Foundry Code — Code Interpreter (Responses API)

You have access to Python code execution via the Foundry Responses API code_interpreter tool. Use this for data analysis, chart generation, complex math, and file processing. No Foundry hosted agent needed — uses direct Responses API.

Why use this instead of local bash/python

The local sandbox has restricted packages. Foundry code_interpreter runs server-side with pre-installed data science libraries (pandas, numpy, matplotlib, scipy, etc.) and can generate visualizations.

Endpoint

All requests: http://localhost:8443 with ?api-version=2025-11-15-preview. Auth is automatic.

Operations

Run Python code

curl -s -X POST 'http://localhost:8443/openai/responses?api-version=2025-11-15-preview' \
  -H 'Content-Type: application/json' \
  -d '{"model":"gpt-4.1","input":"Calculate the first 20 Fibonacci numbers and show them.","tools":[{"type":"code_interpreter","container":{"type":"auto"}}],"store":false}'

The model will write and execute Python code, returning both the code and its output.

Data analysis

curl -s -X POST 'http://localhost:8443/openai/responses?api-version=2025-11-15-preview' \
  -H 'Content-Type: application/json' \
  -d '{"model":"gpt-4.1","input":"Generate 100 random data points, compute mean/median/std, and tell me if normally distributed.","tools":[{"type":"code_interpreter","container":{"type":"auto"}}],"store":false}'

When to use

• User asks for data analysis: "analyze this", "show statistics"
• Charts or visualizations: "plot the trend", "create a chart"
• Complex math: "solve this equation", "calculate correlation"
• Anything requiring Python runtime

When NOT to use

• For simple shell commands (use local bash)
• For web searches (use foundry-web-search)
• For document Q&A (use foundry-knowledge)
• For remembering preferences (use foundry-memory)