308 lines
12 KiB
Python
308 lines
12 KiB
Python
from __future__ import annotations
|
|
|
|
import json
|
|
import sys
|
|
from pathlib import Path
|
|
|
|
import pytest
|
|
|
|
REPO_ROOT = Path(__file__).resolve().parents[1]
|
|
if str(REPO_ROOT) not in sys.path:
|
|
sys.path.insert(0, str(REPO_ROOT))
|
|
|
|
from scripts.lib.model_config import resolve_model_profile
|
|
from scripts.runtime.roles import resolve_runtime_profile
|
|
from scripts.runtime.methods import ResearchMethodRegistry
|
|
from scripts.runtime.phase1 import build_chapter_planning
|
|
from scripts.runtime.skills import SkillRegistry
|
|
from scripts.runtime.tasks import (
|
|
TaskCard,
|
|
detect_dependency_cycles,
|
|
generate_task_cards,
|
|
generate_task_cards_from_research_brief,
|
|
validate_packet,
|
|
validate_task_cards,
|
|
)
|
|
|
|
|
|
def test_skill_registry_uses_agents_skills_as_canonical() -> None:
|
|
registry = SkillRegistry()
|
|
|
|
names = registry.list_names()
|
|
|
|
assert "search-strategy" in names
|
|
assert "search-gateway" in names
|
|
assert "source-quality" in names
|
|
assert "document-ingest" in names
|
|
assert "deep-research" in names
|
|
assert registry.validate()["ok"] is True
|
|
|
|
|
|
def test_model_profile_exposes_task_types_and_role_defaults() -> None:
|
|
resolved = resolve_model_profile(profile="medium")
|
|
|
|
assert resolved["roles"]["dr_pm"]
|
|
assert resolved["task_types"]["source_discovery"] == "dr_searcher"
|
|
assert resolved["task_types"]["chapter_assembly"] == "dr_analyst"
|
|
|
|
|
|
def test_runtime_profile_resolves_task_model_and_skills() -> None:
|
|
runtime = resolve_runtime_profile(profile="medium")
|
|
|
|
worker = runtime.role_for_task("evidence_packet")
|
|
|
|
assert worker.name == "dr_analyst"
|
|
assert worker.model
|
|
assert "evidence-table" in worker.skills
|
|
assert "search-gateway" in worker.skills
|
|
assert worker.max_concurrency >= 1
|
|
|
|
|
|
def test_generate_task_cards_from_chinese_framework() -> None:
|
|
framework = """
|
|
# 研究框架
|
|
|
|
## 第1章 GLP-1 产业链的增量来自适应症扩张
|
|
|
|
研究思路：围绕临床、监管、竞争格局和生产供应链展开。
|
|
|
|
## 第2章 供应链瓶颈决定国产替代窗口
|
|
|
|
研究思路：围绕专利、上游原料、产能和中国市场展开。
|
|
"""
|
|
|
|
cards = generate_task_cards("glp1-test", framework, axes=["clinical", "regulatory"])
|
|
|
|
assert [card.task_id for card in cards] == [
|
|
"ch01-clinical",
|
|
"ch01-regulatory",
|
|
"ch02-clinical",
|
|
"ch02-regulatory",
|
|
]
|
|
assert cards[0].output_packet == "phase2/packets/ch01-clinical.json"
|
|
assert cards[0].chapter_title == "GLP-1 产业链的增量来自适应症扩张"
|
|
assert cards[0].research_goal
|
|
assert "search-gateway" in cards[0].required_skills
|
|
assert cards[0].expected_evidence["min_tier_1_2_sources"] == 2
|
|
assert cards[0].stop_conditions
|
|
|
|
|
|
def test_generate_task_cards_from_research_brief_carries_prompt_and_skills() -> None:
|
|
brief = {
|
|
"research_method": "gmp_quality_operations_diagnosis",
|
|
"work_language": "zh",
|
|
"tone": "面向管理层的事实型整改诊断",
|
|
"task_planning": {
|
|
"required_skills": ["search-gateway", "evidence-table", "source-quality"],
|
|
"search_routes_by_axis": {
|
|
"quality_system_gap": ["general", "news"],
|
|
"counter": ["scholar", "general"],
|
|
},
|
|
"axis_prompt_briefs": {
|
|
"quality_system_gap": "把现场发现映射到质量体系流程缺口和法规要求。",
|
|
"counter": "主动寻找能削弱或限定结论的反方证据。",
|
|
},
|
|
"stop_conditions": ["每张卡至少形成 3 条可追溯证据。"],
|
|
},
|
|
"materials": [{"path": "phase0/extracted/audit.md", "role": "site_evidence"}],
|
|
}
|
|
framework = "## 第1章 质量体系闭环能力决定整改可信度\n\n研究思路。"
|
|
|
|
cards = generate_task_cards_from_research_brief(
|
|
"baifan-test",
|
|
framework,
|
|
brief,
|
|
axes=["quality_system_gap", "counter"],
