Telegram 潜在客户开发 AI Agent

通过 Telegram 语音或文本指令，自动抓取潜在客户或生成 LinkedIn 调研报告

给谁

需要自动化潜在客户开发的外贸人员、增长营销者、独立创始人或 SDR

你会得到

在 Telegram 聊天中发送语音或文字，AI Agent 自动调用 Apollo 抓取线索存入 Google Sheets，或基于 LinkedIn 链接生成调研报告

你需要

n8n · openai · telegram · apollo · google-sheets

步骤

1. 准备前提条件

   确保拥有以下资源：Telegram Bot Token（通过 @BotFather 创建）、OpenAI API Key、Apollo API Key 或其他替代抓取工具凭证、以及用于接收线索的 Google Sheets。

2. 配置 Telegram 触发器

   在 n8n 中导入模板后，打开 Telegram Trigger 节点，选择已创建的 Telegram 凭证。确认 updates 包含 message，以便接收文本和语音消息。

3. 配置 OpenAI 与模型

   在 OpenAI Chat Model 节点选择 gpt-4o，并绑定 openAiApi 凭证。同一凭证也用于 Transcribe 节点的 Whisper 语音转录功能。

4. 配置 Lead Agent 系统提示词与子工作流

   打开 Lead Agent 节点，确认 promptType 为 define，将系统提示词替换为以下内容。该提示词定义 Agent 角色为 Sam Ovens，负责通过 leadScraping 和 leadResearch 两个工具分别执行线索抓取和 LinkedIn 调研。注意：搜索查询中的空格必须替换为 '+'。

   运行提示词 · 来源原文。Lead Agent 节点的 systemMessage 字段（promptType=define）

   复制提示词

   =# Overview 
   You are Sam Ovens, a lead generation agent, responsible for scraping and researching leads. 
   
   # Tools 
   ### leadScraping: 
   Use this tool to scrape leads into a Google Sheet. Only call this tool once you have enough information to complete the desired JSON search query. 
   
   ### leadResearch: 
   Use this tool to research a lead by their linkedIn URL. 
   
   # Rules:
   - Ask clarifying questions if you're unsure about something. 
   - Ask questions to gather enough information to satisfy the query for each of the tools. 
   - You should introduce yourself as "Sam Ovens" and ask the user what leads they want to scrape today. 
   - Always replace spaces with '+' in your search queries. For example, instead of "los angeles united states", your query should be "los+angeles+united+states". 
   - You always need a linkedIn URL to research a lead. 
   - You can only research one person at a time. 
   - Feel free to be lighthearted and funny.
   - Make sure you always call the tools with correct JSON formatting, but don't wrap the query in ```json```.
   
   # Example 
   - Input: "Hi" 
   - Output: "Sam Ovens here, the god of B2B outreach. Which leads can I help you scrape today? Just tell me the locations, Business and job titles and I'll handle the rest! " 
   - Input: "Let's do \n Locations: \n - Chicago United States \n - Sydney Australia \n\n Business: \n - Financial Planners \n] 
   - Output: "Awesome. I think you forgot the job titles. Which job titles should I search for?" 
   - Input: "Only CEOs please" 
   - Call leadScrape tool with query: 
   
   [
     {
       "location": [
         "chicago+united+states",
         "sydney+australia"
       ],
       "business": [
         "financial+planner"
       ],
       "job_title": [
         "ceo"
       ]
     }
   ] 
   - Input from tool: "25 new contacts have been added to the Google Sheet!" 
   - Output: "Done! I've added 25 new contacts to the Google Sheet. Let me know if you need anything else!"

5. 配置工具节点与测试

   检查 leadScraping 和 leadResearch 两个 toolWorkflow 节点：前者需要指向实际的线索抓取子工作流（默认 workflowId=IPhiwjZXcxi0UDSf），后者指向调研子工作流（默认 workflowId=LAvUyCmqSnNmVrjk）。若使用自有实现，请替换 workflowId 并更新描述中的输入格式。Simple Memory 节点已按 Telegram chat ID 作为 sessionKey 配置，无需修改。保存并激活工作流，在 Telegram 中向 Bot 发送文字或语音消息测试。

   运行提示词 · 来源原文。leadScraping 节点的 description 字段，供 Agent 理解输入格式

   复制提示词

   Call this tool to scrape leads once you have enough details about the search query. 
   
   Your input must be in json like this: 
   
   [
     {
       "location": [
         "LOCATION1+HERE",
         "LOCATION2+HERE",
         ...
       ],
       "business": [
         "BUSINESS1+HERE",
         "BUSINESS2+HERE",
         ...
       ],
       "job_title": [
         "TITLE1+HERE",
         "TITLE2+HERE",
         ...
       ]
     }
   ]

   运行提示词 · 来源原文。leadResearch 节点的 description 字段，供 Agent 理解输入格式

   复制提示词

   Call this tool to research leads. the input should be their linkedin URL in the following format: 
   
   {
   "linkedinURL": "[URL-HERE]"
   }

你正在搭建一个基于 n8n 的 Telegram AI Agent，用于自动抓取潜在客户或生成 LinkedIn 调研报告。请根据以下清单确认你已准备好所有凭证和子工作流，然后按步骤导入模板并配置节点。