litellm-api

Auto-generated, fully typed client for the LiteLLM proxy API.

Features

• Fully typed - every operation, parameter and response is generated from the official OpenAPI spec
• Tree-shakable - only the operations you import end up in your bundle
• Runtime agnostic - uses the global fetch, or bring your own implementation

Installation

npm install litellm-api

Quick Start

LiteLLM is self-hosted, so the client requires the baseUrl of your proxy:

import { LiteLLMApi } from "litellm-api";

const client = new LiteLLMApi({
  baseUrl: "http://localhost:4000",
  token: process.env.LITELLM_MASTER_KEY ?? null,
});

const usage = await client.api.billing.checkUsageBillingUsageGet();

Operations are grouped by tag: client.api.<tag>.<operation>(params).

Custom fetch

const client = new LiteLLMApi({
  baseUrl: "http://localhost:4000",
  token: process.env.LITELLM_MASTER_KEY ?? null,
  fetch: myFetchImplementation,
});

Entrypoints

| Import | Contents | | --- | --- | | litellm-api | client class, operations, schemas and types | | litellm-api/components | operations grouped by tag and by path | | litellm-api/schemas | Zod schemas for every operation | | litellm-api/types | TypeScript types for every operation |

License

ISC