Introduction

A typed, provider-independent interface for LLM applications written in TypeScript.

any-llm-ts is a thin, framework-independent layer over official provider SDKs. It gives your application one interface for chat completions, streaming, tools, embeddings, model discovery, responses, images, moderation, and audio—without requiring a hosted proxy.

The library is an independent TypeScript port inspired by Mozilla.ai's Python any-llm. It preserves the original project's central idea while using TypeScript-native primitives such as promises, AsyncIterable, discriminated unions, and overloads.

Choose how you call it

Stateless helpers

Ideal for scripts and isolated calls. Specify a provider or use the provider:model syntax.

Reusable client

Create AnyLLM once and reuse its underlying SDK client across requests.

Provider guide

Inspect supported providers, environment variables, and conservative capability metadata.

API reference

Explore the exported functions, classes, types, and normalized error hierarchy.

Design goals

One application-facing contract. Provider-specific translation stays behind adapters.
Type-safe streaming. stream: true changes the inferred result to an AsyncIterable.
No proxy requirement. Requests go directly from your Node.js application to the provider.
Escape hatches. Use providerOptions, clientOptions, custom endpoints, or custom adapters when the common surface is not enough.
Conservative metadata. Capability flags describe configured support without claiming that every provider and model combination is continuously integration-tested.

Independent project

This package is not an official Mozilla.ai project. See the repository's NOTICE and LICENSE files for attribution and Apache-2.0 licensing details.

Runtime support

The library supports Node.js 20 and newer and ships ESM, CommonJS, and TypeScript declarations. It does not depend on Next.js, React, or Fumadocs at runtime; those packages are used only by this documentation site.