feat: 网页爬虫&总结改成使用MIFY支持

.vscode/settings.json

@@ -24,6 +24,7 @@
     "metas",
     "MIAI",
     "michat",
+    "mify",
     "mina",
     "mindnote",
     "openai",

constant/function.ts

@@ -1,9 +1,14 @@
 const functionMap = {
   egg: {
     xName: "小煎蛋",
-    xAuthor: "zhaoyingbo",
+    xAuthor: "Yingbo",
     xIcon: "🍳",
   },
+  webAgent: {
+    xName: "小煎蛋 简报助手",
+    xAuthor: "Yingbo",
+    xIcon: "📎",
+  },
   groupAgent: {
     xName: "小煎蛋 Group Agent",
     xAuthor: "YIBinary ❤️ Yingbo",
@@ -11,12 +16,12 @@ const functionMap = {
   },
   sheetDB: {
     xName: "小煎蛋 Sheet DB",
-    xAuthor: "zhaoyingbo",
+    xAuthor: "Yingbo",
     xIcon: "🍪",
   },
   gitlabAgent: {
     xName: "小煎蛋 Gitlab Agent",
-    xAuthor: "zhaoyingbo",
+    xAuthor: "Yingbo",
     xIcon: "🐙",
   },
   soupAgent: {

controller/groupAgent/agent.ts

@@ -35,12 +35,6 @@ const agent = async (ctx: Context) => {
       excludedMessageIds: [loadingMessageId, messageId],
       excludeMentions: [appInfo.appName],
     })
-  // 如果没有聊天记录，返回错误信息
-  if (chatHistory.length === 0) {
-    logger.info("No chat history found")
-    await message.updateOrReply(cardGender.genErrorCard("未找到聊天记录"))
-    return
-  }
   logger.debug(`Chat history: ${JSON.stringify(chatHistory)}`)
 
   // 根据Mention，拼装原始消息

controller/intentAgent/index.ts

@@ -0,0 +1,135 @@
+import { parseJsonString } from "@egg/hooks"
+import { z } from "zod"
+import { zodToJsonSchema } from "zod-to-json-schema"
+
+import { Context } from "../../types"
+import llm from "../../utils/llm"
+import { cleanLLMRes } from "../../utils/llm/base"
+
+/**
+ * 基础意图模式
+ */
+const BaseIntentSchema = z.object({
+  intent: z.number().min(1).max(13),
+})
+
+/**
+ * 简报模式
+ */
+const BriefingSchema = BaseIntentSchema.extend({
+  intent: z.literal(5),
+  link: z.string().url(),
+  userDescription: z.string().min(1),
+})
+
+const BriefingLinkSchema = z.object({
+  intent: z.literal(14),
+  link: z.string().url(),
+})
+
+/**
+ * 通用响应模式
+ */
+const CommonResponseSchema = BaseIntentSchema.extend({
+  intent: z.literal(2),
+  message: z.string().min(1),
+})
+
+/**
+ * 意图模式
+ */
+const IntentSchema = z.union([
+  BriefingSchema,
+  CommonResponseSchema,
+  BaseIntentSchema,
+  BriefingLinkSchema,
+])
+
+type BaseIntent = z.infer<typeof BaseIntentSchema>
+type Briefing = z.infer<typeof BriefingSchema>
+type BriefingLink = z.infer<typeof BriefingLinkSchema>
+type CommonResponse = z.infer<typeof CommonResponseSchema>
+export type Intent = z.infer<typeof IntentSchema>
+
+/**
+ * JSON模式
+ */
+const jsonSchema = zodToJsonSchema(IntentSchema)
+
+/**
+ * 代理函数
+ * @param {Context} ctx - 上下文对象
+ * @returns {Promise<Intent>} - 返回意图对象
+ */
+const agent = async (ctx: Context): Promise<Intent> => {
+  const {
+    larkBody: { msgText },
+    logger,
+    requestId,
+  } = ctx
+
+  let attempts = 0
+  while (attempts < 3) {
+    const res = await llm.invoke(
+      "intentRecognitionNext",
+      {
+        userInput: msgText,
+        time: new Date().toLocaleString("zh-CN", { timeZone: "Asia/Shanghai" }),
+        jsonSchema,
+      },
+      requestId,
+      0.5,
+      true
+    )
+
+    const rawIntent = cleanLLMRes(res as string)
+    const parsedIntent = parseJsonString(rawIntent, null)
+
+    if (parsedIntent) {
+      try {
+        IntentSchema.parse(parsedIntent)
+        logger.debug("Intent is valid: " + JSON.stringify(parsedIntent))
+        return parsedIntent
+      } catch (e: any) {
+        logger.error("Invalid intent: " + String(e.errors))
+      }
+    } else {
+      logger.error("Parsed intent is null")
+    }
+
+    attempts++
+  }
+
+  return {
+    intent: 1,
+  }
+}
+
+const isBaseIntent = (intent: Intent): intent is BaseIntent => {
+  return intent.intent !== 5 && intent.intent !== 2
+}
+
+const isBriefing = (intent: Intent): intent is Briefing => {
+  return intent.intent === 5
+}
+
+const isBriefingLink = (intent: Intent): intent is BriefingLink => {
+  return intent.intent === 14
+}
+
+const isCommonResponse = (intent: Intent): intent is CommonResponse => {
+  return intent.intent === 2
+}
+
+/**
+ * 意图代理对象
+ */
+const intentAgent = {
+  agent,
+  isBaseIntent,
+  isBriefing,
+  isCommonResponse,
+  isBriefingLink,
+}
+
+export default intentAgent

controller/reportAgent/index.ts

@@ -0,0 +1,266 @@
+import db from "../../db"
+import { SheetModel } from "../../db/sheet"
+import { Context } from "../../types"
+import llm from "../../utils/llm"
+import { extractSheetIds, validateLink } from "../../utils/string"
+
+/**
+ * 爬取网页内容
+ * @param {Context} ctx - 上下文对象
+ * @param {string} link - 网页链接
+ * @returns {Promise<any>} - 返回爬取结果
+ * @throws {Error} - 当爬取失败时抛出错误
+ */
+export const crawlWebPage = async (
+  ctx: Context,
+  link: string
+): Promise<any> => {
+  const { attachService } = ctx
+  const crawRes = await attachService.crawlWeb(link)
+  if (!crawRes || crawRes?.code) throw new Error("网页抓取失败")
+  return crawRes
+}
+
+/**
+ * 生成网页简报
+ * @param {Context} ctx - 上下文对象
+ * @param {string} userDescription - 用户描述
+ * @param {any} content - 网页内容
+ * @returns {Promise<string>} - 返回简报内容
+ * @throws {Error} - 当生成简报失败时抛出错误
+ */
+export const generateSummary = async (
+  ctx: Context,
+  userDescription: string,
+  content: any
+): Promise<string> => {
+  const { requestId } = ctx
+  const llmRes = (await llm.invoke(
+    "summaryWeb",
+    {
+      description: userDescription,
+      content: content,
+    },
+    requestId,
+    1
+  )) as string
+  if (!llmRes) throw new Error("模型总结失败")
+  return llmRes
+}
+
+/**
+ * 插入到表格
+ * @param {Context} ctx - 上下文对象
+ * @param {string} link - 网页链接
+ * @param {string} userDescription - 用户描述
+ * @param {string} llmRes - 简报内容
+ * @returns {Promise<string>} - 返回表格链接
+ */
+const insert2Sheet = async (
+  ctx: Context,
+  link: string,
+  userDescription: string,
+  llmRes: string
+) => {
+  const {
+    larkBody: { userId },
+    logger,
+    larkService,
+  } = ctx
+  try {
+    const chat = await db.chat.getAndCreate(ctx)
+    if (!chat || !chat?.sheet || !chat?.expand?.sheet) {
+      logger.info("No sheet found, skip insert2Sheet")
+      return ""
+    }
+    const { sheetToken, range, sheetUrl } = chat.expand.sheet as SheetModel
+    await larkService.sheet.insertRows(sheetToken, range, [
+      [
+        userId || "",
+        link,
+        userDescription,
+        llmRes,
+        new Date().toLocaleString("zh-CN", { timeZone: "Asia/Shanghai" }),
+      ],
+    ])
+    return sheetUrl
+  } catch (error: any) {
+    logger.error(`Failed to insert2Sheet: ${error}`)
+    return ""
+  }
+}
+
+/**
+ * 生成简报卡片
+ * @param {Context} ctx - 上下文对象
+ * @param {string} link - 网页链接
+ * @param {string} userDescription - 用户描述
+ * @param {string} llmRes - 简报内容
+ * @param {string} sheetLink - 表格链接
+ * @returns {any} - 返回卡片对象
+ */
+const genReportCard = (
+  ctx: Context,
+  link: string,
+  userDescription: string,
+  llmRes: string,
+  sheetLink: string
+) => {
+  const { larkCard } = ctx
+  const cardGender = larkCard.child("webAgent")
+  const description = userDescription
+    ? `**用户描述📝**\n${userDescription}\n`
+    : ""
+  const sheetLinkMd = sheetLink
+    ? `**小提示🔍**
+您可以在[这里](${sheetLink})查看往期简报哦！`
+    : ""
+  const content = `
+感谢您使用小煎蛋简报🍳！以下是为您总结的简报内容：
+
+**网站地址🌟**
+[${link}](${link})
+
+${description}
+
+**AI简报🌈**
+${llmRes}
+
+${sheetLinkMd}
+`
+  return cardGender.genCard("markdownSuccessCard", {
+    content,
+  })
+}
+
+/**
+ * 生成简报
+ * @param {Context} ctx - 上下文对象
+ * @param {string} link - 网页链接
+ * @param {string} userDescription - 用户描述
+ */
+const agent = async (ctx: Context, link: string, userDescription: string) => {
+  const {
+    larkService: { message },
+    larkCard,
+    logger,
+    attachService,
+  } = ctx
+  const cardGender = larkCard.child("webAgent")
+  try {
+    // 校验链接是否合法
+    validateLink(link)
+    // 发送一个loading卡片
+    await message.updateOrReply(
+      cardGender.genSuccessCard("正在为您收集简报，请稍等片刻~")
