Mistral¶

The mistral component connects to the Mistral AI API. Mistral exposes an OpenAI-compatible endpoint, so this provider reuses the same streaming client as openai — no additional dependency.

Configuration¶

components:
  - name: mistral
    type: mistral
    metadata:
      default_model: mistral-large-latest
      api_key: ...     # or set MISTRAL_API_KEY env var
    defaults:
      temperature: 0.7
      max_tokens: 2048
      top_p: 0.9
      system: "You are a helpful assistant."

Metadata reference¶

Key	Required	Description
`default_model`	no	Model used when the request doesn't specify one. Defaults to `mistral-large-latest`.
`api_key`	no	Falls back to `MISTRAL_API_KEY` environment variable.
`base_url`	no	API base URL. Defaults to `https://api.mistral.ai/v1`.

Supported parameters¶

Parameter	Supported
`temperature`	✓
`max_tokens`	✓
`top_p`	✓
`top_k`	— (ignored)
`stop`	✓
`frequency_penalty`	✓
`presence_penalty`	✓
`seed`	✓
`system`	✓
`tools`	✓

Models¶

Any model available through the Mistral API can be specified in default_model or overridden per-request.

Model	Notes
`mistral-large-latest`	Most capable
`mistral-small-latest`	Faster, lower cost
`open-mistral-nemo`	Open-weight, Apache 2.0
`codestral-latest`	Code-focused

reply = client.llm("mistral").chat("Hello", model="codestral-latest")

API key setup¶

Obtain a key from the Mistral console:

export MISTRAL_API_KEY=...

Multiple instances¶

components:
  - name: mistral-large
    type: mistral
    metadata:
      default_model: mistral-large-latest

  - name: codestral
    type: mistral
    metadata:
      default_model: codestral-latest
    defaults:
      temperature: 0.2
      max_tokens: 4096