Docker-based Code Interpreter

Starting from version v3.4.2, AstrBot supports a code interpreter to enhance LLM capabilities and enable various automated operations.

TIP

This feature is currently in experimental stage and may have some issues. If you encounter any problems, please submit an issue on GitHub. Join our discussion group: 322154837.

To use this feature, ensure that Docker is installed on your machine. This feature requires a dedicated Docker sandbox environment to execute code and prevent malicious code generated by the LLM from harming your machine.

Running AstrBot with Docker on Linux

If you've deployed AstrBot using Docker, some additional setup is required.

1. When starting the Docker container, mount /var/run/docker.sock inside the container. This allows AstrBot to launch sandbox containers.

sudo docker run -itd -p 6180-6200:6180-6200 -p 11451:11451 -v $PWD/data:/AstrBot/data -v /var/run/docker.sock:/var/run/docker.sock --name astrbot soulter/astrbot:latest

2. Use the /pi absdir <absolute-path> command during chat to set the absolute path of AstrBot's data directory on your host machine.

Example:

Running AstrBot from Source on Linux

If your Docker commands require sudo privileges, you need to start AstrBot with sudo, otherwise the code interpreter won't be able to invoke Docker due to insufficient permissions.

sudo -E python3 main.py

Usage

This feature uses the soulter/astrbot-code-interpreter-sandbox image. You can view detailed information about the image on Docker Hub.

The image includes commonly used Python libraries:

• Pillow
• requests
• numpy
• matplotlib
• scipy
• scikit-learn
• beautifulsoup4
• pandas
• opencv-python
• python-docx
• python-pptx
• pymupdf
• mplfonts

Tasks that can be accomplished include:

• Image editing
• Web scraping
• Data analysis and simple machine learning
• Document processing, such as reading and writing Word, PPT, PDF files
• Mathematical calculations, such as plotting graphs and solving equations

Since Docker Hub is inaccessible from mainland China, if you're in that region, use /pi mirror to view/set the mirror source. For example, as of this writing, you can use cjie.eu.org as the mirror source by setting /pi mirror cjie.eu.org.

When the code interpreter is triggered for the first time, AstrBot will automatically pull the image, which may take some time. Please be patient.

The image may be updated periodically to provide more features, so check for updates regularly. If you need to update the image, use the /pi repull command to re-pull it.

TIP

If the feature doesn't start properly initially, after successful startup, execute /tool on python_interpreter to enable this feature. You can use /tool ls to view all tools and their enabled status.

Image and File Input

In addition to recognizing and processing images and text tasks, the code interpreter can also recognize files you send and send files back.

After v3.4.34, use the /pi file command to start uploading files. After uploading, you can use /pi list to view your uploaded files and /pi clean to clear your uploaded files.

Uploaded files will be used as input for the code interpreter.

For example, if you want to add rounded corners to an image, you can upload the image using /pi file, then ask: Please run code to add rounded corners to this image.

Demo