deploy.stack/honcho/stack.yml

# Honcho Self-Hosted 部署栈
# ============================================================
# 部署前准备（仅首次）：
#   mkdir -pv ${Volumes_Path}/{pgdata,redis,source}
#   sudo chown -R 999:999 ${Volumes_Path}/pgdata
#   sudo chown -R 999:999 ${Volumes_Path}/redis
#   git clone https://github.com/plastic-labs/honcho.git ${Volumes_Path}/honcho/source
#   cp env.cfg.example env.cfg && $EDITOR env.cfg   # 填入密码/API Key
#
# pull:: docker compose --env-file ./honcho/env.cfg -f ./honcho/stack.yml pull
# RUN:: docker compose -p honcho --env-file ./honcho/env.cfg -f ./honcho/stack.yml up -d --build
# disc::
#   - api/deriver 从 ${Volumes_Path}/honcho/source 源码 build（不是运行时 bind）
#   - DB/Redis 数据 bind 挂到 WSL/Linux 原生 ext4（不能放 /mnt/9P drvfs）
#   - 端口 8000=API, 5432=PG, 6379=Redis，全部绑定 0.0.0.0 供 LAN 访问
#   - 必须设置 HONCHO_LLM_API_KEY（占位符会导致 deriver 反复重启）
#   - deriver 不可省：没它消息能写但不会产生记忆
#   - PG 5432 与 hindsight/dbSer 撞车，通过 HONCHO_DB_PORT 区分
# ============================================================

services:
  api:
    build:
      context: ${Volumes_Path}/honcho/source
      dockerfile: Dockerfile
    container_name: honcho-api
    restart: unless-stopped
    entrypoint: ["sh", "docker/entrypoint.sh"]
    depends_on:
      database:
        condition: service_healthy
      redis:
        condition: service_healthy
    ports:
      - "${HONCHO_API_PORT:-8000}:8000"
    environment:
      - TZ=Asia/Shanghai
      - DB_CONNECTION_URI=postgresql+psycopg://${HONCHO_DB_USER}:${HONCHO_DB_PASSWORD}@database:5432/${HONCHO_DB_NAME}
      - CACHE_URL=redis://redis:6379/0?suppress=true
      - CACHE_ENABLED=true
      - LOG_LEVEL=${HONCHO_LOG_LEVEL:-INFO}
      - NAMESPACE=${HONCHO_NAMESPACE:-honcho}
      - EMBED_MESSAGES=${HONCHO_EMBED_MESSAGES:-true}
      - EMBEDDING_MODEL_CONFIG__TRANSPORT=openai
      - EMBEDDING_MODEL_CONFIG__MODEL=${HONCHO_EMBEDDING_MODEL:-Qwen/Qwen3-Embedding-0.6B}
      - EMBEDDING_MODEL_CONFIG__OVERRIDES__BASE_URL=${HONCHO_EMBEDDING_BASE_URL:-https://api.siliconflow.cn/v1}
      - EMBEDDING_VECTOR_DIMENSIONS=1024
      - EMBEDDING_API_KEY=${HONCHO_EMBEDDING_API_KEY:-}
      - DERIVER_MODEL_CONFIG__TRANSPORT=openai
      - DERIVER_MODEL_CONFIG__MODEL=${HONCHO_DERIVER_MODEL:-gpt-4o-mini}
      - DERIVER_WORKERS=${HONCHO_DERIVER_WORKERS:-1}
      - LLM_OPENAI_API_KEY=${HONCHO_LLM_API_KEY:?set HONCHO_LLM_API_KEY}
      - MODEL_CONFIG__OVERRIDES__BASE_URL=${HONCHO_LLM_BASE_URL:-https://api.openai.com/v1}
      - AUTH_USE_AUTH=${HONCHO_AUTH_USE_AUTH:-false}
      - VECTOR_STORE_TYPE=pgvector
      - VECTOR_STORE_MIGRATED=false
    volumes:
      - /etc/timezone:/etc/timezone:ro
      - /etc/localtime:/etc/localtime:ro
    healthcheck:
      test:
        - CMD
        - /app/.venv/bin/python
        - -c
        - "import urllib.request; urllib.request.urlopen('http://localhost:8000/health', timeout=2).read()"
      interval: 10s
      timeout: 5s
      retries: 10
      start_period: 30s
    networks:
      - honcho-net

