mistral_client¶

Classes

MistralClient([api_key, base_url, ...])

A minimal Mistral client that inherits from OpenAIClient.

class MistralClient(api_key: str | None = None, base_url: str = 'https://api.mistral.ai/v1', input_type: Literal['text', 'messages'] = 'text', env_api_key_name: str = 'MISTRAL_API_KEY', chat_completion_parser: Callable[[Completion], Any] = None)[source]¶

Bases: OpenAIClient

A minimal Mistral client that inherits from OpenAIClient.

This client is designed to work with Mistral’s API by setting: - The API base URL to https://api.mistral.ai/v1. - The API key is fetched from the environment variable MISTRAL_API_KEY if not provided. - The input format is supported as either "text" or "messages". - The AdalFlow Generator is expected to supply additional model parameters (such as model name, temperature, and max_tokens)

in a single configuration point.

Example usage with the AdalFlow Generator:

import os
from adalflow.core import Generator
from adalflow.components.model_client.mistral_client import MistralClient
from adalflow.utils import setup_env

setup_env()

generator = Generator(
    model_client=MistralClient(),
    model_kwargs={
        "model": "mistral-large-latest",
        "temperature": 0.7,
        "max_tokens": 2000,
    }
)

prompt_kwargs = {"input_str": "Explain the concept of machine learning."}

response = generator(prompt_kwargs)