Mistral

Use Mistral models as the LLM provider in a CompositeVoice pipeline.

Use MistralLLM when you need strong multilingual support, especially for European languages, or want a cost-effective alternative to larger models.

Prerequisites

A Mistral API key or a CompositeVoice proxy server
No additional dependencies required. MistralLLM uses native fetch internally.

Basic setup

import { CompositeVoice, MistralLLM, NativeSTT, NativeTTS } from '@lukeocodes/composite-voice';

const agent = new CompositeVoice({
  providers: [
    new NativeSTT({ language: 'en-US' }),
    new MistralLLM({
      proxyUrl: '/api/proxy/mistral',
      model: 'mistral-small-latest',
      systemPrompt: 'You are a concise voice assistant. Keep answers under two sentences.',
    }),
    new NativeTTS(),
  ],
});

await agent.initialize();
await agent.startListening();

Configuration options

Option	Type	Default	Description
`model`	`string`	`'mistral-small-latest'`	Model identifier. See model variants below.
`systemPrompt`	`string`	—	System-level instructions for the assistant.
`temperature`	`number`	—	Randomness (0 = deterministic, 2 = creative).
`maxTokens`	`number`	—	Maximum tokens per response.
`topP`	`number`	—	Nucleus sampling threshold (0—1).
`stream`	`boolean`	`true`	Stream tokens incrementally.
`proxyUrl`	`string`	—	CompositeVoice proxy endpoint. Recommended for browsers.
`mistralApiKey`	`string`	—	Mistral API key. Convenience alias for `apiKey`.
`apiKey`	`string`	—	Direct API key. `mistralApiKey` takes precedence if both are set.

Model variants

Model	Speed	Notes
`mistral-small-latest`	Fast	Default. Good speed-to-quality ratio for voice.
`mistral-medium-latest`	Moderate	Balanced capability.
`mistral-large-latest`	Slower	Most capable Mistral model.

Complete example

import {
  CompositeVoice,
  MicrophoneInput,
  MistralLLM,
  DeepgramSTT,
  DeepgramTTS,
  BrowserAudioOutput,
} from '@lukeocodes/composite-voice';

const agent = new CompositeVoice({
  providers: [
    new MicrophoneInput(),
    new DeepgramSTT({
      proxyUrl: '/api/proxy/deepgram',
      language: 'en',
      options: { model: 'nova-3', smartFormat: true },
    }),
    new MistralLLM({
      proxyUrl: '/api/proxy/mistral',
      model: 'mistral-small-latest',
      temperature: 0.7,
      maxTokens: 256,
      systemPrompt: 'Tu es un assistant vocal amical. Reponds brievement.',
    }),
    new DeepgramTTS({
      proxyUrl: '/api/proxy/deepgram',
      voice: 'aura-2-thalia-en',
    }),
    new BrowserAudioOutput(),
  ],
  conversationHistory: { enabled: true, maxTurns: 10 },
});

await agent.initialize();
await agent.startListening();

Tips

Mistral excels at multilingual tasks. French and other European languages produce especially good results.
mistral-small-latest is best for voice. It provides the fastest responses while maintaining quality for conversational use cases.
MistralLLM uses native fetch — no @mistralai/mistralai or openai package needed.
Model names use the -latest suffix. This always points to the most recent version of that model tier.

Providers reference — all LLM providers at a glance
API reference — full class documentation
OpenAI Compatible guide — connect custom OpenAI-compatible endpoints