docker-compose.yml

# Deep Research Agent - Docker Compose
#
# Quick Start:
#   1. Copy .env.example to .env and add your API keys
#   2. Run: docker-compose up
#   3. Access UI: http://localhost:8080
#   4. Execute research via API or UI
#
# Required API Keys (set in .env):
#   - OPENROUTER_API_KEY: For LLM inference (https://openrouter.ai/keys)
#     OR use local Ollama: set OLLAMA_BASE_URL and DEFAULT_MODEL=ollama/<model>
#   - At least ONE search provider: JINA_API_KEY, TAVILY_API_KEY, FIRECRAWL_API_KEY, or SERPER_API_KEY

services:
  # Agent Field Control Plane
  agentfield:
    image: agentfield/control-plane:latest
    container_name: agentfield-server
    ports:
      - "8080:8080"
    environment:
      - AGENTFIELD_HTTP_ADDR=0.0.0.0:8080
      - AGENTFIELD_STORAGE_MODE=local
    volumes:
      - agentfield-data:/data
    restart: unless-stopped

  # Deep Research Agent
  deep-research:
    build:
      context: .
      dockerfile: Dockerfile
    container_name: deep-research-agent
    environment:
      # Agent Field connection
      - AGENTFIELD_SERVER=http://agentfield:8080
      - PORT=8001
      # Callback URL for control plane to reach this agent
      - AGENT_CALLBACK_URL=http://deep-research:8001
      # LLM API Key (from .env file)
      - OPENROUTER_API_KEY=${OPENROUTER_API_KEY}
      # Web Search API Keys (configure at least one)
      - JINA_API_KEY=${JINA_API_KEY:-}
      - TAVILY_API_KEY=${TAVILY_API_KEY:-}
      - FIRECRAWL_API_KEY=${FIRECRAWL_API_KEY:-}
      - SERPER_API_KEY=${SERPER_API_KEY:-}
      # Optional: Force specific search provider
      - SEARCH_PROVIDER=${SEARCH_PROVIDER:-}
      # Model configuration
      - DEFAULT_MODEL=${DEFAULT_MODEL:-openrouter/deepseek/deepseek-chat-v3.1}
      - TEMPERATURE=${TEMPERATURE:-0.6}
      # Optional: Local Ollama configuration (set OLLAMA_BASE_URL to use local LLM)
      - OLLAMA_BASE_URL=${OLLAMA_BASE_URL:-}
    ports:
      - "8001:8001"
    depends_on:
      - agentfield
    healthcheck:
      test: ["CMD", "curl", "-f", "http://localhost:8001/health"]
      interval: 30s
      timeout: 10s
      retries: 3
      start_period: 15s
    restart: unless-stopped

volumes:
  agentfield-data: