File size: 2,764 Bytes
3be54c6 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 | # tau-rag production environment variables
# Copy to .env (gitignored) and fill in real values before `docker compose up`.
# ==================================================== CORE
# Preset: mock | no_llm | hebrew_legal | hebrew_legal_prod | hebrew_dense
TAU_RAG_PRESET=hebrew_legal_prod
# Bind address / port (inside container; map via compose)
TAU_RAG_HOST=0.0.0.0
TAU_RAG_PORT=8000
# Number of uvicorn workers (rule of thumb: 2 × cores)
TAU_RAG_WORKERS=4
# Log level: DEBUG | INFO | WARNING | ERROR
TAU_RAG_LOG_LEVEL=INFO
# Request timeout (seconds) — hard kill above this
TAU_RAG_REQUEST_TIMEOUT_S=30
# ==================================================== AUTH
# Require API keys on all /v1/* endpoints. Highly recommended in prod.
TAU_RAG_AUTH_REQUIRED=true
# Seed an admin API key on first boot (delete after rotating in-app).
# Leave empty to disable auto-seed.
TAU_RAG_SEED_ADMIN_KEY=
# HMAC signing secret (32+ chars, hex or random). Required if hmac enabled.
TAU_RAG_HMAC_SECRET=
# ==================================================== DATA
# Root of the legal corpus. Mount as read-only volume.
LAWDBHEB_ROOT=/data/LawDBHeb
# Which subfolder under LAWDBHEB_ROOT to load (legal_rag_index_hybrid, etc.)
LAWDBHEB_INDEX_SUBFOLDER=legal_rag_index_hybrid
# Persistent runtime directory (signals, snapshots, metrics). Mount as volume.
TAU_RAG_RUNTIME_DIR=/app/runtime
# ==================================================== LLM PROVIDERS
# Only set the one(s) you actually use.
ANTHROPIC_API_KEY=
OPENAI_API_KEY=
# Default generation model (override via preset)
TAU_RAG_LLM_MODEL=claude-sonnet-4-6
TAU_RAG_LLM_TEMPERATURE=0.2
TAU_RAG_LLM_MAX_TOKENS=800
# ==================================================== OBSERVABILITY
# Prometheus scrape endpoint exposed at /metrics when true
TAU_RAG_PROMETHEUS_ENABLED=true
# OpenTelemetry OTLP endpoint (empty = disabled)
OTEL_EXPORTER_OTLP_ENDPOINT=
OTEL_SERVICE_NAME=tau-rag
# Sentry DSN for error tracking (empty = disabled)
SENTRY_DSN=
# ==================================================== HUGGING FACE
# Cache dir for transformers/sentence-transformers (mount as named volume)
HF_HOME=/root/.cache/huggingface
# Optional HF token for gated models
HF_TOKEN=
# ==================================================== RATE LIMITING
# Per-key QPS (0 = unlimited)
TAU_RAG_RATE_LIMIT_QPS=10
TAU_RAG_RATE_LIMIT_BURST=20
# ==================================================== FEATURE FLAGS
# Toggle entire middleware families (per v3.x module)
TAU_RAG_FEAT_SEMANTIC_CACHE=true
TAU_RAG_FEAT_DIVERSITY_RANKER=true
TAU_RAG_FEAT_ANSWER_GROUNDING=true
TAU_RAG_FEAT_ANSWER_HEDGING_DETECTOR=true
TAU_RAG_FEAT_ANSWER_NUMERIC_CONSISTENCY=true
TAU_RAG_FEAT_RANK_STABILITY=true
TAU_RAG_FEAT_QUERY_THROUGHPUT=true
|