最近调研了几个RAG+知识库方向的开源产品。自己实操完成了知识库的私有化部署。透过产品了解RAG技术原理，对未来大模型应用开发也将起到事半功倍的效果。

虽然网络上有很多此类技术文章，这里自己也进行一次总结，加深印象的同时给小伙伴做一个参考，多多交流。

mkdir fastgpt
cd fastgpt
curl -O https://raw.githubusercontent.com/labring/FastGPT/main/files/deploy/fastgpt/docker-compose.yml
curl -O https://raw.githubusercontent.com/labring/FastGPT/main/projects/app/data/config.json

Shell
# 在 docker-compose.yml 同级目录下执行
docker-compose pull
docker-compose up -d

Shell
version: '3.3'
services:
  pg:
    #image: ankane/pgvector:v0.5.0 # git
    image: registry.cn-hangzhou.aliyuncs.com/fastgpt/pgvector:v0.5.0 # 阿里云
    container_name: pg
    restart: always
    ports: # 生产环境建议不要暴露
      - 5432:5432
    networks:
      - fastgpt
    environment:
      # 这里的配置只有首次运行生效。修改后，重启镜像是不会生效的。需要把持久化数据删除再重启，才有效果
      - POSTGRES_USER=username
      - POSTGRES_PASSWORD=password
      - POSTGRES_DB=postgres
    volumes:
      - ./pg/data:/var/lib/postgresql/data
  mongo:
    #image: mongo:5.0.18
    image: registry.cn-hangzhou.aliyuncs.com/fastgpt/mongo:5.0.18 # 阿里云
    container_name: mongo
    restart: always
    ports: # 生产环境建议不要暴露
      - 27017:27017
    networks:
      - fastgpt
    environment:
      # 这里的配置只有首次运行生效。修改后，重启镜像是不会生效的。需要把持久化数据删除再重启，才有效果
      - MONGO_INITDB_ROOT_USERNAME=username
      - MONGO_INITDB_ROOT_PASSWORD=password
    volumes:
      - ./mongo/data:/data/db
  fastgpt:
    container_name: fastgpt
    #image: ghcr.io/labring/fastgpt:latest # git
    image: registry.cn-hangzhou.aliyuncs.com/fastgpt/fastgpt:latest # 阿里云
    ports:
      - 3000:3000
    networks:
      - fastgpt
    depends_on:
      - mongo
      - pg
    restart: always
    environment:
      # root 密码，用户名为: root
       # root 密码，用户名为: root
      - DEFAULT_ROOT_PSW=1234
      # 中转地址，如果是用官方号，不需要管。务必加 /v1
      - OPENAI_BASE_URL=http://ip:端口/v1
      - CHAT_API_KEY=sk-xxxx
      - DB_MAX_LINK=5 # database max link
      - TOKEN_KEY=any
      - ROOT_KEY=root_key
      - FILE_TOKEN_KEY=filetoken
      # mongo 配置，不需要改. 如果连不上，可能需要去掉 ?authSource=admin
      - MONGODB_URI=mongodb://username:password@mongo:27017/fastgpt?authSource=admin
      # pg配置. 不需要改
      - PG_URL=postgresql://username:password@pg:5432/postgres
    volumes:
      - ./config.json:/app/data/config.json

Shell
#拷贝至指定目录
mkdir /oneapi
cp docker-compose.yml /oneapi
cd /oneapi
#文件同级目录执行命启动容器
docker-compose up -d

