docker-compose.yml

# RECAP backend - one-command local run.
#
#   docker compose up --build      # start (first build downloads ML wheels)
#   docker compose down            # stop
#
# Then load the extension in Chrome (chrome://extensions → Load unpacked →
# extension/) - it talks to the backend at http://localhost:8000.
services:
  recap:
    build: .
    image: recap-backend:latest
    container_name: recap-backend
    # Publish ONLY to the host loopback - the backend is never exposed to the LAN.
    ports:
      - "127.0.0.1:8000:8000"
    # Secrets/config come from .env at runtime (never baked into the image).
    # Optional: you can also set keys via the extension Options page instead.
    env_file:
      - path: .env
        required: false
    environment:
      # Bind inside the container; the loopback-only publish above keeps it private.
      HOST: "0.0.0.0"
      PORT: "8000"
    volumes:
      # The index (safe to delete to reset) and the downloaded model cache persist
      # across restarts. ./data mirrors the manual-setup layout.
      - ./data:/app/data
      - recap-hf-cache:/app/.cache/huggingface
    # Lets the container reach an Ollama running on the HOST. When using Ollama,
    # set in .env:  LLM_BASE_URL=http://host.docker.internal:11434/v1  (provider "custom"),
    # and for local Ollama embeddings: EMBEDDING_BASE_URL=http://host.docker.internal:11434/v1
    extra_hosts:
      - "host.docker.internal:host-gateway"
    healthcheck:
      test: ["CMD", "curl", "-fsS", "http://localhost:8000/health"]
      interval: 30s
      timeout: 5s
      retries: 3
      start_period: 120s
    restart: unless-stopped

volumes:
  recap-hf-cache: