FastGPT 知识库问答系统：从部署到应用

mkdir fastgpt && cd fastgpt

version: '3.3'
services:
  # MongoDB
  mongo:
    image: mongo:5.0.18
    container_name: fastgpt-mongo
    ports:
      - "27017:27017"
    environment:
      MONGO_INITDB_ROOT_USERNAME: myusername
      MONGO_INITDB_ROOT_PASSWORD: mypassword
    volumes:
      - ./mongo/data:/data/db
    restart: unless-stopped
    command: mongod --quiet

  # PostgreSQL with PgVector
  pg:
    image: ankane/pgvector:v0.5.0
    container_name: fastgpt-pg
    ports:
      - "5432:5432"
    environment:
      POSTGRES_USER: username
      POSTGRES_PASSWORD: password
      POSTGRES_DB: postgres
    volumes:
      - ./pg/data:/var/lib/postgresql/data
    restart: unless-stopped

  # FastGPT
  fastgpt:
    image: ghcr.io/labring/fastgpt:latest
    container_name: fastgpt
    ports:
      - "3000:3000"
    depends_on:
      - mongo
      - pg
    environment:
      # MongoDB连接
      MONGODB_URI: mongodb://myusername:mypassword@mongo:27017/fastgpt?authSource=admin
      # PgVector连接
      PG_URL: postgresql://username:password@pg:5432/postgres
      # 向量模型（这里用OpenAI，也可以换成本地）
      VECTOR_MAX_PROCESS_LEN: 512
      OPENAI_BASE_URL: https://api.openai.com/v1
      # 初始化Root账号密码
      DEFAULT_ROOT_PSW: your-admin-password
      # 加密密钥
      TOKEN_KEY: your-random-token-key
      ROOT_KEY: your-random-root-key
      FILE_TOKEN_KEY: your-file-token-key
    volumes:
      - ./config.json:/app/data/config.json
    restart: unless-stopped

{
  "feConfigs": {
    "lafEnv": "https://laf.dev",
    "show_emptyChat": true,
    "show_contact": false,
    "show_git": true,
    "show_register": false,
    "show_appStore": false,
    "isPlus": false,
    "show_openai_account": true
  },
  "systemEnv": {
    "openapiPrefix": "fastgpt",
    "vectorMaxProcess": 15,
    "qaMaxProcess": 15,
    "pgHNSWEfSearch": 100
  },
  "llmModels": [
    {
      "model": "gpt-4o",
      "name": "GPT-4o",
      "avatar": "/imgs/model/openai.svg",
      "maxContext": 128000,
      "maxResponse": 16000,
      "quoteMaxToken": 100000,
      "maxTemperature": 1.2,
      "vision": true,
      "toolChoice": true,
      "functionCall": false,
      "defaultSystemChatPrompt": ""
    },
    {
      "model": "gpt-4o-mini",
      "name": "GPT-4o-mini",
      "avatar": "/imgs/model/openai.svg",
      "maxContext": 128000,
      "maxResponse": 16000,
      "quoteMaxToken": 100000,
      "maxTemperature": 1.2,
      "vision": true,
      "toolChoice": true,
      "functionCall": false,
      "defaultSystemChatPrompt": ""
    }
  ],
  "vectorModels": [
    {
      "model": "text-embedding-3-small",
      "name": "Embedding-3-small",
      "avatar": "/imgs/model/openai.svg",
      "charsPointsPrice": 0,
      "defaultToken": 512,
      "maxToken": 3000,
      "weight": 100
    }
  ],
  "audioSpeechModels": [],
  "whisperModel": {},
  "reRankModels": []
}

docker compose up -d

# 查看启动日志
docker compose logs fastgpt -f

# 看到 "Server started" 说明启动完成
# 访问 http://your-server:3000

curl -X POST 'https://your-fastgpt-domain/api/v1/chat/completions' \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer app-your-api-key" \
  -d '{
    "chatId": "unique-session-id",
    "stream": false,
    "messages": [{"role": "user", "content": "Pod一直OOMKill怎么办？"}]
  }'

from fastapi import FastAPI, Request
import httpx
import json

app = FastAPI()

FASTGPT_URL = "https://your-fastgpt-domain/api/v1/chat/completions"
FASTGPT_KEY = "app-your-api-key-here"
DINGTALK_TOKEN = "your-dingtalk-outgoing-token"

@app.post("/webhook/dingtalk")
async def dingtalk_webhook(request: Request):
    body = await request.json()
    
    # 验证token
    if body.get("token") != DINGTALK_TOKEN:
        return {"errcode": 403, "errmsg": "Forbidden"}
    
    user_question = body.get("text", {}).get("content", "").strip()
    sender_id = body.get("senderId", "unknown")
    
    # 调用FastGPT
    async with httpx.AsyncClient(timeout=60) as client:
        response = await client.post(
            FASTGPT_URL,
            headers={"Authorization": f"Bearer {FASTGPT_KEY}"},
            json={
                "chatId": f"dingtalk-{sender_id}",
                "stream": False,
                "messages": [{"role": "user", "content": user_question}]
            }
        )
    
    result = response.json()
    answer = result["choices"][0]["message"]["content"]
    
    # 返回给钉钉
    return {
        "msgtype": "markdown",
        "markdown": {
            "title": "运维助手",
            "text": f"**你的问题**：{user_question}\n\n{answer}"
        }
    }

# AlertName: KubePodCrashLooping

## 含义
某个Pod在过去一段时间内多次重启，触发了CrashLoopBackOff状态。

## 可能原因
1. 应用启动失败（配置错误、依赖服务不可达）
2. OOMKill（内存不足）
3. 代码bug（panic、未处理的异常）
4. 健康检查配置过于严格

## 排查步骤
1. 查看Pod状态和重启次数：
   `kubectl get pod <pod-name> -n <namespace>`

2. 查看最近的崩溃日志：
   `kubectl logs <pod-name> -n <namespace> --previous`

3. 查看Pod详情（关注Events部分）：
   `kubectl describe pod <pod-name> -n <namespace>`

4. 如果是OOMKill，查看内存使用：
   `kubectl top pod <pod-name> -n <namespace>`

## 处理方法
- 配置错误：修复ConfigMap或环境变量，重新部署
- OOMKill：调高memory limits，或排查内存泄漏
- 代码bug：联系开发修复，临时回滚版本

## 升级条件
- 影响生产流量的核心服务 → 立即通知on-call
- 非核心服务重启>10次 → 创建P3工单

// MongoDB中执行，清理30天前的对话记录
db.chatItems.deleteMany({
  updateTime: { $lt: new Date(Date.now() - 30 * 24 * 60 * 60 * 1000) }
})