JSON
{
  "SystemParams": {
    "pluginBaseUrl": "",
    "vectorMaxProcess": 15,
    "qaMaxProcess": 15,
    "pgHNSWEfSearch": 100
  },
  "ChatModels": [
    {
      "model": "gpt-3.5-turbo",
      "name": "GPT35-turbo",
      "price": 0,
      "maxContext": 16000,
      "maxResponse": 4000,
      "quoteMaxToken": 2000,
      "maxTemperature": 1.2,
      "censor": false,
      "vision": false,
      "defaultSystemChatPrompt": ""
    },
    {
      "model": "gpt-4-turbo",
      "name": "gpt-4-turbo",
      "maxContext": 8000,
      "maxResponse": 8000,
      "price": 0,
      "quoteMaxToken": 4000,
      "maxTemperature": 1.2,
      "censor": false,
      "vision": false,
      "defaultSystemChatPrompt": ""
    }
  ],
  "QAModels": [
    {
      "model": "gpt-3.5-turbo",
      "name": "GPT35-turbo",
      "maxContext": 16000,
      "maxResponse": 16000,
      "price": 0
    },
    {
      "model": "gpt-4-turbo",
      "name": "GPT4-turbo",
      "maxContext": 16000,
      "maxResponse": 16000,
      "price": 0
    }
  ],
  "CQModels": [
    {
      "model": "gpt-3.5-turbo",
      "name": "GPT35",
      "maxContext": 4000,
      "maxResponse": 4000,
      "price": 0,
      "functionCall": true,
      "functionPrompt": ""
    },
    {
      "model": "gpt-4-turbo",
      "name": "GPT4-turbo",
      "maxContext": 8000,
      "maxResponse": 8000,
      "price": 0,
      "functionCall": true,
      "functionPrompt": ""
    }
  ],
  "ExtractModels": [
    {
      "model": "gpt-3.5-turbo",
      "name": "GPT35",
      "maxContext": 16000,
      "maxResponse": 4000,
      "price": 0,
      "functionCall": true,
      "functionPrompt": ""
    }
  ],
  "QGModels": [
    {
      "model": "gpt-4-turbo",
      "name": "GPT4-turbo",
      "maxContext": 1600,
      "maxResponse": 4000,
      "price": 0
    }
  ],
  "VectorModels": [
    {
      "model": "text-embedding-ada-002",
      "name": "Embedding-2",
      "price": 0.0,
      "defaultToken": 700,
      "maxToken": 3000
    },
     {
      "model": "text2vec-large-chinese",
      "name": "text2vec-large-chinese",
      "price": 0.0,
      "defaultToken": 700,
      "maxToken": 3000
    },
     {
      "model": "m3e",
      "name": "m3e",
      "price": 0.0,
      "defaultToken": 700,
      "maxToken": 3000
    }

  ],
  "ReRankModels": [],
  "AudioSpeechModels": [
    {
      "model": "tts-1",
      "name": "OpenAI TTS1",
      "price": 0,
      "voices": [
        { "label": "Alloy", "value": "alloy", "bufferId": "openai-Alloy" },
        { "label": "Echo", "value": "echo", "bufferId": "openai-Echo" },
        { "label": "Fable", "value": "fable", "bufferId": "openai-Fable" },
        { "label": "Onyx", "value": "onyx", "bufferId": "openai-Onyx" },
        { "label": "Nova", "value": "nova", "bufferId": "openai-Nova" },
        { "label": "Shimmer", "value": "shimmer", "bufferId": "openai-Shimmer" }
      ]
    }
  ],
  "WhisperModel": {
    "model": "whisper-1",
    "name": "Whisper1",
    "price": 0
  }
}

Shell
#执行命令重启fastgpt服务
cd /fastgpt
docker-compose restart

JSON
version: '1'
services:
  m3e:
    image: registry.cn-hangzhou.aliyuncs.com/fastgpt_docker/m3e-large-api:latest
    restart: no
    ports:
      - "6008:6008"

JSON
"VectorModels": [
    {
      "model": "text-embedding-ada-002",
      "name": "Embedding-2",
      "price": 0.2,
      "defaultToken": 500,
      "maxToken": 3000
    },
    {
      "model": "m3e",
      "name": "M3E（测试使用）",
      "price": 0.1,
      "defaultToken": 500,
      "maxToken": 1800
    }
]