TokenLab JavaScript SDK

Lightweight JavaScript client for TokenLab discovery, OpenAI-compatible APIs, and native endpoint families.

Install from npm:

npm install @tokenlabai/sdk

Usage

import { createTokenLabClient } from '@tokenlabai/sdk';

const tokenlab = createTokenLabClient({
  apiKey: process.env.TOKENLAB_API_KEY,
});

const models = await tokenlab.listModels({ category: 'chat' });
const response = await tokenlab.createResponse({
  model: 'gpt-5.5',
  input: 'Hello from TokenLab',
});

Native Endpoints

await tokenlab.createAnthropicMessage({
  model: 'claude-sonnet-5',
  max_tokens: 512,
  messages: [{ role: 'user', content: 'Hello' }],
});

await tokenlab.createGeminiContent('gemini-3.5-flash', {
  contents: [{ role: 'user', parts: [{ text: 'Hello' }] }],
});

Discovery Helpers

await tokenlab.getModelsJson();
await tokenlab.getPricingJson();
await tokenlab.getDiscovery();

System One decisions

Discover await tokenlab.listModels({ category: 'decision' }) and inspect the selected model with getModel() first. Decision models declare the systemone public operation, not a chat format.

const decision = await tokenlab.evaluateDecisions({
  model: 'jev-1.13',
  state: { ticket: 'Please refund my duplicate charge.' },
  questions: { refund: { type: 'noul', instructions: 'Is a refund requested?' } },
}, { signal: AbortSignal.timeout(60_000) });
console.log(decision.answers, decision.usage);

This synchronous /v1/systemone call preserves typed answers, probabilities, optional confidence, and usage. It is not Chat, streaming, or Batch. A decision does not authorize a refund or another side effect. Errors are returned without automatic resubmission. Available in SDK 0.1.2+.

Base URLs

• API: https://api.tokenlab.sh
• OpenAI-compatible SDK base URL: https://api.tokenlab.sh/v1
• OpenAPI: https://tokenlab.sh/docs/openapi.json