Features / Collections

Ask your documents.

Collections index PDF, text, Markdown and web pages. An agent finds the passages that answer you.

In development builds, coming in the next release.

What it does

  • A collection is a named set of documents. LocalAI splits them into chunks and stores them as embeddings.
  • Add files, or a web address as a source.
  • A search finds passages by meaning. With the default engine it does not match exact words or codes.
  • A saved agent lists the collections it may search. An agent searches a collection when you name it.
Good to know
  • Collections need an embedding model. Install one first.
  • On a server with user accounts, an admin turns Collections on for each person.