docker-compose.dev.yml

# ============================================================
# Code Documentation Assistant — Docker Compose (dev branch)
# ============================================================
# Changes from main docker-compose.yml:
#   - MLflow tracking server (always-on, not behind profile)
#   - vLLM service (GPU required, behind 'vllm' profile)
#   - HITL_ENABLED, OUTPUT_REVIEW_MODE, QUALITY_GATE_THRESHOLD
#   - INFERENCE_BACKEND: "ollama" (default) | "vllm"
#   - CONFIDENCE_THRESHOLD, MAX_RETRIEVAL_ATTEMPTS configurable
#   - src/ bind-mount for live-reload during development
#
# Usage:
#   docker compose -f docker-compose.dev.yml up --build
#   HITL_ENABLED=false docker compose -f docker-compose.dev.yml up
#   OUTPUT_REVIEW_MODE=supervisor docker compose -f docker-compose.dev.yml up
#   INFERENCE_BACKEND=vllm docker compose -f docker-compose.dev.yml --profile vllm up
#   MODEL_TIER=lightweight docker compose -f docker-compose.dev.yml up
#
# Access:
#   Streamlit UI:  http://localhost:8501
#   MLflow UI:     http://localhost:5000
# ============================================================

services:

  # --- Ollama (default inference backend) ---
  ollama:
    image: ollama/ollama:latest
    container_name: code-doc-ollama
    ports:
      - "11434:11434"
    volumes:
      - ollama_data:/root/.ollama
    environment:
      - OLLAMA_HOST=0.0.0.0
    healthcheck:
      test: ["CMD", "ollama", "list"]
      interval: 10s
      timeout: 5s
      retries: 5
      start_period: 15s
    # GPU: docker compose -f docker-compose.dev.yml -f docker-compose.gpu.yml up

  # --- Model bootstrap ---
  ollama-bootstrap:
    image: ollama/ollama:latest
    container_name: code-doc-bootstrap
    depends_on:
      ollama:
        condition: service_healthy
    environment:
      - OLLAMA_HOST=ollama:11434
      - MODEL_TIER=${MODEL_TIER:-full}
      - EMBEDDING_MODEL=${EMBEDDING_MODEL:-nomic-embed-text}
    entrypoint: >
      bash -c '
        case "$${MODEL_TIER}" in
          full)        MODEL="mistral-nemo" ;;
          balanced)    MODEL="qwen2.5-coder:7b" ;;
          lightweight) MODEL="phi3.5" ;;
          *)           MODEL="mistral-nemo" ;;
        esac
        echo "Pulling LLM: $${MODEL}"
        ollama pull $${MODEL}
        echo "Pulling embedding model: $${EMBEDDING_MODEL}"
        ollama pull $${EMBEDDING_MODEL}
        echo "Bootstrap complete."
      '
    restart: "no"

  # --- vLLM (alternative inference backend, GPU required) ---
  # Enable with: INFERENCE_BACKEND=vllm docker compose -f docker-compose.dev.yml --profile vllm up
  vllm:
    image: vllm/vllm-openai:latest
    container_name: code-doc-vllm
    profiles: ["vllm"]
    ports:
      - "8080:8000"
    volumes:
      - vllm_model_cache:/root/.cache/huggingface
    environment:
      - HUGGING_FACE_HUB_TOKEN=${HF_TOKEN:-}
    command:
      - "--model"
      - "${VLLM_MODEL:-mistralai/Mistral-7B-Instruct-v0.3}"
      - "--tensor-parallel-size"
      - "${VLLM_TP:-1}"
      - "--max-model-len"
      - "32768"
      - "--gpu-memory-utilization"
      - "0.90"
    healthcheck:
      test: ["CMD", "python3", "-c", "import urllib.request; urllib.request.urlopen('http://localhost:8000/health')"]
      interval: 30s
      timeout: 10s
      retries: 5
      start_period: 60s
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: all
              capabilities: [gpu]

  # --- ChromaDB ---
  chromadb:
    image: chromadb/chroma:0.6.3
    container_name: code-doc-chromadb
    ports:
      - "8000:8000"
    volumes:
      - chroma_data:/chroma/chroma
    environment:
      - IS_PERSISTENT=TRUE
      - PERSIST_DIRECTORY=/chroma/chroma
      - ANONYMIZED_TELEMETRY=false
    healthcheck:
      test: ["CMD", "python3", "-c", "import urllib.request; urllib.request.urlopen('http://localhost:8000/api/v1/heartbeat')"]
      interval: 10s
      timeout: 5s
      retries: 5
      start_period: 10s

  # --- MLflow tracking server (always-on in dev) ---
  mlflow:
    image: ghcr.io/mlflow/mlflow:latest
    container_name: code-doc-mlflow
    ports:
      - "5000:5000"
    volumes:
      - mlflow_data:/mlflow
    command:
      - mlflow
      - server
      - --host=0.0.0.0
      - --port=5000
      - --backend-store-uri=sqlite:////mlflow/mlflow.db
      - --default-artifact-root=/mlflow/artifacts
    healthcheck:
      test: ["CMD", "python3", "-c", "import urllib.request; urllib.request.urlopen('http://localhost:5000/health')"]
      interval: 15s
      timeout: 5s
      retries: 5
      start_period: 10s

  # --- Application (dev branch — LangGraph agent + Streamlit) ---
  app:
    build:
      context: .
      dockerfile: Dockerfile
    container_name: code-doc-app
    ports:
      - "8501:8501"
    depends_on:
      chromadb:
        condition: service_healthy
      mlflow:
        condition: service_healthy
    environment:
      # Inference backend
      - INFERENCE_BACKEND=${INFERENCE_BACKEND:-ollama}
      - OLLAMA_HOST=http://ollama:11434
      - VLLM_HOST=http://vllm:8000
      - OLLAMA_MODEL=${OLLAMA_MODEL:-mistral-nemo}

      # Vector DB
      - CHROMA_HOST=http://chromadb:8000

      # MLflow
      - MLFLOW_TRACKING_URI=http://mlflow:5000

      # Model / embedding
      - MODEL_TIER=${MODEL_TIER:-full}
      - EMBEDDING_MODEL=${EMBEDDING_MODEL:-nomic-embed-text}

      # Agent behaviour
      - HITL_ENABLED=${HITL_ENABLED:-true}
      - OUTPUT_REVIEW_MODE=${OUTPUT_REVIEW_MODE:-human}
      - QUALITY_GATE_THRESHOLD=${QUALITY_GATE_THRESHOLD:-6.0}
      - CONFIDENCE_THRESHOLD=${CONFIDENCE_THRESHOLD:-0.45}
      - MAX_RETRIEVAL_ATTEMPTS=${MAX_RETRIEVAL_ATTEMPTS:-3}
      - MAX_GENERATION_ATTEMPTS=${MAX_GENERATION_ATTEMPTS:-3}
      - MAX_CONTEXT_TOKENS=${MAX_CONTEXT_TOKENS:-8000}

      - LOG_LEVEL=${LOG_LEVEL:-info}
    volumes:
      - ${REPO_PATH:-./repos}:/data/repos:ro
      - ./src:/app/src   # live-reload during development

volumes:
  ollama_data:
    driver: local
  vllm_model_cache:
    driver: local
  chroma_data:
    driver: local
  mlflow_data:
    driver: local