🤖 AI Document Bot

PDF Analysis Telegram Bot with Dual AI Support (GROQ API + Ollama Models)

📖 Overview

AI Document Bot is Telegram Bot that transforms PDF Documents into interactive, queryable knowledge bases. Upload any PDF Document and ask questions about its content using either cloud-based GROQ API or Local Ollama Models.

✨ Key Features

• 📄 PDF Document Processing - extract and analyze text from pdf files (up to 20mb)

• 🧠 Dual AI Support - choose between groq (cloud/fast) or ollama (local/private)

• 🔍 Advanced Vector Search - multiple search strategies for accurate content retrieval

• 🛠️ Debug Mode - inspect search results and understand how the bot finds relevant information

• ⚙️ Flexible Configuration - switch between AI Services and Models on the fly

🚀 Quick Start

📋 Prerequisites

• Python 3.8+
• Telegram Bot Token (get from @botfather)
• Groq API Key (optional, get from groq.com)
• Ollama (optional, for local ai - ollama.ai)

Note

To experience the Bot in Action, access it on Telegram via @AI_Docz_Bot. No GROQ API Key or Ollama Model is required.

⚙️ Installation

1. Clone the Repository:

git clone https://github.com/jafarbekyusupov/ai-docs-tgbot.git
cd ai-docs-tgbot

2. Create virtual environment

python -m venv venv
source venv/bin/activate

Tip

On Windows:

python -m venv venv
venv\Scripts\activate

3. Install Dependencies:

pip install -r requirements.txt

4. Configure Environment Variables:

Create a .env file in the root directory:

TELEGRAM_BOT_TOKEN=your_telegram_bot_token_here
GROQ_API_KEY=your_groq_api_key_here  # optional

5. Setup Ollama (Optional):

Install Ollama from https://ollama.ai

ollama pull llama3.2  # or any other model you prefer

6. Run the Bot:

python run.py

🎯 Usage

🔧 Basic Workflow

1. Start the Bot - send /start to see welcome message
2. Configure AI Service (Optional) - use /settings to choose between Groq or Ollama (GROQ is set as default)
3. Upload PDF - send any PDF Document (Max 20MB - telegram's limit)
4. Ask Questions - start asking Questions about your Document Content

📱 Available Commands

Command	Description
/start	show welcome message and basic instructions
/settings	choose between groq and ollama ai services
/models	list and switch between available ollama models
/status	check ai services status and current configuration
/clear	clear current document and start over
/debug <query>	see detailed search results for debugging
/help	show help information

🤖 AI Service Options

☁️ Groq (Cloud AI/LLM)

• pros: fast response times, no local setup required
• cons: requires api key, sends data to cloud
• models: llama3-8b-8192 (default)

🏠 Ollama (Local AI/LLM)

• pros: completely private, no api costs, multiple model choices
• cons: slower response times, requires local installation
• models: llama3.2, mistral, codellama, and many more

🏗️ Architecture & Structure

📁 Project Structure

ai-docs-tgbot/
├── config.py                 # configuration and environment variables
├── run.py                    # application entry point
├── document_bot.py           # main bot class and setup
├── bot_handlers.py           # telegram message and callback handlers
├── document_processor.py     # pdf text extraction and segmentation
├── ai_processor.py           # groq and ollama ai integration
├── vector_search.py          # faiss-based semantic search
├── ollama.py                 # ollama client implementation
├── requirements.txt          # python dependencies
└── .env                      # environment variables (CREATE THIS FILE ON UR OWN MACHINE)

🔄 Core Components

📑 Document Processor

• extracts text from pdf files using pypdf2
• analyzes document structure and identifies headers
• segments text into meaningful chunks for better retrieval
• supports both advanced and simple segmentation strategies

🔍 Vector Search

• Creates embeddings using sentence-transformers
• Implements multiple search strategies:
• • Smantic Search - finds content based on meaning
• • Keyword Search - matches specific terms
• • Fuzzy Search - handles partial matches
• • Section Search - searches within document sections
• Uses faiss for efficient similarity search

🤖 AI Processor

• supports both groq and ollama apis
• handles model selection and switching
• manages api calls and error handling
• provides consistent interface for different ai services

🎛️ Bot Handlers

• processes telegram messages and commands
• manages user sessions and preferences
• handles file uploads and user interactions
• provides inline keyboards for easy configuration

🛠️ Advanced Features

🔍 Multi-Strategy Search

Bot uses Four Different Search Strategies to find the Most Relevant Content:

Important

1. Semantic Search - understands the meaning of your question
2. Adaptive Keyword Search - matches important document terms
3. Fuzzy Matching - finds partial word matches
4. Section-Based Search - searches within specific document sections

📊 Debug Mode

use /debug <your question> to see exactly how the bot finds relevant information:

• view search strategies and their results
• see similarity scores for different content segments
• understand why certain answers were selected

⚡ Intelligent Segmentation

the document processor automatically:

• identifies document headers and sections
• creates logical text segments
• preserves context across segment boundaries
• handles various document formats and structures

🔧 Configuration

🌍 Environment Variables

Warning

Required

TELEGRAM_BOT_TOKEN=your_bot_token

Optional (for Groq Support)

GROQ_API_KEY=your_groq_api_key

Optional (Customize Ollama URL)

OLLAMA_BASE_URL=http://localhost:11434 # or the port you set it to

🎛️ Runtime Configuration

Note

Users can configure the Bot via Telegram Commands:

• switch between groq and ollama
• select different ollama models
• view service status and availability

🧪 Technical Details

📚 Dependencies

Package	Purpose
pyTelegramBotAPI	telegram bot framework
PyPDF2	pdf text extraction
groq	groq ai api client
sentence-transformers	text embeddings
faiss-cpu	vector similarity search
numpy	numerical computations

🔒 Security Features

• no persistent storage of document content
• user sessions are memory-based only
• api keys are environment-based
• local ollama option for complete privacy

⚡ Performance Optimizations

• efficient text segmentation algorithms
• normalized vector embeddings for better search
• combined search strategies for improved accuracy
• fallback mechanisms for robust operation

---

AI Docs TGBot @ jafarbekyusupov

⭐ Star this Repo • 🐛 Report Bug • 💡 Request Feature