Mistral AI Python Client

2.3.0 · active · verified Sun Apr 05

The `mistralai` library is the official Python Client SDK for interacting with the Mistral AI API, providing programmatic access to Mistral's large language models and other AI services. It is currently at version 2.3.0 and actively maintained with regular updates and feature enhancements.

Warnings

breaking The import path for the `Mistral` client class changed significantly from v1.x to v2.x. Old code importing `from mistralai import Mistral` will fail.
Fix: Update your imports to `from mistralai.client import Mistral`.
breaking The structure of response objects, specifically `response.choices[]`, changed in v2.x for methods like `mistral.chat.complete()`, `mistral.fim.complete()`, and `mistral.agents.complete()`.
Fix: Refer to the official migration guide (e.g., `MIGRATION.md` on GitHub) to adapt your code to the new response payload structure. For chat completions, content is now often accessed via `response.choices[0].message.content`.
gotcha API Key (`MISTRAL_API_KEY`) is crucial for authentication. Not setting it as an environment variable or passing it directly to the client will lead to authentication errors.
Fix: Set `export MISTRAL_API_KEY='your_api_key_here'` in your environment or explicitly pass `api_key='your_api_key_here'` during client initialization.
gotcha Incorrect data serialization when sending requests to the API can lead to 'Data Serialization Error'.
Fix: Ensure that the data you send in API requests (e.g., messages, tool definitions) strictly adheres to the format specified in the Mistral AI API documentation.
gotcha Common API errors include 401 (Unauthorized), 413 (Payload Too Large/Context Window Overflow), and 429 (Rate Limit Exceeded).
Fix: Implement robust error handling, including retries with exponential backoff for transient errors (429). For 413, review prompt length and model context window limits. Ensure API key is valid for 401 errors.
deprecated Several older models, such as `open-mistral-7b`, `mistral-tiny`, `mistral-medium-2312`, and certain dated model aliases, have been deprecated.
Fix: Update to newer, recommended models for continued support and improved performance. Check the latest model availability via the API or official documentation.

Install

pip install mistralai Install with pip

Imports

Mistral
```
from mistralai.client import Mistral
```
Starting with v2.x, the main client class was moved into the `client` submodule. Direct import from `mistralai` will result in an `AttributeError`.

Quickstart

This quickstart initializes the Mistral client and performs a simple chat completion request. It assumes the `MISTRAL_API_KEY` environment variable is set for authentication. The response content is then printed.

import os
from mistralai.client import Mistral

# Ensure MISTRAL_API_KEY environment variable is set
api_key = os.environ.get('MISTRAL_API_KEY', '')

if not api_key:
    print("Error: MISTRAL_API_KEY environment variable not set.")
    print("Please set it using: export MISTRAL_API_KEY='your_api_key_here'")
else:
    try:
        client = Mistral(api_key=api_key)
        chat_response = client.chat.complete(
            model="mistral-small-latest",
            messages=[
                {"role": "user", "content": "What is the capital of France?"}
            ]
        )
        print(chat_response.choices[0].message.content)
    except Exception as e:
        print(f"An error occurred: {e}")

view raw JSON →