@m3chat SDK

@m3chat/sdk is the official TypeScript/JavaScript client SDK for M3 Chat — a powerful AI chat platform powered by multiple models including LLaMA, Mistral, Qwen, and more. This SDK provides a simple and flexible interface to interact with M3 Chat's API for sending chat messages, streaming responses, batch requests, and model management.

Table of Contents

• Introduction
• Table of Contents
• Features
• Installation
• Quick Start
• Usage
  • Creating a Client
  • Getting a Response
  • Batch Requests
• Available Models
• Error Handling
• Development
• Contribution
• License
• Contact
• Example Project

Features

• Send chat requests to M3 Chat API
• Support for streaming and non-streaming responses
• Batch request support for multiple messages
• Validation of available models
• Easy to integrate in Node.js and browser environments
• Written in TypeScript with types included

Installation

npm install @m3chat/sdk
# or
yarn add @m3chat/sdk

Quick Start

import { M3ChatClient } from "@m3chat/sdk";

async function main() {
  const client = new M3ChatClient({ stream: false });

  try {
    const response = await client.getResponse({
      model: "mistral",
      content: "Hello, how are you?",
    });
    console.log("Response:", response);
  } catch (error) {
    console.error("Error fetching response:", error);
  }
}

main();

Usage

Creating a Client

import { M3ChatClient } from "@m3chat/sdk";

const client = new M3ChatClient({
  stream: true, // Set to true to enable streaming response
});

Getting a Response

const response = await client.getResponse({
  model: "llama3:8b",
  content: "Explain quantum computing in simple terms.",
});

console.log(response);

• If stream option is enabled, the response is streamed and printed chunk-by-chunk in real time.
• If disabled, the full response is returned as a string.

Batch Requests

Send multiple messages in sequence:

const messages = [
  "Hello, who won the world cup in 2022?",
  "What is the capital of France?",
  "Tell me a joke.",
];

const results = await client.batchRequests(messages, {
  model: "gemma",
});

results.forEach((res, idx) => {
  console.log(`Response ${idx + 1}: ${res}`);
});

Available Models

The SDK internally validates models against this list:

[
  "llama3:8b",
  "llama2-uncensored",
  "gemma3",
  "gemma",
  "phi3:mini",
  "mistral",
  "gemma:2b",
  "gemma:7b",
  "qwen:7b",
  "qwen2.5-coder",
  "qwen3",
  "deepseek-coder:6.7b",
  "deepseek-v2:16b",
  "dolphin-mistral:7b",
  "dolphin3",
  "starcoder2:7b",
  "magistral",
  "devstral",
];

If an unsupported model is passed, the SDK throws an error.

Error Handling

• Throws if the API returns non-OK status.
• Throws if an invalid model is provided.
• Throws if streaming response body is missing.

Development

Clone the repository, install dependencies, and build:

git clone https://github.com/m3-chat/sdk.git
cd sdk
npm install
npm run build

Contribution

Contributions are welcome! Please open issues and pull requests on the GitHub repository.

License

This project is licensed under the Apache License 2.0 - see the LICENSE file for details.

Contact

For questions or support, join M3 Chat discussions or open an issue on GitHub.

Example Project

import { M3ChatClient } from "@m3chat/sdk";

(async () => {
  const client = new M3ChatClient({ stream: false });

  try {
    const response = await client.getResponse({
      model: "gemma",
      content: "Write a poem about the sea.",
    });
    console.log("Chat response:", response);
  } catch (error) {
    console.error(error);
  }
})();

Thank you for using M3 Chat SDK — build awesome AI chat experiences with ease!