Mistral AI¶

Use the mistral provider to call Mistral with an API key. It supports chat and embeddings. Chat uses Chat Completions.

Get started¶

Install Republic using the quickstart, set REPUBLIC_MISTRAL_API_KEY, and replace MODEL_ID with a model available to your account:

import asyncio

import republic


async def main():
    model = republic.get_model("mistral:MODEL_ID")
    response = await model.chat("Say hello in one sentence.")
    print(response.text)


asyncio.run(main())

Authentication¶

Republic reads REPUBLIC_MISTRAL_API_KEY when the provider is created and sends the key in the Authorization: Bearer header. You can pass a key loaded by your application as api_key=; see API-key authentication.

Model support¶

Chat requests send max_tokens, map seed to random_seed, and omit stream_options because Mistral reports usage in the last chunk. top_k raises errors.UnsupportedFeatureError.

With reasoning_effort, reasoning models return thinking chunks, which Republic exposes as response.reasoning. Earlier thinking is sent back as a thinking chunk; keep response.message in the conversation.

For embedding models, use get_embedding_model(); dimensions= is sent as output_dimension.

See configuration for endpoint overrides and request options.