Live demo
See the engine think.
Ask a question. Watch it embed the query, search a small built-in corpus, and answer with its sources attached. This is the difference between an engine and a chatbot: every answer can be traced back.
// This runs a lightweight retrieval over a built-in corpus, entirely in your browser, with no network calls. It mirrors how Manifest works end to end: embed, search, and answer with sources. The production engine uses neural embeddings and pgvector over real documents at scale.