/openai-o1-code-interpreter

Code interpreter support for o1

Primary LanguagePython

O1-Claude Code Interpreter

A powerful visualization tool that combines OpenAI's O1 for planning and Claude 3.5 Sonnet for code generation. This tool can create various data visualizations based on natural language prompts.

How It Works

  1. O1 Model: Creates a detailed execution plan from your prompt
  2. Claude 3.5 Sonnet: Generates and debugs Python code based on O1's plan
  3. E2B: Executes the code in a secure sandbox environment
  4. Output: Generates visualizations and saves them automatically

All code execution runs on e2b for safety. While local execution is possible by modifying code_interpret, it's not recommended.

Setup

  1. Install required dependencies: 
    pip install -r requirements.txt

Set up your API keys in .env:

  • OPENAI_API_KEY=your_openai_key
  • ANTHROPIC_API_KEY=your_anthropic_key
  • E2B_API_KEY =your_e2b_key (Use .env.sample as a template)

Usage

Run with a custom prompt:

python main.py "Create a visualization of market shares of self-driving vehicle manufacturers in 2022 vs 2023 vs 2024"

VSCode

Click the "play" button in main.py to run with the default prompt.

Try these example prompts to see different visualization capabilities:

  1. Historical Data: Create a visualization of the growth of the Roman empire population and land under its control. Mark every important historical event along the way.

  2. Market Analysis: Create a visualization of market shares of self-driving vehicle manufacturers in 2022 vs 2023 vs 2024.

  3. Economic Trends: Create a report on the evolution of labor productivity in the United States, and what major technological shifts led to it.

Output

  • Generated visualizations are saved in the output/charts directory
  • Each prompt creates its own cache file for faster subsequent runs
  • Markdown reports with embedded charts are saved in the output directory

Features

📊 Automated chart generation and saving 🔄 Smart caching system (prompt-specific) 📝 Markdown report generation 🎨 Support for various visualization types 🛠 Automatic error handling and code fixes

Technical Details

  • Uses OpenAI's O1 model for execution planning
  • Uses Claude 3.5 Sonnet for code generation and debugging
  • Supports matplotlib, pandas, and other Python visualization libraries
  • Secure code execution via e2b sandbox