.env.example

# LLM provider order is implemented by the app:
# Groq -> Gemini -> Docker Model Runner Gemma -> deterministic template fallback.
LLM_PROVIDER=auto
USE_LLM_GENERATION=false

# Cloud LLMs. Leave blank if unavailable.
GROQ_API_KEY=
GROQ_MODEL=llama-3.3-70b-versatile

GEMINI_API_KEY=
GEMINI_MODEL=gemini-2.5-flash

# Docker Model Runner local fallback.
# Enable host TCP access in Docker Desktop Model Runner settings.
DMR_BASE_URL=http://localhost:12434/engines/v1
DMR_MODEL=gemma4:4B-Q4_K_XL

# Qdrant Docker service.
VECTOR_BACKEND=qdrant
QDRANT_URL=http://localhost:6333
QDRANT_COLLECTION=support_corpus
QDRANT_IMAGE=qdrant/qdrant:latest
QDRANT_HTTP_PORT=6333
QDRANT_GRPC_PORT=6334
EMBEDDING_MODEL=BAAI/bge-small-en

# Local cache for BM25/index metadata built by the app.
CACHE_DIR=code/.cache
FASTEMBED_CACHE_PATH=code/.cache/fastembed