@llm4ts/js

The fastest way to try llm4ts: a Promise-based client for plain JavaScript and TypeScript consumers — no Effect knowledge required.

Install

npm install @llm4ts/js

Requires Node.js >= 22.

Usage

Works out of the box with the built-in mock provider — no credentials needed:

import { createClient } from "@llm4ts/js"

const client = createClient({ provider: "mock", model: "mock" })
const response = await client.complete("Say hello")
console.log(response.content)

Swap in a real provider by name (openai, anthropic, gemini, lm-studio, ollama) — API keys are read from the standard environment variables (OPENAI_API_KEY, ANTHROPIC_API_KEY, GEMINI_API_KEY):

const client = createClient({
  provider: "anthropic",
  model: "claude-sonnet-4-5"
})
const answer = await client.complete("Summarize the Effect library in one line")

The client also exposes health() for checking provider availability and credentials.

For full flows — plans, review loops, Git integration, CLI coding agents — use @llm4ts/runner.

License

MIT