|
|
)
|
|
|
|
assert [card.task_id for card in cards] == ["ch01-quality_system_gap", "ch01-counter"]
|
|
assert cards[0].prompt_brief == "把现场发现映射到质量体系流程缺口和法规要求。"
|
|
assert cards[0].research_method == "gmp_quality_operations_diagnosis"
|
|
assert cards[0].allowed_materials == ["phase0/extracted/audit.md"]
|
|
assert cards[0].preferred_model_role == "dr_analyst"
|
|
assert cards[1].preferred_model_role == "dr_verifier"
|
|
assert "search-gateway" in cards[0].required_skills
|
|
|
|
|
|
def test_gmp_task_cards_include_fda_enforcement_route() -> None:
|
|
brief = {
|
|
"research_method": "gmp_quality_operations_diagnosis",
|
|
"task_planning": {
|
|
"search_routes_by_axis": {
|
|
"quality_system_gap": ["fda", "general"],
|
|
},
|
|
},
|
|
}
|
|
framework = "## 第1章 偏差和 CAPA 闭环能力决定质量体系可信度\n\n研究思路。"
|
|
|
|
cards = generate_task_cards_from_research_brief(
|
|
"baifan-test",
|
|
framework,
|
|
brief,
|
|
axes=["quality_system_gap"],
|
|
)
|
|
|
|
assert cards[0].search_routes == ["fda", "general"]
|
|
assert "FDA Warning Letters" in " ".join(cards[0].questions)
|
|
assert "fda_warning_letter_or_meeting_record" in cards[0].expected_evidence["preferred_evidence_types"]
|
|
|
|
|
|
def test_integrated_chapter_mode_is_method_driven_not_gmp_hardcoded() -> None:
|
|
brief = {
|
|
"research_method": "mckinsey_market",
|
|
"phase2_mode": "chapter_integrated",
|
|
"task_planning": {},
|
|
}
|
|
framework = "## 第1章 市场需求正在被支付政策重塑\n\n研究思路。"
|
|
|
|
cards = generate_task_cards_from_research_brief(
|
|
"market-test",
|
|
framework,
|
|
brief,
|
|
)
|
|
|
|
assert [card.task_id for card in cards] == ["ch01-chapter_integrated"]
|
|
assert "literature evidence" in " ".join(cards[0].questions)
|
|
assert "FDA Warning Letters" not in " ".join(cards[0].questions)
|
|
|
|
|
|
def test_integrated_task_card_uses_phase1_chapter_planning() -> None:
|
|
brief = {
|
|
"research_method": "gmp_quality_operations_diagnosis",
|
|
"phase2_mode": "chapter_integrated",
|
|
"chapter_planning": [
|
|
{
|
|
"chapter_id": "ch01",
|
|
"title": "审计发现应先转化为商业化阶段门缺口",
|
|
"core_question": "本章要判断审计发现是否反映阶段门缺口。",
|
|
"bold_hypothesis": "大胆假设：风险项计数低估了商业化 readiness 缺口。",
|
|
"verification_plan": ["提取现场材料原文", "检索官方法规和执法案例"],
|
|
"evidence_lanes": ["site audit findings", "official baseline"],
|
|
"minimum_evidence": {"local_material_quotes": 2},
|
|
"phase2_prompt_context": "章节：ch01\n必须围绕阶段门缺口求证。",
|
|
}
|
|
],
|
|
"task_planning": {"phase2_mode": "chapter_integrated"},
|
|
}
|
|
framework = "## 第1章 审计发现应先转化为商业化阶段门缺口\n\n研究思路。"
|
|
|
|
cards = generate_task_cards_from_research_brief("baifan-test", framework, brief)
|
|
|
|
assert cards[0].prompt_brief == "章节：ch01\n必须围绕阶段门缺口求证。"
|
|
assert cards[0].research_goal == "本章要判断审计发现是否反映阶段门缺口。"
|
|
assert "大胆假设：风险项计数低估了商业化 readiness 缺口。" in cards[0].questions
|
|
assert cards[0].expected_evidence["phase1_minimum_evidence"] == {"local_material_quotes": 2}
|
|
assert cards[0].expected_evidence["must_address_phase1_hypothesis"] is True
|
|
assert any("Phase1 的大胆假设" in item for item in cards[0].stop_conditions)
|
|
|
|
|
|
def test_phase1_gmp_hypotheses_are_not_title_restatements(tmp_path: Path) -> None:
|
|
project = tmp_path / "baifan"
|
|
project.mkdir()
|
|
(project / "phase0/extracted").mkdir(parents=True)
|
|
material = project / "phase0/extracted/audit.md"
|
|
material.write_text(
|
|
"人员培训记录齐全，但无菌操作动作违反 First Air 原则，需要进一步培训。\n"
|
|
"复盘显示 owner、关闭证据和问题升级机制仍需补齐。\n",
|
|
encoding="utf-8",
|
|
)
|
|
manifest = {
|
|
"topic": "白帆生物 GMP 与运营诊断",
|
|
"material_inventory": [{"extracted_to": "phase0/extracted/audit.md"}],
|
|
}
|
|
method = ResearchMethodRegistry().get("gmp_quality_operations_diagnosis")
