Code Interpreter

Enable LLMs to execute Python code in a secure sandbox.

Code Interpreter is a built-in tool that allows LLMs to execute Python code in a secure sandboxed environment. When enabled, the model can write and run code for computations, data analysis, plotting, file processing, and more.

Pricing

Code Interpreter is currently free to use. Each execution runs in an isolated sandbox with no additional cost beyond token usage.

How it works

Include { type: 'code_interpreter' } in your tools array to give the model the ability to execute Python code. The model will autonomously decide when to write and run code to answer the user's question.

import OpenAI from 'openai';

const client = new OpenAI({
  baseURL: 'https://api.aivene.com/v1',
  apiKey: process.env.AIVENE_API_KEY
});

const response = await client.chat.completions.create({
  model: 'gpt-4o',
  messages: [
    { role: 'user', content: 'Calculate the first 20 Fibonacci numbers' }
  ],
  tools: [
    { type: 'code_interpreter' }
  ]
});

console.log(response.choices[0].message.content);

Configuration options

typestringrequired

Must be "code_interpreter".

timeout_msnumberoptional

Maximum execution time for the code in milliseconds. The execution will be terminated if it exceeds this limit.

sandbox_timeout_msnumberoptional

Maximum lifetime of the sandbox environment in milliseconds.

Example with configuration

const response = await client.chat.completions.create({
  model: 'gpt-4o',
  messages: [
    { role: 'user', content: 'Generate a sine wave plot' }
  ],
  tools: [
    {
      type: 'code_interpreter',
      timeout_ms: 30000,
      sandbox_timeout_ms: 60000
    }
  ]
});

Sandbox environment

Each code execution runs in a fresh, isolated sandbox:

Ephemeral — No state is shared between calls
Secure — Isolated from other executions and the host system
Python environment — Common libraries like NumPy, Pandas, Matplotlib are available
Output capture — stdout, stderr, and results (including charts/images) are returned

Use cases

Mathematical computations — Complex calculations, statistics, simulations
Data analysis — Process and analyze datasets
Visualization — Generate charts, plots, and graphs
File processing — Parse, transform, or generate files
Algorithm implementation — Run custom algorithms on demand

Response format

The code execution returns a JSON object with:

Field	Description
`stdout`	Standard output from the code
`stderr`	Standard error output
`results`	Array of results including charts/images
`error`	Error object if execution failed

Example: Data visualization

const response = await client.chat.completions.create({
  model: 'gpt-4o',
  messages: [
    { 
      role: 'user', 
      content: 'Create a bar chart showing the population of the top 5 most populous countries' 
    }
  ],
  tools: [
    { type: 'code_interpreter' }
  ]
});

// The model will generate and execute Python code using matplotlib
// The chart image will be included in the response

Combining with other tools

Code Interpreter works well alongside Web Search for data-driven tasks:

const response = await client.chat.completions.create({
  model: 'gpt-4o',
  messages: [
    { role: 'user', content: 'Find the current stock prices of FAANG companies and create a comparison chart' }
  ],
  tools: [
    { type: 'web_search' },
    { type: 'code_interpreter' }
  ]
});

The model will search for current stock prices, then use Code Interpreter to create a visualization.