Telegram RAG AI Bot using APIs

A Telegram bot that integrates with CustomGPT.ai to provide AI-powered responses using your custom knowledge base.

More information on GitHub - https://github.com/Poll-The-People/customgpt-integrations

Get your CustomGPT.ai RAG API key here, needed to use this integration.

Features

AI-Powered Responses: Uses CustomGPT.ai's API to answer questions from your knowledge base
Conversation Management: Maintains context within chat sessions
Rate Limiting: Built-in daily (100) and per-minute (5) message limits
Starter Questions: Interactive buttons with example queries
Usage Statistics: Track your daily usage with /stats
Session Management: 30-minute conversation timeout with auto-cleanup
Security: SSL certificate handling and secure API communication

Quick Start

Prerequisites

Python 3.8+
Telegram Bot Token (from @BotFather)
CustomGPT.ai API Key and Project ID
SSL certificates (handled automatically with certifi)

Installation

Clone the repository:
```
cd Telegram-Bot/
```
Install dependencies:
```
pip install -r requirements.txt
```

Set up environment variables: Create a .env file:

TELEGRAM_BOT_TOKEN=your_bot_token_here
CUSTOMGPT_API_KEY=your_api_key_here
CUSTOMGPT_PROJECT_ID=your_project_id_here

# Optional configuration
DAILY_LIMIT=100
MINUTE_LIMIT=5
SESSION_TIMEOUT_MINUTES=30

Running the Bot

Local Development (Polling Mode)

python bot.py

This runs the bot in polling mode - perfect for development and testing.

Bot Commands

/start - Welcome message with example question buttons
/help - Show available commands and tips
/examples - Display example questions you can ask
/stats - View your usage statistics
/clear - Clear conversation history and start fresh

Deployment Options

Option 1: Vercel (Webhook Mode) - Free

Best for: Simple bots with quick responses

Pros:

Completely free
Auto-scaling
HTTPS included
Easy deployment

Cons:

10-second timeout limit
No persistent storage
Cold starts
No rate limiting

Vercel Setup Instructions

Prerequisites

Telegram Bot Token (from @BotFather)
CustomGPT API Key and Project ID
Vercel account (free at vercel.com)
Git installed locally

Deployment Steps

Option A: Deploy via Vercel CLI

# Install Vercel CLI
npm i -g vercel

# Login to Vercel
vercel login

# Navigate to vercel-bot directory
cd vercel-bot

# Deploy (follow prompts)
vercel

# Set environment variables
vercel env add TELEGRAM_BOT_TOKEN production
vercel env add CUSTOMGPT_API_KEY production
vercel env add CUSTOMGPT_PROJECT_ID production

# Redeploy to apply env vars
vercel --prod

Option B: Deploy via GitHub

Push code to GitHub repository
Connect repository to Vercel
Add environment variables in Vercel dashboard
Deploy

Set Telegram Webhook

After deployment, set your bot's webhook:

import requests

BOT_TOKEN = "your_bot_token_here"
WEBHOOK_URL = "https://your-project.vercel.app/api/webhook"

response = requests.post(
    f"https://api.telegram.org/bot{BOT_TOKEN}/setWebhook",
    json={"url": WEBHOOK_URL}
)
print(response.json())

Or use curl:

curl -X POST "https://api.telegram.org/bot<BOT_TOKEN>/setWebhook" \
     -H "Content-Type: application/json" \
     -d '{"url": "https://your-project.vercel.app/api/webhook"}'

Test Your Bot

Open Telegram and search for your bot
Send /start to begin
Ask any question!

Vercel Limitations

10-Second Timeout

Vercel serverless functions have a 10-second execution timeout on the free tier. This means:

Each webhook request (Telegram message) must complete within 10 seconds
This includes: receiving message → calling CustomGPT API → sending response
If CustomGPT takes too long to respond, the function will timeout
The bot uses 8-second timeouts for API calls to stay within limits

Other Limitations

No persistent storage: Sessions reset on cold starts
No rate limiting: Each request is independent
Cold starts: First message after inactivity may be slower
No background tasks: Can't do async processing

How Vercel Deployment Works

Webhook Reception: Telegram sends updates to /api/webhook
Message Processing: Bot processes commands or forwards to CustomGPT
Session Management: Maintains conversation context in memory
Response Delivery: Sends CustomGPT response back to user

Option 2: Replit (Polling Mode) - Free with Limitations

Best for: Development and testing

Pros:

Free tier available
Persistent storage
Web IDE
Easy setup

Cons:

Sleeps after inactivity
Requires pinging to stay alive
Limited resources

Replit Setup Instructions

Step 1: Create Replit Account

Go to replit.com
Sign up for free

Step 2: Create New Repl

Click "Create Repl"
Choose "Python" template
Name it "customgpt-telegram-bot"

Step 3: Upload Files

Upload these files to your Repl:

bot.py
customgpt_client.py
simple_cache.py
requirements.txt

Step 4: Set Environment Variables

Click "Secrets" (lock icon) in left sidebar
Add these secrets:
- TELEGRAM_BOT_TOKEN = your bot token
- CUSTOMGPT_API_KEY = your API key
- CUSTOMGPT_PROJECT_ID = your project ID

Step 5: Update bot.py for Replit

Add this at the top of bot.py:

import os
# Replit uses Secrets instead of .env
BOT_TOKEN = os.environ.get('TELEGRAM_BOT_TOKEN')
CUSTOMGPT_API_KEY = os.environ.get('CUSTOMGPT_API_KEY')
CUSTOMGPT_PROJECT_ID = os.environ.get('CUSTOMGPT_PROJECT_ID')

Step 6: Create main.py

from bot import main

if __name__ == '__main__':
    main()

Step 7: Run the Bot

Click "Run" button
Your bot should start!

