/codeinterpreter-codebox

Easy to deploy.A cloud service for python code interpreter sandbox for Code-Interpreter.

Primary LanguagePythonMIT LicenseMIT

Codeinterpreter-Codebox

Welcome to Codeinterpreter-Codebox, a cutting-edge cloud service tailored for executing Code Interpreter code.

📜 Table of Contents

🎯 Key Features

  1. 🐍 Independent Jupyter Sandbox Environment
    A completely isolated Jupyter sandbox environment that supports state-preserving code execution. Supports file uploads and downloads.
  2. 🐳 Dockerized One-click Deployment
  3. 🌐 Completely Free and Open Source
  4. 🛡️ Information Security
    Supports completely private deployments without the need to upload files to external servers.
  5. 🚀 More Flexible
    Supports the opening of more ports to achieve more custom functionality requirements, such as: connecting to databases, connecting to the internet, connecting to other servers.

💡 What Can You Do?

  1. Execute Python code.
  2. Supports reading of uploaded files (Upload and Analyze Files).
if name=='main':
session=CodeinterpreterSession()
try:
session.upload_files(['./../data/test_data.csv'])
session.chat('How many columns are in the uploaded file?')
finally:
session.close()
  1. Supports downloading files from the sandbox.
  2. Supports dynamic scaling for commercial deployments.
  3. Supports custom feature modifications, open ports, network connections, etc., supports web crawlers.

📋 Related Projects & Their Differences

  1. Code Interpreter api
  2. open-interpreter
  3. E2B
Feature/Project Codeinterpreter-Codebox★ Code Interpreter api open-interpreter E2B
Private Commercial Deployment ✅ ❌ ❌ ❌
Remote Invocation ✅ ✅ ❌ ✅
Fully Open Source ✅ ❌ ✅ ❌
Free of Charge ✅ ❌ ✅ ❌
Customization ✅ ❌ ❌ ❌
Information Security ✅ ❌ ✅ ❌
Multiple Programming Languages ❌ ❌ ✅ ✅
Local Execution ✅ ✅ ✅ ❌
Direct Invocation w/o Deployment ❌ ✅ ✅ ✅

★ Indicates our project

💻 Sample Codes

  1. Server-Client Interaction: jupyter server communicate showcases the details of interaction between the server and the Jupyter container.
  2. Remote Execution: jupyter call demonstrates how to invoke a web service via an HTTP request, execute code, and retrieve results.
test_code="""
import docker
print(docker.version)

"""
execute(test_code)

#output
Execute Result= {"output_type":"text/plain","content":"6.1.3\n","files":null}
  1. Client Session Demonstration: client session invocation shows how to allow LLM to invoke and execute code in a project through session and HTTP request.
session=CodeinterpreterSession()
try:
session.upload_files(['./../data/test_data.csv'])
session.chat('How many columns are in the uploaded file?')
finally:
session.close()

📦 Deployment

  1. Install Docker
    Linux: Install Docker by terminal
    Mac os: Install Docker desktop for mac
    Windows: Install Docker desktop for windows
  2. Modify Docker config file
    Head over to docker compose and substitute 'CODEBOX_ROOT_PATH' and 'YOUR_MNT_PATH:/codebox' with your file path to mount 'YOUR_MNT_PATH'.
  3. Launch server
  • Build custom jupyter image
cd docker
docker build -t scipy-notebook:custom -f Dockerfile .
  • Build web server image
cd app
docker-compose -f docker_dev.yml build
  • Launch server
docker-compose -f docker_dev.yml up

📧 Contact

WeChat: zjajzzj1996
Email: zhangzhehian@gmail.com