A Telegram bot framework for experimenting with LLM agents. BotLab provides a flexible XML-based configuration system for defining agent behaviors, conversation momentum, and interaction patterns.
- XML-based agent configuration
- Momentum-based conversation management
- Topic-based filtering
- Rate limiting and timing control
- Conversation history tracking
- DTD validation for configurations
pip install -e ".[dev]" # Install with development dependencies
# or
pip install -e . # Install only runtime dependencies- Create a
.envfile with your API keys:
ALLOWED_TOPIC_NAME=your_topic
ANTHROPIC_API_VERSION=2023-06-01
SPEAKER_MODEL=claude-3-sonnet-latest
INHIBITOR_MODEL=claude-3-haiku-latest
ANTHROPIC_API_KEY=your_key
TELEGRAM_TOKEN=your_token
SPEAKER_PROMPT_FILE=config/agents/odv.xml
INHIBITOR_PROMPT_FILE=config/agents/inhibitor.xml- Configure your agents in
config/agents/:
odv.xml- Main speaker agent configurationinhibitor.xml- Conversation flow control agent- Additional agents as needed
- Start the bot:
python -m botlab.bot-
Add the bot to a Telegram group
-
Create a topic named as specified in
ALLOWED_TOPIC_NAME -
Interact with the bot in the designated topic
Run tests:
pytestbotlab/
├── src/
│ └── botlab/
│ ├── __init__.py
│ ├── bot.py
│ └── xml_handler.py
├── config/
│ ├── agents/
│ │ ├── odv.xml
│ │ └── inhibitor.xml
│ └── dtd/
│ ├── agent.dtd
│ └── messages.dtd
├── tests/
│ ├── integration/
│ │ └── test_bot.py
│ └── unit/
│ └── test_xml_handler.py
├── setup.py
└── README.md
MIT License
