NPM AI Client for OpenRouter

A lightweight, easy-to-use Node.js client for the OpenRouter API that simplifies sending chat completion requests to various AI models and handling structured JSON responses.

Installation

```bash npm install npmai-client ```

Basic Usage

```javascript const NpmAiClient = require('npmai-client'); const ai = new NpmAiClient(process.env.OPENROUTER_API_KEY);

// Define chat messages const messages = [ { role: "system", content: "You are a helpful assistant. Output your response in JSON format: { "answer": "string" }" }, { role: "user", content: "What is JavaScript?" } ];

// Send a request to a specific model ai.ask("google/gemini-pro", messages) .then(response => { // Access the parsed JSON response const answer = response.get("answer"); console.log("Answer:", answer); }) .catch(error => { console.error("Error:", error); }); ```

Features

• Simple API for sending chat completion requests
• Automatic JSON parsing of AI responses
• Helper methods for accessing structured data
• Support for all OpenRouter models
• Fallback model routing
• Comprehensive error handling
• Structured Outputs - Enforce JSON Schema validation on model responses

Structured Outputs

For models that support it (like OpenAI's GPT-4o), you can enforce a specific JSON schema:

```javascript // Define your JSON schema const schema = { name: "weather_info", strict: true, schema: { type: "object", properties: { location: { type: "string", description: "City name" }, temperature: { type: "number", description: "Temperature in Celsius" } }, required: ["location", "temperature"] } };

// Use the askForJson method ai.askForJson("openai/gpt-4o", messages, schema) .then(response => { const data = response.getRawJson(); console.log(It's ${data.temperature}°C in ${data.location}); }); ```

API Reference

NpmAiClient

```javascript // Initialize with API key const ai = new NpmAiClient(apiKey, options);

// Send a request const response = await ai.ask(modelName, messages, options); // or const response = await ai.chat(modelName, messages, options); // or with structured output const response = await ai.askForJson(modelName, messages, jsonSchema, options); ```

ApiResponse

```javascript // Get a value by key const value = response.get("key");

// Get an array by key const list = response.getList("items");

// Get the full parsed JSON const json = response.getRawJson();

// Get the raw API response const rawResponse = response.getRawResponse();

// Check if JSON was parsed successfully const isParsed = response.isJsonParsed();

// Get the model used const model = response.getModel();

// Get token usage const usage = response.getUsage(); ```

For full documentation, see doc.md.

License

MIT