+    )
+    // // 抓取网页
+    // const crawRes = await crawlWebPage(ctx, link)
+    // // 调用模型生成简报
+    // const llmRes = await generateSummary(ctx, userDescription, crawRes)
+    // 调用mify服务生成简报
+    const llmRes = await attachService.mifyCrawler(link, userDescription)
+    // 插入到表格
+    const sheetLink = await insert2Sheet(ctx, link, userDescription, llmRes)
+    // 发送简报卡片
+    await message.updateOrReply(
+      genReportCard(ctx, link, userDescription, llmRes, sheetLink)
+    )
+  } catch (error: any) {
+    logger.error(`Failed gen report: ${error}`)
+    await message.updateOrReply(
+      cardGender.genErrorCard(`简报生成失败: ${error.message}`)
+    )
+  }
+}
+
+/**
+ * 解析表格信息
+ * @param {Context} ctx - 上下文对象
+ * @param {string} sheetUrl - 表格链接
+ * @returns {Promise<{sheetToken: string, range: string}>} - 返回表格信息
+ * @throws {Error} - 当解析失败时抛出错误
+ */
+const parseSheetInfo = async (
+  ctx: Context,
+  sheetUrl: string
+): Promise<{ sheetToken: string; range: string }> => {
+  const { larkService } = ctx
+  const token = extractSheetIds(sheetUrl)
+  if (!token || !token.sheetToken) throw new Error("链接格式错误")
+
+  const getRange = async (sheetToken: string) => {
+    const res = await larkService.sheet.getMetaInfo(sheetToken)
+    if (!res || !res?.data?.sheets?.length)
+      throw new Error("获取工作表信息失败")
+    return res.data.sheets[0].sheetId
+  }
+  // 如果不是Wiki链接，直接提取sheetToken和range
+  if (!sheetUrl.includes("wiki")) {
+    if (token.range) return token as { sheetToken: string; range: string }
+    // 获取第一个工作表作为range
+    token.range = await getRange(token.sheetToken)
+    return token as { sheetToken: string; range: string }
+  }
+  // 如果是Wiki链接，需要先获取sheetToken
+  const res = await larkService.wiki.getNodeInfo(token?.sheetToken)
+  if (!res || !res?.data?.node?.obj_token) throw new Error("获取Wiki信息失败")
+  token.sheetToken = res.data.node.obj_token
+  if (token.range) return token as { sheetToken: string; range: string }
+  token.range = await getRange(token.sheetToken)
+  return token as { sheetToken: string; range: string }
+}
+
+/**
+ * 设置简报表格
+ * @param {Context} ctx - 上下文对象
+ * @param {string} sheetUrl - 表格链接
+ */
+const setSheet = async (ctx: Context, sheetUrl: string) => {
+  const { larkCard, larkService, logger } = ctx
+  const cardGender = larkCard.child("webAgent")
+  try {
+    // 获取chat信息
+    const chat = await db.chat.getAndCreate(ctx)
+    if (!chat) throw new Error("获取聊天信息失败")
+    // 获取是否已经有存在的sheet信息
+    const existSheet = await db.sheet.getByUrl(sheetUrl)
+    // 如果已经存在，则直接写入chat表
+    if (existSheet) {
+      const updateRes = await db.chat.update(chat.id, { sheet: existSheet.id })
+      if (!updateRes) throw new Error("更新chat记录失败")
+      await larkService.message.updateOrReply(
+        cardGender.genSuccessCard("简报汇总表设置成功")
+      )
+      return
+    }
+
+    // 获取sheet的token和range
+    const sheetInfo = await parseSheetInfo(ctx, sheetUrl)
+
+    // 创建sheet记录
+    const createRes = await db.sheet.create({
+      sheetToken: sheetInfo.sheetToken,
+      range: sheetInfo.range,
+      sheetUrl,
+    })
+    if (!createRes) throw new Error("创建sheet记录失败")
+    const updateRes = await db.chat.update(chat.id, { sheet: createRes.id })
+    if (!updateRes) throw new Error("更新chat记录失败")
+    await larkService.message.updateOrReply(
+      cardGender.genSuccessCard("简报汇总表创建成功")
+    )
+  } catch (error: any) {
+    logger.error(`Failed setSheet: ${error}`)
+    await larkService.message.updateOrReply(
+      cardGender.genErrorCard(`简报汇总表设置失败: ${error.message}`)
+    )
+  }
+}
+
+const reportAgent = {
+  agent,
+  setSheet, // 添加 setSheet 到导出的对象中
+}
+
+export default reportAgent

