Getting Started

Follow these steps to set up the project and run your first example.

Prerequisites

  • Node.js (v22.x or newer recommended)

  • An active JetBrains AI Platform account and a JSON Web Token (JWT)

Installation

For Your Own Project (Recommended)

Merci SDK is a zero-dependency library. To use it in your own project, simply download the desired file from the /lib directory on GitHub and import it.

  • merci.2.14.0.mjs: The full library file, great for development.
  • merci.2.14.0.d.ts: TypeScript definitions for autocompletion.

To Run the Examples

If you want to run the tutorial lessons locally, clone the entire repository.

git clone https://github.com/mobeetle/merci-sdk.git
cd merci-sdk

Setup

1. Configure Your Authentication Token

Create a new folder named secret, and inside it, a new file named token.mjs. Add the following content, replacing `"YOUR_JWT_HERE"` with your actual token.

// File: secret/token.mjs
// This file securely exports your authentication token.

export const token = "YOUR_JWT_HERE";
⚠️ Security Warning: The `secret` directory is already included in the project's `.gitignore` file to prevent you from accidentally committing your token. Never share this file or commit it to a public repository.

⚡ Quick Start: Your First API Call

The examples/lesson_1_basic_usage.mjs file is the best place to start. It demonstrates the simplest possible interaction: sending a prompt to the model and streaming the response.

Run the example from your terminal:

node examples/lesson_1_basic_usage.mjs

You should see output like this:

--- Merci SDK Tutorial: Lesson 1 - Basic Usage (Model: google-chat-gemini-flash-2.5) ---
[STEP 1] Initializing MerciClient...
[STEP 2] Preparing prompt and input data...
[STEP 3] Configuring the chat session...
[STEP 4] Creating the message payload...
[STEP 5] Sending request and processing stream...
🤖 Assistant > Stardust Brew: Your cosmic cup of inspiration.

[INFO] Stream finished. Response fully received.