waken-mistral

A Mistral Target adapter for Waken — "nginx for AI agents." Routes waken Events to Mistral AI's chat completion API and returns the result as a waken Response.

Install

pip install waken-mistral

Usage

from waken import Runtime
from waken_mistral import MistralAdapter

runtime = Runtime()
runtime.target("mistral", MistralAdapter())
runtime.run()

MistralAdapter reads MISTRAL_API_KEY from the environment — that's the mistralai SDK's own convention, nothing waken-specific. Pass api_key=... (or any other Mistral client keyword, e.g. server_url=...) to override:

runtime.target("mistral", MistralAdapter(model="mistral-small-latest", api_key="..."))

Sessions

Unlike Claude or OpenAI, Mistral's chat completion API has no server-side resumable session id — it's a stateless endpoint that expects the full conversation history resent on every call. So MistralAdapter maps a waken session_id to a plain, growing list of {"role": ..., "content": ...} messages, appends the new user prompt and the assistant's reply to that list on every turn, and resends the whole thing each time. History is kept client-side, in this process's memory only — there's no persistence layer, and it's lost on restart. That's the same tradeoff waken itself makes for sessions in general: the runtime only stores the session id mapping, never conversation content, so any content-level durability is the adapter's responsibility, not the runtime's. An Event with no session_id sends a fresh one-message list each call and is never stored.

Development

git clone https://github.com/WakenHQ/waken-mistral
cd waken-mistral
pip install -e ".[dev]"
pytest

Tests mock mistralai.client.Mistral entirely — no API key or network access needed to run the suite.

License

MIT