You remember the conversation, not where you put it. Was the discount rule in the deal review, in a note you took after the call, or in the PDF the customer sent? Search looks through all of it at once and takes you to the place it was said.
Search from any app#
Press Ctrl Shift F wherever you are on your Mac. Azynote's card opens in search, with the field ready.
Type what you are looking for. Results come as you type, about six of them, one per session:
- the session's name and date,
- where it matched ("Note · Agenda", "Transcript · 14:32", "contract.pdf · page 3", "Artifact · Q3 pipeline", "Chat thread · Azynote"),
- the passage itself, with the words you typed in bold. An image shows as a small thumbnail.
Use the up and down arrows to move through the results and Enter to open the one you want. It opens in Azynote right where it matched: the OnePager or the note, or the transcript scrolled to the line that has your words, outlined for a moment. A document, a picture or an artifact opens its session with the file over it: an artifact or a web page in its own window, anything else in the Mac app it belongs to. A chat thread opens in the chat window, scrolled to the message that matched, which is outlined for a moment; when the thread belongs to a session, the session opens behind it. See all opens the same search in the Timeline, with every result.
From the card you already use to ask Azynote (Ctrl Shift C), type a few words and pick Search for to search instead of asking.
Narrow it to one kind#
Under the field, the chips OnePagers, Notes, Transcripts, Documents, Images, Artifacts and Chat narrow the results to what you pick. Pick none to search everything.
Words or meaning, both at once#
Search finds a passage that has your words, and one that is about the same thing in other words. "How do we answer the competitor's price comparison" finds the meeting where the lakehouse pricing came up, even if nobody said "comparison".
- Small words do not count. "Pricing discussion with the customer" looks for pricing, discussion and customer, in English and in French.
- Put a phrase in quotes to find it exactly as written.
- It works across languages. A question in French finds an English note, and the other way round.
- When nothing is close enough, it says "Nothing found", rather than showing the least bad results.
Search a list of sessions#
The search field on Your sessions and on a subject's Sessions tab first filters on what each row shows: title, summary, subjects, date. A moment later, sessions whose content has your words join the list, each with the passage that has them. A few sessions close in meaning follow under Related. A session found in a file or in its chat opens with that file or the chat over it, as from the search card.
Search a picture#
Screenshots, photos and scanned pages are found by what they show: "the architecture diagram with the three regions" finds the whiteboard photo in last week's session. This needs a second download, the image encoder (555 MB). Azynote offers it once your sessions have images. Until you download it, images are not searched and the search card says how many are waiting.
A PDF page with no text, like a scan, is searched as a picture.
What is searched#
Every file you can see in a session:
- OnePagers and notes,
- transcripts, once the meeting is over,
- documents: PDF, Word, Markdown, text, CSV, code and web pages (only the text the page shows, never its code),
- images: PNG, JPEG, HEIC and WebP,
- artifacts you saved to a session, by their name and the text they show,
- and your chat threads: every question you asked and every answer, not just the thread's name and summary.
Every document is also found by its name: "pricing deck" finds pricing-deck-v2.html. An artifact is found by the name you gave it, or by the agent's if you never renamed it.
Audio and video are found through their transcript. Files and folders whose name starts with a dot are left out.
Search stays up to date as you work: a note you save or a file you drop in a session can be found a moment later.
Everything stays on your Mac#
Search runs on a model on your Mac, EmbeddingGemma 2 (310 MB). Azynote asks to download it at first launch. You can say Later: everything else works without it, and Azynote asks again at the next launch.
The index is built on your Mac and never leaves it, whatever AI engines you use for the rest. The first time, indexing a large workspace takes a few minutes. Meanwhile the search card shows how far it has got ("Indexing: 120 of 410") and finds what is done so far.
Settings > Search#
Settings > Search shows the state of the index (Up to date, Indexing, or Search model missing) and offers the downloads you have not made yet.
If search seems to miss things it should find, Rebuild the index deletes it and builds it again from your files. Search finds less until the rebuild is done.