Step 8: Keep Bot Alive (Optional)

For free tier with better uptime:

Set up UptimeRobot to ping your Repl every 5 minutes
Your Repl URL: https://your-repl-name.your-username.repl.co

Replit-Specific Files

.replit

run = "python bot.py"
language = "python3"

[packager]
language = "python3"

replit.nix

{ pkgs }: {
  deps = [
    pkgs.python39
    pkgs.python39Packages.pip
  ];
}

Option 3: Railway (Polling/Webhook) - Paid

Best for: Production bots

Pros:

No timeout limits
Persistent storage options
Better performance
Supports both modes

Cons:

Requires payment
More complex setup

Railway Setup

Install Railway CLI:

npm install -g @railway/cli

Deploy:

railway login
railway init
railway up

Add environment variables in Railway dashboard:
- TELEGRAM_BOT_TOKEN
- CUSTOMGPT_API_KEY
- CUSTOMGPT_PROJECT_ID

Option 4: VPS/Cloud (Any Mode) - Varies

Best for: Full control

Options:

AWS EC2 (free tier)
Google Cloud (free tier)
DigitalOcean ($5/month)
Any Linux VPS

VPS Setup

Install Python and dependencies:

sudo apt update
sudo apt install python3 python3-pip
pip3 install -r requirements.txt

Set up environment variables:

export TELEGRAM_BOT_TOKEN="your_token"
export CUSTOMGPT_API_KEY="your_key"
export CUSTOMGPT_PROJECT_ID="your_project_id"

Run bot:

python3 bot.py

Set up systemd service (optional):

[Unit]
Description=Telegram CustomGPT Bot
After=network.target

[Service]
Type=simple
User=youruser
WorkingDirectory=/path/to/bot
Environment="TELEGRAM_BOT_TOKEN=your_token"
Environment="CUSTOMGPT_API_KEY=your_key"
Environment="CUSTOMGPT_PROJECT_ID=your_project_id"
ExecStart=/usr/bin/python3 bot.py
Restart=always

[Install]
WantedBy=multi-user.target

Deployment Comparison

Platform	Best For	Cost	Persistent Storage	Rate Limiting	Timeout
Vercel	Simple bots	Free	No	No	10s
Replit	Development	Free*	Yes	Yes	None
Railway	Production	Paid	Yes	Yes	None
VPS	Full control	Varies	Yes	Yes	None

*Free tier sleeps after inactivity

Recommended Strategy

Development/Testing: Replit or local polling
Simple production: Vercel (if responses are fast)
Production with complex queries: Railway or VPS
Enterprise/High traffic: VPS with load balancing

Technical Details

Rate Limiting

Daily Limit: 100 messages per user
Minute Limit: 5 messages per minute
Implementation: In-memory cache (resets on restart)

Session Management

Timeout: 30 minutes of inactivity
Storage: In-memory (non-persistent)
Cleanup: Automatic for expired sessions

API Integration

CustomGPT API: RESTful API with streaming support
SSL Handling: Uses certifi for certificate verification
Error Handling: Graceful degradation with user-friendly messages

Common Issues & Solutions

SSL Certificate Error (macOS)

# Already fixed in customgpt_client.py
import certifi
import ssl
ssl_context = ssl.create_default_context(cafile=certifi.where())

Bot Not Responding

Check bot token is correct
Verify API credentials
Ensure bot is running (python bot.py)
Check network connectivity

For webhook mode:

Check webhook is set correctly: https://api.telegram.org/bot<TOKEN>/getWebhookInfo
Verify environment variables in hosting platform dashboard
Check function logs for errors

Rate Limit Exceeded

Wait for the timeout period
Daily limits reset at midnight
Minute limits reset after 60 seconds

Session Expired

Use /clear to start a new conversation
Sessions timeout after 30 minutes of inactivity

Vercel Timeout Errors

CustomGPT API might be slow - the bot has 8-second timeout
Consider implementing a "thinking..." message for long responses
For complex queries, responses might timeout
Consider using Railway or VPS for complex bots

Sessions Reset Frequently (Vercel/Serverless)

This is normal behavior on serverless platforms
Each cold start resets the in-memory session cache
Consider using a database for persistent sessions (requires paid tier)

Development

Testing Locally

Use polling mode for easier debugging
Set lower rate limits for testing
Use /stats to monitor usage

Adding Features

Extend handle_message in bot.py for new commands
Modify simple_cache.py for persistence
Update customgpt_client.py for API changes

Debugging

# Enable debug logging
import logging
logging.basicConfig(level=logging.DEBUG)

Security Considerations

Never commit .env file - It contains sensitive credentials
Use environment variables for all secrets

Implement user allowlisting if needed:

ALLOWED_USERS = [123456789]  # Telegram user IDs

Monitor usage with /stats command
Set appropriate rate limits based on your needs

Support

CustomGPT Links

Video Tutorial

Telegram Bot Setup Video

Platform Documentation

Telegram Bot Documentation: core.telegram.org/bots/api

Troubleshooting

For issues:

Check the Common Issues section
Review logs for error messages
Ensure all prerequisites are met
Verify API credentials are correct

For CustomGPT API issues, check the resources above. For Telegram Bot API issues, check Telegram Bot Documentation.

License

MIT

Ready to deploy? Choose your hosting option above and follow the deployment guide!