📐 Diagrams, lists, tables and mind maps, drawn by Azynote#
Ask for a diagram of what's being described and Azynote now draws it itself from what the agent sends: boxes, flows and groups, laid out cleanly. The same goes for lists, tables and mind maps. A live one updates step by step as the meeting goes on, and when you edit a name or remove a part, your change wins over later updates. In a narrow window a diagram goes top to bottom so every label stays readable, and the title bar keeps the time of the last update. A deck opens its presenter view, Open in browser shows any artifact in your browser, and ⌘R reloads the one in front.
🧠 Chats that remember#
A long thread now remembers how it started, even when the model has to start fresh. On a local model, a long chat makes room for itself before the model's memory fills. Type /compact to have a thread squeeze its memory, or /clear to start it over, and a line in the chat tells you what happened. A question you ask the orb over a session continues that session's thread.
💻 Local models that take turns#
Azynote now loads local models only as far as your Mac's memory allows. What you're waiting for goes first, and when one model has to make way for another, you see the swap. Switch a capability to another model and the old one is released right away. Quitting is quick too, even while a model is loading or answering.
🐛 Notable Fixes#
8 fixes
- Stopping a transcription keeps the last words said before Stop
- A transcription started right after a stop stays in its own session
- OnePagers are written in the meeting's language, read from what was actually said
- Dictating into a note lands where your caret is
- An orb action on text selected in a note pastes its result back into the note
- Read aloud works again on a speaker you chose
- A narrow Settings window stacks its rows instead of breaking words
- Refusing notifications no longer stops the first run
Have great meetings! 🎙️