  deriver:
    build:
      context: ${Volumes_Path}/honcho/source
      dockerfile: Dockerfile
    container_name: honcho-deriver
    restart: unless-stopped
    entrypoint: ["/app/.venv/bin/python", "-m", "src.deriver"]
    depends_on:
      api:
        condition: service_healthy
      database:
        condition: service_healthy
      redis:
        condition: service_healthy
    environment:
      - TZ=Asia/Shanghai
      - DB_CONNECTION_URI=postgresql+psycopg://${HONCHO_DB_USER}:${HONCHO_DB_PASSWORD}@database:5432/${HONCHO_DB_NAME}
      - CACHE_URL=redis://redis:6379/0?suppress=true
      - CACHE_ENABLED=true
      - NAMESPACE=${HONCHO_NAMESPACE:-honcho}
      - DERIVER_MODEL_CONFIG__TRANSPORT=openai
      - DERIVER_MODEL_CONFIG__MODEL=${HONCHO_DERIVER_MODEL:-gpt-5.4-mini}
      - DERIVER_WORKERS=${HONCHO_DERIVER_WORKERS:-1}
      - EMBEDDING_MODEL_CONFIG__TRANSPORT=openai
      - EMBEDDING_MODEL_CONFIG__MODEL=${HONCHO_EMBEDDING_MODEL:-Qwen/Qwen3-Embedding-0.6B}
      - EMBEDDING_MODEL_CONFIG__OVERRIDES__BASE_URL=${HONCHO_EMBEDDING_BASE_URL:-https://api.siliconflow.cn/v1}
      - EMBEDDING_VECTOR_DIMENSIONS=1024
      - LLM_OPENAI_API_KEY=${HONCHO_LLM_API_KEY:?set HONCHO_LLM_API_KEY}
      - EMBEDDING_API_KEY=${HONCHO_EMBEDDING_API_KEY:-}
      - MODEL_CONFIG__OVERRIDES__BASE_URL=${HONCHO_LLM_BASE_URL:-https://api.openai.com/v1}
    volumes:
      - /etc/timezone:/etc/timezone:ro
      - /etc/localtime:/etc/localtime:ro
    networks:
      - honcho-net

  database:
    image: ${HONCHO_DB_IMAGE:-pgvector/pgvector}:pg${HONCHO_DB_VERSION:-18}
    container_name: honcho-database
    restart: unless-stopped
    command: ["postgres", "-c", "max_connections=200"]
    environment:
      - TZ=Asia/Shanghai
      - POSTGRES_USER=${HONCHO_DB_USER}
      - POSTGRES_PASSWORD=${HONCHO_DB_PASSWORD:?set HONCHO_DB_PASSWORD}
      - POSTGRES_DB=${HONCHO_DB_NAME}
      - POSTGRES_HOST_AUTH_METHOD=${POSTGRES_HOST_AUTH_METHOD:-scram-sha-256}
      - POSTGRES_INITDB_ARGS=--encoding=UTF8 --locale=C --auth-host=${POSTGRES_HOST_AUTH_METHOD:-scram-sha-256}
      - PGDATA=/var/lib/postgresql/data/pgdata
    volumes:
      - /etc/timezone:/etc/timezone:ro
      - /etc/localtime:/etc/localtime:ro
      - ${Volumes_Path}/pgdata:/var/lib/postgresql/data
    ports:
      - "${HONCHO_DB_PORT:-5432}:5432"
    healthcheck:
      test:
        - CMD-SHELL
        - pg_isready -U ${HONCHO_DB_USER} -d ${HONCHO_DB_NAME}
      interval: 10s
      timeout: 5s
      retries: 10
    networks:
      - honcho-net

  redis:
    image: ${HONCHO_REDIS_IMAGE:-redis}:${HONCHO_REDIS_VERSION:-8}
    container_name: honcho-redis
    restart: unless-stopped
    command: ["redis-server", "--save", "60", "1", "--appendonly", "yes"]
    environment:
      - TZ=Asia/Shanghai
    volumes:
      - /etc/timezone:/etc/timezone:ro
      - /etc/localtime:/etc/localtime:ro
      - ${Volumes_Path}/redis:/data
    ports:
      - "${HONCHO_REDIS_PORT:-6379}:6379"
    healthcheck:
      test: ["CMD-SHELL", "redis-cli ping"]
      interval: 10s
      timeout: 5s
      retries: 5
    networks:
      - honcho-net

networks:
  honcho-net:
    driver: bridge