Files
infra-docker-configs/servers/hindsight/hindsight/.env.example
T

89 lines
3.7 KiB
Bash

TZ=Asia/Shanghai
# PostgreSQL. The password and the derived database URL stay in Arcane only.
HINDSIGHT_DB_USER=hindsight_user
HINDSIGHT_DB_PASSWORD=replace-with-a-random-url-safe-password
HINDSIGHT_DB_NAME=hindsight_db
# Hindsight API and control-plane UI.
HINDSIGHT_ENABLE_API=true
HINDSIGHT_ENABLE_CP=true
HINDSIGHT_API_HOST=0.0.0.0
HINDSIGHT_API_PORT=8888
HINDSIGHT_API_LOG_LEVEL=info
HINDSIGHT_API_WORKERS=1
HINDSIGHT_API_VECTOR_EXTENSION=pgvector
HINDSIGHT_API_TEXT_SEARCH_EXTENSION=native
HINDSIGHT_API_TEXT_SEARCH_EXTENSION_NATIVE_LANGUAGE=simple
HINDSIGHT_API_ENABLE_AUTO_CONSOLIDATION=true
HINDSIGHT_API_MCP_ENABLED=true
# ZenMux, using its OpenAI-compatible endpoint. Store the real key in Arcane.
HINDSIGHT_API_LLM_PROVIDER=openai
HINDSIGHT_API_LLM_BASE_URL=https://zenmux.ai/api/v1
HINDSIGHT_API_LLM_API_KEY=replace-with-zenmux-subscription-key
HINDSIGHT_API_LLM_MODEL=openai/gpt-6-luna
HINDSIGHT_API_LLM_MAX_CONCURRENT=4
HINDSIGHT_API_LLM_MAX_RETRIES=4
HINDSIGHT_API_LLM_INITIAL_BACKOFF=2
HINDSIGHT_API_LLM_MAX_BACKOFF=120
HINDSIGHT_API_LLM_TIMEOUT=300
HINDSIGHT_API_LLM_TEMPERATURE=none
HINDSIGHT_API_LLM_TEMPERATURE_VERIFICATION=none
HINDSIGHT_API_LLM_TEMPERATURE_RETAIN=none
HINDSIGHT_API_LLM_TEMPERATURE_CONSOLIDATION=none
HINDSIGHT_API_LLM_TEMPERATURE_REFLECT=none
HINDSIGHT_API_LLM_PROMPT_CACHE_ENABLED=true
HINDSIGHT_API_LLM_SEND_BANK_AS_USER=true
# Retain: fast structured extraction.
HINDSIGHT_API_RETAIN_LLM_PROVIDER=openai
HINDSIGHT_API_RETAIN_LLM_BASE_URL=https://zenmux.ai/api/v1
HINDSIGHT_API_RETAIN_LLM_MODEL=openai/gpt-6-luna
HINDSIGHT_API_RETAIN_LLM_MAX_CONCURRENT=2
HINDSIGHT_API_RETAIN_LLM_TIMEOUT=300
HINDSIGHT_API_RETAIN_LLM_REASONING_EFFORT=low
# Consolidation: stronger cross-memory organization, quota-aware concurrency.
HINDSIGHT_API_CONSOLIDATION_LLM_PROVIDER=openai
HINDSIGHT_API_CONSOLIDATION_LLM_BASE_URL=https://zenmux.ai/api/v1
HINDSIGHT_API_CONSOLIDATION_LLM_MODEL=x-ai/grok-4.7
HINDSIGHT_API_CONSOLIDATION_LLM_MAX_CONCURRENT=1
HINDSIGHT_API_CONSOLIDATION_LLM_TIMEOUT=600
HINDSIGHT_API_CONSOLIDATION_LLM_BATCH_SIZE=10
HINDSIGHT_API_CONSOLIDATION_LLM_REASONING_EFFORT=low
# Reflect: highest-quality synthesis and belief evolution.
HINDSIGHT_API_REFLECT_LLM_PROVIDER=openai
HINDSIGHT_API_REFLECT_LLM_BASE_URL=https://zenmux.ai/api/v1
HINDSIGHT_API_REFLECT_LLM_MODEL=anthropic/claude-sonnet-5.5
HINDSIGHT_API_REFLECT_LLM_MAX_CONCURRENT=1
HINDSIGHT_API_REFLECT_LLM_TIMEOUT=600
# Local TEI embeddings. BGE-M3 emits 1024-dimensional multilingual vectors.
# OpenRouter settings remain available for a manual rollback; Hindsight does not
# provide a native embedding-provider fallback chain.
HINDSIGHT_API_OPENROUTER_API_KEY=replace-with-openrouter-key
HINDSIGHT_API_EMBEDDINGS_PROVIDER=tei
HINDSIGHT_API_EMBEDDINGS_TEI_URL=http://192.168.50.100:18083
HINDSIGHT_API_EMBEDDINGS_OPENROUTER_MODEL=baai/bge-m3
HINDSIGHT_API_EMBEDDINGS_OPENAI_BATCH_SIZE=32
HINDSIGHT_API_EMBEDDINGS_MAX_INPUT_TOKENS=8192
# Google tested with synthetic text only; keep production reranking local.
HINDSIGHT_API_RERANKER_PROVIDER=tei
HINDSIGHT_API_RERANKER_LITELLM_API_BASE=http://192.168.50.49:4000
HINDSIGHT_API_RERANKER_LITELLM_API_KEY=replace-with-litellm-client-key
HINDSIGHT_API_RERANKER_LITELLM_MODEL=google-rerank-default
HINDSIGHT_API_RERANKER_LITELLM_TIMEOUT=10
HINDSIGHT_API_RERANKER_TEI_URL=http://192.168.50.100:18084
HINDSIGHT_API_RERANKER_TEI_BATCH_SIZE=32
HINDSIGHT_API_RERANKER_TEI_MAX_CONCURRENT=4
HINDSIGHT_API_RERANKER_TEI_HTTP_TIMEOUT=30
HINDSIGHT_API_RERANKER_MAX_CANDIDATES=100
HINDSIGHT_API_RERANKER_1_PROVIDER=openrouter
HINDSIGHT_API_RERANKER_1_OPENROUTER_API_KEY=replace-with-openrouter-key
HINDSIGHT_API_RERANKER_1_OPENROUTER_MODEL=cohere/rerank-4-fast
HINDSIGHT_API_RERANKER_1_OPENROUTER_TIMEOUT=30
HINDSIGHT_API_RERANKER_2_PROVIDER=rrf