapter.title]
|
||||
|
||||
|
||||
def render_framework(
|
||||
project_root: Path,
|
||||
*,
|
||||
method_key: str | None = None,
|
||||
chapter_count: int = 10,
|
||||
preserve_existing_outline: bool = False,
|
||||
) -> Path:
|
||||
manifest = load_manifest(project_root)
|
||||
registry = ResearchMethodRegistry()
|
||||
method = registry.get(method_key or manifest.get("research_method"))
|
||||
if method_key:
|
||||
manifest["research_method"] = method.key
|
||||
existing_titles = _existing_chapter_titles(project_root) if preserve_existing_outline else []
|
||||
titles = existing_titles or CHAPTER_TEMPLATES.get(method.key) or CHAPTER_TEMPLATES["mckinsey_market"]
|
||||
chapter_count = max(8, min(15, chapter_count))
|
||||
selected = titles[:chapter_count] if not existing_titles else titles
|
||||
quota = max(800, int(manifest.get("target_words", 30000)) // len(selected))
|
||||
chapter_planning = build_chapter_planning(project_root, manifest, method, selected, quota=quota)
|
||||
sections = "\n".join(f"- {item}" for item in method.framework_sections)
|
||||
axes = "、".join(method.task_axes)
|
||||
material_text = render_material_inventory(manifest.get("material_inventory") or [])
|
||||
lines = [
|
||||
f"# {manifest.get('report_title') or manifest['topic']}：研究框架",
|
||||
"",
|
||||
f"research_method: {method.key}",
|
||||
f"method_name: {method.name}",
|
||||
f"work_language: 中文主写作；检索关键词、证据摘录、source title、raw notes 可保留英文。",
|
||||
f"target_words: {manifest.get('target_words', 30000)}",
|
||||
"",
|
||||
"## 方法选择",
|
||||
"",
|
||||
f"本项目采用 `{method.key}`，因为其结构原则是：{method.structure_principle}",
|
||||
"",
|
||||
"框架模块：",
|
||||
sections,
|
||||
"",
|
||||
"Phase 2 任务轴：",
|
||||
f"- {axes}",
|
||||
"",
|
||||
"## 输入材料与使用边界",
|
||||
"",
|
||||
material_text,
|
||||
"",
|
||||
"这些材料作为现场问题线索和内部事实起点使用；正式结论仍需结合 NMPA、FDA、EMA、ICH、WHO 等权威法规、指南和最佳实践进行验证。",
|
||||
"",
|
||||
"## 中心假设",
|
||||
"",
|
||||
_central_thesis(manifest, method),
|
||||
"",
|
||||
"Phase1 的职责是大胆假设：基于材料、访谈和初步搜索定下主基调、章节命题和求证路线。Phase2 的职责是小心求证：验证、证伪、补证，而不是重新发明报告方向。Phase3 则检查 Phase1 假设与 Phase2 证据是否自洽。",
|
||||
"",
|
||||
]
|
||||
for item in chapter_planning:
|
||||
lines.extend(
|
||||
[
|
||||
f"## 第{int(item['chapter_id'][2:])}章 {item['title']}",
|
||||
"",
|
||||
f"建议字数：约 {item['suggested_words']} 字。",
|
||||
f"本章要解决的问题：{item['core_question']}",
|
||||
f"大胆假设：{item['bold_hypothesis']}",
|
||||
f"写作主张：{item['writing_claim']}",
|
||||
f"证据线：{'；'.join(item['evidence_lanes'])}",
|
||||
"",
|
||||
"材料起点：",
|
||||
*[f"- {line}" for line in item["material_starting_points"]],
|
||||
"",
|
||||
"求证计划：",
|
||||
*[f"- {line}" for line in item["verification_plan"]],
|
||||
"",
|
||||
"必须寻找的反方/边界：",
|
||||
*[f"- {line}" for line in item["counter_evidence_to_seek"]],
|
||||
"",
|
||||
f"最低证据要求：`{json.dumps(item['minimum_evidence'], ensure_ascii=False)}`",
|
||||
"",
|
||||
]
|
||||
)
|
||||
lines.extend(
|
||||
[
|
||||
"## 暂停点",
|
||||
"",
|
||||
"请先确认 `phase1/material_brief.md` 的材料解读和访谈问题，再确认本框架后进入 Phase 2。若章节逻辑、方法框架或字数配额需要调整，应先修改本文件。",
|
||||
"",
|
||||
"确认后运行：`uv run python scripts/dr.py approve <project>`；未批准时 `research` 默认会拒绝推进，可用 `--force` 临时覆盖。",
|
||||
"",
|
||||
]
|
||||
)
|
||||
out = project_root / "phase1" / "framework.md"
|
||||
out.parent.mkdir(parents=True, exist_ok=True)
|
||||
out.write_text("\n".join(lines), encoding="utf-8")
|
||||
research_brief_md, research_brief_json = write_research_brief(project_root, manifest, method, chapter_planning=chapter_planning)
|
||||
manifest["phase1"] = {
|
||||
"status": "completed",
|
||||
"approved": False,
|
||||
"framework_path": "phase1/framework.md",
|
||||
"research_brief_path": str(research_brief_md.relative_to(project_root)),
|
||||
"research_brief_json_path": str(research_brief_json.relative_to(project_root)),
|
||||
"requires_user_interview": True,
|
||||
"research_method": method.key,
|
||||
"updated_at": utc_now_iso(),
|
||||
}
|
||||
manifest["updated_at"] = utc_now_iso()
|
||||
write_manifest(project_root, manifest)
|
||||
return out
|
||||
@@ -0,0 +1,328 @@
|
||||
"""Phase 3 review checks for the Python core."""
|
||||
|
||||
from __future__ import annotations
|
||||
|
||||
import json
|
||||
import re
|
||||
from datetime import datetime, timezone
|
||||
from pathlib import Path
|
||||
from typing import Any
|
||||
|
||||
from scripts.runtime.artifacts import load_manifest, write_manifest
|
||||
from scripts.runtime.tasks import validate_packet
|
||||
|
||||
|
||||
def utc_now_iso() -> str:
|
||||
return datetime.now(timezone.utc).replace(microsecond=0).isoformat()
|
||||
|
||||
|
||||
def _source_ids_from_jsonl(path: Path) -> set[str]:
|
||||
ids: set[str] = set()
|
||||
if not path.exists():
|
||||
return ids
|
||||
for line in path.read_text(encoding="utf-8").splitlines():
|
||||
if not line.strip():
|
||||
continue
|
||||
try:
|
||||
obj = json.loads(line)
|
||||
except json.JSONDecodeError:
|
||||
continue
|
||||
source_id = obj.get("id") or obj.get("source_id")
|
||||
if source_id:
|
||||
ids.add(str(source_id))
|
||||
return ids
|
||||
|
||||
|
||||
def _draft_citations(drafts: list[Path]) -> set[str]:
|
||||
cited: set[str] = set()
|
||||
for draft in drafts:
|
||||
cited.update(re.findall(r"\[(src_[A-Za-z0-9_-]+)\]", draft.read_text(encoding="utf-8")))
|
||||
return cited
|
||||
|
||||
|
||||
def _ready_packet_stems(project_root: Path) -> set[str]:
|
||||
ready: set[str] = set()
|
||||
for path in sorted((project_root / "phase2" / "packets").glob("*.json")):
|
||||
try:
|
||||
packet = json.loads(path.read_text(encoding="utf-8"))
|
||||
validate_packet(packet)
|
||||
except Exception:
|
||||
continue
|
||||
ready.add(path.stem)
|
||||
return ready
|
||||
|
||||
|
||||
def _read_text_if_exists(path: Path, *, max_chars: int | None = None) -> str:
|
||||
if not path.exists():
|
||||
return ""
|
||||
text = path.read_text(encoding="utf-8", errors="ignore")
|
||||
return text[:max_chars] if max_chars is not None else text
|
||||
|
||||
|
||||
def _json_if_exists(path: Path, *, max_chars: int | None = None) -> str:
|
||||
if not path.exists():
|
||||
return ""
|
||||
try:
|
||||
data = json.loads(path.read_text(encoding="utf-8"))
|
||||
text = json.dumps(data, ensure_ascii=False, indent=2)
|
||||
except Exception:
|
||||
text = path.read_text(encoding="utf-8", errors="ignore")
|
||||
return text[:max_chars] if max_chars is not None else text
|
||||
|
||||
|
||||
def _draft_quality_findings(drafts: list[Path]) -> list[dict[str, Any]]:
|
||||
findings: list[dict[str, Any]] = []
|
||||
generic_markers = [
|
||||
"需要进一步完善",
|
||||
"应当加强",
|
||||
"持续改进",
|
||||
"系统性",
|
||||
"闭环管理",
|
||||
"质量文化",
|
||||
]
|
||||
for draft in drafts:
|
||||
text = draft.read_text(encoding="utf-8")
|
||||
zh_chars = sum(1 for char in text if "\u4e00" <= char <= "\u9fff")
|
||||
citations = re.findall(r"\[(src_[A-Za-z0-9_-]+)\]", text)
|
||||
evidence_table_present = "证据落点" in text and "待补证据" in text
|
||||
if zh_chars >= 1200 and len(set(citations)) < 5:
|
||||
findings.append({"severity": "P1", "message": f"{draft.name} 引用来源过少，可能未充分使用 evidence packet。"})
|
||||
if zh_chars >= 1200 and not evidence_table_present:
|
||||
findings.append({"severity": "P1", "message": f"{draft.name} 缺少“证据落点与待补证据”小节，难以判断 evidence 是否真正落到纸面。"})
|
||||
generic_count = sum(text.count(marker) for marker in generic_markers)
|
||||
if zh_chars >= 1200 and generic_count >= 18:
|
||||
findings.append({"severity": "P1", "message": f"{draft.name} 泛化管理表述过多，需要回炉为具体审计发现、风险影响和整改动作。"})
|
||||
return findings
|
||||
|
||||
|
||||
def build_phase3_model_review_context(project_root: Path, *, max_chars: int = 650_000) -> str:
|
||||
"""Build a structured, bounded context packet for an independent model review."""
|
||||
deterministic_path = build_phase3_critique(project_root)
|
||||
deterministic_copy = project_root / "phase3" / "critique_deterministic.md"
|
||||
deterministic_copy.write_text(deterministic_path.read_text(encoding="utf-8"), encoding="utf-8")
|
||||
|
||||
manifest = load_manifest(project_root)
|
||||
parts: list[str] = [
|
||||
f"# Phase 3 Model Review Context: {manifest.get('topic', project_root.name)}",
|
||||
"",
|
||||
"## Review Contract",
|
||||
"",
|
||||
"- 这是给非 Codex 模型的独立总编审校上下文，不要求重写正文。",
|
||||
"- 请判断 Phase2 草稿能否进入 Phase4，或必须回炉补证据/重写。",
|
||||
"- 重点关注：证据是否落纸面、并发 packet 是否造成碎片化、法规/最佳实践覆盖是否足够、整改建议是否具体可执行。",
|
||||
"",
|
||||
"## Manifest",
|
||||
"",
|
||||
"```json",
|
||||
json.dumps(manifest, ensure_ascii=False, indent=2),
|
||||
"```",
|
||||
"",
|
||||
"## Deterministic Review Baseline",
|
||||
"",
|
||||
_read_text_if_exists(deterministic_copy),
|
||||
"",
|
||||
"## Phase 1 Framework",
|
||||
"",
|
||||
_read_text_if_exists(project_root / "phase1" / "framework.md", max_chars=50_000),
|
||||
"",
|
||||
"## Phase 1 Research Brief",
|
||||
"",
|
||||
_read_text_if_exists(project_root / "phase1" / "research_brief.md", max_chars=30_000),
|
||||
"",
|
||||
"## Phase 2 Brief Warnings",
|
||||
"",
|
||||
_json_if_exists(project_root / "phase2" / "brief_warnings.json", max_chars=30_000) or "无",
|
||||
"",
|
||||
"## Phase 2 Packet Errors",
|
||||
"",
|
||||
]
|
||||
errors = sorted((project_root / "phase2" / "packet_errors").glob("*.json"))
|
||||
if errors:
|
||||
for path in errors[:40]:
|
||||
parts.extend([f"### {path.name}", "", _json_if_exists(path, max_chars=2_000), ""])
|
||||
else:
|
||||
parts.append("无")
|
||||
|
||||
parts.extend(["", "## Source Registry Summary", ""])
|
||||
source_lines = []
|
||||
sources_path = project_root / "phase2" / "sources.jsonl"
|
||||
if sources_path.exists():
|
||||
for line in sources_path.read_text(encoding="utf-8").splitlines()[:260]:
|
||||
if not line.strip():
|
||||
continue
|
||||
try:
|
||||
source = json.loads(line)
|
||||
except json.JSONDecodeError:
|
||||
continue
|
||||
source_lines.append(
|
||||
"- {id} | {tier} | {title} | {url} | cached={cached}".format(
|
||||
id=source.get("id", ""),
|
||||
tier=source.get("tier", ""),
|
||||
title=str(source.get("title", ""))[:120],
|
||||
url=source.get("url", ""),
|
||||
cached=source.get("cached_text_path", ""),
|
||||
)
|
||||
)
|
||||
parts.append("\n".join(source_lines) or "无")
|
||||
|
||||
parts.extend(["", "## Compressed Findings", ""])
|
||||
for path in sorted((project_root / "phase2" / "compressed_findings").glob("ch*.json")):
|
||||
parts.extend([f"### {path.name}", "", "```json", _json_if_exists(path, max_chars=35_000), "```", ""])
|
||||
|
||||
parts.extend(["", "## Chapter Drafts", ""])
|
||||
for path in sorted((project_root / "phase2" / "drafts").glob("ch*.md")):
|
||||
parts.extend([f"### {path.name}", "", _read_text_if_exists(path, max_chars=55_000), ""])
|
||||
|
||||
context = "\n".join(parts)
|
||||
if len(context) > max_chars:
|
||||
context = context[:max_chars] + "\n\n[Context truncated by max_chars; review should flag if truncation limits confidence.]\n"
|
||||
|
||||
out = project_root / "phase3" / "review_context_opus_4_7.md"
|
||||
out.parent.mkdir(parents=True, exist_ok=True)
|
||||
out.write_text(context, encoding="utf-8")
|
||||
return context
|
||||
|
||||
|
||||
def phase3_model_review_system_prompt() -> str:
|
||||
return (
|
||||
"你是 Deep Research Phase 3 的独立总编审校模型，本次由 ZenMux Claude Opus 4.7 执行，用于避免 Codex/OpenAI 模型偏见。\n"
|
||||
"你的任务是审校，不是润色或重写。必须用中文输出，英文仅可保留 source title、URL、法规缩写和原文短摘录。\n"
|
||||
"请严格检查：1) 研究目标与 Phase1 框架是否契合；2) Phase2 并发 evidence packets 是否被章节真正吸收，还是造成碎片化；"
|
||||
"3) FDA/NMPA/EMA/ICH/WHO/EU GMP 等权威来源是否足以支撑关键判断；4) 用户材料是否被正确作为起点且被权威来源交叉验证；"
|
||||
"5) 运营管理与团队能力章节是否具体，不得泛泛咨询腔；6) CAPA 建议是否包含 owner、期限、关闭证据、QA verification、复核窗口和升级阈值；"
|
||||
"7) 引用链和 source_id 是否可追踪；8) 是否仍有明显 AI 味、中英文混杂或空泛表达。\n\n"
|
||||
"输出必须使用以下 Markdown 结构：\n"
|
||||
"# Phase 3 Opus 4.7 独立审校\n"
|
||||
"## 总体判定\n"
|
||||
"给出：通过 / 有条件通过 / 回炉 Phase2，并说明最核心理由。\n"
|
||||
"## P0/P1 阻断问题\n"
|
||||
"列出必须修复的问题；每条写明章节/文件、问题、为什么阻断、建议动作。\n"
|
||||
"## 章节级审校表\n"
|
||||
"用表格覆盖 ch01-ch11：主线质量、证据密度、法规覆盖、整改可执行性、是否需要回炉。\n"
|
||||
"## 证据与信源质量\n"
|
||||
"单独评价 FDA warning letters、ICH Q9/Q10、EU GMP Annex 1、本地缓存信源、第三方低质信源的使用情况。\n"
|
||||
"## 碎片化与叙事连贯性\n"
|
||||
"判断并发研究是否造成割裂，并给出具体整合建议。\n"
|
||||
"## Phase2 回炉任务清单\n"
|
||||
"如果需要回炉，列出可执行任务卡级别的补证据/重写要求。\n"
|
||||
"## Phase4 准入条件\n"
|
||||
"明确进入 final 前必须满足的条件。\n"
|
||||
)
|
||||
|
||||
|
||||
def build_phase3_model_critique(
|
||||
project_root: Path,
|
||||
*,
|
||||
client: Any,
|
||||
model: str = "zenmux-anthropic/claude-opus-4-7",
|
||||
max_context_chars: int = 650_000,
|
||||
) -> Path:
|
||||
context = build_phase3_model_review_context(project_root, max_chars=max_context_chars)
|
||||
content = client.chat_complete(
|
||||
model=model,
