A FastAPI-based code interpreter service that provides code execution and file management capabilities. This service is compatible with the LibreChat Code Interpreter API specification.
Note
This project is mostly a proof of concept and is not intended to be used in production.
For production ready solution and to support the amazing work of LibreChat maintainers use the LibreChat Code Interpreter.
- Easy deployment with single docker compose file
- Code is executed in a isolated Docker container sandbox, custom images supported
- Supports file upload and download
- Supports concurrent code execution
- Supports Python and R languages (possibility to extend to other languages)
- RESTful API with OpenAPI documentation
code-interpreter-demo.mp4
Run the project with docker compose using docker compose -f compose.prod.yml up
It's possible to overwrite the default environment variables defined in ./app/shared/config.py by creating a .env file in the root directory.
By default the project will create two directories in the root directory: ./config and ./uploads.
config directory will hold the sqlite database and temp uploaded files.
uploads directory will hold the files uploaded by the users. All files uploaded by the users will be, by default, deleted after 24 hours.
LibreChat is configured to use the code interpreter API by default.
To configure LibreChat to use the local code interpreter, set the following environment variables in LibreChat:
LIBRECHAT_CODE_API_KEY=<any-value-here>
LIBRECHAT_CODE_BASEURL=http(s)://host:port/v1/librechat # for local testing use to point to host IP http://host.docker.internal:8000/v1/librechat- Install dependencies using uv:
uv sync --all-extras
source .venv/bin/activate- Start the development server:
docker compose upThe API will be available at http://localhost:8000. The OpenAPI documentation can be accessed at http://localhost:8000/docs.
pytest