mistralai/client-python

Mistral Python Client – Quick Overview

What it is: An official Python SDK for calling Mistral AI’s cloud services – chat‑completion, embeddings, audio, file handling, agents, and a host of management APIs. It lets you talk to Mistral models from your own Python code, both synchronously and with asyncio.

Key features

• Chat & agents – simple mistral.chat.complete(...) and mistral.agents.complete(...) calls, with optional streaming.
• Embeddings – mistral.embeddings.create(...) for vectorising text.
• File uploads – upload workspace files via mistral.files.upload(...).
• Audio – speech synthesis and transcription endpoints.
• Batch jobs, libraries, connectors, conversations – full‑featured management of Mistral resources (e.g., document libraries for RAG).
• Async support – every operation has an *_async counterpart.
• Server‑sent events – streaming responses via SSE.
• Built‑in pagination, retries, and error handling.
• Cloud‑provider wrappers – ready‑made clients for Azure AI and Google Cloud (Vertex AI) that automatically handle auth and endpoint construction.

How to get started

1. Install – pip install mistralai (or uv add mistralai).
2. Set your API key – export MISTRAL_API_KEY or point MISTRAL_SA_TOKEN_PATH to a service‑account token file.
3. Write code – import Mistral and call the desired method, e.g.:

   from mistralai.client import Mistral
   import os
   
   with Mistral(api_key=os.getenv("MISTRAL_API_KEY")) as client:
       resp = client.chat.complete(
           model="mistral-large-latest",
           messages=[{"role": "user", "content": "Hello!"}],
           stream=False,
       )
       print(resp)
   

4. Run examples – the repo ships an examples/ folder you can execute with uv run.

Why it matters: The SDK abstracts away HTTP details, token rotation, pagination, and streaming, letting developers focus on building applications (chatbots, RAG pipelines, voice assistants, etc.) that leverage Mistral’s large language models.