# Anthropic Claude API key (required only when using Claude API provider)
ANTHROPIC_API_KEY=sk-ant-...

# Ollama host URL
# Docker Compose (app + Ollama both in Docker): use the service name
# OLLAMA_HOST=http://ollama:11434
# Host run (streamlit run app.py with Ollama in Docker port-mapped):
OLLAMA_HOST=http://localhost:11434

# Skip LiteLLM's startup network call to GitHub (fetches model cost map).
# Without this, import takes ~2-5s or hangs on restricted networks.
LITELLM_LOCAL_MODEL_COST_MAP=True