GitHub - m0wer/aibot: Telegram bot powered by Ollama, capable of handling text and voice messages, with configurable language models and system prompts.

2 min read Original article ↗

Ollama Telegram Bot

This project implements a Telegram bot powered by Ollama, capable of handling text and voice messages, with configurable language models and system prompts.

Features

  • Handles text and voice messages
  • Uses Ollama for natural language processing
  • Configurable Ollama URL and model
  • Text-to-Speech (TTS) and Speech-to-Text (STT) capabilities
  • Uses RQ (Redis Queue) for processing messages and audio conversion
  • Different priority queues and a special GPU queue for Whisper STT using CUDA
  • Allows users to adjust their system prompt
  • Uses recent messages as context for responses
  • Stores all interactions in SQLite using SQLModel
  • Dockerized setup with configurable resource limits

Prerequisites

  • Docker and Docker Compose
  • Ollama server running (can be on the same machine or a remote server)
  • Telegram Bot Token (obtainable from BotFather on Telegram)

Running the Bot

  1. Build and start the services:

    cp .env.example .env
# Edit the .env file and set your TELEGRAM_TOKEN and OLLAMA_URL
docker compose up --build

  2. The bot should now be running and responding to messages on Telegram.

Configuration

You can configure the following environment variables in your .env file:

  • TELEGRAM_TOKEN: Your Telegram Bot Token (required)
  • OLLAMA_URL: URL of your Ollama server (default: http://host.docker.internal:11434)
  • OLLAMA_MODEL: Ollama model to use (default: llama3.1)

Usage

  • Start a chat with your bot on Telegram
  • Send text messages or voice notes to interact with the bot
  • Use the /prompt command to change the system prompt for the Ollama model

Development

To modify the bot or add new features:

  1. Edit the bot.py file to change the bot's behavior
  2. Modify the Dockerfile if you need to add new system dependencies
  3. Update the requirements.in file if you add or remove Python dependencies
  4. Rebuild the requirements.txt file using pip-compile requirements.in
  5. Rebuild and restart the services using docker compose up --build

Troubleshooting

  • If the bot can't connect to Ollama, make sure the OLLAMA_URL is correctly set and the Ollama server is running and accessible.
  • For GPU support, ensure you have the necessary GPU drivers and Docker GPU runtime set up. Otherwise, disable the GPU reservations from the docker compose.yaml file with sed -i 's/^.*devices:.*$/#&/' docker-compose.yaml.
  • Check the Docker logs for any error messages: docker compose logs.

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

License

This project is licensed under the MIT License - see the LICENSE file for details.