Clawdbot: AI Assistant on Telegram

🤖 Why Telegram?

Telegram is arguably the best platform for bots. Its API is robust, fast, and free. By connecting your self-hosted LLM (Clawdbot) to Telegram, you get:

1. Access Everywhere: Chat with your AI from any device.
2. Voice Messages: Send audio, get text back (Speech-to-Text).
3. Media: Send images for analysis (Multi-modal models).

🏗️ Architecture

1. Telegram Bot API: Webhook or Long Polling.
2. Middleware (Node.js): Receives message -> Sends to LLM -> Formats response -> Sends back to Telegram.
3. LLM Backend: Ollama / LocalAI running Llama 3.

🛠️ Step-by-Step Guide

1. Create a Bot

Talk to @BotFather on Telegram.

• /newbot -> Name: MyPrivateAI -> Username: MyPrivateAI_bot.
• Copy the API Token.

2. Set Up Node.js Project

mkdir clawdbot-telegram
cd clawdbot-telegram
npm init -y
npm install telegraf axios dotenv

3. Write the Bot Logic (index.js)

require('dotenv').config();
const { Telegraf } = require('telegraf');
const axios = require('axios');

const bot = new Telegraf(process.env.TELEGRAM_TOKEN);

bot.start((ctx) => ctx.reply('Hello! I am your private AI.'));

bot.on('text', async (ctx) => {
    const userMessage = ctx.message.text;
    ctx.sendChatAction('typing'); // Show "typing..." status

    try {
        // Call local LLM (Ollama)
        const response = await axios.post('http://localhost:11434/api/generate', {
            model: 'llama3',
            prompt: userMessage,
            stream: false
        });

        ctx.reply(response.data.response);
    } catch (error) {
        console.error(error);
        ctx.reply('Error talking to the brain 🧠');
    }
});

bot.launch();
console.log('Bot is running...');

4. Run It

node index.js

🧠 Advanced Features

Maintaining Context

The simple example above has no memory. To fix this:

1. Store chat_id and message history in a simple array or Redis.
2. Send the last N messages as context to Ollama.

Voice Notes (Whisper)

Telegram sends audio as .ogg.

1. Download file.
2. Convert to .wav (ffmpeg).
3. Send to OpenAI Whisper (or local Faster-Whisper).
4. Feed text to LLM.

🚀 Deployment

Run this Node.js script on a Raspberry Pi, a spare laptop, or a cheap VPS (if you expose your Ollama port via tunnel like Ngrok).

🏁 Conclusion

Building a Telegram bot for your AI is a weekend project with massive utility. It democratizes access to your personal intelligence stack.