controller/soupAgent/index.ts

@@ -131,10 +131,8 @@ const chat2Soup = async (ctx: Context) => {
     larkBody: { msgText, chatId, messageId },
     logger,
     attachService,
-    larkCard,
     larkService: { message },
   } = ctx
-  const cardGender = larkCard.child("soupAgent")
   message.setReplyMessage(messageId, "text")
   const activeGame = await db.soupGame.getActiveOneByChatId(chatId)
   if (!activeGame) {
@@ -150,9 +148,7 @@ const chat2Soup = async (ctx: Context) => {
   })
   if (!res) {
     logger.error(`chatId: ${chatId} failed to get soup result`)
-    await message.updateOrReply(
+    await message.updateOrReply(SoupGameMessage.chatFailed)
-      cardGender.genErrorCard(SoupGameMessage.chatFailed)
-    )
     return
   }
   // 用户答对了

db/chat/index.ts

@@ -3,6 +3,7 @@ import { RecordModel } from "pocketbase"
 import { Context } from "../../types"
 import { managePbError } from "../../utils/pbTools"
 import pbClient from "../pbClient"
+import { SheetModel } from "../sheet"
 
 const DB_NAME = "chat"
 
@@ -13,18 +14,29 @@ export interface Chat {
   mode: "group" | "p2p" | "topic"
   weeklySummary: boolean
   dailySummary: boolean
+  sheet: string
 }
 
 export type ChatModel = Chat & RecordModel
 
+export interface ChatModelWithExpand extends ChatModel {
+  expand: {
+    sheet: SheetModel
+  }
+}
+
 /**
  * 获取单个群组信息
  * @param id
  * @returns
  */
 const getOneByChatId = (chatId: string) =>
-  managePbError<ChatModel>(() =>
+  managePbError<ChatModelWithExpand>(() =>
-    pbClient.collection(DB_NAME).getFirstListItem(`chatId = "${chatId}"`)
+    pbClient
+      .collection<ChatModelWithExpand>(DB_NAME)
+      .getFirstListItem(`chatId = "${chatId}"`, {
+        expand: "sheet",
+      })
   )
 
 /**
@@ -60,10 +72,20 @@ const getAndCreate = async ({ larkService, logger, larkBody }: Context) => {
     mode: chat_mode,
     weeklySummary: false,
     dailySummary: false,
+    sheet: "",
   }
   return create(newChat)
 }
 
+/**
+ * 更新群组
+ * @param id
+ * @param chat
+ * @returns
+ */
+const update = (id: string, chat: Partial<Chat>) =>
+  managePbError<ChatModel>(() => pbClient.collection(DB_NAME).update(id, chat))
+
 /**
  * 更新群组总结
  * @param id
@@ -97,6 +119,7 @@ const getNeedSummaryChats = async (timeScope: "daily" | "weekly" | "all") => {
 }
 
 const chat = {
+  update,
   getAndCreate,
   getOneByChatId,
   updateSummary,

db/index.ts

@@ -4,6 +4,7 @@ import gitlabProject from "./gitlabProject/index."
 import grpSumLog from "./grpSumLog"
 import log from "./log"
 import receiveGroup from "./receiveGroup"
+import sheet from "./sheet"
 import soupGame from "./soupGame"
 import user from "./user"
 
@@ -16,6 +17,7 @@ const db = {
   grpSumLog,
   gitlabProject,
   soupGame,
+  sheet,
 }
 
 export default db

db/sheet/index.ts

@@ -0,0 +1,44 @@
+import { RecordModel } from "pocketbase"
+
+import { managePbError } from "../../utils/pbTools"
+import pbClient from "../pbClient"
+
+const DB_NAME = "sheet"
+
+export interface Sheet {
+  /***
+   * 表格区间
+   */
+  range: string
+  /***
+   * 表格Token
+   */
+  sheetToken: string
+  /***
+   * 表格链接
+   */
+  sheetUrl: string
+}
+
+export type SheetModel = Sheet & RecordModel
+
+const getByUrl = (sheetUrl: string) =>
+  managePbError<SheetModel>(() =>
+    pbClient.collection(DB_NAME).getFirstListItem(`sheetUrl = "${sheetUrl}"`)
+  )
+
+const update = (id: string, sheet: Partial<SheetModel>) =>
+  managePbError<SheetModel>(() =>
+    pbClient.collection(DB_NAME).update(id, sheet)
+  )
+
+const create = (sheet: Sheet) =>
+  managePbError<SheetModel>(() => pbClient.collection(DB_NAME).create(sheet))
+
+const sheet = {
+  getByUrl,
+  update,
+  create,
+}
+
+export default sheet

index.ts

@@ -12,8 +12,10 @@ initSchedule()
 
 await initAppConfig()
 
-const server = Bun.serve({
+const bunServer = Bun.serve({
-  async fetch(req) {
+  fetch: async (req, server) => {
+    // 设置超时时间
+    server.timeout(req, 30)
     // 生成上下文
     const ctx = await genContext(req)
     const { path, genResp, logger } = ctx
@@ -43,6 +45,7 @@ const server = Bun.serve({
       return genResp.healthCheck("hello, there is egg, glade to serve you!")
     } catch (error: any) {
       // 错误处理
+      logger.error(error.message)
       return genResp.serverError(error.message || "server error")
     }
   },
@@ -54,4 +57,4 @@ const server = Bun.serve({
   port: 3000,
 })
 