|
||||
system=phase3_model_review_system_prompt(),
|
||||
user=context,
|
||||
temperature=0.2,
|
||||
max_tokens=20_000,
|
||||
tag="phase3:opus-review",
|
||||
)
|
||||
out = project_root / "phase3" / "critique.md"
|
||||
out.parent.mkdir(parents=True, exist_ok=True)
|
||||
out.write_text(content.rstrip() + "\n", encoding="utf-8")
|
||||
|
||||
manifest = load_manifest(project_root)
|
||||
phase3 = manifest.setdefault("phase3", {})
|
||||
phase3.update(
|
||||
{
|
||||
"status": "completed",
|
||||
"review_mode": "model",
|
||||
"review_model": model,
|
||||
"critique_path": "phase3/critique.md",
|
||||
"context_path": "phase3/review_context_opus_4_7.md",
|
||||
"deterministic_critique_path": "phase3/critique_deterministic.md",
|
||||
"updated_at": utc_now_iso(),
|
||||
}
|
||||
)
|
||||
manifest["updated_at"] = utc_now_iso()
|
||||
write_manifest(project_root, manifest)
|
||||
return out
|
||||
|
||||
|
||||
def build_phase3_critique(project_root: Path) -> Path:
|
||||
manifest = load_manifest(project_root)
|
||||
drafts = sorted((project_root / "phase2" / "drafts").glob("ch*.md"))
|
||||
packets = sorted((project_root / "phase2" / "packets").glob("*.json"))
|
||||
ready_stems = _ready_packet_stems(project_root)
|
||||
packet_errors = [
|
||||
path for path in sorted((project_root / "phase2" / "packet_errors").glob("*.json"))
|
||||
if path.stem not in ready_stems
|
||||
]
|
||||
chapter_errors = sorted((project_root / "phase2" / "chapter_errors").glob("*.json"))
|
||||
sources = _source_ids_from_jsonl(project_root / "phase2" / "sources.jsonl")
|
||||
cited = _draft_citations(drafts)
|
||||
missing_sources = sorted(cited - sources) if sources else sorted(cited)
|
||||
uncited_sources = sorted(sources - cited) if cited else sorted(sources)
|
||||
|
||||
findings: list[dict[str, Any]] = []
|
||||
if not drafts:
|
||||
findings.append({"severity": "P1", "message": "Phase 2 drafts 缺失，尚不能进入 Phase 4 成稿。"})
|
||||
if packet_errors:
|
||||
findings.append({"severity": "P1", "message": f"存在 {len(packet_errors)} 个 packet 失败，需要回炉补证据。"})
|
||||
if chapter_errors:
|
||||
findings.append({"severity": "P1", "message": f"存在 {len(chapter_errors)} 个章节组装失败，需要修复引用或重写该章。"})
|
||||
quality_holds = manifest.get("quality_holds") or []
|
||||
if quality_holds:
|
||||
findings.append({"severity": "P1", "message": "存在质量暂停标记：" + ", ".join(quality_holds)})
|
||||
findings.extend(_draft_quality_findings(drafts))
|
||||
if missing_sources:
|
||||
findings.append({"severity": "P1", "message": f"正文引用未在 sources.jsonl 中登记：{', '.join(missing_sources)}"})
|
||||
if not findings:
|
||||
findings.append({"severity": "P2", "message": "基础产物完整；仍需人工或大上下文模型审校逻辑链、反方证据和章节叙事。"})
|
||||
|
||||
lines = [
|
||||
"# Phase 3 审校 critique",
|
||||
"",
|
||||
f"- 项目：{manifest.get('topic', project_root.name)}",
|
||||
f"- 运行时：python-core-v0.20",
|
||||
f"- drafts：{len(drafts)}",
|
||||
f"- packets：{len(packets)}",
|
||||
f"- sources：{len(sources)}",
|
||||
f"- cited_source_ids：{', '.join(sorted(cited)) if cited else '无'}",
|
||||
"",
|
||||
"## Findings",
|
||||
"",
|
||||
]
|
||||
for item in findings:
|
||||
lines.append(f"- [{item['severity']}] {item['message']}")
|
||||
lines.extend(
|
||||
[
|
||||
"",
|
||||
"## Residual Risks",
|
||||
"",
|
||||
"- 本 deterministic review 只做结构、引用和错误包检查；深层逻辑审校仍建议交给 `phase3_review` 角色执行。",
|
||||
"- 若 sources 为空，本审校会把所有正文引用视为待登记来源。",
|
||||
"",
|
||||
"## Next",
|
||||
"",
|
||||
"- 若存在 P1，先回到 Phase 2 修复 packet/chapter 错误。",
|
||||
"- 若仅有 P2，可进入 `dr.py finalize` 的中文原生成稿路径。",
|
||||
"",
|
||||
]
|
||||
)
|
||||
out = project_root / "phase3" / "critique.md"
|
||||
out.parent.mkdir(parents=True, exist_ok=True)
|
||||
out.write_text("\n".join(lines), encoding="utf-8")
|
||||
phase3 = manifest.setdefault("phase3", {})
|
||||
phase3.update(
|
||||
{
|
||||
"status": "completed",
|
||||
"critique_path": "phase3/critique.md",
|
||||
"findings_total": len(findings),
|
||||
"missing_sources": missing_sources,
|
||||
"uncited_sources": uncited_sources,
|
||||
"updated_at": utc_now_iso(),
|
||||
}
|
||||
)
|
||||
manifest["updated_at"] = utc_now_iso()
|
||||
write_manifest(project_root, manifest)
|
||||
return out
|
||||
@@ -0,0 +1,149 @@
|
||||
"""Runtime role and task-model resolution."""
|
||||
|
||||
from __future__ import annotations
|
||||
|
||||
from dataclasses import dataclass
|
||||
|
||||
from scripts.lib.model_config import resolve_model_profile
|
||||
|
||||
|
||||
ROLE_DEFAULTS = {
|
||||
"dr_plan": {
|
||||
"skills": ["document-ingest", "search-gateway", "search-strategy", "source-quality", "length-budget", "mckinsey-method"],
|
||||
"temperature": 0.4,
|
||||
"max_tokens": 12000,
|
||||
"max_concurrency": 1,
|
||||
},
|
||||
"dr_pm": {
|
||||
"skills": ["length-budget", "evidence-table", "mckinsey-method"],
|
||||
"temperature": 0.2,
|
||||
"max_tokens": 8000,
|
||||
"max_concurrency": 1,
|
||||
},
|
||||
"dr_searcher": {
|
||||
"skills": ["search-gateway", "search-strategy", "source-quality"],
|
||||
"temperature": 0.1,
|
||||
"max_tokens": 6000,
|
||||
"max_concurrency": 6,
|
||||
},
|
||||
"dr_analyst": {
|
||||
"skills": ["search-gateway", "search-strategy", "source-quality", "evidence-table", "mckinsey-method"],
|
||||
"temperature": 0.3,
|
||||
"max_tokens": 14000,
|
||||
"max_concurrency": 6,
|
||||
},
|
||||
"dr_verifier": {
|
||||
"skills": ["search-gateway", "search-strategy", "source-quality", "evidence-table"],
|
||||
"temperature": 0.2,
|
||||
"max_tokens": 10000,
|
||||
"max_concurrency": 4,
|
||||
},
|
||||
"dr_chief_editor": {
|
||||
"skills": ["mckinsey-method", "evidence-table", "output-hygiene"],
|
||||
"temperature": 0.2,
|
||||
"max_tokens": 16000,
|
||||
"max_concurrency": 1,
|
||||
},
|
||||
"dr_editor_in_chief": {
|
||||
"skills": ["mckinsey-method", "citation-manager", "humanizer-cn", "output-hygiene"],
|
||||
"temperature": 0.4,
|
||||
"max_tokens": 20000,
|
||||
"max_concurrency": 1,
|
||||
},
|
||||
"dr_reporter": {
|
||||
"skills": ["pdf-reportlab", "citation-manager", "output-hygiene"],
|
||||
"temperature": 0.1,
|
||||
"max_tokens": 6000,
|
||||
"max_concurrency": 1,
|
||||
},
|
||||
}
|
||||
|
||||
|
||||
ROLE_IDENTITIES = {
|
||||
"dr_plan": (
|
||||
"你是 Deep Research 的 Phase1 研究架构师。你的工作不是列目录，而是先消化材料、访谈和初步搜索，"
|
||||
"形成可被证伪的主判断、章节命题和求证路线。你要大胆假设，但必须给 Phase2 留下清晰的验证和推翻条件。"
|
||||
),
|
||||
"dr_pm": (
|
||||
"你是 Deep Research 的研究项目经理。你的职责是把研究意图转化为可并发执行、可回收校验的任务，"
|
||||
"控制碎片化、重复检索和上下文污染。"
|
||||
),
|
||||
"dr_searcher": (
|
||||
"你是 Deep Research 的信源发现员。你的职责是用短英文关键词和轴向词找到高质量入口，"
|
||||
"优先官方、法规、学术和一手材料；你不写结论，只交付可追溯来源。"
|
||||
),
|
||||
"dr_analyst": (
|
||||
"你是 Deep Research 的章节证据分析师。你的职责不是写一篇像样的空泛文章，而是围绕 Phase1 命题"
|
||||
"小心求证：提取材料原文、检索权威证据、寻找反方边界，并把证据整理成可审计的结构化 packet。"
|
||||
),
|
||||
"dr_verifier": (
|
||||
"你是 Deep Research 的独立反方审校员。你的默认姿态是质疑：找证据缺口、适用边界、反例和过度推断，"
|
||||
"并指出哪些结论必须降级或回炉。"
|
||||
),
|
||||
"dr_chief_editor": (
|
||||
"你是 Deep Research 的 Phase3 总编审校。你的职责是通读 Phase1 假设与 Phase2 证据，判断二者是否自洽，"
|
||||
"优先指出结构性失败、证据不足和需要回炉的章节。"
|
||||
),
|
||||
"dr_editor_in_chief": (
|
||||
"你是 Deep Research 的终稿主编。你的职责是把已验证证据组织成客户可读的中文报告，"
|
||||
"保持观点清晰、证据密实、表达克制，避免翻译腔和 AI 味。"
|
||||
),
|
||||
"dr_reporter": (
|
||||
"你是 Deep Research 的报告制作负责人。你的职责是把已定稿内容可靠渲染为 PDF/DOCX，"
|
||||
"确保引用、排版、中文字体、表格和输出卫生可交付。"
|
||||
),
|
||||
}
|
||||
|
||||
|
||||
@dataclass(frozen=True)
|
||||
class RoleDefinition:
|
||||
name: str
|
||||
model: str
|
||||
skills: list[str]
|
||||
temperature: float
|
||||
max_tokens: int
|
||||
max_concurrency: int
|
||||
identity: str = ""
|
||||
|
||||
|
||||
class RuntimeProfile:
|
||||
def __init__(self, *, profile: str, roles: dict[str, RoleDefinition], task_types: dict[str, str]) -> None:
|
||||
self.profile = profile
|
||||
self.roles = roles
|
||||
self.task_types = task_types
|
||||
|
||||
def role_for_task(self, task_type: str) -> RoleDefinition:
|
||||
role_name = self.task_types.get(task_type)
|
||||
if not role_name:
|
||||
raise KeyError(f"unknown task_type: {task_type}")
|
||||
if role_name not in self.roles:
|
||||
raise KeyError(f"task_type {task_type} maps to missing role {role_name}")
|
||||
return self.roles[role_name]
|
||||
|
||||
|
||||
def resolve_runtime_profile(
|
||||
*,
|
||||
profile: str | None = None,
|
||||
overrides: dict[str, str] | None = None,
|
||||
) -> RuntimeProfile:
|
||||
resolved = resolve_model_profile(profile=profile, overrides=overrides)
|
||||
role_models = resolved["roles"]
|
||||
roles: dict[str, RoleDefinition] = {}
|
||||
for name, defaults in ROLE_DEFAULTS.items():
|
||||
model = role_models.get(name)
|
||||
if not model:
|
||||
continue
|
||||
roles[name] = RoleDefinition(
|
||||
name=name,
|
||||
model=model,
|
||||
skills=list(defaults["skills"]),
|
||||
temperature=float(defaults["temperature"]),
|
||||
max_tokens=int(defaults["max_tokens"]),
|
||||
max_concurrency=int(defaults["max_concurrency"]),
|
||||
identity=ROLE_IDENTITIES.get(name, ""),
|
||||
)
|
||||
return RuntimeProfile(
|
||||
profile=resolved["profile"],
|
||||
roles=roles,
|
||||
task_types=dict(resolved.get("task_types") or {}),
|
||||
)
|
||||
@@ -0,0 +1,108 @@
|
||||
"""Canonical skill registry and adapter sync helpers."""
|
||||
|
||||
from __future__ import annotations
|
||||
|
||||
import shutil
|
||||
from dataclasses import dataclass
|
||||
from pathlib import Path
|
||||
|
||||
|
||||
REPO_ROOT = Path(__file__).resolve().parents[2]
|
||||
CANONICAL_SKILLS_DIR = REPO_ROOT / ".agents" / "skills"
|
||||
PROJECT_SKILLS_DIR = REPO_ROOT / "skills"
|
||||
REQUIRED_SKILLS = {
|
||||
"search-strategy",
|
||||
"search-gateway",
|
||||
"source-quality",
|
||||
"length-budget",
|
||||
"evidence-table",
|
||||
"citation-manager",
|
||||
"mckinsey-method",
|
||||
}
|
||||
|
||||
|
||||
@dataclass(frozen=True)
|
||||
class SkillInfo:
|
||||
name: str
|
||||
path: Path
|
||||
|
||||
|
||||
class SkillRegistry:
|
||||
"""Reads skills from the canonical cross-adapter registry."""
|
||||
|
||||
def __init__(self, canonical_dir: Path | None = None) -> None:
|
||||
self.canonical_dir = canonical_dir or CANONICAL_SKILLS_DIR
|
||||
|
||||
def roots(self) -> list[Path]:
|
||||
roots = []
|
||||
if self.canonical_dir == CANONICAL_SKILLS_DIR and PROJECT_SKILLS_DIR.exists():
|
||||
roots.append(PROJECT_SKILLS_DIR)
|
||||
roots.append(self.canonical_dir)
|
||||
return roots
|
||||
|
||||
def list(self) -> list[SkillInfo]:
|
||||
seen: set[str] = set()
|
||||
out: list[SkillInfo] = []
|
||||
for root in self.roots():
|
||||
if not root.exists():
|
||||
continue
|
||||
for path in sorted(root.glob("*/SKILL.md")):
|
||||
name = path.parent.name
|
||||
if name in seen:
|
||||
continue
|
||||
seen.add(name)
|
||||
out.append(SkillInfo(name=name, path=path))
|
||||
return out
|
||||
|
||||
def list_names(self) -> list[str]:
|
||||
return [item.name for item in self.list()]
|
||||
|
||||
def read(self, name: str) -> str:
|
||||
for root in self.roots():
|
||||
path = root / name / "SKILL.md"
|
||||
if path.exists():
|
||||
return path.read_text(encoding="utf-8")
|
||||
raise FileNotFoundError(f"skill not found: {name}")
|
||||
|
||||
def validate(self, required: set[str] | None = None) -> dict[str, object]:
|
||||
names = set(self.list_names())
|
||||
required_names = required or REQUIRED_SKILLS
|
||||
missing = sorted(required_names - names)
|
||||
malformed: list[str] = []
|
||||
for item in self.list():
|
||||
text = item.path.read_text(encoding="utf-8")
|
||||
if "name:" not in text[:300]:
|
||||
malformed.append(item.name)
|
||||
return {
|
||||
"ok": not missing and not malformed,
|
||||
"canonical_dir": str(self.canonical_dir),
|
||||
"count": len(names),
|
||||
"missing": missing,
|
||||
"malformed": malformed,
|
||||
}
|
||||
|
||||
def sync_to(self, targets: list[Path], *, force: bool = True) -> int:
|
||||
"""Copy canonical skills into adapter skill directories.
|
||||
|
||||
Returns the number of skill directories copied across all targets.
|
||||
"""
|
||||
copied = 0
|
||||
for target in targets:
|
||||
if target.resolve() == self.canonical_dir.resolve():
|
||||
continue
|
||||
target.mkdir(parents=True, exist_ok=True)
|
||||
for item in self.list():
|
||||
dst = target / item.name
|
||||
if dst.exists() and force:
|
||||
shutil.rmtree(dst)
|
||||
if not dst.exists():
|
||||
shutil.copytree(item.path.parent, dst)
|
||||
copied += 1
|
||||
return copied
|
||||
|
||||
|
||||
def default_adapter_skill_dirs() -> list[Path]:
|
||||
return [
|
||||
REPO_ROOT / ".opencode" / "skills",
|
||||
REPO_ROOT / ".agents" / "skills",
|
||||
]
|
||||
@@ -0,0 +1,230 @@
|
||||
"""Cache important external sources as local Markdown snapshots."""
