All projects

Telegram Gemini Chatbot

Experiment

Telegram chatbot powered by Gemini with real-time search capabilities.

View on GitHub

🤖 Telegram Gemini 2.5 Flash Chatbot with Internet Search

GitHub Repository: vieteri/chatbot

A powerful Telegram chatbot powered by Google's Gemini 2.5 Flash with real-time internet search capabilities.

✨ Features

  • 🔍 Built-in Google Search - Native search grounding with automatic citations
  • 🧠 Gemini 2.5 Flash - Google's latest thinking model with enhanced reasoning
  • ⚡ Automatic Search Detection - Searches when current information is needed
  • 🚫 Unfiltered Responses - No content policy restrictions
  • 📱 HTML Formatting - Rich text with bold, italic, code, links
  • 💬 Conversation Memory - Remembers 20 messages of context
  • 📊 Message Chunking - Handles long responses automatically
  • 🔧 Serverless - Webhook-based architecture

🚀 Technology Stack

  • Backend: Node.js with serverless architecture
  • AI Model: Google Gemini 2.5 Flash
  • Platform: Telegram Bot API
  • Search: Built-in Google Search (no additional API needed)
  • Deployment: Vercel/Netlify compatible
  • Architecture: Webhook-based for real-time responses

🎯 Key Capabilities

Intelligent Search Integration

  • Smart Detection: Automatically identifies when current information is needed
  • Function Calling: Uses Google's function calling API to trigger web searches
  • Real-time Results: Fetches up to 5 search results with titles, snippets, and links
  • Intelligent Synthesis: Combines search results with AI reasoning for comprehensive answers

Advanced Chat Features

  • Conversation Memory: Maintains context across 20 messages per user
  • HTML Formatting Support: Rich text formatting with Telegram HTML tags
  • Message Chunking: Automatically handles long responses within Telegram's limits
  • Unfiltered Responses: Provides unrestricted AI interactions

💡 Usage Examples

Automatic Current Information

User: "What's the latest news about AI?"
Bot: 🔍 [Automatically searches Google and provides current AI news with citations]

User: "What's the Bitcoin price today?"
Bot: 🔍 [Searches for current Bitcoin price and provides up-to-date information]

Conversational AI

User: "Explain quantum computing"
Bot: [Provides detailed explanation, may search for latest developments]

🔧 Technical Implementation

API Architecture

  • POST /api/webhook - Telegram webhook handler
  • POST /api/set-webhook - Set Telegram webhook URL
  • GET / - Bot status and setup page

Environment Configuration

# Telegram Bot Token (from @BotFather)
TELEGRAM_BOT_TOKEN=your_telegram_bot_token_here

# Google AI API Key (from https://ai.google.dev/)
GEMINI_API_KEY=your_gemini_api_key_here

Deployment Process

  1. Vercel Deployment (Recommended):

    npm install
vercel --prod

  2. Webhook Setup:

    curl -X POST https://your-domain.com/api/set-webhook \
  -H "Content-Type: application/json" \
  -d '{"webhookUrl": "https://your-domain.com/api/webhook"}'

🎨 Bot Commands

CommandDescription
/startWelcome message and feature overview
/clearClear conversation history

🔒 Security & Privacy

  • Serverless Architecture: No persistent server storage
  • Memory Management: Conversation history cleared on server restart
  • API Security: Secure webhook validation
  • Environment Variables: Secure credential management

🚀 Performance Features

Message Handling

  • Chunking Algorithm: Respects Telegram's 4096 character limit
  • Format Preservation: Maintains HTML formatting across message splits
  • Response Optimization: Efficient processing of long AI responses

Search Optimization

  • Citation Integration: Automatic source attribution
  • Result Filtering: Intelligent selection of most relevant search results
  • Response Synthesis: Combines multiple sources for comprehensive answers

📱 HTML Formatting Support

The bot supports rich Telegram HTML formatting:

  • <b>bold</b> - Bold text
  • <i>italic</i> - Italic text
  • <code>code</code> - Inline code
  • <u>underline</u> - Underlined text
  • <s>strikethrough</s> - Strikethrough text
  • <a href="url">link</a> - Clickable links
  • <pre>code block</pre> - Code blocks

🔮 Advanced Features

Function Calling Implementation

  • Google Search Integration: Native search function calling
  • Context Awareness: Understands when to search vs. use training data
  • Result Processing: Intelligent parsing and synthesis of search results

Conversation Management

  • Memory Persistence: Maintains context throughout conversations
  • User Isolation: Separate conversation threads per user
  • Context Optimization: Efficient memory usage with 20-message limit

🌟 Development Highlights

AI Integration Challenges

  • Implementing Gemini 2.5 Flash function calling for search
  • Balancing search frequency with response speed
  • Managing conversation context and memory efficiently

Telegram API Optimization

  • Webhook-based architecture for real-time responses
  • Message chunking algorithm for long responses
  • HTML formatting preservation across message splits

Serverless Architecture

  • Stateless design for scalability
  • Environment-based configuration management
  • Efficient API endpoint organization

🎯 Use Cases

  • Research Assistant: Get current information on any topic
  • News Updates: Latest developments in specific fields
  • Technical Support: Programming and technical questions with current solutions
  • General Knowledge: Comprehensive answers combining AI knowledge and current data
  • Conversational AI: Natural language interactions with memory

📊 Project Statistics

  • Languages: JavaScript/Node.js
  • APIs: Telegram Bot API, Google Gemini API
  • Architecture: Serverless, Webhook-based
  • Features: 8+ core features including search, memory, formatting
  • Deployment: Production-ready on Vercel/Netlify

This project demonstrates advanced integration of modern AI APIs with messaging platforms, showcasing expertise in serverless architecture, real-time communication, and intelligent search integration.