About LoremLLM

LoremLLM started as the world's most overengineered lorem ipsum generator. Building chat interfaces meant constantly waiting on, and paying for, a real model just to see how a streamed reply looks in the UI. LoremLLM replaces that model with something predictable.

The @loremllm/transport package plugs into the Vercel AI SDK's useChat and streams whatever message parts you describe — text, reasoning, tool calls — with realistic chunk timing and no network. The hosted platform does the same over HTTP: record input/output pairs in a collection, and the API answers each incoming question with the closest stored output, matched by embedding similarity.

It is useful for prototyping an AI product before the backend is ready, for deterministic end-to-end tests, for demos where you need to control exactly what appears on screen, and for teaching without an API bill.

LoremLLM is built and maintained by Kaiyu Hsu and is open source on GitHub. The stack is Next.js, React, the Vercel AI SDK, oRPC and better-auth, with Turso (libSQL) for storage and vector search.

Where to go next