JavaScript / TypeScript SDK

@responsible-ai-labs/rail-scorev2.3.0 on npm

Official JavaScript/TypeScript client for the RAIL Score API. Full TypeScript definitions, LLM provider wrappers, policy engine, session tracking, and middleware.

Quick Start — 30 seconds

npm install @responsible-ai-labs/rail-score

import { RailScore, getScoreLabel } from '@responsible-ai-labs/rail-score';

const client = new RailScore({ apiKey: "YOUR_RAIL_API_KEY" });
const result = await client.eval({ content: "Your AI-generated text here...", mode: "basic" });

console.log(result.rail_score.score);   // 8.4
console.log(getScoreLabel(8.4));        // "Excellent"

Installation

npm / yarn

npm install @responsible-ai-labs/rail-score
yarn add @responsible-ai-labs/rail-score

With LLM provider wrappers (optional peer deps)

npm install openai                    # For RAILOpenAI wrapper
npm install @anthropic-ai/sdk         # For RAILAnthropic wrapper
npm install @google/generative-ai     # For RAILGemini wrapper
npm install langfuse                  # For RAILLangfuse observability

Requires: Node.js ≥ 16.0.0

Modules: CommonJS + ESM (dual build)

Types: Full TypeScript definitions included

Client Initialization

import { RailScore } from '@responsible-ai-labs/rail-score';

const client = new RailScore({
  apiKey: 'YOUR_RAIL_API_KEY',
  baseUrl: 'https://api.responsibleailabs.ai',  // Optional (default)
  timeout: 30000,                                // Optional (ms)
  cache: true,    // Cache eval results for 5 min (default true)
  retry: true,    // Retry on 429/500/502/503 (default true)
});

// Health check — no auth required
const health = await client.health();
console.log(health.status);   // "healthy"

In this SDK

Evaluation

client.eval() — basic, deep, selective dimensions, custom weights

Safe Regeneration

client.safeRegenerate() — iterate until quality thresholds are met

Compliance

client.complianceCheck() — GDPR, HIPAA, EU AI Act, and more

Sessions & Policy

RAILSession, PolicyEngine, Middleware

Providers & Reference

LLM wrappers, Langfuse, error types, utilities, TypeScript types

Start with Evaluation →View on npm