TinyAgent
Example LLM agent for academic purposes. It has a very simple architecture:
- The LLM is given a mission and some tools
- It is called repeatedly until it emits a final message
This LLM is Read-Only and attempts to stay inside the Working Directory as it lacks safety features for writing.
- Note is it possible the Agent can break out of the working directory and send ANY file on your computer to the API.
It supports LM Studio (recommended) or any other OpenAI compatible API.
Usage
go run github.com/dans-stuff/tinyagent@main
Example
Contributing
Contributions are welcome. Please fork the repository and submit pull requests for improvements or bug fixes.
License
This project is open source and available under the MIT License.