Documentation

Notes & Files

Two simple, everyday places to keep things: free-form Notes and a searchable Files library. Both are also reachable by the assistant — it can write a note for you, or search your documents to answer a question.

Notes

The Notes page (in the sidebar) is a two-pane notebook: your notes on the left, the editor on the right.

  • New creates a note; give it a title and a body. The body is Markdown (including Mermaid diagrams) edited in a Monaco editor, with a formatting toolbar. The view toggle switches between edit, split, and preview (split falls back to single pane on phones).
  • Notes save automatically as you type (“Saving…” → “Saved”), so you never lose work.
  • Tags organize notes: add or remove them on the open note, and click a tag chip in the list to filter by it.
  • Pin keeps important notes at the top of the list.
  • Search filters by title and content.
  • Share a note with other users or groups at read or write access; a writable note edits live (real-time collaboration) when more than one person has it open.
  • The note’s “more” menu also exports it as Markdown or prints it (to PDF via your browser’s print dialog).

Your assistant can create and update notes too — ask it to “jot this down” or “draft a note summarizing this thread”, and it appears here. Notes you own can be deleted; deletion is permanent. A note shared with you at write access can be edited but not deleted.

Files

The Files page (in the sidebar) is “all your files in one place” — both documents you upload and files on your connected devices.

Uploaded documents

  • Upload a PDF or text file (plain text, Markdown, CSV, JSON; up to 12 MB, 100 documents per user). Its text is extracted and indexed for semantic search (RAG), so the assistant can find and cite it in answers via its search_documents tool. While indexing, it shows “indexing…”; once done, you’ll see how many chunks it was split into. If no embedding model is configured the upload is stored as pending, ready to index later.
  • Mark a document Confidential to restrict it to models cleared for the internal (data classification) trust tier, so the passages never reach a provider that isn’t cleared for that data. Toggling confidential re-stamps the document’s chunks immediately; no re-indexing is needed.
  • Re-index re-processes a file - useful if you replaced its contents or after an embedding model was configured (it appears on documents still pending).

Files on your devices

Switch the source from Uploaded to one of your connected devices (those that expose a home index) and type a name to locate files on that machine. This is a name search of the device’s read-only home directory - it finds where things are; it doesn’t copy their contents into the index. In a regular (non-coding) chat the assistant can then read those files for you with its search_files / read_file tools; for editing and running code, use a coding chat instead.

!!! tip “Search across everything” The All source searches your uploaded documents and every connected device at once, so you don’t have to remember where a file lives.