|
||||
|
||||
from __future__ import annotations
|
||||
|
||||
import hashlib
|
||||
import json
|
||||
import re
|
||||
from dataclasses import dataclass
|
||||
from pathlib import Path
|
||||
from urllib.parse import urlparse
|
||||
|
||||
import httpx
|
||||
from lxml import html
|
||||
|
||||
|
||||
IMPORTANT_DOMAINS = (
|
||||
"fda.gov",
|
||||
"ema.europa.eu",
|
||||
"nmpa.gov.cn",
|
||||
"cde.org.cn",
|
||||
"ich.org",
|
||||
"who.int",
|
||||
"edqm.eu",
|
||||
"pmda.go.jp",
|
||||
"ec.europa.eu",
|
||||
"health.ec.europa.eu",
|
||||
)
|
||||
|
||||
|
||||
@dataclass(frozen=True)
|
||||
class CacheResult:
|
||||
source_id: str
|
||||
url: str
|
||||
cached_text_path: str
|
||||
raw_path: str
|
||||
status: str
|
||||
chars: int
|
||||
|
||||
|
||||
def _safe_stem(source: dict) -> str:
|
||||
source_id = str(source.get("id") or "source")
|
||||
digest = hashlib.sha1(str(source.get("url") or source_id).encode("utf-8")).hexdigest()[:10]
|
||||
safe_id = re.sub(r"[^A-Za-z0-9_-]+", "_", source_id).strip("_") or "source"
|
||||
return f"{safe_id}-{digest}"
|
||||
|
||||
|
||||
def _domain(url: str) -> str:
|
||||
return urlparse(url).netloc.lower()
|
||||
|
||||
|
||||
def is_important_source(source: dict) -> bool:
|
||||
url = str(source.get("url") or "")
|
||||
if not url.startswith(("http://", "https://")):
|
||||
return False
|
||||
domain = _domain(url)
|
||||
if any(domain.endswith(item) for item in IMPORTANT_DOMAINS):
|
||||
return True
|
||||
tier = str(source.get("tier") or "").lower()
|
||||
if "tier 1" in tier or tier in {"1", "1.0"}:
|
||||
return True
|
||||
title = str(source.get("title") or "").lower()
|
||||
return any(term in title for term in ("ich q9", "ich q10", "annex 1", "fda guidance", "who guideline"))
|
||||
|
||||
|
||||
def load_sources(path: Path) -> list[dict]:
|
||||
if not path.exists():
|
||||
return []
|
||||
rows: list[dict] = []
|
||||
for line in path.read_text(encoding="utf-8").splitlines():
|
||||
if not line.strip():
|
||||
continue
|
||||
rows.append(json.loads(line))
|
||||
return rows
|
||||
|
||||
|
||||
def write_sources(path: Path, rows: list[dict]) -> None:
|
||||
path.write_text("".join(json.dumps(row, ensure_ascii=False) + "\n" for row in rows), encoding="utf-8")
|
||||
|
||||
|
||||
def _response_ext(url: str, content_type: str) -> str:
|
||||
lowered = url.lower()
|
||||
if "pdf" in content_type or lowered.endswith(".pdf"):
|
||||
return ".pdf"
|
||||
if "html" in content_type or lowered.endswith((".html", ".htm", "/")):
|
||||
return ".html"
|
||||
return ".bin"
|
||||
|
||||
|
||||
def _html_to_text(content: bytes) -> str:
|
||||
doc = html.fromstring(content)
|
||||
for bad in doc.xpath("//script|//style|//noscript"):
|
||||
bad.drop_tree()
|
||||
return "\n".join(line.strip() for line in doc.text_content().splitlines() if line.strip())
|
||||
|
||||
|
||||
def _pdf_to_text(path: Path) -> str:
|
||||
try:
|
||||
import fitz
|
||||
except Exception:
|
||||
return ""
|
||||
doc = fitz.open(path)
|
||||
parts: list[str] = []
|
||||
for index, page in enumerate(doc, start=1):
|
||||
text = page.get_text("text").strip()
|
||||
if text:
|
||||
parts.append(f"## Page {index}\n\n{text}")
|
||||
return "\n\n".join(parts)
|
||||
|
||||
|
||||
def _bytes_to_text(*, raw_path: Path, content: bytes, content_type: str, url: str) -> str:
|
||||
if raw_path.suffix == ".pdf" or "pdf" in content_type or url.lower().endswith(".pdf"):
|
||||
return _pdf_to_text(raw_path)
|
||||
if raw_path.suffix in {".html", ".htm"} or "html" in content_type:
|
||||
return _html_to_text(content)
|
||||
try:
|
||||
return content.decode("utf-8")
|
||||
except UnicodeDecodeError:
|
||||
return content.decode("utf-8", errors="ignore")
|
||||
|
||||
|
||||
def cache_source(
|
||||
project_root: Path,
|
||||
source: dict,
|
||||
*,
|
||||
client: httpx.Client | None = None,
|
||||
force: bool = False,
|
||||
timeout: float = 45.0,
|
||||
) -> CacheResult:
|
||||
url = str(source.get("url") or "")
|
||||
if not url.startswith(("http://", "https://")):
|
||||
raise ValueError(f"source URL is not remote: {url}")
|
||||
cache_dir = project_root / "phase2" / "source_cache"
|
||||
raw_dir = cache_dir / "raw"
|
||||
text_dir = cache_dir / "md"
|
||||
raw_dir.mkdir(parents=True, exist_ok=True)
|
||||
text_dir.mkdir(parents=True, exist_ok=True)
|
||||
|
||||
stem = _safe_stem(source)
|
||||
md_path = text_dir / f"{stem}.md"
|
||||
if md_path.exists() and not force:
|
||||
return CacheResult(
|
||||
source_id=str(source.get("id") or ""),
|
||||
url=url,
|
||||
cached_text_path=str(md_path.relative_to(project_root)),
|
||||
raw_path=str(source.get("cached_raw_path") or ""),
|
||||
status="cached",
|
||||
chars=len(md_path.read_text(encoding="utf-8")),
|
||||
)
|
||||
|
||||
owns_client = client is None
|
||||
http = client or httpx.Client(trust_env=False, follow_redirects=True, timeout=timeout)
|
||||
try:
|
||||
response = http.get(url)
|
||||
response.raise_for_status()
|
||||
content_type = response.headers.get("content-type", "").lower()
|
||||
ext = _response_ext(str(response.url), content_type)
|
||||
raw_path = raw_dir / f"{stem}{ext}"
|
||||
raw_path.write_bytes(response.content)
|
||||
text = _bytes_to_text(raw_path=raw_path, content=response.content, content_type=content_type, url=str(response.url))
|
||||
lines = [
|
||||
f"# Source Snapshot: {source.get('title') or source.get('id') or url}",
|
||||
"",
|
||||
f"- source_id: {source.get('id', '')}",
|
||||
f"- original_url: {url}",
|
||||
f"- fetched_url: {response.url}",
|
||||
f"- content_type: {content_type}",
|
||||
f"- raw_path: {raw_path.relative_to(project_root)}",
|
||||
"",
|
||||
"## Extracted Text",
|
||||
"",
|
||||
text.strip() or "[No extractable text. Keep raw file for manual review.]",
|
||||
"",
|
||||
]
|
||||
md_path.write_text("\n".join(lines), encoding="utf-8")
|
||||
return CacheResult(
|
||||
source_id=str(source.get("id") or ""),
|
||||
url=url,
|
||||
cached_text_path=str(md_path.relative_to(project_root)),
|
||||
raw_path=str(raw_path.relative_to(project_root)),
|
||||
status="fetched",
|
||||
chars=len(text),
|
||||
)
|
||||
finally:
|
||||
if owns_client:
|
||||
http.close()
|
||||
|
||||
|
||||
def cache_sources(
|
||||
project_root: Path,
|
||||
*,
|
||||
sources_rel: str = "phase2/sources.jsonl",
|
||||
important_only: bool = True,
|
||||
limit: int | None = None,
|
||||
force: bool = False,
|
||||
) -> list[CacheResult]:
|
||||
sources_path = project_root / sources_rel
|
||||
rows = load_sources(sources_path)
|
||||
results: list[CacheResult] = []
|
||||
selected_indexes = [
|
||||
index
|
||||
for index, row in enumerate(rows)
|
||||
if row.get("url")
|
||||
and (not row.get("cached_text_path") or force)
|
||||
and (not important_only or is_important_source(row))
|
||||
]
|
||||
if limit is not None:
|
||||
selected_indexes = selected_indexes[:limit]
|
||||
|
||||
with httpx.Client(trust_env=False, follow_redirects=True, timeout=45.0) as client:
|
||||
for index in selected_indexes:
|
||||
row = rows[index]
|
||||
try:
|
||||
result = cache_source(project_root, row, client=client, force=force)
|
||||
except Exception as exc:
|
||||
row["cache_status"] = "failed"
|
||||
row["cache_error"] = str(exc)[:300]
|
||||
continue
|
||||
row["cached_text_path"] = result.cached_text_path
|
||||
row["cached_raw_path"] = result.raw_path
|
||||
row["cache_status"] = result.status
|
||||
row["cached_text_chars"] = result.chars
|
||||
results.append(result)
|
||||
write_sources(sources_path, rows)
|
||||
manifest = project_root / "phase2" / "source_cache" / "manifest.json"
|
||||
manifest.parent.mkdir(parents=True, exist_ok=True)
|
||||
manifest.write_text(
|
||||
json.dumps([result.__dict__ for result in results], ensure_ascii=False, indent=2) + "\n",
|
||||
encoding="utf-8",
|
||||
)
|
||||
return results
|
||||
@@ -0,0 +1,83 @@
|
||||
"""Source registry helpers for Phase 2 packets."""
|
||||
|
||||
from __future__ import annotations
|
||||
|
||||
import json
|
||||
from pathlib import Path
|
||||
from typing import Any
|
||||
|
||||
|
||||
def _source_key(source: dict[str, Any]) -> str:
|
||||
return (source.get("id") or source.get("source_id") or source.get("doi") or source.get("url") or "").strip()
|
||||
|
||||
|
||||
def append_packet_sources(sources_path: Path, packet: dict[str, Any]) -> int:
|
||||
"""Append packet sources to sources.jsonl, preserving every citeable source_id."""
|
||||
sources_path.parent.mkdir(parents=True, exist_ok=True)
|
||||
existing: set[str] = set()
|
||||
if sources_path.exists():
|
||||
for line in sources_path.read_text(encoding="utf-8").splitlines():
|
||||
if not line.strip():
|
||||
continue
|
||||
try:
|
||||
existing.add(_source_key(json.loads(line)))
|
||||
except json.JSONDecodeError:
|
||||
continue
|
||||
|
||||
written = 0
|
||||
with sources_path.open("a", encoding="utf-8") as f:
|
||||
for source in packet.get("sources") or []:
|
||||
key = _source_key(source)
|
||||
if not key or key in existing:
|
||||
continue
|
||||
existing.add(key)
|
||||
f.write(json.dumps(source, ensure_ascii=False) + "\n")
|
||||
written += 1
|
||||
return written
|
||||
|
||||
|
||||
def rebuild_sources_from_packets(project_root: Path) -> int:
|
||||
"""Rebuild phase2/sources.jsonl from packet-level source metadata.
|
||||
|
||||
The registry is keyed by source_id, not URL. Two packet sources may point to
|
||||
the same URL but have different source_ids already cited in drafts; dropping
|
||||
either row would break citation traceability.
|
||||
"""
|
||||
packets_dir = project_root / "phase2" / "packets"
|
||||
sources_path = project_root / "phase2" / "sources.jsonl"
|
||||
sources_path.parent.mkdir(parents=True, exist_ok=True)
|
||||
existing_by_key: dict[str, dict[str, Any]] = {}
|
||||
if sources_path.exists():
|
||||
for line in sources_path.read_text(encoding="utf-8").splitlines():
|
||||
if not line.strip():
|
||||
continue
|
||||
try:
|
||||
row = json.loads(line)
|
||||
except json.JSONDecodeError:
|
||||
continue
|
||||
key = _source_key(row)
|
||||
if key:
|
||||
existing_by_key[key] = row
|
||||
seen: set[str] = set()
|
||||
rows: list[dict[str, Any]] = []
|
||||
|
||||
for packet_path in sorted(packets_dir.glob("*.json")):
|
||||
try:
|
||||
packet = json.loads(packet_path.read_text(encoding="utf-8"))
|
||||
except json.JSONDecodeError:
|
||||
continue
|
||||
for source in packet.get("sources") or []:
|
||||
if not isinstance(source, dict):
|
||||
continue
|
||||
key = _source_key(source)
|
||||
if not key or key in seen:
|
||||
continue
|
||||
seen.add(key)
|
||||
previous = existing_by_key.get(key, {})
|
||||
rows.append({**source, **{k: v for k, v in previous.items() if k.startswith("cache") or k.startswith("cached_")}})
|
||||
|
||||
sources_path.write_text(
|
||||
"".join(json.dumps(row, ensure_ascii=False) + "\n" for row in rows),
|
||||
encoding="utf-8",
|
||||
)
|
||||
return len(rows)
|
||||
@@ -0,0 +1,472 @@
|
||||
"""Task-card and evidence-packet primitives for v0.20 Phase 2."""
|
||||
|
||||
from __future__ import annotations
|
||||
|
||||
import json
|
||||
import re
|
||||
from dataclasses import asdict, dataclass, field
|
||||
from pathlib import Path
|
||||
from typing import Any
|
||||
|
||||
from scripts.runtime.methods import ResearchMethod
|
||||
|
||||
|
||||
VALID_ROUTES = {"general", "evidence", "scholar", "patents", "news", "fda"}
|
||||
DEFAULT_AXES = ["literature", "regulatory", "patents", "market", "counter"]
|
||||
AXIS_ROUTES = {
|
||||
"literature": ["scholar", "evidence", "general"],
|
||||
"clinical": ["scholar", "evidence", "general"],
|
||||
"regulatory": ["fda", "evidence", "general", "news"],
|
||||
"patents": ["patents", "evidence", "general"],
|
||||
"market": ["news", "general"],
|
||||
"china": ["news", "general"],
|
||||
"counter": ["fda", "scholar", "evidence", "general"],
|
||||
"regulatory_gap": ["fda", "evidence", "general", "news"],
|
||||
"risk_classification": ["evidence", "general", "scholar"],
|
||||
"capa_design": ["evidence", "general", "news"],
|
||||
"ownership_timeline": ["general"],
|
||||
"verification_evidence": ["fda", "evidence", "general", "scholar"],
|
||||
"process_flow": ["scholar", "evidence", "general"],
|
||||
"cqa_cpp": ["scholar", "evidence", "general"],
|
||||
"scale_up_risk": ["scholar", "evidence", "general"],
|
||||
"control_strategy": ["scholar", "evidence", "general"],
|
||||
"supply_chain": ["news", "general"],
|
||||
"scientific_rationale": ["scholar", "evidence", "general"],
|
||||
"poc_evidence": ["scholar", "evidence", "general"],
|
||||
"ip_fto": ["patents", "evidence", "general"],
|
||||
"development_path": ["scholar", "evidence", "general"],
|
||||
"commercial_window": ["news", "general"],
|
||||
"current_state": ["evidence", "general"],
|
||||
"capability_gap": ["evidence", "general"],
|
||||
"operating_model": ["evidence", "general"],
|
||||
"governance": ["evidence", "general"],
|
||||
"implementation_roadmap": ["evidence", "general"],
|
||||
"nmpa_fda_ema_ich_who_baseline": ["fda", "evidence", "general", "news"],
|
||||
"quality_system_gap": ["fda", "evidence", "general"],
|
||||
"manufacturing_process_risk": ["fda", "scholar", "evidence", "general"],
|
||||
"operations_management_gap": ["fda", "evidence", "general"],
|
||||
"team_capability": ["evidence", "general", "news"],
|
||||
"capa_roadmap": ["fda", "evidence", "general"],
|
||||
"input_material_findings": ["evidence", "general"],
|
||||
"fda_enforcement_precedents": ["fda"],
|
||||
"chapter_integrated": ["fda", "scholar", "evidence", "general"],
|
||||
}
|
||||
|
||||
|
||||
@dataclass
|
||||
class Chapter:
|
||||
chapter_id: str
|
||||
index: int
|
||||
title: str
|
||||
notes: str = ""
|
||||
|
||||
|
||||
@dataclass
|
||||
class TaskCard:
|
||||
task_id: str
|
||||
chapter_ids: list[str]
|
||||
topic_axis: str
|
||||
questions: list[str]
|
||||
search_routes: list[str]
|
||||
output_packet: str
|
||||
chapter_title: str = ""
|
||||
preferred_model_role: str = "dr_analyst"
|
||||
status: str = "pending"
|
||||
dependencies: list[str] = field(default_factory=list)
|
||||
research_goal: str = ""
|
||||
research_method: str = ""
|
||||
prompt_brief: str = ""
|
||||
required_skills: list[str] = field(default_factory=list)
|
||||
allowed_materials: list[str] = field(default_factory=list)
|
||||
expected_evidence: dict[str, Any] = field(default_factory=dict)
|
||||
stop_conditions: list[str] = field(default_factory=list)
|
||||
model_hint: str = ""
|
||||
|
||||
def to_dict(self) -> dict[str, Any]:
|
||||
return asdict(self)
|
||||
|
||||
|
||||
def parse_framework_chapters(framework_text: str) -> list[Chapter]:
|
||||
"""Extract Chinese or English chapter headings from a framework markdown."""