-logger.info(`Listening on ${server.hostname}:${server.port}`)
+logger.info(`Listening on ${bunServer.hostname}:${bunServer.port}`)

package.json

@@ -22,15 +22,15 @@
     "@eslint/js": "^9.19.0",
     "@types/node-schedule": "^2.1.7",
     "@types/uuid": "^10.0.0",
-    "bun-types": "^1.2.0",
+    "bun-types": "^1.2.1",
     "eslint": "^9.19.0",
     "eslint-plugin-simple-import-sort": "^12.1.1",
     "eslint-plugin-unused-imports": "^4.1.4",
     "husky": "^9.1.7",
-    "lint-staged": "^15.4.2",
+    "lint-staged": "^15.4.3",
     "oxlint": "^0.13.2",
     "prettier": "^3.4.2",
-    "typescript-eslint": "^8.21.0"
+    "typescript-eslint": "^8.22.0"
   },
   "peerDependencies": {
     "typescript": "^5.5.4"
@@ -39,7 +39,7 @@
     "@egg/hooks": "^1.2.0",
     "@egg/lark-msg-tool": "^1.21.0",
     "@egg/logger": "^1.6.0",
-    "@egg/net-tool": "^1.28.2",
+    "@egg/net-tool": "^1.31.1",
     "@egg/path-tool": "^1.4.1",
     "@langchain/core": "^0.3.36",
     "@langchain/langgraph": "^0.2.41",

routes/bot/eventMsg.ts

@@ -1,8 +1,8 @@
 import groupAgent from "../../controller/groupAgent"
+import intentAgent from "../../controller/intentAgent"
+import reportAgent from "../../controller/reportAgent"
 import soupAgent from "../../controller/soupAgent"
 import { Context } from "../../types"
-import llm from "../../utils/llm"
-import { cleanLLMRes } from "../../utils/llm/base"
 import { isNotP2POrAtBot } from "../../utils/message"
 
 /**
@@ -49,111 +49,88 @@ const filterIllegalMsg = async (ctx: Context): Promise<boolean> => {
   return true
 }
 
-/**
- * 发送ID消息
- * @param {Context} ctx - 上下文数据
- */
-const manageIdMsg = ({
-  larkBody: { chatId },
-  larkCard,
-  larkService,
-}: Context) =>
-  larkService.message.sendCard2Chat(
-    chatId,
-    larkCard.genTempCard("chatId", { chat_id: chatId })
-  )
-
 /**
  * 处理意图消息
  * @param {Context} ctx - 上下文数据
  */
 const manageIntent = async (ctx: Context) => {
   const {
-    body,
     logger,
     larkService: { message },
     attachService,
     larkBody: { msgText, chatId },
     larkCard,
-    requestId,
   } = ctx
   logger.info(`bot req text: ${msgText}`)
   await message.updateOrReply(
     larkCard.genPendingCard("正在理解您的意图，请稍等...")
   )
   try {
-    const llmRes = (await llm.invoke(
+    const intentRes = await intentAgent.agent(ctx)
-      "intentRecognition",
+    logger.info(`intentRes: ${JSON.stringify(intentRes)}`)
-      {
+    if (intentAgent.isBriefing(intentRes)) {
-        userInput: msgText,
+      reportAgent
-        time: new Date().toLocaleString("zh-CN", { timeZone: "Asia/Shanghai" }),
+        .agent(ctx, intentRes.link, intentRes.userDescription)
-      },
+        .catch(() => null)
-      requestId
-    )) as string
-    const cleanedLlmRes = cleanLLMRes(llmRes)
-    logger.info(`intentRecognition llm res: ${cleanedLlmRes}`)
-    // 返回值不是数字，说明是通识回答
-    const intent = Number(cleanedLlmRes)
-    if (isNaN(intent)) {
-      await message.updateOrReply(larkCard.genSuccessCard(cleanedLlmRes))
       return
     }
-
+    if (intentAgent.isBriefingLink(intentRes)) {
-    switch (intent) {
+      reportAgent.setSheet(ctx, intentRes.link).catch(() => null)
-      // 获取聊天ID
+      return
-      case 1:
+    }
-        await manageIdMsg(ctx)
+    if (intentAgent.isCommonResponse(intentRes)) {
-        break
+      await message.updateOrReply(larkCard.genSuccessCard(intentRes.message))
-      // CI监控
+      return
-      case 2:
+    }
-        await attachService.ciMonitor(chatId)
+    if (intentAgent.isBaseIntent(intentRes)) {
-        break
+      switch (intentRes.intent) {
-      // 生成简报
+        case 3:
-      case 3:
+          await message.updateOrReply(
-        await message.updateOrReply(
+            larkCard.genTempCard("chatId", { chat_id: chatId }) as string
-          larkCard.genSuccessCard("正在为您收集简报，请稍等片刻~")
+          )
-        )
+          break
-        await attachService.reportCollector(body)
+        case 4:
-        break
+          await attachService.ciMonitor(chatId)
-      // 获取帮助
+          break
-      case 4:
+        case 6:
-        await message.updateOrReply(
+          await message.updateOrReply(
-          larkCard.genTempCard("eggGuide", { chat_id: chatId }) as string
+            larkCard.genTempCard("eggGuide", { chat_id: chatId }) as string
-        )
+          )
-        break
+          break
-      // 开启日报订阅
+        case 7:
-      case 5:
+          groupAgent.report
-        groupAgent.report.setSubscription(ctx, "daily", true)
+            .setSubscription(ctx, "daily", true)
-        break
+            .catch(() => null)
-      // 开启周报订阅
+          break
-      case 6:
+        case 8:
-        groupAgent.report.setSubscription(ctx, "weekly", true)
+          groupAgent.report
-        break
+            .setSubscription(ctx, "weekly", true)
-      // 关闭日报订阅
+            .catch(() => null)
-      case 7:
+          break
-        groupAgent.report.setSubscription(ctx, "daily", false)
+        case 9:
-        break
+          groupAgent.report
-      // 关闭周报订阅
+            .setSubscription(ctx, "daily", false)
-      case 8:
+            .catch(() => null)
-        groupAgent.report.setSubscription(ctx, "weekly", false)
+          break
-        break
+        case 10:
-      // 立即发送日报
+          groupAgent.report
-      case 9:
+            .setSubscription(ctx, "weekly", false)
-        groupAgent.report.gen4Test(ctx, "daily")
+            .catch(() => null)
-        break
+          break
-      // 立即发送周报
+        case 11:
-      case 10:
+          groupAgent.report.gen4Test(ctx, "daily").catch(() => null)
-        groupAgent.report.gen4Test(ctx, "weekly")
+          break
-        break
+        case 12:
-      // 开始海龟汤游戏
+          groupAgent.report.gen4Test(ctx, "weekly").catch(() => null)
-      case 11:
+          break
-        soupAgent.startOrStopGame(ctx, true, "manual")
+        case 13:
-        break
+          soupAgent.startOrStopGame(ctx, true, "manual").catch(() => null)
-      // 通识回答
+          break
-      case 12:
+        case 1:
-      default:
+        default:
-        groupAgent.agent(ctx)
+          groupAgent.agent(ctx).catch(() => null)
-        break
+          break
+      }
     }
   } catch (error) {
     logger.error(`manageIntent error: ${error}`)

