@tokenlens/fetch

@tokenlens/fetch

Tiny typed client for https://models.dev/api.json with friendly error codes.

Install

• npm: npm i @tokenlens/fetch
• pnpm: pnpm add @tokenlens/fetch
• yarn: yarn add @tokenlens/fetch

API

• fetchModels({ provider?, model?, fetch?, signal?, baseUrl? })
• getModelMeta(providers, ...)

Usage

import { fetchModels, getModelMeta, FetchModelsError } from '@tokenlens/fetch';

const catalog = await fetchModels();
const openai = await fetchModels({ provider: 'openai' });
const gpt4o = await fetchModels({ provider: 'openai', model: 'gpt-4o' });

// Pick raw metadata directly from the fetched catalog
const prov = getModelMeta({ providers: catalog, provider: 'openai' });
const model = getModelMeta({ providers: catalog, id: 'openai:gpt-4o' });

try {
  await fetchModels();
} catch (err) {
  if (err instanceof FetchModelsError) {
    // 'UNAVAILABLE' | 'NETWORK' | 'HTTP' | 'PARSE'
    console.error(err.code, err.status, err.message);
  }
}

License MIT