A privacy-focused Telegram bot that runs 100% locally using open-source AI models. Supports text, voice, and image processing without sending data to external servers.

| Category | Feature | Description |
|---|---|---|
| Core | Local Execution | Runs entirely on your machine using Ollama |
| User Authentication | Password-protected access with personalized sessions | |
| AI Capabilities | Model Switching | Choose between Gemma3 (image support) and Qwen3 (advanced reasoning) |
| Context Memory | Maintain conversation history (2-50 message pairs) | |
| Voice Processing | Transcribe voice messages using Whisper | |
| Multimedia | Image Analysis | Describe images using Gemma3 |
| Image Generation | Create images via Stable Diffusion integration | |
| Customization | Temperature Control | Adjust creativity level (0.0-1.0) |
| Thinking Modes | Toggle advanced reasoning mode for Qwen3 | |
| Security | Secret Management | Store token/password in .env file |
- Python 3.10+
- At least 16 GB RAM, 32 Is reccomended
- Ollama running locally (installation guide)
- Optional: Stable Diffusion WebUI for image generation
ollama pull qwen3:14b
📦 Dependencies:
pip install python-telegram-bot ollama faster-whisper pydub pillow python-dotenv aiohttp nest_asyncio🧪 Configuration Fill .env file with : TOKEN=your_telegram_bot_token PASSWORD=your_secure_password
🚀 Usage:
python telebot.pyThis project uses open-source components:
Ollama (MIT)
Faster-Whisper (MIT)
stable-diffusion-webui (AGPL-3.0 license)