services/attach/index.ts

@@ -1,6 +1,9 @@
 import type { LarkEvent } from "@egg/lark-msg-tool"
 import { NetToolBase } from "@egg/net-tool"
 
+import { APP_CONFIG } from "../../constant/config"
+import { LarkServer } from "../../types"
+
 interface Chat2SoupParams {
   user_query: string
   soup_id: string
@@ -84,6 +87,53 @@ class AttachService extends NetToolBase {
     const URL = "https://lark-egg.ai.xiaomi.com/soup/chat"
     return this.post<Chat2SoupResp>(URL, body).catch(() => null)
   }
+
+  /**
+   * 抓取网页内容
+   * @param {string} url - 网页URL。
+   * @returns {Promise<string>} 返回抓取的网页内容。
+   */
+  async crawlWeb(url: string) {
+    const URL = "https://lark-egg.ai.xiaomi.com/tools/web/crawler"
+    return this.get<LarkServer.BaseRes<string>>(URL, { url }).catch(() => null)
+  }
+
+  /**
+   * 使用mify爬虫抓取网页内容
+   * @param {string} link - 网页链接
+   * @param {string} userDescription - 用户描述
+   * @returns {Promise<any>} 返回爬取结果
+   */
+  async mifyCrawler(link: string, userDescription: string) {
+    const URL = "https://mify-be.pt.xiaomi.com/api/v1/workflows/run"
+    return this.post(
+      URL,
+      {
+        inputs: {
+          link,
+          userDescription,
+        },
+        response_mode: "blocking",
+        user: "egg-server",
+      },
+      {},
+      {
+        Authorization: `Bearer ${APP_CONFIG.MIFY_CRAWLER_TOKEN}`,
+      }
+    )
+      .then((res) => {
+        const llmRes = res.data.outputs.content
+        if (!llmRes) throw new Error("模型总结失败")
+        if (llmRes === "crawlerErr") throw new Error("网页抓取失败")
+        return llmRes as string
+      })
+      .catch((error) => {
+        if (["网页抓取失败", "模型总结失败"].includes(error.message)) {
+          throw error
+        }
+        throw new Error("MIFY爬虫请求失败")
+      })
+  }
 }
 
 export default AttachService

test/archive/wiki.ts

@@ -0,0 +1,8 @@
+import initAppConfig from "../../constant/config"
+import genLarkService from "../../utils/genLarkService"
+
+await initAppConfig()
+
+const service = genLarkService("egg", "test")
+
+service.wiki.getNodeInfo("V4ZkwhDR8iRCqIk7X81k1rBj4Sc").then(console.log)

test/llm/intentRecognition.ts

@@ -1,15 +1,59 @@
+import { parseJsonString } from "@egg/hooks"
+import { z } from "zod"
+import { zodToJsonSchema } from "zod-to-json-schema"
+
 import initAppConfig from "../../constant/config"
 import llm from "../../utils/llm"
+import { cleanLLMRes } from "../../utils/llm/base"
 
 await initAppConfig()
 