|
||||
lines = framework_text.splitlines()
|
||||
chapters: list[Chapter] = []
|
||||
current: Chapter | None = None
|
||||
note_lines: list[str] = []
|
||||
heading_re = re.compile(
|
||||
r"^#{1,3}\s*(?:第\s*)?(\d{1,2})\s*(?:章|[.)、:-])?\s*(.+?)\s*$",
|
||||
re.IGNORECASE,
|
||||
)
|
||||
english_re = re.compile(r"^#{1,3}\s*chapter\s+(\d{1,2})[:.)\s-]+(.+?)\s*$", re.IGNORECASE)
|
||||
for line in lines:
|
||||
match = heading_re.match(line.strip()) or english_re.match(line.strip())
|
||||
if match:
|
||||
if current:
|
||||
current.notes = "\n".join(note_lines).strip()
|
||||
chapters.append(current)
|
||||
index = int(match.group(1))
|
||||
title = match.group(2).strip(" #")
|
||||
current = Chapter(chapter_id=f"ch{index:02d}", index=index, title=title)
|
||||
note_lines = []
|
||||
elif current:
|
||||
note_lines.append(line)
|
||||
if current:
|
||||
current.notes = "\n".join(note_lines).strip()
|
||||
chapters.append(current)
|
||||
return chapters
|
||||
|
||||
|
||||
def _questions_for_axis(chapter: Chapter, axis: str, method: ResearchMethod | None = None) -> list[str]:
|
||||
if axis == "chapter_integrated":
|
||||
lanes = "；".join(method.integrated_lanes if method else [])
|
||||
return [
|
||||
f"围绕《{chapter.title}》形成章节级综合证据包，不再拆成孤立小轴。",
|
||||
f"必须按当前 research_method 的 evidence lanes 组织证据：{lanes or '本地材料、权威来源、反方证据、可执行建议'}。",
|
||||
"若项目有用户材料，必须先读取本地材料证据并提取原文；再用本方法适用的权威来源交叉验证。",
|
||||
"必须形成：材料/事实基线、外部权威证据、差距或机会判断、反方/限制条件、可执行建议和待补证据。",
|
||||
]
|
||||
questions = [
|
||||
f"围绕《{chapter.title}》从 {axis} 角度提炼可证伪的核心结论。",
|
||||
"至少寻找两个 Tier 1-2 来源支撑主要结论；不足时标注待验证。",
|
||||
"主动检索反方证据、限制条件或失败案例。",
|
||||
]
|
||||
if axis in {
|
||||
"nmpa_fda_ema_ich_who_baseline",
|
||||
"quality_system_gap",
|
||||
"manufacturing_process_risk",
|
||||
"operations_management_gap",
|
||||
"capa_roadmap",
|
||||
"verification_evidence",
|
||||
"counter",
|
||||
"fda_enforcement_precedents",
|
||||
}:
|
||||
questions.append(
|
||||
"必须检索并优先评估 FDA Warning Letters、inspection/enforcement 页面、会议纪要或 meeting materials，作为 GMP 缺陷严重度和整改优先级的佐证。"
|
||||
)
|
||||
return questions
|
||||
|
||||
|
||||
def _default_required_skills(axis: str) -> list[str]:
|
||||
skills = ["search-gateway", "search-strategy", "source-quality", "evidence-table"]
|
||||
if axis == "counter":
|
||||
skills.append("mckinsey-method")
|
||||
return skills
|
||||
|
||||
|
||||
def _default_expected_evidence(axis: str) -> dict[str, Any]:
|
||||
expected = {
|
||||
"min_tier_1_2_sources": 2,
|
||||
"must_include_counter_evidence": True,
|
||||
"must_include_source_metadata": True,
|
||||
"preferred_evidence_types": [
|
||||
"regulatory_or_best_practice_requirement",
|
||||
"fda_warning_letter_or_meeting_record",
|
||||
"site_or_material_finding",
|
||||
"quantitative_fact_or_record",
|
||||
"implementation_or_verification_evidence",
|
||||
],
|
||||
"axis": axis,
|
||||
}
|
||||
if axis == "chapter_integrated":
|
||||
expected.update(
|
||||
{
|
||||
"min_local_material_evidence": 2,
|
||||
"min_official_sources": 2,
|
||||
"min_fda_or_regulatory_precedents": 1,
|
||||
"min_capa_actions": 3,
|
||||
"preferred_evidence_types": [
|
||||
"local_audit_or_recap_quote",
|
||||
"official_regulatory_requirement",
|
||||
"fda_warning_letter_or_meeting_record",
|
||||
"gap_analysis",
|
||||
"capa_action_with_owner_and_verification",
|
||||
"counter_evidence_or_boundary_condition",
|
||||
],
|
||||
}
|
||||
)
|
||||
return expected
|
||||
|
||||
|
||||
def _default_stop_conditions() -> list[str]:
|
||||
return [
|
||||
"已形成至少 3 条可追溯 evidence_items，且每条关键 claim 有 source_id。",
|
||||
"已主动记录 counter_evidence 或明确说明未找到反方证据的检索路径。",
|
||||
"candidate_sources 不足以支撑结论时停止写作，并把缺口写入 open_questions。",
|
||||
]
|
||||
|
||||
|
||||
def _integrated_prompt_brief(chapter: Chapter, method: ResearchMethod | None) -> str:
|
||||
lanes = "；".join(method.integrated_lanes if method else [])
|
||||
return (
|
||||
f"本任务是《{chapter.title}》的章节级综合证据包。不要把多条窄轴 packet 机械拼贴；"
|
||||
f"必须围绕当前研究方法的 lanes 一次性收束主线：{lanes or '事实材料、权威证据、反方证据、行动建议'}。"
|
||||
"输出必须让章节作者能直接写出判断、证据落点和可执行建议。"
|
||||
)
|
||||
|
||||
|
||||
def _task_card_for_chapter_axis(
|
||||
*,
|
||||
chapter: Chapter,
|
||||
axis: str,
|
||||
routes: list[str],
|
||||
method_key: str,
|
||||
required_skills: list[str] | None = None,
|
||||
allowed_materials: list[str] | None = None,
|
||||
prompt_brief: str | None = None,
|
||||
questions: list[str] | None = None,
|
||||
research_goal: str | None = None,
|
||||
expected_evidence: dict[str, Any] | None = None,
|
||||
stop_conditions: list[str] | None = None,
|
||||
method: ResearchMethod | None = None,
|
||||
) -> TaskCard:
|
||||
return TaskCard(
|
||||
task_id=f"{chapter.chapter_id}-{axis}",
|
||||
chapter_ids=[chapter.chapter_id],
|
||||
topic_axis=axis,
|
||||
questions=questions or _questions_for_axis(chapter, axis, method),
|
||||
search_routes=routes,
|
||||
output_packet=f"phase2/packets/{chapter.chapter_id}-{axis}.json",
|
||||
chapter_title=chapter.title,
|
||||
preferred_model_role="dr_verifier" if axis == "counter" else "dr_analyst",
|
||||
research_goal=research_goal or f"为《{chapter.title}》收集并验证 {axis} 轴证据，形成可写入章节的具体判断与证据落点。",
|
||||
research_method=method_key,
|
||||
prompt_brief=prompt_brief or (_integrated_prompt_brief(chapter, method) if axis == "chapter_integrated" else f"围绕《{chapter.title}》的 {axis} 轴，优先形成可证伪、可引用、可落地的证据包。"),
|
||||
required_skills=required_skills or _default_required_skills(axis),
|
||||
allowed_materials=allowed_materials or [],
|
||||
expected_evidence=expected_evidence or _default_expected_evidence(axis),
|
||||
stop_conditions=stop_conditions or _default_stop_conditions(),
|
||||
model_hint="use_cross_model_verifier" if axis == "counter" else "use_cost_effective_research_worker",
|
||||
)
|
||||
|
||||
|
||||
def generate_task_cards(
|
||||
slug: str,
|
||||
framework_text: str,
|
||||
*,
|
||||
axes: list[str] | None = None,
|
||||
method: ResearchMethod | None = None,
|
||||
) -> list[TaskCard]:
|
||||
del slug # slug is kept for call-site clarity and future namespacing.
|
||||
chapters = parse_framework_chapters(framework_text)
|
||||
selected_axes = axes or (method.task_axes if method else DEFAULT_AXES)
|
||||
cards: list[TaskCard] = []
|
||||
for chapter in chapters:
|
||||
for axis in selected_axes:
|
||||
routes = AXIS_ROUTES.get(axis, ["general"])
|
||||
cards.append(
|
||||
_task_card_for_chapter_axis(
|
||||
chapter=chapter,
|
||||
axis=axis,
|
||||
routes=routes,
|
||||
method_key=method.key if method else "",
|
||||
method=method,
|
||||
)
|
||||
)
|
||||
validate_task_cards(cards)
|
||||
return cards
|
||||
|
||||
|
||||
def generate_task_cards_from_research_brief(
|
||||
slug: str,
|
||||
framework_text: str,
|
||||
research_brief: dict[str, Any],
|
||||
*,
|
||||
axes: list[str] | None = None,
|
||||
method: ResearchMethod | None = None,
|
||||
) -> list[TaskCard]:
|
||||
del slug
|
||||
chapters = parse_framework_chapters(framework_text)
|
||||
planning = research_brief.get("task_planning") or {}
|
||||
method_key = research_brief.get("research_method") or (method.key if method else "")
|
||||
if method is None and method_key:
|
||||
from scripts.runtime.methods import ResearchMethodRegistry
|
||||
|
||||
method = ResearchMethodRegistry().get(method_key)
|
||||
phase2_mode = planning.get("phase2_mode") or research_brief.get("phase2_mode")
|
||||
if axes:
|
||||
selected_axes = axes
|
||||
elif phase2_mode == "chapter_integrated":
|
||||
selected_axes = ["chapter_integrated"]
|
||||
else:
|
||||
selected_axes = (method.task_axes if method else None) or list(planning.get("search_routes_by_axis") or []) or DEFAULT_AXES
|
||||
routes_by_axis = planning.get("search_routes_by_axis") or {}
|
||||
prompt_by_axis = planning.get("axis_prompt_briefs") or {}
|
||||
base_skills = list(planning.get("required_skills") or [])
|
||||
stop_conditions = list(planning.get("stop_conditions") or [])
|
||||
allowed_materials = [
|
||||
str(item.get("path"))
|
||||
for item in research_brief.get("materials", [])
|
||||
if item.get("path")
|
||||
]
|
||||
if not allowed_materials:
|
||||
material_digest = (research_brief.get("phase1_inputs") or {}).get("material_digest")
|
||||
if material_digest:
|
||||
allowed_materials.append(str(material_digest))
|
||||
chapter_plan_by_id = {
|
||||
str(item.get("chapter_id")): item
|
||||
for item in research_brief.get("chapter_planning", [])
|
||||
if item.get("chapter_id")
|
||||
}
|
||||
cards: list[TaskCard] = []
|
||||
for chapter in chapters:
|
||||
for axis in selected_axes:
|
||||
routes = list(routes_by_axis.get(axis) or AXIS_ROUTES.get(axis, ["general"]))
|
||||
skills = base_skills or _default_required_skills(axis)
|
||||
if "search-gateway" not in skills:
|
||||
skills = ["search-gateway", *skills]
|
||||
chapter_plan = chapter_plan_by_id.get(chapter.chapter_id) if axis == "chapter_integrated" else None
|
||||
prompt_brief = prompt_by_axis.get(axis)
|
||||
questions = None
|
||||
research_goal = None
|
||||
expected_evidence = None
|
||||
card_stop_conditions = stop_conditions or None
|
||||
if chapter_plan:
|
||||
prompt_brief = chapter_plan.get("phase2_prompt_context") or prompt_brief
|
||||
research_goal = chapter_plan.get("core_question")
|
||||
questions = [
|
||||
chapter_plan.get("core_question", ""),
|
||||
chapter_plan.get("bold_hypothesis", ""),
|
||||
"按 Phase1 求证计划逐条收集支持证据、反方证据和待补证据。",
|
||||
"不得绕开 Phase1 主基调另起炉灶；若证据推翻假设，必须明确写出修正建议。",
|
||||
]
|
||||
questions.extend(str(item) for item in chapter_plan.get("verification_plan", []))
|
||||
expected_evidence = _default_expected_evidence(axis)
|
||||
expected_evidence.update(
|
||||
{
|
||||
"phase1_minimum_evidence": chapter_plan.get("minimum_evidence") or {},
|
||||
"evidence_lanes": chapter_plan.get("evidence_lanes") or [],
|
||||
"must_address_phase1_hypothesis": True,
|
||||
}
|
||||
)
|
||||
card_stop_conditions = [
|
||||
*(stop_conditions or _default_stop_conditions()),
|
||||
"已经逐条回应 Phase1 的大胆假设：支持、修正或推翻，并说明依据。",
|
||||
"已经把本地材料原文、外部证据、反方边界和行动建议分开记录。",
|
||||
]
|
||||
cards.append(
|
||||
_task_card_for_chapter_axis(
|
||||
chapter=chapter,
|
||||
axis=axis,
|
||||
routes=routes,
|
||||
method_key=method_key,
|
||||
required_skills=skills,
|
||||
allowed_materials=allowed_materials,
|
||||
prompt_brief=prompt_brief,
|
||||
questions=questions,
|
||||
research_goal=research_goal,
|
||||
expected_evidence=expected_evidence,
|
||||
stop_conditions=card_stop_conditions,
|
||||
method=method,
|
||||
)
|
||||
)
|
||||
validate_task_cards(cards)
|
||||
return cards
|
||||
|
||||
|
||||
def detect_dependency_cycles(cards: list[TaskCard]) -> None:
|
||||
graph = {card.task_id: card.dependencies for card in cards}
|
||||
visiting: set[str] = set()
|
||||
visited: set[str] = set()
|
||||
|
||||
def visit(node: str) -> None:
|
||||
if node in visiting:
|
||||
raise ValueError(f"dependency cycle detected at {node}")
|
||||
if node in visited:
|
||||
return
|
||||
visiting.add(node)
|
||||
for dep in graph.get(node, []):
|
||||
visit(dep)
|
||||
visiting.remove(node)
|
||||
visited.add(node)
|
||||
|
||||
for task_id in graph:
|
||||
visit(task_id)
|
||||
|
||||
|
||||
def validate_task_cards(cards: list[TaskCard]) -> None:
|
||||
seen: set[str] = set()
|
||||
for card in cards:
|
||||
if not card.research_goal:
|
||||
card.research_goal = f"围绕 {card.topic_axis} 轴收集并验证结构化证据。"
|
||||
if not card.prompt_brief:
|
||||
card.prompt_brief = f"按 {card.topic_axis} 轴形成证据包，避免泛泛结论。"
|
||||
if not card.required_skills:
|
||||
card.required_skills = _default_required_skills(card.topic_axis)
|
||||
if "search-gateway" not in card.required_skills:
|
||||
card.required_skills = ["search-gateway", *card.required_skills]
|
||||
if not card.expected_evidence:
|
||||
card.expected_evidence = _default_expected_evidence(card.topic_axis)
|
||||
if not card.stop_conditions:
|
||||
card.stop_conditions = _default_stop_conditions()
|
||||
if card.task_id in seen:
|
||||
raise ValueError(f"duplicate task_id: {card.task_id}")
|
||||
seen.add(card.task_id)
|
||||
if not card.chapter_ids:
|
||||
raise ValueError(f"{card.task_id}: chapter_ids required")
|
||||
if not card.chapter_title:
|
||||
card.chapter_title = card.chapter_ids[0]
|
||||
if not card.questions:
|
||||
raise ValueError(f"{card.task_id}: questions required")
|
||||
if not card.output_packet.endswith(".json"):
|
||||
raise ValueError(f"{card.task_id}: output_packet must be json")
|
||||
invalid_routes = sorted(set(card.search_routes) - VALID_ROUTES)
|
||||
if invalid_routes:
|
||||
raise ValueError(f"{card.task_id}: invalid search_routes {invalid_routes}")
|
||||
missing_deps = sorted({dep for card in cards for dep in card.dependencies} - seen)
|
||||
if missing_deps:
|
||||
raise ValueError(f"unknown dependencies: {missing_deps}")
|
||||
detect_dependency_cycles(cards)
|
||||
|
||||
|
||||
def write_task_cards(path: Path, cards: list[TaskCard]) -> None:
|
||||
path.parent.mkdir(parents=True, exist_ok=True)
|
||||
path.write_text(
|
||||
json.dumps([card.to_dict() for card in cards], ensure_ascii=False, indent=2) + "\n",
|
||||
encoding="utf-8",
|
||||
)
|
||||
|
||||
|
||||
def load_task_cards(path: Path) -> list[TaskCard]:
|
||||
data = json.loads(path.read_text(encoding="utf-8"))
|
||||
cards = [TaskCard(**item) for item in data]
|
||||
validate_task_cards(cards)
|
||||
return cards
|
||||
|
||||
|
||||
def validate_packet(packet: dict[str, Any]) -> None:
|
||||
required = {
|
||||
"task_id",
|
||||
"claims",
|
||||
"evidence_items",
|
||||
"counter_evidence",
|
||||
"source_ids",
|
||||
"source_quality_notes",
|
||||
"open_questions",
|
||||
"raw_quotes_or_notes",
|
||||
}
|
||||
missing = sorted(required - set(packet))
|
||||
if missing:
|
||||
raise ValueError(f"packet missing fields: {missing}")
|
||||
if not packet["claims"]:
|
||||
raise ValueError("packet claims must not be empty")
|
||||
if not packet["evidence_items"]:
|
||||
raise ValueError("packet evidence_items must not be empty")
|
||||
if not packet["counter_evidence"]:
|
||||
raise ValueError("packet counter_evidence must not be empty")
|
||||
declared = set(packet.get("source_ids") or [])
|
||||
referenced: set[str] = set()
|
||||
for section in ("claims", "counter_evidence"):
|
||||
for item in packet.get(section) or []:
|
||||
referenced.update(item.get("source_ids") or [])
|
||||
for item in packet.get("evidence_items") or []:
|
||||
if item.get("source_id"):
|
||||
referenced.add(item["source_id"])
|
||||
undeclared = sorted(referenced - declared)
|
||||
if undeclared:
|
||||
raise ValueError(f"packet source_ids referenced but not declared: {undeclared}")
|
||||
packet_sources = packet.get("sources") or []
|
||||
if not packet_sources:
|
||||
raise ValueError("packet sources must not be empty")
|
||||
known_source_ids = {source.get("id") for source in packet_sources}
|
||||
missing_sources = sorted(declared - known_source_ids)
|
||||
if missing_sources:
|
||||
raise ValueError(f"packet source_ids missing source metadata: {missing_sources}")
|
||||
@@ -0,0 +1,497 @@
|
||||
"""Python role workers for task-card execution."""
