Mistral AI

Mistral AI Provider

The Mistral provider enables you to use Mistral AI's powerful language models in your Life.js agent.

Configuration

import { Agent } from "life/agent";

const agent = new Agent({
  llm: {
    provider: "mistral",
    apiKey: "sk-...", // Or set MISTRAL_API_KEY environment variable
    model: "mistral-small-latest", // Default model
    temperature: 0.5, // Controls randomness (0.0 to 1.0)
  },
  // ... other config
});

Environment Variables

Set your Mistral API key as an environment variable:

export MISTRAL_API_KEY="sk-..."

Available Models

Model	Description
`mistral-large-latest`	Most capable model for complex tasks
`mistral-large-2411`	Specific version of Mistral Large (Nov 2024)
`mistral-large-2407`	Previous version of Mistral Large (Jul 2024)
`mistral-small-latest`	Cost-efficient model for most tasks (default)
`mistral-small-2501`	Specific version of Mistral Small (Jan 2025)
`mistral-small-2503`	Alternative version of Mistral Small (Mar 2025)
`mistral-medium-latest`	Balanced performance and cost
`mistral-medium-2505`	Specific version of Mistral Medium (May 2025)
`pixtral-large-latest`	Multimodal model with vision capabilities
`pixtral-large-2411`	Specific version of Pixtral Large (Nov 2024)
`codestral-latest`	Specialized for code generation
`codestral-2501`	Specific version of Codestral (Jan 2025)
`codestral-2405`	Previous version of Codestral (May 2024)
`ministral-3b-latest`	World's best edge model (3B parameters)
`ministral-8b-latest`	Powerful edge model (8B parameters)
`open-mistral-7b`	Open-source 7B model
`open-mixtral-8x7b`	Open-source Mixture of Experts model
`open-mixtral-8x22b`	Large open-source MoE model

Configuration Schema

{
  apiKey: string;        // Mistral API key (required)
  model: string;         // Model name (see table above)
  temperature: number;   // 0.0 to 1.0, controls randomness
}

Features

✅ Chat completions
✅ Streaming responses
✅ Function/tool calling
✅ JSON mode (structured output)
✅ Vision models (Pixtral series)
✅ Code generation (Codestral series)
✅ Edge models (Ministral series)

Getting Started

Get your API key: Visit console.mistral.ai to obtain your Mistral API key
Set environment variable: export MISTRAL_API_KEY="your-key-here"
Configure your agent: Use the configuration example above
Start building: Your agent is now ready to use Mistral models!

Example Usage

import { Agent } from "life/agent";

const agent = new Agent({
  llm: {
    provider: "mistral",
    model: "mistral-small-latest",
    temperature: 0.7,
  },
});

// The agent will now use Mistral AI models for all LLM operations

Model Selection Guide

mistral-large: Best for complex reasoning, analysis, and creative tasks
mistral-small: Ideal for most use cases with excellent cost/performance ratio
mistral-medium: Good balance between capability and cost
pixtral-large: Use when you need vision/image understanding capabilities
codestral: Optimized for code generation and programming tasks
ministral: Perfect for edge deployment and resource-constrained environments
open-mistral/mixtral: Great for self-hosted deployments

Mistral AI

On this page