+const BaseIntentSchema = z.object({
+  intent: z.number().min(1).max(13),
+})
+
+const BriefingSchema = BaseIntentSchema.extend({
+  intent: z.literal(3),
+  link: z.string().url(),
+  userDescription: z.string().min(1),
+})
+
+const CommonResponseSchema = BaseIntentSchema.extend({
+  intent: z.literal(12),
+  message: z.string().min(1),
+})
+
+const IntentSchema = z.union([
+  BriefingSchema,
+  CommonResponseSchema,
+  BaseIntentSchema,
+])
+
+const jsonSchema = zodToJsonSchema(IntentSchema)
+
 const res = await llm.invoke(
-  "intentRecognition",
+  "intentRecognitionNext",
   {
-    userInput: "你是干嘛的",
+    userInput:
+      "https://mp.weixin.qq.com/s/-0J8XbXJU6Bu-UihRtgGAQ Airbnb死磕React Native惨败，微软却玩出花！Office、Outlook全线接入，Copilot成最大赢家 推荐大家看一下，rn助力微软copilot 跨平台实现",
     time: new Date().toLocaleString("zh-CN", { timeZone: "Asia/Shanghai" }),
+    jsonSchema,
   },
-  "test"
+  "test",
+  0,
+  true
 )
 
-console.log(res)
+const rawIntent = cleanLLMRes(res as string)
+
+const parsedIntent = parseJsonString(rawIntent, null)
+console.log("🚀 ~ parsedIntent:", parsedIntent)
+
+try {
+  IntentSchema.parse(parsedIntent)
+  console.log("Intent is valid:", parsedIntent)
+} catch (e: any) {
+  console.error("Invalid intent:", e.errors)
+}
+
+console.log(cleanLLMRes(res as string))

test/llm/mifyCrawler.ts

@@ -0,0 +1,11 @@
+import initAppConfig from "../../constant/config"
+import { AttachService } from "../../services"
+
+await initAppConfig()
+
+const server = new AttachService()
+
+server
+  .mifyCrawler("https://lacus.site", "详细介绍alpine")
+  .then(console.log)
+  .catch(console.error)

test/reportAgent/agent.http

@@ -0,0 +1,4 @@
+POST http://localhost:3000/bot?app=egg HTTP/1.1
+content-type: application/json
+
+{"schema":"2.0","header":{"event_id":"c94518fbcb9d66cc93b4144cc69e4f0c","token":"tV9djUKSjzVnekV7xTg2Od06NFTcsBnj","create_time":"1732612280153","event_type":"im.message.receive_v1","tenant_key":"2ee61fe50f4f1657","app_id":"cli_a1eff35b43b89063"},"event":{"message":{"chat_id":"oc_8c789ce8f4ecc6695bb63ca6ec4c61ea","chat_type":"group","content":"{\"text\":\"@_user_1  https://juejin.cn/post/7463301526800826404 openai agent实现，大概是云端起个chrome用playwright搞得，感觉vm技术是ai重要基建之一\"}","create_time":"1732612279943","mentions":[{"id":{"open_id":"ou_032f507d08f9a7f28b042fcd086daef5","union_id":"on_7111660fddd8302ce47bf1999147c011","user_id":""},"key":"@_user_1","name":"小煎蛋","tenant_key":"2ee61fe50f4f1657"}],"message_id":"om_4ab57cb6e889eeb81ca061b137238189","message_type":"text","update_time":"1732612279943"},"sender":{"sender_id":{"open_id":"ou_470ac13b8b50fc472d9d8ee71e03de26","union_id":"on_9dacc59a539023df8b168492f5e5433c","user_id":"zhaoyingbo"},"sender_type":"user","tenant_key":"2ee61fe50f4f1657"}}}

test/reportAgent/setUrl.http

@@ -0,0 +1,4 @@
+POST http://localhost:3000/bot?app=egg HTTP/1.1
+content-type: application/json
+
+{"schema":"2.0","header":{"event_id":"c94518fbcb9d66cc93b4144cc69e4f0c","token":"tV9djUKSjzVnekV7xTg2Od06NFTcsBnj","create_time":"1732612280153","event_type":"im.message.receive_v1","tenant_key":"2ee61fe50f4f1657","app_id":"cli_a1eff35b43b89063"},"event":{"message":{"chat_id":"oc_8c789ce8f4ecc6695bb63ca6ec4c61ea","chat_type":"group","content":"{\"text\":\"@_user_1  https://xiaomi.f.mioffice.cn/wiki/V4ZkwhDR8iRCqIk7X81k1rBj4Sc?sheet=a7a0f7\"}","create_time":"1732612279943","mentions":[{"id":{"open_id":"ou_032f507d08f9a7f28b042fcd086daef5","union_id":"on_7111660fddd8302ce47bf1999147c011","user_id":""},"key":"@_user_1","name":"小煎蛋","tenant_key":"2ee61fe50f4f1657"}],"message_id":"om_4ab57cb6e889eeb81ca061b137238189","message_type":"text","update_time":"1732612279943"},"sender":{"sender_id":{"open_id":"ou_470ac13b8b50fc472d9d8ee71e03de26","union_id":"on_9dacc59a539023df8b168492f5e5433c","user_id":"zhaoyingbo"},"sender_type":"user","tenant_key":"2ee61fe50f4f1657"}}}