|
||||
|
||||
from __future__ import annotations
|
||||
|
||||
import json
|
||||
import re
|
||||
from concurrent.futures import ThreadPoolExecutor, as_completed
|
||||
from pathlib import Path
|
||||
from typing import Callable, Protocol, Any
|
||||
|
||||
from scripts.runtime.roles import RoleDefinition, RuntimeProfile
|
||||
from scripts.runtime.skills import SkillRegistry
|
||||
from scripts.runtime.tasks import TaskCard, validate_packet
|
||||
from scripts.runtime.sources import append_packet_sources
|
||||
|
||||
|
||||
class ChatClient(Protocol):
|
||||
def chat_complete(self, **kwargs) -> str:
|
||||
...
|
||||
|
||||
|
||||
class SearchProvider(Protocol):
|
||||
def search(self, *, query: str, route: str, num_results: int) -> list[dict[str, Any]]:
|
||||
...
|
||||
|
||||
|
||||
class ProjectSearchProvider:
|
||||
"""Thin adapter over the project-owned search client."""
|
||||
|
||||
def __init__(self, *, strict_specialized: bool = True) -> None:
|
||||
from scripts.lib.search_client import SearchClient
|
||||
|
||||
self.client = SearchClient(strict_specialized=strict_specialized)
|
||||
|
||||
def search(self, *, query: str, route: str, num_results: int) -> list[dict[str, Any]]:
|
||||
if route == "scholar":
|
||||
hits = self.client.scholar(query, num_results=num_results, year_low=2020)
|
||||
elif route == "patents":
|
||||
hits = self.client.patents(query, num_results=num_results)
|
||||
elif route == "news":
|
||||
hits = self.client.news(query, num_results=num_results, time_range="y")
|
||||
elif route == "fda":
|
||||
hits = self.client.fda(query, num_results=num_results)
|
||||
elif route == "evidence":
|
||||
hits = self.client.evidence(query, num_results=num_results)
|
||||
else:
|
||||
hits = self.client.search(query, num_results=num_results)
|
||||
return [
|
||||
{
|
||||
"title": hit.title,
|
||||
"url": hit.url,
|
||||
"snippet": hit.snippet,
|
||||
"route": route,
|
||||
}
|
||||
for hit in hits
|
||||
]
|
||||
|
||||
def close(self) -> None:
|
||||
self.client.close()
|
||||
|
||||
|
||||
def _extract_json_object(text: str) -> dict:
|
||||
stripped = text.strip()
|
||||
if stripped.startswith("```"):
|
||||
stripped = stripped.strip("`")
|
||||
if stripped.startswith("json"):
|
||||
stripped = stripped[4:].strip()
|
||||
start = stripped.find("{")
|
||||
end = stripped.rfind("}")
|
||||
if start == -1 or end == -1 or end < start:
|
||||
raise ValueError("worker response does not contain a JSON object")
|
||||
return json.loads(stripped[start : end + 1])
|
||||
|
||||
|
||||
def _safe_source_stem(task_id: str) -> str:
|
||||
return re.sub(r"[^a-zA-Z0-9]+", "_", task_id).strip("_").lower()
|
||||
|
||||
|
||||
def contains_cjk(text: str) -> bool:
|
||||
return any("\u4e00" <= char <= "\u9fff" for char in text)
|
||||
|
||||
|
||||
def strip_cjk(text: str) -> str:
|
||||
return re.sub(r"[\u3400-\u9fff]+", " ", text)
|
||||
|
||||
|
||||
def validate_packet_against_allowed_context(
|
||||
packet: dict,
|
||||
search_context: dict[str, Any] | None,
|
||||
material_context: dict[str, Any] | None,
|
||||
) -> None:
|
||||
"""Ensure the model did not invent source IDs or URLs beyond candidates."""
|
||||
if not search_context and not material_context:
|
||||
return
|
||||
candidates = (search_context or {}).get("candidate_sources") or []
|
||||
materials = (material_context or {}).get("materials") or []
|
||||
if not candidates and not materials:
|
||||
return
|
||||
candidate_ids = {source.get("id") for source in candidates}
|
||||
candidate_ids.update(item.get("source_id") for item in materials)
|
||||
candidate_urls = {source.get("url") for source in candidates if source.get("url")}
|
||||
candidate_urls.update(item.get("path") for item in materials if item.get("path"))
|
||||
packet_sources = packet.get("sources") or []
|
||||
unknown_ids = sorted(
|
||||
source.get("id")
|
||||
for source in packet_sources
|
||||
if source.get("id") and source.get("id") not in candidate_ids
|
||||
)
|
||||
unknown_urls = sorted(
|
||||
source.get("url")
|
||||
for source in packet_sources
|
||||
if source.get("url") and source.get("url") not in candidate_urls
|
||||
)
|
||||
if (candidates or materials) and not packet_sources:
|
||||
raise ValueError("packet must include source metadata from candidate_sources or local materials")
|
||||
if unknown_ids:
|
||||
raise ValueError(f"packet sources include non-candidate source IDs: {unknown_ids}")
|
||||
if unknown_urls:
|
||||
raise ValueError(f"packet sources include non-candidate URLs: {unknown_urls}")
|
||||
|
||||
|
||||
def normalize_packet_against_context(
|
||||
packet: dict[str, Any],
|
||||
search_context: dict[str, Any] | None,
|
||||
material_context: dict[str, Any] | None,
|
||||
) -> dict[str, Any]:
|
||||
"""Deterministically fill schema metadata the model often omits."""
|
||||
packet = dict(packet)
|
||||
referenced: set[str] = set(packet.get("source_ids") or [])
|
||||
for section in ("claims", "counter_evidence"):
|
||||
for item in packet.get(section) or []:
|
||||
referenced.update(item.get("source_ids") or [])
|
||||
for item in packet.get("evidence_items") or []:
|
||||
if item.get("source_id"):
|
||||
referenced.add(item["source_id"])
|
||||
if "source_ids" not in packet or not packet.get("source_ids"):
|
||||
packet["source_ids"] = sorted(referenced)
|
||||
|
||||
available_sources: dict[str, dict[str, Any]] = {}
|
||||
for source in (search_context or {}).get("candidate_sources") or []:
|
||||
if source.get("id"):
|
||||
available_sources[source["id"]] = source
|
||||
for material in (material_context or {}).get("materials") or []:
|
||||
source_id = material.get("source_id")
|
||||
if source_id:
|
||||
available_sources[source_id] = {
|
||||
"id": source_id,
|
||||
"title": material.get("title") or Path(material.get("path", "")).name,
|
||||
"url": material.get("path") or "",
|
||||
"tier": "local_material",
|
||||
"score": 8,
|
||||
}
|
||||
|
||||
existing_sources = {
|
||||
source.get("id"): source
|
||||
for source in packet.get("sources") or []
|
||||
if source.get("id")
|
||||
}
|
||||
for source_id in packet.get("source_ids") or []:
|
||||
if source_id not in existing_sources and source_id in available_sources:
|
||||
existing_sources[source_id] = available_sources[source_id]
|
||||
if existing_sources:
|
||||
packet["sources"] = [existing_sources[source_id] for source_id in packet.get("source_ids", []) if source_id in existing_sources]
|
||||
return packet
|
||||
|
||||
|
||||
FDA_AXIS_TERMS = {
|
||||
"nmpa_fda_ema_ich_who_baseline": "CGMP pharmaceutical quality system process validation aseptic processing data integrity",
|
||||
"quality_system_gap": "CGMP CAPA deviation change control data integrity quality unit pharmaceutical",
|
||||
"manufacturing_process_risk": "aseptic processing sterile drug manufacturing process validation PPQ cleaning validation water system",
|
||||
"operations_management_gap": "pharmaceutical quality system quality metrics management review senior management FDA",
|
||||
"capa_roadmap": "CGMP CAPA effectiveness remediation warning letter close-out pharmaceutical",
|
||||
"verification_evidence": "FDA 483 response CAPA effectiveness verification EIR pharmaceutical quality",
|
||||
"counter": "FDA warning letter CGMP pharmaceutical quality data integrity remediation limitations",
|
||||
"fda_enforcement_precedents": "FDA warning letter CGMP pharmaceutical aseptic processing data integrity CAPA process validation",
|
||||
}
|
||||
|
||||
|
||||
FDA_CHAPTER_TERMS = {
|
||||
"ch01": "commercial readiness phase gate remediation governance",
|
||||
"ch02": "regulatory baseline CGMP EU GMP Annex 1 ICH Q9 ICH Q10",
|
||||
"ch03": "aseptic processing RABS first air media fill visual inspection depyrogenation tunnel",
|
||||
"ch04": "biologics drug substance WFI clean utilities SCADA EMS single-use system",
|
||||
"ch05": "process validation master batch record CPP CQA PPQ cleaning validation technology transfer",
|
||||
"ch06": "deviation change control CAPA document control training data integrity quality unit",
|
||||
"ch07": "training effectiveness quality culture operator qualification human factors",
|
||||
"ch08": "quality metrics management review escalation cross-functional governance operations",
|
||||
"ch09": "CDMO quality organization technology transfer project governance capability matrix",
|
||||
"ch10": "CAPA remediation plan effectiveness check owner due date verification evidence",
|
||||
"ch11": "regulatory mapping CAPA tracker closure evidence quality assurance verification",
|
||||
}
|
||||
|
||||
|
||||
ROUTE_CHAPTER_TERMS = {
|
||||
**FDA_CHAPTER_TERMS,
|
||||
}
|
||||
|
||||
ROUTE_SUFFIX_TERMS = {
|
||||
"scholar": "pharmaceutical GMP review validation risk management quality system",
|
||||
"patents": "biologics manufacturing patent process formulation device",
|
||||
"news": "pharmaceutical quality operations CDMO quality governance",
|
||||
"evidence": "pharmaceutical GMP evidence guidance enforcement best practice quality operations",
|
||||
"general": "pharmaceutical GMP best practice guidance quality operations remediation",
|
||||
}
|
||||
|
||||
INTERNAL_QUERY_TOKENS = {
|
||||
"chapter_integrated",
|
||||
"input_material_findings",
|
||||
}
|
||||
|
||||
|
||||
def _compact_english_query(*parts: str, max_terms: int = 16) -> str:
|
||||
text = strip_cjk(" ".join(part for part in parts if part))
|
||||
text = re.sub(r"[^A-Za-z0-9./+-]+", " ", text)
|
||||
terms: list[str] = []
|
||||
seen: set[str] = set()
|
||||
for raw in text.split():
|
||||
term = raw.strip(" ./+-").lower()
|
||||
if not term or term in INTERNAL_QUERY_TOKENS:
|
||||
continue
|
||||
key = term.casefold()
|
||||
if key in seen:
|
||||
continue
|
||||
seen.add(key)
|
||||
terms.append(term)
|
||||
if len(terms) >= max_terms:
|
||||
break
|
||||
return " ".join(terms)
|
||||
|
||||
|
||||
def _chapter_terms(card: TaskCard) -> str:
|
||||
mapped = " ".join(ROUTE_CHAPTER_TERMS.get(chapter_id, "") for chapter_id in card.chapter_ids)
|
||||
if mapped.strip():
|
||||
return mapped
|
||||
return strip_cjk(card.chapter_title)
|
||||
|
||||
|
||||
def build_route_query(card: TaskCard, route: str) -> str:
|
||||
"""Build short, route-aware queries instead of sending whole task cards."""
|
||||
if route == "fda":
|
||||
terms = FDA_AXIS_TERMS.get(card.topic_axis, "FDA warning letter CGMP pharmaceutical quality")
|
||||
chapter_terms = " ".join(FDA_CHAPTER_TERMS.get(chapter_id, "") for chapter_id in card.chapter_ids)
|
||||
query = f"{terms} {chapter_terms}".strip()
|
||||
if contains_cjk(query):
|
||||
raise ValueError(f"FDA route query must not contain Chinese text: {query}")
|
||||
return query
|
||||
if route == "scholar":
|
||||
return _compact_english_query(_chapter_terms(card), ROUTE_SUFFIX_TERMS["scholar"])
|
||||
if route == "patents":
|
||||
return _compact_english_query(_chapter_terms(card), ROUTE_SUFFIX_TERMS["patents"])
|
||||
if route == "news":
|
||||
return _compact_english_query(_chapter_terms(card), ROUTE_SUFFIX_TERMS["news"])
|
||||
if route == "evidence":
|
||||
return _compact_english_query(_chapter_terms(card), ROUTE_SUFFIX_TERMS["evidence"])
|
||||
return _compact_english_query(_chapter_terms(card), ROUTE_SUFFIX_TERMS["general"])
|
||||
|
||||
|
||||
def _material_excerpt(project_root: Path | None, rel_path: str, *, max_chars: int = 6000) -> dict[str, str] | None:
|
||||
if project_root is None:
|
||||
return None
|
||||
path = project_root / rel_path
|
||||
if not path.exists() or not path.is_file():
|
||||
return None
|
||||
text = path.read_text(encoding="utf-8", errors="ignore")
|
||||
return {
|
||||
"path": rel_path,
|
||||
"source_id": f"src_local_{_safe_source_stem(Path(rel_path).stem)}",
|
||||
"title": Path(rel_path).name,
|
||||
"excerpt": text[:max_chars],
|
||||
}
|
||||
|
||||
|
||||
def build_material_context(card: TaskCard, project_root: Path | None, *, max_chars_per_material: int = 6000) -> dict[str, Any]:
|
||||
materials = []
|
||||
seen: set[str] = set()
|
||||
for rel in card.allowed_materials:
|
||||
if rel in seen:
|
||||
continue
|
||||
seen.add(rel)
|
||||
item = _material_excerpt(project_root, rel, max_chars=max_chars_per_material)
|
||||
if item:
|
||||
materials.append(item)
|
||||
return {"materials": materials}
|
||||
|
||||
|
||||
def build_search_context(
|
||||
card: TaskCard,
|
||||
search_provider: SearchProvider,
|
||||
*,
|
||||
num_results_per_route: int = 5,
|
||||
) -> dict[str, Any]:
|
||||
candidate_sources: list[dict[str, Any]] = []
|
||||
routes_used: list[str] = []
|
||||
source_stem = _safe_source_stem(card.task_id)
|
||||
idx = 1
|
||||
for route in card.search_routes:
|
||||
routes_used.append(route)
|
||||
query = build_route_query(card, route)
|
||||
hits = search_provider.search(query=query, route=route, num_results=num_results_per_route)
|
||||
for hit in hits:
|
||||
candidate_sources.append(
|
||||
{
|
||||
"id": f"src_{source_stem}_{idx:03d}",
|
||||
"title": hit.get("title", ""),
|
||||
"url": hit.get("url", ""),
|
||||
"snippet": hit.get("snippet", ""),
|
||||
"route": hit.get("route", route),
|
||||
"tier": "Tier 2",
|
||||
"score": 6,
|
||||
}
|
||||
)
|
||||
idx += 1
|
||||
return {"routes_used": routes_used, "candidate_sources": candidate_sources}
|
||||
|
||||
|
||||
def build_packet_user_prompt(
|
||||
card: TaskCard,
|
||||
search_context: dict[str, Any] | None = None,
|
||||
material_context: dict[str, Any] | None = None,
|
||||
) -> str:
|
||||
context = search_context or {"routes_used": [], "candidate_sources": []}
|
||||
materials = material_context or {"materials": []}
|
||||
return (
|
||||
"请根据以下 task card 产出一个证据包 JSON。\n"
|
||||
"正式结论、summary、open_questions 用中文；英文原文摘录、source title、DOI/URL 可以保留英文。\n"
|
||||
"必须主动包含 counter_evidence，且所有引用的 source_id 必须出现在 source_ids 中。\n\n"
|
||||
"只能使用 candidate_sources 或 Local material context 中的来源，不得编造 URL、DOI、trial ID 或 source_id。\n"
|
||||
"输出 JSON 必须包含 sources 字段；sources 只能来自 candidate_sources 或 Local material context。\n"
|
||||
"如 Local material context 非空，必须至少提取 1 条本地材料原文证据；如果与本章无关，必须在 open_questions 说明为什么无关。\n\n"
|
||||
f"{json.dumps(card.to_dict(), ensure_ascii=False, indent=2)}\n\n"
|
||||
f"Local material context:\n{json.dumps(materials, ensure_ascii=False, indent=2)}\n\n"
|
||||
f"Search context:\n{json.dumps(context, ensure_ascii=False, indent=2)}\n\n"
|
||||
"只输出 JSON，不要输出 Markdown 解释。"
|
||||
)
|
||||
|
||||
|
||||
def build_packet_repair_prompt(
|
||||
*,
|
||||
card: TaskCard,
|
||||
raw_response: str,
|
||||
error: Exception,
|
||||
search_context: dict[str, Any] | None = None,
|
||||
material_context: dict[str, Any] | None = None,
|
||||
) -> str:
|
||||
context = search_context or {"routes_used": [], "candidate_sources": []}
|
||||
materials = material_context or {"materials": []}
|
||||
return (
|
||||
