Search
Noted has two search surfaces: a fast switcher for jumping to a note by name, and a full-text search across the whole vault.
Quick Open
Press ⌘K to open the switcher.
- With an empty query, it lists your 20 most recently modified notes.
- As you type, it fuzzy-matches note names and their aliases (the
aliases:of a note's frontmatter; the alias that matched is shown beside the name) first, then appends full-text content matches below them, so a note whose body (but not title) mentions your query still shows up. - Type
/to switch to commands: New note, Daily note, Settings, Templates, and Shortcuts. - If nothing matches, a Create "…" row lets you make a new note with that title on the spot.
Use the arrow keys to move, Enter to open, and Esc to close.
Full-text search
Press ⌘⇧F to search the content of every note. Results show the note title, a highlighted snippet around the match, and the folder it lives in, and are fully keyboard-navigable.
If your vault is large enough to exceed the index limits (below), a note tells you the results were truncated.
How the index works
Full-text search is backed by an in-memory BM25 inverted index built from a scan of your vault. To stay fast and bounded, the scan visits files newest-first and applies these limits:
| Limit | Value |
|---|---|
| Maximum files indexed | 20,000 |
| Maximum total content | 200 MB |
| Maximum size per file | 20 MB |
The index updates incrementally as you save notes and as new notes arrive from Quick Capture, so results stay current without a full rescan.
Search vs. AI retrieval
This full-text index powers Quick Open and full-text search. The AI assistant uses a separate retrieval system (RAG) to decide which notes to send as context — the two are independent.
How fast it is
Measured on a 10,000-note vault of 25 MB (notes of about 350 words), on a laptop:
| Vault index (links, tags, headings, aliases), cold start | about 0.3 s |
| Full-text index, built on the first search | about 0.6 s |
| A search, 95th percentile | under 10 ms |
| Memory the two indexes take | about 75 MB |
Cost grows in line with the vault: 20,000 notes take about 1.5 s to index and 150 MB. Because that is quick, the indexes are rebuilt in memory when the app starts rather than saved to disk: a saved copy would be one more thing to keep in step with the notes. The benchmark (npm run bench:index) runs in CI against a budget of 2 s to index and 50 ms to search, and fails a change that makes either more than 20% slower.
Next steps
- AI assistant — completions, commands, chat, and retrieval.