docker-compose.yml

services:
  ollama:
    image: ollama/ollama:latest
    container_name: skn-ollama
    ports:
      - "11434:11434"
    volumes:
      - ollama_data:/root/.ollama
    healthcheck:
      test: ["CMD", "curl", "-f", "http://localhost:11434/api/tags"]
      interval: 10s
      timeout: 5s
      retries: 5
    networks:
      - agentic-rag-net

  backend:
    build:
      context: .
      dockerfile: backend/Dockerfile
    container_name: skn-backend
    depends_on:
      ollama:
        condition: service_healthy
    ports:
      - "8000:8000"
    environment:
      OLLAMA_BASE_URL: ${OLLAMA_BASE_URL:-http://ollama:11434}
      OLLAMA_CHAT_MODEL: ${OLLAMA_CHAT_MODEL:-qwen3.5:4b}
      OLLAMA_EMBEDDING_MODEL: ${OLLAMA_EMBEDDING_MODEL:-nomic-embed-text:latest}
      CHROMA_COLLECTION_NAME: ${CHROMA_COLLECTION_NAME:-knowledge_base}
      CHROMA_PERSIST_DIRECTORY: /data/chroma
      RETRIEVAL_TOP_K: ${RETRIEVAL_TOP_K:-4}
      RETRIEVAL_PER_QUERY_K: ${RETRIEVAL_PER_QUERY_K:-4}
      RETRIEVAL_BM25_K: ${RETRIEVAL_BM25_K:-6}
      RETRIEVAL_MIN_SCORE: ${RETRIEVAL_MIN_SCORE:-0.42}
      RETRIEVAL_MIN_CHUNKS: ${RETRIEVAL_MIN_CHUNKS:-3}
      RETRIEVAL_MIN_SOURCE_DIVERSITY: ${RETRIEVAL_MIN_SOURCE_DIVERSITY:-2}
      RERANK_ENABLED: ${RERANK_ENABLED:-true}
      DEBUG_TRACE_ENABLED: ${DEBUG_TRACE_ENABLED:-false}
      MAX_RETRIEVAL_RETRIES: ${MAX_RETRIEVAL_RETRIES:-1}
      MAX_VALIDATION_RETRIES: ${MAX_VALIDATION_RETRIES:-1}
      RUNTIME_PROFILE: ${RUNTIME_PROFILE:-balanced}
      LOW_LATENCY_SKIP_OVERLAP_CHECK: ${LOW_LATENCY_SKIP_OVERLAP_CHECK:-true}
      MODEL_TEMPERATURE: ${MODEL_TEMPERATURE:-1.0}
      MODEL_TOP_P: ${MODEL_TOP_P:-0.95}
      MODEL_TOP_K: ${MODEL_TOP_K:-20}
      MODEL_MIN_P: ${MODEL_MIN_P:-0.0}
      MODEL_PRESENCE_PENALTY: ${MODEL_PRESENCE_PENALTY:-1.5}
      MODEL_REPETITION_PENALTY: ${MODEL_REPETITION_PENALTY:-1.0}
      MODEL_REQUEST_TIMEOUT_SECONDS: ${MODEL_REQUEST_TIMEOUT_SECONDS:-45}
      LOW_LATENCY_MODEL_REQUEST_TIMEOUT_SECONDS: ${LOW_LATENCY_MODEL_REQUEST_TIMEOUT_SECONDS:-25}
      LOW_LATENCY_MAX_OUTPUT_TOKENS: ${LOW_LATENCY_MAX_OUTPUT_TOKENS:-300}
      MODEL_MAX_OUTPUT_TOKENS: ${MODEL_MAX_OUTPUT_TOKENS:-420}
      CHUNK_SIZE: ${CHUNK_SIZE:-1200}
      CHUNK_OVERLAP: ${CHUNK_OVERLAP:-200}
      ALLOWED_SOURCE_DOMAINS: ${ALLOWED_SOURCE_DOMAINS:-support.atlassian.com,confluence.atlassian.com,docs.langchain.com,python.langchain.com,langchain.com,atlassian.com,www.atlassian.com,www.langchain.com,arxiv.org,cdn.openai.com,openai.com,www.openai.com,anthropic.com,www.anthropic.com,ai.google.dev,developers.google.com,research.google,paperswithcode.com,langchain-ai.github.io,github.io}
      PUBLIC_SOURCES_ONLY: ${PUBLIC_SOURCES_ONLY:-true}
    volumes:
      - chroma_data:/data/chroma
    healthcheck:
      test: ["CMD", "curl", "-f", "http://localhost:8000/health"]
      interval: 10s
      timeout: 5s
      retries: 5
      start_period: 30s
    networks:
      - agentic-rag-net

  frontend:
    image: node:20-alpine
    container_name: skn-frontend
    depends_on:
      backend:
        condition: service_healthy
    working_dir: /app/react-frontend
    ports:
      - "5173:5173"
    volumes:
      - .:/app
      - /app/react-frontend/node_modules
    environment:
      VITE_BACKEND_URL: http://localhost:8000
      VITE_BACKEND_STREAM_URL: http://localhost:8000/chat/stream
      NODE_ENV: development
    command: sh -c "npm install && npm run dev -- --host"
    networks:
      - agentic-rag-net

volumes:
  chroma_data:
  ollama_data:

networks:
  agentic-rag-net:
    driver: bridge