"请修复上一次 evidence packet 输出，使其成为合法且通过 schema 校验的 JSON。\n"
|
||||
"只输出 JSON 对象，不要输出 Markdown、解释或代码块。\n"
|
||||
"保留中文主写作；英文只允许出现在来源标题、URL、DOI、原文摘录或检索笔记中。\n"
|
||||
"不得编造 candidate_sources 或 Local material context 以外的来源、URL、DOI、trial ID 或 source_id。\n\n"
|
||||
f"Schema error:\n{error}\n\n"
|
||||
f"Task card:\n{json.dumps(card.to_dict(), ensure_ascii=False, indent=2)}\n\n"
|
||||
f"Search context:\n{json.dumps(context, ensure_ascii=False, indent=2)}\n\n"
|
||||
f"Local material context:\n{json.dumps(materials, ensure_ascii=False, indent=2)}\n\n"
|
||||
f"Previous raw response:\n{raw_response[:12000]}"
|
||||
)
|
||||
|
||||
|
||||
class PacketWorker:
|
||||
def __init__(
|
||||
self,
|
||||
*,
|
||||
role: RoleDefinition,
|
||||
client: ChatClient,
|
||||
project_root: Path | None = None,
|
||||
search_provider: SearchProvider | None = None,
|
||||
skill_registry: SkillRegistry | None = None,
|
||||
num_results_per_route: int = 5,
|
||||
) -> None:
|
||||
self.role = role
|
||||
self.client = client
|
||||
self.project_root = project_root
|
||||
self.search_provider = search_provider
|
||||
self.skill_registry = skill_registry or SkillRegistry()
|
||||
self.num_results_per_route = num_results_per_route
|
||||
|
||||
def _system_prompt(self) -> str:
|
||||
skill_texts = []
|
||||
for name in self.role.skills:
|
||||
try:
|
||||
skill_texts.append(f"# Skill: {name}\n\n{self.skill_registry.read(name)}")
|
||||
except FileNotFoundError:
|
||||
skill_texts.append(f"# Skill: {name}\n\n[missing skill: {name}]")
|
||||
return (
|
||||
f"{self.role.identity}\n\n"
|
||||
"你是 Deep Research v0.20 Python runtime 的证据包 worker。\n"
|
||||
"你的唯一任务是把一个 task card 转换为结构化 evidence packet。\n"
|
||||
"遵循中文主写作原则；不要写章节正文；不要编造 URL、DOI、trial ID 或 source_id。\n\n"
|
||||
"搜索只能走项目 Python search gateway 或调用方提供的 search_context；不要直接使用 Tavily MCP、browser MCP、平台 web search 或任何需要用户权限确认的外部搜索工具。\n\n"
|
||||
+ "\n\n".join(skill_texts)
|
||||
)
|
||||
|
||||
def run(self, card: TaskCard) -> dict:
|
||||
search_context = None
|
||||
if self.search_provider:
|
||||
search_context = build_search_context(
|
||||
card,
|
||||
self.search_provider,
|
||||
num_results_per_route=self.num_results_per_route,
|
||||
)
|
||||
material_context = build_material_context(card, self.project_root)
|
||||
raw = self.client.chat_complete(
|
||||
model=self.role.model,
|
||||
system=self._system_prompt(),
|
||||
user=build_packet_user_prompt(card, search_context, material_context),
|
||||
temperature=self.role.temperature,
|
||||
max_tokens=self.role.max_tokens,
|
||||
tag=f"packet:{card.task_id}",
|
||||
)
|
||||
try:
|
||||
packet = normalize_packet_against_context(
|
||||
_extract_json_object(raw),
|
||||
search_context,
|
||||
material_context,
|
||||
)
|
||||
validate_packet(packet)
|
||||
validate_packet_against_allowed_context(packet, search_context, material_context)
|
||||
return packet
|
||||
except Exception as error:
|
||||
repaired = self.client.chat_complete(
|
||||
model=self.role.model,
|
||||
system=self._system_prompt(),
|
||||
user=build_packet_repair_prompt(
|
||||
card=card,
|
||||
raw_response=raw,
|
||||
error=error,
|
||||
search_context=search_context,
|
||||
material_context=material_context,
|
||||
),
|
||||
temperature=0,
|
||||
max_tokens=self.role.max_tokens,
|
||||
tag=f"packet-repair:{card.task_id}",
|
||||
)
|
||||
packet = normalize_packet_against_context(
|
||||
_extract_json_object(repaired),
|
||||
search_context,
|
||||
material_context,
|
||||
)
|
||||
validate_packet(packet)
|
||||
validate_packet_against_allowed_context(packet, search_context, material_context)
|
||||
return packet
|
||||
|
||||
|
||||
def _write_packet_error(project_root: Path, card: TaskCard, error: Exception) -> None:
|
||||
path = project_root / "phase2" / "packet_errors" / f"{card.task_id}.json"
|
||||
path.parent.mkdir(parents=True, exist_ok=True)
|
||||
payload = {
|
||||
"task_id": card.task_id,
|
||||
"status": "failed",
|
||||
"error": str(error),
|
||||
"output_packet": card.output_packet,
|
||||
}
|
||||
path.write_text(json.dumps(payload, ensure_ascii=False, indent=2) + "\n", encoding="utf-8")
|
||||
|
||||
|
||||
def run_packet_workers(
|
||||
*,
|
||||
project_root: Path,
|
||||
cards: list[TaskCard],
|
||||
runtime: RuntimeProfile,
|
||||
client_factory: Callable[[RoleDefinition], ChatClient],
|
||||
search_provider_factory: Callable[[], SearchProvider] | None = None,
|
||||
workers: int,
|
||||
) -> int:
|
||||
role = runtime.role_for_task("evidence_packet")
|
||||
max_workers = max(1, min(workers, role.max_concurrency))
|
||||
|
||||
def run_one(card: TaskCard) -> tuple[TaskCard, dict | None, Exception | None]:
|
||||
search_provider = search_provider_factory() if search_provider_factory else None
|
||||
try:
|
||||
worker = PacketWorker(role=role, client=client_factory(role), project_root=project_root, search_provider=search_provider)
|
||||
return card, worker.run(card), None
|
||||
except Exception as error:
|
||||
return card, None, error
|
||||
finally:
|
||||
close = getattr(search_provider, "close", None)
|
||||
if close:
|
||||
close()
|
||||
|
||||
written = 0
|
||||
with ThreadPoolExecutor(max_workers=max_workers) as pool:
|
||||
futures = [pool.submit(run_one, card) for card in cards]
|
||||
for future in as_completed(futures):
|
||||
card, packet, error = future.result()
|
||||
if error is not None:
|
||||
_write_packet_error(project_root, card, error)
|
||||
continue
|
||||
if packet is None:
|
||||
_write_packet_error(project_root, card, RuntimeError("packet worker returned no packet"))
|
||||
continue
|
||||
path = project_root / card.output_packet
|
||||
path.parent.mkdir(parents=True, exist_ok=True)
|
||||
path.write_text(json.dumps(packet, ensure_ascii=False, indent=2) + "\n", encoding="utf-8")
|
||||
append_packet_sources(project_root / "phase2" / "sources.jsonl", packet)
|
||||
written += 1
|
||||
return written
|
||||
+16
-5
@@ -25,17 +25,19 @@ from scripts.lib.zenmux_client import load_secrets
|
||||
|
||||
|
||||
ROUTE_HELP = {
|
||||
"general": "Exa -> Tavily generic web discovery",
|
||||
"general": "Tavily -> Exa -> Brave generic web discovery",
|
||||
"evidence": "Exa highlights -> Tavily -> Brave controlled evidence discovery",
|
||||
"scholar": "Serper Scholar -> generic fallback",
|
||||
"patents": "Serper Google Patents -> site:patents.google.com fallback",
|
||||
"news": "Serper News -> generic fallback",
|
||||
"fda": "FDA-focused discovery for warning letters, enforcement pages, and meeting materials",
|
||||
}
|
||||
|
||||
PROFILE_ROUTES = {
|
||||
"biomed_literature": ["scholar", "general"],
|
||||
"patent_heavy": ["patents", "general"],
|
||||
"china_market": ["news", "general"],
|
||||
"investment": ["news", "general"],
|
||||
"biomed_literature": ["scholar", "evidence", "general"],
|
||||
"patent_heavy": ["patents", "evidence", "general"],
|
||||
"china_market": ["news", "evidence", "general"],
|
||||
"investment": ["news", "evidence", "general"],
|
||||
}
|
||||
|
||||
PROFILE_QUERY_PREFIX = {
|
||||
@@ -46,12 +48,16 @@ PROFILE_QUERY_PREFIX = {
|
||||
def search_route(client: SearchClient, route: str, query: str, args: argparse.Namespace) -> list[SearchHit]:
|
||||
if route == "general":
|
||||
return client.search(query, num_results=args.num_results)
|
||||
if route == "evidence":
|
||||
return client.evidence(query, num_results=args.num_results, category=args.exa_category)
|
||||
if route == "scholar":
|
||||
return client.scholar(query, num_results=args.num_results, year_low=args.year_low)
|
||||
if route == "patents":
|
||||
return client.patents(query, num_results=args.num_results)
|
||||
if route == "news":
|
||||
return client.news(query, num_results=args.num_results, time_range=args.time_range)
|
||||
if route == "fda":
|
||||
return client.fda(query, num_results=args.num_results)
|
||||
raise SystemExit(f"unknown route: {route}")
|
||||
|
||||
|
||||
@@ -107,6 +113,11 @@ def build_parser() -> argparse.ArgumentParser:
|
||||
help="Run a strategy profile instead of a single route",
|
||||
)
|
||||
parser.add_argument("--num-results", type=int, default=10)
|
||||
parser.add_argument(
|
||||
"--exa-category",
|
||||
choices=["research paper", "news", "company", "financial report", "github", "tweet", "personal site", "pdf"],
|
||||
help="Optional Exa category for the evidence route",
|
||||
)
|
||||
parser.add_argument("--year-low", type=int, help="Lower year bound for scholar searches")
|
||||
parser.add_argument("--time-range", choices=["d", "w", "m", "y"], help="Serper news time range")
|
||||
parser.add_argument("--json", action="store_true", help="Emit JSON instead of Markdown")
|
||||
|
||||
@@ -0,0 +1,96 @@
|
||||
#!/usr/bin/env python3
|
||||
"""v0.20 Python-core regression checks."""
|
||||
|
||||
from __future__ import annotations
|
||||
|
||||
import json
|
||||
import subprocess
|
||||
import sys
|
||||
import tempfile
|
||||
from pathlib import Path
|
||||
|
||||
|
||||
REPO_ROOT = Path(__file__).resolve().parent.parent
|
||||
|
||||
|
||||
def run(cmd: list[str]) -> None:
|
||||
print("$ " + " ".join(cmd))
|
||||
result = subprocess.run(cmd, cwd=REPO_ROOT, check=False, text=True, capture_output=True)
|
||||
if result.stdout:
|
||||
print(result.stdout.rstrip())
|
||||
if result.stderr:
|
||||
print(result.stderr.rstrip(), file=sys.stderr)
|
||||
if result.returncode != 0:
|
||||
raise SystemExit(result.returncode)
|
||||
|
||||
|
||||
def make_fixture(root: Path) -> Path:
|
||||
project = root / "v020-fixture"
|
||||
(project / "phase4").mkdir(parents=True, exist_ok=True)
|
||||
(project / "phase4" / "final_zh.md").write_text(
|
||||
"# v0.20 回归测试报告\n\n正文引用占位。\n",
|
||||
encoding="utf-8",
|
||||
)
|
||||
return project
|
||||
|
||||
|
||||
def write_fixture_packet(project: Path) -> None:
|
||||
cards = json.loads((project / "phase2" / "task_cards.json").read_text(encoding="utf-8"))
|
||||
for index, card in enumerate(cards, start=1):
|
||||
source_a = f"src_{index:03d}_a"
|
||||
source_b = f"src_{index:03d}_b"
|
||||
packet = {
|
||||
"task_id": card["task_id"],
|
||||
"claims": [{"claim": f"{card['task_id']} 回归证据支持章节主线", "source_ids": [source_a]}],
|
||||
"evidence_items": [{"source_id": source_a, "summary": "权威来源支持该判断"}],
|
||||
"counter_evidence": [{"claim": "证据仍需更多来源交叉验证", "source_ids": [source_b]}],
|
||||
"source_ids": [source_a, source_b],
|
||||
"source_quality_notes": [f"{source_a} Tier 1", f"{source_b} Tier 2"],
|
||||
"open_questions": ["需要在真实项目中补充更多来源。"],
|
||||
"raw_quotes_or_notes": ["English raw note can remain here."],
|
||||
"sources": [
|
||||
{"id": source_a, "title": f"Regression Source {index}A", "url": f"https://example.com/source-{index}-a", "tier": 1},
|
||||
{"id": source_b, "title": f"Regression Source {index}B", "url": f"https://example.com/source-{index}-b", "tier": 2},
|
||||
],
|
||||
}
|
||||
packet_path = project / card["output_packet"]
|
||||
packet_path.write_text(json.dumps(packet, ensure_ascii=False, indent=2) + "\n", encoding="utf-8")
|
||||
|
||||
|
||||
def main() -> int:
|
||||
python = sys.executable
|
||||
run([python, "scripts/dr.py", "skills", "validate"])
|
||||
run([python, "scripts/dr.py", "models", "--profile", "medium", "--json"])
|
||||
with tempfile.TemporaryDirectory(prefix="deep-research-v020-") as tmp:
|
||||
tmp_root = Path(tmp)
|
||||
run(
|
||||
[
|
||||
python,
|
||||
"scripts/dr.py",
|
||||
"init",
|
||||
"v0.20 fixture",
|
||||
"--slug",
|
||||
"v020-fixture",
|
||||
"--projects-dir",
|
||||
str(tmp_root),
|
||||
"--method",
|
||||
"mckinsey_market",
|
||||
]
|
||||
)
|
||||
project = make_fixture(tmp_root)
|
||||
run([python, "scripts/dr.py", "frame", str(project)])
|
||||
assert (project / "phase1" / "research_brief.json").exists()
|
||||
run([python, "scripts/dr.py", "approve", str(project)])
|
||||
run([python, "scripts/dr.py", "research", str(project), "--workers", "2", "--axis", "literature", "--dry-run"])
|
||||
run([python, "scripts/dr.py", "research", str(project), "--workers", "2", "--axis", "literature"])
|
||||
write_fixture_packet(project)
|
||||
run([python, "scripts/dr.py", "research", str(project), "--workers", "2", "--axis", "literature", "--build-briefs"])
|
||||
assert (project / "phase2" / "compressed_findings" / "ch01.json").exists()
|
||||
run([python, "scripts/dr.py", "review", str(project)])
|
||||
run([python, "scripts/dr.py", "finalize", str(project), "--dry-run"])
|
||||
print("v0.20 regression PASS")
|
||||
return 0
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
raise SystemExit(main())
|
||||
@@ -0,0 +1,60 @@
|
||||
---
|
||||
name: deep-research
|
||||
description: Use when a user wants to run Deep Research, create a client-ready research report, analyze supplied audit/material files, compare evidence against regulations or best practices, or continue a phase-based research project.
|
||||
---
|
||||
|
||||
# Deep Research
|
||||
|
||||
## Principle
|
||||
|
||||
Deep Research is driven by the repository Python core, not by chat context. Treat Codex, OpenCode, Claude Code, Antigravity, and Gemini CLI as surface interfaces that call `scripts/dr.py`; all durable state must be written under `projects/<slug>/`.
|
||||
|
||||
## Required Flow
|
||||
|
||||
1. Work from the repository root.
|
||||
2. Validate the skill/runtime registry before a serious run:
|
||||
`uv run python scripts/dr.py skills validate`
|
||||
3. For user-supplied materials, stop after Phase 0/1 material brief and interview:
|
||||
`uv run python scripts/dr.py init "<topic>" --slug <slug> --method <method> --input-material <path>`
|
||||
Review `phase1/material_brief.md` with the user before generating or approving the framework.
|
||||
4. If model calls will be used, probe the selected profile first:
|
||||
`uv run python scripts/dr.py models --profile medium --probe`
|
||||
5. Run Phase 1 and pause for framework review:
|
||||
`uv run python scripts/dr.py frame <slug> --method <method>`
|
||||
After user approval: `uv run python scripts/dr.py approve <slug>`
|
||||
6. Run Phase 2 with file-backed task cards and packets:
|
||||
`uv run python scripts/dr.py research <slug> --workers 6 --execute-packets --allow-search-fallback`
|
||||
7. Build briefs and chapters only from persisted packets:
|
||||
`uv run python scripts/dr.py sources cache <slug> --limit 50`
|
||||
`uv run python scripts/dr.py research <slug> --build-briefs`
|
||||
`uv run python scripts/dr.py research <slug> --assemble-chapters --workers 4`
|
||||
8. Review and finalize through Python:
|
||||
`uv run python scripts/dr.py review <slug>`
|
||||
`uv run python scripts/dr.py finalize <slug> --report-engine reportlab`
|
||||
|
||||
## Research Rules
|
||||
|
||||
- Chinese is the formal thinking and writing language by default; English is allowed for search keywords, source titles, abstracts, and raw excerpts.
|
||||
- Do not invent evidence when model/API access fails. Stop at the last durable artifact and report the exact blocker.
|
||||
- Phase 2 concurrency must use task cards and packet files, not platform subagents as the default mechanism.
|
||||
- Search must use the project Python gateway (`scripts/search.py` / `scripts.lib.search_client`) by default. Do not use Tavily MCP, browser MCP, or platform-native web search in subagents unless the user explicitly requests that escape hatch.
|
||||
- Key Tier 1-2 sources such as ICH Q9/Q10, EU GMP Annex 1, FDA guidance/warning letters, EMA/NMPA/WHO pages, and pharmacopeia materials should be cached as local Markdown snapshots under `phase2/source_cache/` before chapter assembly.
|
||||
- User materials are starting evidence, not final truth. Cross-check against authoritative sources such as NMPA, FDA, EMA, ICH, WHO, pharmacopeias, and recognized best-practice references.
|
||||
- For GMP/quality/operations diagnosis, prefer `--method gmp_quality_operations_diagnosis`.
|
||||
- Chapter drafts are not acceptable if they merely summarize principles. Each section must turn evidence into concrete findings, risk implications, and整改动作；otherwise return to Phase 2 enrichment.
|
||||
|
||||
## Useful Commands
|
||||
|
||||
- Status: `uv run python scripts/dr.py status <slug>`
|
||||
- List methods: `uv run python scripts/dr.py methods list`
|
||||
- Show method: `uv run python scripts/dr.py methods show <method>`
|
||||
- Dry-run task cards: `uv run python scripts/dr.py research <slug> --dry-run`
|
||||
- Sync adapter skills: `uv run python scripts/dr.py skills sync`
|
||||
|
||||
## Common Failures
|
||||
|
||||
- If Codex cannot write `~/.codex`, run `uv run python scripts/deploy_adapters.py codex --force` outside sandboxed mode.
|
||||
- The Codex deploy script does not copy `config.toml` by default; use `--include-config` only when the user explicitly wants the bundled Codex profile.
|
||||
- If `models --probe` returns subscription/model errors, do not launch packet workers; switch profile/key/provider first.
|
||||
- If a PDF has little embedded text, Phase 0 should call FireRed OCR. Default endpoint: `http://192.168.50.100:8001`.
|
||||
- If a subagent asks for MCP/web permissions during research, stop it and reroute the task through `search-gateway`.
|
||||
@@ -0,0 +1,34 @@
|
||||
---
|
||||
name: document-ingest
|
||||
description: Use when Deep Research starts from user-provided PDFs, scanned audit reports, DOCX/PPTX files, images, or internal notes that must become phase0 persisted inputs.
|
||||
---
|
||||
|
||||
# Document Ingest
|
||||
|
||||
## Core Rule
|
||||
|
||||
User-provided materials are evidence leads, not final evidence. Always persist the original file, extracted text, OCR status, and limitations under `phase0/` before Phase 1 framing depends on them.
|
||||
|
||||
## Required Artifacts
|
||||
|
||||
- `phase0/inputs/<original-file>` stores the source file copy.
|
||||
- `phase0/extracted/<stem>.md` stores direct text extraction plus OCR text when available.
|
||||
- `phase0/extracted/<stem>.ocr.md` stores OCR-only output for scanned PDFs.
|
||||
- `manifest.json.material_inventory[]` records `copied_to`, `extracted_to`, `ocr_required`, `ocr_status`, and OCR errors if any.
|
||||
|
||||
## PDF Handling
|
||||
|
||||
- Text PDFs should use direct extraction first.
|
||||
- If direct extraction is too sparse, run FireRed OCR through the configured LAN endpoint.
|
||||
- Default endpoint: `http://192.168.50.100:8001`.
|
||||
- Override endpoint with `DEEP_RESEARCH_OCR_ENDPOINT`.
|
||||
- Limit page count with `DEEP_RESEARCH_OCR_MAX_PAGES` when testing or when documents are very long.
|
||||
|
||||
## Current Boundary
|
||||
|
||||
PDF text extraction and scanned-PDF OCR are supported. DOCX, PPTX, image-only batches, table reconstruction, and layout-aware evidence mapping should remain explicit next-step work unless implemented in Python core.
|
||||
|
||||
## Quality Notes
|
||||
|
||||
- OCR text may contain spacing or line-break errors. Treat it as internal material evidence and verify formal findings against NMPA, FDA, EMA, ICH, WHO, or other authoritative sources.
|
||||
- If OCR fails, do not block project initialization. Record the failure and continue Phase 1 with a clear limitation.
|
||||
@@ -0,0 +1,62 @@
|
||||
---
|
||||
name: search-gateway
|
||||
description: Use when Deep Research agents or subagents need web, scholar, patent, news, regulatory, or source-discovery search without using platform MCP tools or browser search directly.
|
||||
---
|
||||
|
||||
# Search Gateway
|
||||
|
||||
## Rule
|
||||
|
||||
Use the project Python search gateway as the only default search interface. Do not call Tavily MCP, browser MCP, generic web tools, or platform-native search from a subagent unless the user explicitly asks for that escape hatch.
|
||||
|
||||
## Commands
|
||||
|
||||
Run searches from the repository root:
|
||||
|
||||
```bash
|
||||
uv run python scripts/search.py "<query>" --route general --json --trace
|
||||
uv run python scripts/search.py "<query>" --route evidence --json --trace
|
||||
uv run python scripts/search.py "<query>" --route scholar --year-low 2020 --json --trace
|
||||
uv run python scripts/search.py "<query>" --route news --time-range y --json --trace
|
||||
uv run python scripts/search.py "<query>" --route patents --json --trace
|
||||
uv run python scripts/search.py "<query>" --profile biomed_literature --json --trace
|
||||
```
|
||||
|
||||
If `uv` cannot use the user cache in a sandbox, set a local cache:
|
||||
|
||||
```bash
|
||||
UV_CACHE_DIR=/private/tmp/deep_research_uv_cache uv run python scripts/search.py "<query>" --route general --json --trace
|
||||
```
|
||||
|
||||
## Routing
|
||||
|
||||
- `general`: Tavily first, Exa fallback, Brave fallback; use for broad discovery and gap filling.
|
||||
- `evidence`: Exa highlights first, Tavily fallback, Brave fallback; use when a task card needs concise, source-level candidate evidence for an evidence packet.
|
||||
- `scholar`: Serper Scholar first; use for papers, reviews, technical literature, and academic validation only.
|
||||
- `news`: Serper News first; use for recent industry/current information.
|
||||
- `patents`: Serper Google Patents first.
|
||||
- `biomed_literature`: scholar plus general discovery.
|
||||
|
||||
Serper is not the default general web search source. Keep it mainly for Scholar, Google Patents, News, and targeted `site:` searches where Google coverage matters.
|
||||
|
||||
Tavily Research is a phase-level scan tool, not a packet-writing shortcut. Use it for Phase 1 initial landscape scans, Phase 2 gap-fill after a chapter is thin, or Phase 3回炉补证据；its output must be saved, source-scored, deduplicated, and converted into candidate evidence before citation.
|
||||
|
||||
Exa is the preferred controlled evidence discovery route for agents because it can return short highlights/text per URL. Treat Exa hits as candidate sources unless the URL itself is an original Tier 1-2 source.
|
||||
|
||||
API keys are loaded from `secrets.env` by `scripts/search.py`; do not ask the user to authorize MCP calls when the env keys are available.
|
||||
|
||||
## Subagent Protocol
|
||||
|
||||
For evidence packets:
|
||||
|
||||
1. Search through `scripts/search.py`, save or summarize the returned JSON in the packet’s `raw_quotes_or_notes`.
|
||||
2. Use search hits only as candidate sources; whenever possible, cite the original regulator, guideline, paper, or official document.
|
||||
3. Put every used source in `sources` with `id`, `title`, `url`, `tier`, and `score`.
|
||||
4. Do not write a final chapter during search; produce structured evidence only.
|
||||
5. For repeatedly used Tier 1-2 sources, run `uv run python scripts/dr.py sources cache <project>` so later phases can cite a local Markdown snapshot rather than only a URL.
|
||||
|
||||
For chapter assembly:
|
||||
|
||||
1. Do not search. Use only `phase2/chapter_briefs`, `phase2/packets`, `phase2/sources.jsonl`, `phase0/extracted`, and `phase1/framework.md`.
|
||||
2. Do not create new `source_id`.
|
||||
3. If evidence is thin, mark the chapter as needing Phase 2 enrichment instead of filling with generic prose.
|
||||
@@ -0,0 +1,76 @@
|
||||
---
|
||||
name: search-strategy
|
||||
description: 生物医药深度研究的统一检索策略。规定信源优先级、检索轮次、关键词构造、API 路由，以及何时切换到专业信源。所有做信息收集的 worker/agent 必须加载此技能。
|
||||
---
|
||||
|
||||
# Search Strategy
|
||||
|
||||
## Core Rule
|
||||
|
||||
Do not send Chinese chapter titles, interview paragraphs, or full task-card text directly to search APIs. For formal search, first convert the task into short English query terms plus axis terms, then add source/domain constraints when useful.
|
||||
|
||||
## Query Construction
|
||||
|
||||
Build every query from three parts:
|
||||
|
||||
- `entity/domain`: the object or field, such as `pharmaceutical`, `biologics`, `sterile drug`, `CDMO`, `quality system`.
|
||||
- `axis`: the research axis, such as `CAPA deviation change control`, `aseptic processing process validation PPQ`, `quality metrics management review`.
|
||||
- `evidence type`: the evidence to retrieve, such as `Warning Letter`, `meeting materials`, `guidance`, `systematic review`, `patent`, `best practices`.
|
||||
|
||||
Default English query length is 5-12 keywords. Chinese terms are useful for NMPA, local industry sources, and internal-material matching, but Chinese long sentences must not be the default query form.
|
||||
|
||||
For route-specific searches, do not append the original Chinese chapter title after the English query. If chapter context is needed, map the chapter to short English concept terms first, such as `commercial readiness phase gate`, `aseptic processing`, `quality metrics management review`, or `CAPA effectiveness check`.
|
||||
|
||||
## Route Patterns
|
||||
|
||||
- `fda`: use `site:fda.gov` plus `Warning Letter`, `inspection`, `enforcement`, `meeting materials`, or `meeting minutes`, then add the axis terms.
|
||||
- `scholar`: use technical/scientific terms plus `review`, `validation`, `risk management`, `quality system`, or disease/mechanism terms.
|
||||
- `evidence`: use Exa highlights for controlled evidence discovery when a packet needs concise source-level excerpts; still trace important hits back to original Tier 1-2 sources.
|
||||
- `patents`: use technology route plus material, target, process, formulation, device, or manufacturing terms.
|
||||
- `news`: use company/industry plus event type and recency terms.
|
||||
- `general`: use Tavily/Exa/Brave for discovery and gap filling; trace useful hits back to Tier 1-2 original sources before citing. Do not route generic web discovery through Serper by default.
|
||||
- `tavily_research` conceptually means a phase-level scan, not a normal packet route. Save the research result, score/deduplicate sources, then convert it into candidate evidence before writing claims.
|
||||
|
||||
## GMP/FDA Examples
|
||||
|
||||
Bad query:
|
||||
|
||||
```text
|
||||
围绕《审计发现应先转化为商业化阶段门缺口，而不是停留在风险项计数》从质量体系角度提炼可证伪的核心结论
|
||||
```
|
||||
|
||||
Good queries:
|
||||
|
||||
```text
|
||||
site:fda.gov "Warning Letter" CGMP CAPA deviation change control data integrity pharmaceutical
|
||||
site:fda.gov "meeting materials" "pharmaceutical quality" "quality metrics"
|
||||
site:fda.gov/inspections-compliance-enforcement-and-criminal-investigations "Warning Letter" aseptic processing process validation
|
||||
```
|
||||
|
||||
## Source Priority
|
||||
|
||||
- Tier 1: regulator, guideline, pharmacopeia, primary literature, trial registry, patent original, company filing.
|
||||
- Tier 2: systematic review, recognized consulting or industry association report, professional database/media.
|
||||
- Tier 3: conference abstract, broker report, preprint, vendor white paper.
|
||||
- Tier 4: generic web search result; discovery only, not conclusion support.
|
||||
|
||||
## Four-Round Search Discipline
|
||||
|
||||
1. Tier 1 direct hit: regulator, PubMed/Scholar, trial registry, patent original, or official filing.
|
||||
2. Tier 2 synthesis: recognized review, guideline interpretation, consulting/association report.
|
||||
3. Counter-evidence: limitations, failures, enforcement actions, contradictory interpretations.
|
||||
4. Gap fill: Exa evidence discovery or Tavily/Brave general discovery, then trace back to original sources. Use Serper here only for Google-specific needs such as `site:` targeting, Scholar, Patents, or News.
|
||||
|
||||
## Tavily Research vs Exa Evidence
|
||||
|
||||
- Tavily Research is best for Phase 1 initial landscape scans, thin-chapter补证据, and Phase 3回炉. Prompt in English, specify source priority, counter-evidence, and structured output. Do not cite its synthesized prose directly.
|
||||
- Exa evidence discovery is best for Phase 2 packet work because highlights/text are compact enough for source-quality scoring and evidence-table mapping.
|
||||
- Serper remains preferred for Scholar, Google Patents, News, and Google-specific `site:` targeting.
|
||||
- Brave remains a cross-check and mixed-language fallback, not the first evidence route.
|
||||
|
||||
## Required Packet Behavior
|
||||
|
||||
- Put search keywords or route notes in `raw_quotes_or_notes` when evidence is weak or no suitable source was found.
|
||||
- FDA/GMP tasks must explicitly check Warning Letters, inspection/enforcement pages, and meeting materials/minutes.
|
||||
- Do not cite search snippets as final evidence when an original regulator, guideline, paper, or official PDF can be reached.
|
||||
- If the candidate sources are not sufficient, stop and record the gap in `open_questions` instead of writing generic prose.
|
||||
@@ -0,0 +1,104 @@
|
||||
from __future__ import annotations
|
||||
|
||||
import sys
|
||||
from pathlib import Path
|
||||
|
||||
REPO_ROOT = Path(__file__).resolve().parents[1]
|
||||
if str(REPO_ROOT) not in sys.path:
|
||||
sys.path.insert(0, str(REPO_ROOT))
|
||||
|
||||
from scripts.deploy_adapters import default_codex_home, deploy_antigravity, deploy_codex
|
||||
|
||||
|
||||
def test_default_codex_home_is_external_to_project(tmp_path: Path) -> None:
|
||||
home = default_codex_home(env={}, user_home=tmp_path)
|
||||
|
||||
assert home == tmp_path / ".codex"
|
||||
assert not home.is_relative_to(REPO_ROOT)
|
||||
|
||||
|
||||
def test_deploy_codex_writes_adapter_to_external_home(tmp_path: Path) -> None:
|
||||
target = tmp_path / "codex-home"
|
||||
|
||||
result = deploy_codex(target=target, force=True, repo_root=REPO_ROOT)
|
||||
|
||||
assert result.written
|
||||
assert not (target / "config.toml").exists()
|
||||
assert (target / "commands" / "dr-run.md").exists()
|
||||
assert (target / "agents" / "dr-pm.toml").exists()
|
||||
assert (target / "skills" / "search-strategy" / "SKILL.md").exists()
|
||||
assert (target / "skills" / "search-gateway" / "SKILL.md").exists()
|
||||
assert (target / "skills" / "document-ingest" / "SKILL.md").exists()
|
||||
assert (target / "skills" / "deep-research" / "SKILL.md").exists()
|
||||
assert result.target == target
|
||||
|
||||
|
||||
def test_deploy_codex_config_is_explicit_opt_in(tmp_path: Path) -> None:
|
||||
target = tmp_path / "codex-home"
|
||||
|
||||
result = deploy_codex(target=target, force=True, repo_root=REPO_ROOT, include_config=True)
|
||||
|
||||
assert result.written
|
||||
assert (target / "config.toml").exists()
|
||||
|
||||
|
||||
def test_deploy_codex_dry_run_does_not_write(tmp_path: Path) -> None:
|
||||
target = tmp_path / "codex-home"
|
||||
|
||||
result = deploy_codex(target=target, force=True, repo_root=REPO_ROOT, dry_run=True)
|
||||
|
||||
assert result.planned
|
||||
assert not target.exists()
|
||||
|
||||
|
||||
def test_deploy_antigravity_writes_workspace_rules_and_skills(tmp_path: Path) -> None:
|
||||
target = tmp_path / "workspace"
|
||||
|
||||
result = deploy_antigravity(target=target, repo_root=REPO_ROOT)
|
||||
|
||||
assert result.written
|
||||
assert (target / ".agents" / "agents.md").exists()
|
||||
assert (target / ".agents" / "rules" / "deep-research-antigravity.md").exists()
|
||||
assert (target / ".agents" / "workflows" / "deep-research-native.md").exists()
|
||||
assert (target / ".agents" / "skills" / "antigravity-surface-adapter" / "SKILL.md").exists()
|
||||
assert (target / ".agents" / "skills" / "search-strategy" / "SKILL.md").exists()
|
||||
assert result.target == target
|
||||
|
||||
|
||||
def test_deploy_antigravity_skips_existing_files_by_default(tmp_path: Path) -> None:
|
||||
target = tmp_path / "workspace"
|
||||
existing = target / ".agents" / "rules" / "deep-research-antigravity.md"
|
||||
existing.parent.mkdir(parents=True)
|
||||
existing.write_text("custom rule\n", encoding="utf-8")
|
||||
|
||||
result = deploy_antigravity(target=target, repo_root=REPO_ROOT)
|
||||
|
||||
assert existing.read_text(encoding="utf-8") == "custom rule\n"
|
||||
assert existing in result.skipped
|
||||
assert not existing.with_name("deep-research-antigravity.md.bak").exists()
|
||||
|
||||
|
||||
def test_deploy_antigravity_skips_existing_agents_file_by_default(tmp_path: Path) -> None:
|
||||
target = tmp_path / "workspace"
|
||||
existing = target / ".agents" / "agents.md"
|
||||
existing.parent.mkdir(parents=True)
|
||||
existing.write_text("custom agents\n", encoding="utf-8")
|
||||
|
||||
result = deploy_antigravity(target=target, repo_root=REPO_ROOT)
|
||||
|
||||
assert existing.read_text(encoding="utf-8") == "custom agents\n"
|
||||
assert existing in result.skipped
|
||||
assert not existing.with_name("agents.md.bak").exists()
|
||||
|
||||
|
||||
def test_deploy_antigravity_force_backs_up_existing_files(tmp_path: Path) -> None:
|
||||
target = tmp_path / "workspace"
|
||||
existing = target / ".agents" / "rules" / "deep-research-antigravity.md"
|
||||
existing.parent.mkdir(parents=True)
|
||||
existing.write_text("custom rule\n", encoding="utf-8")
|
||||
|
||||
result = deploy_antigravity(target=target, repo_root=REPO_ROOT, force=True)
|
||||
|
||||
assert existing.read_text(encoding="utf-8") != "custom rule\n"
|
||||
assert existing.with_name("deep-research-antigravity.md.bak").read_text(encoding="utf-8") == "custom rule\n"
|
||||
assert existing.with_name("deep-research-antigravity.md.bak") in result.backups
|
||||
@@ -0,0 +1,292 @@
|
||||
from __future__ import annotations
|
||||
|
||||
import json
|
||||
import sys
|
||||
from pathlib import Path
|
||||
|
||||
REPO_ROOT = Path(__file__).resolve().parents[1]
|
||||
if str(REPO_ROOT) not in sys.path:
|
||||
sys.path.insert(0, str(REPO_ROOT))
|
||||
|
||||
from scripts.runtime.assembly import (
|
||||
ChapterAssemblyWorker,
|
||||
build_chapter_briefs,
|
||||
build_compressed_findings,
|
||||
build_chapter_user_prompt,
|
||||
run_chapter_assembly_workers,
|
||||
validate_chapter_markdown_citations,
|
||||
validate_chapter_brief,
|
||||
validate_compressed_finding,
|
||||
)
|
||||
from scripts.runtime.roles import resolve_runtime_profile
|
||||
|
||||
|
||||
class FakeClient:
|
||||
def __init__(self, response: str) -> None:
|
||||
self.response = response
|
||||
self.calls: list[dict[str, object]] = []
|
||||
|
||||
def chat_complete(self, **kwargs) -> str:
|
||||
self.calls.append(kwargs)
|
||||
return self.response
|
||||
|
||||
|
||||
class TaggedClient:
|
||||
def __init__(self, responses_by_tag: dict[str, str]) -> None:
|
||||
self.responses_by_tag = responses_by_tag
|
||||
self.calls: list[dict[str, object]] = []
|
||||
|
||||
def chat_complete(self, **kwargs) -> str:
|
||||
self.calls.append(kwargs)
|
||||
return self.responses_by_tag[str(kwargs["tag"])]
|
||||
|
||||
|
||||
def chapter_brief(chapter_id: str, source_ids: list[str]) -> dict:
|
||||
return {
|
||||
"chapter_id": chapter_id,
|
||||
"chapter_title": "临床证据正在重塑需求判断",
|
||||
"packet_ids": [f"{chapter_id}-clinical"],
|
||||
"core_claims": [{"claim": "临床证据支持核心判断", "source_ids": source_ids[:1]}],
|
||||
"evidence_items": [{"source_id": source_ids[0], "summary": "III 期数据支持主要终点"}],
|
||||
"counter_evidence": [{"claim": "长期安全性仍需随访", "source_ids": source_ids[-1:]}],
|
||||
"source_ids": source_ids,
|
||||
"open_questions": [],
|
||||
"assembly_notes": ["按金字塔结构组织。"],
|
||||
}
|
||||
|
||||
|
||||
def write_packet(path: Path, task_id: str, claim: str, source_id: str) -> None:
|
||||
packet = {
|
||||
"task_id": task_id,
|
||||
"claims": [{"claim": claim, "source_ids": [source_id]}],
|
||||
"evidence_items": [{"source_id": source_id, "summary": f"{claim} 的证据"}],
|
||||
"counter_evidence": [{"claim": "仍需关注样本量和外推限制", "source_ids": ["src_counter"]}],
|
||||
"source_ids": [source_id, "src_counter"],
|
||||
"sources": [
|
||||
{"id": source_id, "title": "来源", "url": f"https://example.com/{source_id}"},
|
||||
{"id": "src_counter", "title": "反方来源", "url": "https://example.com/counter"},
|
||||
],
|
||||
"source_quality_notes": [f"{source_id} Tier 1"],
|
||||
"open_questions": ["还需要补充中国市场数据"],
|
||||
"raw_quotes_or_notes": ["English note can remain as source material."],
|
||||
}
|
||||
path.parent.mkdir(parents=True, exist_ok=True)
|
||||
path.write_text(json.dumps(packet, ensure_ascii=False, indent=2) + "\n", encoding="utf-8")
|
||||
|
||||
|
||||
def test_build_chapter_briefs_aggregates_packets_by_chapter(tmp_path: Path) -> None:
|
||||
project = tmp_path / "project"
|
||||
cards = [
|
||||
{
|
||||
"task_id": "ch01-clinical",
|
||||
"chapter_ids": ["ch01"],
|
||||
"chapter_title": "临床证据正在重塑需求判断",
|
||||
"topic_axis": "clinical",
|
||||
"questions": ["q"],
|
||||
"search_routes": ["scholar"],
|
||||
"output_packet": "phase2/packets/ch01-clinical.json",
|
||||
},
|
||||
{
|
||||
"task_id": "ch01-market",
|
||||
"chapter_ids": ["ch01"],
|
||||
"chapter_title": "临床证据正在重塑需求判断",
|
||||
"topic_axis": "market",
|
||||
"questions": ["q"],
|
||||
"search_routes": ["news"],
|
||||
"output_packet": "phase2/packets/ch01-market.json",
|
||||
},
|
||||
]
|
||||
(project / "phase2").mkdir(parents=True)
|
||||
(project / "phase2" / "task_cards.json").write_text(json.dumps(cards, ensure_ascii=False), encoding="utf-8")
|
||||
write_packet(project / "phase2/packets/ch01-clinical.json", "ch01-clinical", "临床证据支持核心判断", "src_001")
|
||||
write_packet(project / "phase2/packets/ch01-market.json", "ch01-market", "市场数据支持需求增长", "src_002")
|
||||
|
||||
briefs = build_chapter_briefs(project)
|
||||
|
||||
assert len(briefs) == 1
|
||||
brief = briefs[0]
|
||||
validate_chapter_brief(brief)
|
||||
assert brief["chapter_id"] == "ch01"
|
||||
assert brief["chapter_title"] == "临床证据正在重塑需求判断"
|
||||
assert brief["packet_ids"] == ["ch01-clinical", "ch01-market"]
|
||||
assert "src_001" in brief["source_ids"]
|
||||
assert "src_002" in brief["source_ids"]
|
||||
assert (project / "phase2/chapter_briefs/ch01.json").exists()
|
||||
|
||||
compressed = build_compressed_findings(project)
|
||||
|
||||
assert len(compressed) == 1
|
||||
finding = compressed[0]
|
||||
validate_compressed_finding(finding)
|
||||
assert finding["chapter_id"] == "ch01"
|
||||
assert "chapter_thesis" in finding
|
||||
assert "evidence_landings" in finding
|
||||
assert "src_001" in finding["source_ids"]
|
||||
assert (project / "phase2/compressed_findings/ch01.json").exists()
|
||||
|
||||
|
||||
def test_build_chapter_briefs_skips_placeholder_packets(tmp_path: Path) -> None:
|
||||
project = tmp_path / "project"
|
||||
cards = [
|
||||
{
|
||||
"task_id": "ch01-good",
|
||||
"chapter_ids": ["ch01"],
|
||||
"chapter_title": "临床证据正在重塑需求判断",
|
||||
"topic_axis": "clinical",
|
||||
"questions": ["q"],
|
||||
"search_routes": ["scholar"],
|
||||
"output_packet": "phase2/packets/ch01-good.json",
|
||||
},
|
||||
{
|
||||
"task_id": "ch01-empty",
|
||||
"chapter_ids": ["ch01"],
|
||||
"chapter_title": "临床证据正在重塑需求判断",
|
||||
"topic_axis": "clinical",
|
||||
"questions": ["q"],
|
||||
"search_routes": ["scholar"],
|
||||
"output_packet": "phase2/packets/ch01-empty.json",
|
||||
},
|
||||
]
|
||||
(project / "phase2").mkdir(parents=True)
|
||||
(project / "phase2/task_cards.json").write_text(json.dumps(cards, ensure_ascii=False), encoding="utf-8")
|
||||
write_packet(project / "phase2/packets/ch01-good.json", "ch01-good", "临床证据支持核心判断", "src_001")
|
||||
empty = {
|
||||
"task_id": "ch01-empty",
|
||||
"claims": [],
|
||||
"evidence_items": [],
|
||||
"counter_evidence": [],
|
||||
"source_ids": [],
|
||||
"source_quality_notes": [],
|
||||
"open_questions": [],
|
||||
"raw_quotes_or_notes": [],
|
||||
}
|
||||
(project / "phase2/packets/ch01-empty.json").write_text(json.dumps(empty, ensure_ascii=False), encoding="utf-8")
|
||||
|
||||
briefs = build_chapter_briefs(project)
|
||||
|
||||
assert len(briefs) == 1
|
||||
warnings = json.loads((project / "phase2/brief_warnings.json").read_text(encoding="utf-8"))
|
||||
assert warnings[0]["task_id"] == "ch01-empty"
|
||||
|
||||
|
||||
def test_chapter_prompt_contains_brief_and_fragmentation_guard() -> None:
|
||||
brief = {
|
||||
"chapter_id": "ch01",
|
||||
"chapter_title": "临床证据正在重塑需求判断",
|
||||
"packet_ids": ["ch01-clinical"],
|
||||
"core_claims": [{"claim": "临床证据支持核心判断", "source_ids": ["src_001"]}],
|
||||
"evidence_items": [{"source_id": "src_001", "summary": "III 期数据支持主要终点"}],
|
||||
"counter_evidence": [{"claim": "长期安全性仍需随访", "source_ids": ["src_002"]}],
|
||||
"source_ids": ["src_001", "src_002"],
|
||||
"open_questions": [],
|
||||
"assembly_notes": ["避免重复 packet 原文；按金字塔结构组织。"],
|
||||
}
|
||||
|
||||
prompt = build_chapter_user_prompt(brief)
|
||||
|
||||
assert "临床证据正在重塑需求判断" in prompt
|
||||
assert "避免碎片化" in prompt
|
||||
assert "compressed finding" in prompt
|
||||
assert "只输出 Markdown" in prompt
|
||||
|
||||
|
||||
def test_chapter_assembly_worker_writes_markdown(tmp_path: Path) -> None:
|
||||
runtime = resolve_runtime_profile(profile="medium")
|
||||
role = runtime.role_for_task("chapter_assembly")
|
||||
fake = FakeClient("# 第1章 临床证据正在重塑需求判断\n\n结论先行。[src_001]\n")
|
||||
worker = ChapterAssemblyWorker(role=role, client=fake)
|
||||
brief = {
|
||||
"chapter_id": "ch01",
|
||||
"chapter_title": "临床证据正在重塑需求判断",
|
||||
"packet_ids": ["ch01-clinical"],
|
||||
"core_claims": [{"claim": "临床证据支持核心判断", "source_ids": ["src_001"]}],
|
||||
"evidence_items": [{"source_id": "src_001", "summary": "III 期数据支持主要终点"}],
|
||||
"counter_evidence": [{"claim": "长期安全性仍需随访", "source_ids": ["src_002"]}],
|
||||
"source_ids": ["src_001", "src_002"],
|
||||
"open_questions": [],
|
||||
"assembly_notes": ["按金字塔结构组织。"],
|
||||
}
|
||||
|
||||
output = worker.write_chapter(project_root=tmp_path, brief=brief)
|
||||
|
||||
assert output == tmp_path / "phase2/drafts/ch01.md"
|
||||
assert "结论先行" in output.read_text(encoding="utf-8")
|
||||
assert fake.calls[0]["model"] == role.model
|
||||
assert "章节证据分析师" in fake.calls[0]["system"]
|
||||
assert "中文章节组装 worker" in fake.calls[0]["system"]
|
||||
|
||||
|
||||
def test_validate_chapter_markdown_rejects_unknown_source_ids() -> None:
|
||||
brief = {
|
||||
"chapter_id": "ch01",
|
||||
"chapter_title": "临床证据正在重塑需求判断",
|
||||
"packet_ids": ["ch01-clinical"],
|
||||
"core_claims": [{"claim": "临床证据支持核心判断", "source_ids": ["src_001"]}],
|
||||
"evidence_items": [{"source_id": "src_001", "summary": "III 期数据支持主要终点"}],
|
||||
"counter_evidence": [{"claim": "长期安全性仍需随访", "source_ids": ["src_002"]}],
|
||||
"source_ids": ["src_001", "src_002"],
|
||||
"open_questions": [],
|
||||
"assembly_notes": ["按金字塔结构组织。"],
|
||||
}
|
||||
|
||||
try:
|
||||
validate_chapter_markdown_citations("结论引用了不存在的来源。[src_fake]", brief)
|
||||
except ValueError as exc:
|
||||
assert "unknown citation ids" in str(exc)
|
||||
assert "src_fake" in str(exc)
|
||||
else:
|
||||
raise AssertionError("unknown source id should fail validation")
|
||||
|
||||
|
||||
def test_chapter_assembly_worker_refuses_to_write_unknown_citations(tmp_path: Path) -> None:
|
||||
runtime = resolve_runtime_profile(profile="medium")
|
||||
role = runtime.role_for_task("chapter_assembly")
|
||||
fake = FakeClient("# 第1章 临床证据正在重塑需求判断\n\n结论先行。[src_fake]\n")
|
||||
worker = ChapterAssemblyWorker(role=role, client=fake)
|
||||
brief = {
|
||||
"chapter_id": "ch01",
|
||||
"chapter_title": "临床证据正在重塑需求判断",
|
||||
"packet_ids": ["ch01-clinical"],
|
||||
"core_claims": [{"claim": "临床证据支持核心判断", "source_ids": ["src_001"]}],
|
||||
"evidence_items": [{"source_id": "src_001", "summary": "III 期数据支持主要终点"}],
|
||||
"counter_evidence": [{"claim": "长期安全性仍需随访", "source_ids": ["src_002"]}],
|
||||
"source_ids": ["src_001", "src_002"],
|
||||
"open_questions": [],
|
||||
"assembly_notes": ["按金字塔结构组织。"],
|
||||
}
|
||||
|
||||
try:
|
||||
worker.write_chapter(project_root=tmp_path, brief=brief)
|
||||
except ValueError as exc:
|
||||
assert "unknown citation ids" in str(exc)
|
||||
else:
|
||||
raise AssertionError("chapter with unknown citation should not be written")
|
||||
assert not (tmp_path / "phase2/drafts/ch01.md").exists()
|
||||
|
||||
|
||||
def test_run_chapter_assembly_workers_records_errors_without_aborting_batch(tmp_path: Path) -> None:
|
||||
runtime = resolve_runtime_profile(profile="medium")
|
||||
fake = TaggedClient(
|
||||
{
|
||||
"chapter:ch01": "# 第1章 临床证据正在重塑需求判断\n\n结论先行。[src_001]\n",
|
||||
"chapter:ch02": "# 第2章 临床证据存在不确定性\n\n错误引用。[src_fake]\n",
|
||||
}
|
||||
)
|
||||
|
||||
count = run_chapter_assembly_workers(
|
||||
project_root=tmp_path,
|
||||
briefs=[chapter_brief("ch01", ["src_001", "src_002"]), chapter_brief("ch02", ["src_003", "src_004"])],
|
||||
runtime=runtime,
|
||||
client_factory=lambda _role: fake,
|
||||
workers=2,
|
||||
)
|
||||
|
||||
assert count == 1
|
||||
assert (tmp_path / "phase2/drafts/ch01.md").exists()
|
||||
assert not (tmp_path / "phase2/drafts/ch02.md").exists()
|
||||
error_path = tmp_path / "phase2/chapter_errors/ch02.json"
|
||||
assert error_path.exists()
|
||||
error = json.loads(error_path.read_text(encoding="utf-8"))
|
||||
assert error["chapter_id"] == "ch02"
|
||||
assert error["status"] == "failed"
|
||||
assert "src_fake" in error["error"]
|
||||
Some files were not shown because too many files have changed in this diff Show More
Reference in New Issue
Block a user