utils/llm/index.ts

@@ -1,3 +1,5 @@
+import loggerIns from "@egg/logger"
+import { JsonOutputParser } from "@langchain/core/output_parsers"
 import { PromptTemplate } from "@langchain/core/prompts"
 
 import { adjustTimeRange, getSpecificTime, getTimeRange } from "../time"
@@ -15,24 +17,51 @@ const invoke = async (
   promptName: string,
   variables: Record<string, any>,
   requestId: string,
-  temperature = 0
+  temperature = 0,
+  jsonMode = false
 ) => {
-  const { langfuse, langfuseHandler } = await getLangfuse("invoke", requestId)
+  const logger = loggerIns.child({ requestId })
-  const prompt = await langfuse.getPrompt(promptName)
+  const attemptInvoke = async () => {
+    const { langfuse, langfuseHandler } = await getLangfuse("invoke", requestId)
+    const prompt = await langfuse.getPrompt(promptName)
 
-  const langchainTextPrompt = PromptTemplate.fromTemplate(
+    const langchainTextPrompt = PromptTemplate.fromTemplate(
-    prompt.getLangchainPrompt()
+      prompt.getLangchainPrompt()
-  ).withConfig({
+    ).withConfig({
-    metadata: { langfusePrompt: prompt },
+      metadata: { langfusePrompt: prompt },
-  })
+    })
 
-  const chain = langchainTextPrompt.pipe(await getModel(temperature))
+    const chain = langchainTextPrompt.pipe(await getModel(temperature))
 
-  const { content } = await chain.invoke(variables, {
+    if (jsonMode) {
-    callbacks: [langfuseHandler],
+      chain.pipe(new JsonOutputParser())
-  })
+    }
 
-  return content
+    const { content } = await chain.invoke(variables, {
+      callbacks: [langfuseHandler],
+    })
+
+    return content
+  }
+
+  let result
+  let attempts = 0
+  do {
+    try {
+      result = await attemptInvoke()
+      break
+    } catch (e) {
+      logger.error(`🚀 ~ invoke ~ attemptInvoke ~ e: ${e}`)
+      attempts++
+    }
+  } while (attempts < 3)
+
+  if (!result) {
+    logger.error("Failed to invoke after 3 attempts")
+    return ""
+  }
+
+  return result
 }
 
 /**
@@ -42,6 +71,7 @@ const invoke = async (
  * @returns
  */
 const timeParser = async (userInput: string, requestId: string) => {
+  const logger = loggerIns.child({ requestId })
   const time = new Date().toLocaleString("zh-CN", { timeZone: "Asia/Shanghai" })
   const weekDay = `星期${"日一二三四五六"[new Date().getDay()]}`
   const invokeParser = async () => {
@@ -63,7 +93,7 @@ const timeParser = async (userInput: string, requestId: string) => {
       )) as string
       return JSON.parse(res.replaceAll("`", ""))
     } catch (e) {
-      console.error("🚀 ~ timeParser ~ invokeParser ~ e", e)
+      logger.error(`🚀 ~ timeParser ~ invokeParser ~ e: ${e}`)
       // 如果解析失败，则返回空字符串
       return { s: "", e: "" }
     }

utils/string.ts

@@ -0,0 +1,33 @@
+/**
+ * 从小米链接中提取 sheetToken 和 range。
+ *
+ * @param {string} url - 要提取数据的URL。
+ * @returns {{sheetToken: string, range: string | null} | null} - 包含 sheetToken 和 range 的对象，如果没有匹配则返回 null。
+ */
+export const extractSheetIds = (url: string) => {
+  // 定义匹配 wiki 和 sheets 两种URL格式的正则表达式
+  const pattern = /(?:wiki|sheets)\/([\w\d]+)(?:\?sheet=([\w\d]+))?/
+  const match = url.match(pattern)
+  console.log("🚀 ~ extractSheetIds ~ match:", match)
+  if (match) {
+    return {
+      sheetToken: match[1], // sheetToken 在组1
+      range: match[2] || null, // range 在组2，如果没有则返回 null
+    }
+  }
+  return null // 如果没有匹配，则返回 null
+}
+
+/**
+ * 校验链接是否合法
+ * @param {string} link - 网页链接
+ * @throws {Error} - 当链接为空或格式不正确时抛出错误
+ */
+export const validateLink = (link: string): void => {
+  if (!link) throw new Error("链接不能为空")
+  try {
+    new URL(link)
+  } catch {
+    throw new Error("链接格式不正确")
+  }
+}