A LangChain implementation of the ChatGPT Code Interpreter. Using CodeBoxes as backend for sandboxed python code execution. CodeBox is the simplest cloud infrastructure for your LLM Apps. You can run everything local except the LLM using your own OpenAI API Key.

• Dataset Analysis, Stock Charting, Image Manipulation, ....
• Internet access and auto Python package installation
• Input text + files -> Receive text + files
• Conversation Memory: respond based on previous inputs
• Run everything local except the OpenAI API (OpenOrca or others maybe soon)
• Use CodeBox API for easy scaling in production (coming soon)

Get your OpenAI API Key here and install the package.

pip install "codeinterpreterapi[all]"

Everything for local experiments are installed with the all extra. For deployments, you can use pip install codeinterpreterapi instead which does not install the additional dependencies.

Make sure to set the OPENAI_API_KEY environment variable (or use a .env file)

from codeinterpreterapi import CodeInterpreterSession


async def main():
    # create a session
    session = CodeInterpreterSession()
    await session.astart()

    # generate a response based on user input
    response = await session.generate_response(
        "Plot the bitcoin chart of 2023 YTD"
    )

    # output the response (text + image)
    print("AI: ", response.content)
    for file in response.files:
        file.show_image()

    # terminate the session
    await session.astop()


if __name__ == "__main__":
    import asyncio
    # run the async function
    asyncio.run(main())

Bitcoin YTD Chart Output

from codeinterpreterapi import CodeInterpreterSession, File


async def main():
    # context manager for auto start/stop of the session
    async with CodeInterpreterSession() as session:
        # define the user request
        user_request = "Analyze this dataset and plot something interesting about it."
        files = [
            File.from_path("examples/assets/iris.csv"),
        ]

        # generate the response
        response = await session.generate_response(
            user_request, files=files
        )

        # output to the user
        print("AI: ", response.content)
        for file in response.files:
            file.show_image()


if __name__ == "__main__":
    import asyncio

    asyncio.run(main())

Iris Dataset Analysis Output

In case you want to deploy to production, you can utilize the CodeBox API for seamless scalability.

Please contact me if you are interested in this, as it is still in the early stages of development.

There are some remaining TODOs in the code. So, if you want to contribute, feel free to do so. You can also suggest new features. Code refactoring is also welcome. Just open an issue or pull request and I will review it.

Please also submit any bugs you find as an issue with a minimal code example or screenshot. This helps me a lot in improving the code.

Thanks!

To start the web application created with streamlit:

streamlit run frontend/app.py

MIT

You can contact me at contact@shroominic.com. But I prefer to use Twitter or Discord DMs.

If you would like to help this project with a donation, you can click here. Thanks, this helps a lot! ❤️