|
|
|
|
plans = build_chapter_planning(
|
|
project,
|
|
manifest,
|
|
method,
|
|
["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"],
|
|
quota=2000,
|
|
)
|
|
|
|
assert "关键解释变量" not in plans[0]["bold_hypothesis"]
|
|
assert "不缺培训台账" in plans[0]["bold_hypothesis"]
|
|
assert "固定节奏和可视化管理系统" in plans[1]["bold_hypothesis"]
|
|
assert plans[0]["core_question"] != plans[0]["title"]
|
|
assert "章节成稿应围绕这一观点展开" not in plans[1]["writing_claim"]
|
|
|
|
|
|
def test_research_brief_without_materials_falls_back_to_material_digest() -> None:
|
|
brief = {
|
|
"research_method": "mckinsey_market",
|
|
"phase2_mode": "chapter_integrated",
|
|
"phase1_inputs": {"material_digest": "phase1/material_digest.md"},
|
|
}
|
|
framework = "## 第1章 市场需求正在被支付政策重塑\n\n研究思路。"
|
|
|
|
cards = generate_task_cards_from_research_brief("market-test", framework, brief)
|
|
|
|
assert cards[0].allowed_materials == ["phase1/material_digest.md"]
|
|
|
|
|
|
def test_task_card_validation_rejects_duplicates_and_cycles() -> None:
|
|
cards = [
|
|
TaskCard(task_id="a", chapter_ids=["ch01"], topic_axis="clinical", questions=["q"], search_routes=["scholar"], output_packet="phase2/packets/a.json", dependencies=["b"]),
|
|
TaskCard(task_id="b", chapter_ids=["ch01"], topic_axis="regulatory", questions=["q"], search_routes=["general"], output_packet="phase2/packets/b.json", dependencies=["a"]),
|
|
]
|
|
|
|
with pytest.raises(ValueError, match="dependency cycle"):
|
|
detect_dependency_cycles(cards)
|
|
|
|
with pytest.raises(ValueError, match="duplicate task_id"):
|
|
validate_task_cards([cards[0], cards[0]])
|
|
|
|
|
|
def test_packet_validation_requires_sources_and_counter_evidence() -> None:
|
|
packet = {
|
|
"task_id": "ch01-clinical",
|
|
"claims": [{"claim": "结论", "source_ids": ["src_001"]}],
|
|
"evidence_items": [{"source_id": "src_001", "summary": "证据"}],
|
|
"counter_evidence": [],
|
|
"source_ids": ["src_001"],
|
|
"source_quality_notes": ["Tier 1"],
|
|
"open_questions": [],
|
|
"raw_quotes_or_notes": ["Original English excerpt allowed."],
|
|
}
|
|
|
|
with pytest.raises(ValueError, match="counter_evidence"):
|
|
validate_packet(packet)
|
|
|
|
packet["counter_evidence"] = [{"claim": "限制", "source_ids": ["src_002"]}]
|
|
with pytest.raises(ValueError, match="not declared"):
|
|
validate_packet(packet)
|
|
|
|
packet["source_ids"].append("src_002")
|
|
packet["sources"] = [
|
|
{"id": "src_001", "title": "来源1", "url": "https://example.com/1"},
|
|
{"id": "src_002", "title": "来源2", "url": "https://example.com/2"},
|
|
]
|
|
validate_packet(packet)
|
|
|
|
|
|
def test_skill_sync_copies_to_adapter_dirs(tmp_path: Path) -> None:
|
|
canonical = tmp_path / "skills"
|
|
target = tmp_path / "adapter" / "skills"
|
|
source_skill = canonical / "demo"
|
|
source_skill.mkdir(parents=True)
|
|
(source_skill / "SKILL.md").write_text("---\nname: demo\n---\n\nBody\n", encoding="utf-8")
|
|
|
|
registry = SkillRegistry(canonical_dir=canonical)
|
|
copied = registry.sync_to([target])
|
|
|
|
assert copied == 1
|
|
assert (target / "demo" / "SKILL.md").read_text(encoding="utf-8").endswith("Body\n")
|
|
|
|
|
|
def test_skill_sync_skips_canonical_dir_to_avoid_deleting_source(tmp_path: Path) -> None:
|
|
canonical = tmp_path / "skills"
|
|
source_skill = canonical / "demo"
|
|
source_skill.mkdir(parents=True)
|
|
(source_skill / "SKILL.md").write_text("---\nname: demo\n---\n\nBody\n", encoding="utf-8")
|
|
|
|
registry = SkillRegistry(canonical_dir=canonical)
|
|
copied = registry.sync_to([canonical])
|
|
|
|
assert copied == 0
|
|
assert (source_skill / "SKILL.md").exists()
|