Open chud

Documentation

Library, files and knowledge

Where work is kept, and how asking your own files actually works.

Where things are kept#

Library
Everything you have produced in one place — documents, images, quizzes, decks, reports.
Files
Files you have uploaded, with a preview and the option to make one searchable.
Archive
Chats you have archived. They leave the recents list and stay readable.
Saved insights
Individual answers you starred, kept apart from the conversation they came from. Pro.
Knowledge base
Notes and facts you keep deliberately, editable by hand. Pro.
Books
Books you add, with passage-level questions answered against the text. Pro.

Asking your own files#

A file has to be indexed before it can be asked about. Indexing uploads a copy to a retrieval store and answers are then drawn from the indexed text with the passage cited — this is retrieval over your documents, not the model recalling them.

  • Indexable: PDF, TXT, Markdown, DOC and DOCX, PPTX, TeX, and common code files — JS and TS, Python, Java, C, C++, C#, Go, Ruby, PHP, shell, HTML, CSS, JSON.
  • Not indexable: images and spreadsheets. There is no searchable text for the index to hold, so they are not offered.
  • An index is scoped: files indexed inside a workspace answer questions in that workspace, and your personal index is separate.
  • Removing a file from the index stops it being searched.

Indexed is not the same as attached

Attaching a file to a message sends that document with that message. Indexing makes a file findable across later questions without re-attaching it. Both exist because they answer different needs, and a file being in your Files list means neither until you choose one.

Written from the code that implements it. Verify against src/lib/m/fileSearch.js, src/pages/m/Library.jsx, src/pages/m/KnowledgeBase.jsx.