Chatting
The conversation window — sending, stopping, the thinking timeline and the sidebar.
A chat is a conversation with the model you picked, carried out on your computer with your own API key. The model can answer from what it knows or reach for tools — search the web, read a file, run a command — and every step it takes is shown as it happens.
Sending a message
Type into Ask anything on the home page and press Enter. Shift+Enter starts a new line, and while an input method is composing (Chinese, Japanese, Korean) Enter picks the candidate instead of sending. The reply streams in place; a title for the chat appears in the sidebar a few seconds later.
- Stop ends a reply at once. What was written so far is kept, and you can send the next message straight away.
- While a reply is still being written, a second Enter shows Please wait instead of sending twice.
- You can leave a chat while it is answering. It carries on in the background, and a Response ready notice with View tells you when it is done.
The thinking timeline
Above each answer, a timeline shows what the model did to get there: its thinking, every search, file read and command, each as a row you can open. While the run is going it says Thinking… or Working…; afterwards it folds into a single line such as Thought for 12 seconds. A tool that failed is a line of the timeline, not an error box.
Answers
Answers are Markdown: headings, lists, tables, code blocks with a copy button
and syntax highlighting, quotes, links and maths ($$…$$). Prose stays prose —
$200 - $300 is money, not a formula, and 19~32°C is a range, not
strikethrough.
When a reply stops because the provider failed (a used-up quota, a bad key), what was written is kept and the reason shows at its foot: This reply stopped with an error: …. See Answers & sources for copying, citations, regenerating and reading aloud.
Long chats
A chat opens on its newest messages. Scroll towards the top and earlier ones load in without your reading position jumping. On a long chat an outline on the right lists your questions — including ones not loaded yet — and a click jumps there. Scroll to bottom brings you back to the newest message.
The sidebar and the header
- The sidebar's title says where you are — Chat or Philharmonic — and a click switches between them.
- New chat (⌘N), Search chats (⌘⇧F) and Artifacts sit at the top of the sidebar; ⌘B shows or hides it.
- The menu at its foot has Settings, Get the apps (the iPhone app, Health and the command-line tool), Help & docs, What's new and, once a PIN is set, Lock Actovia.
- Each chat's menu has Rename, Favorite (favourites get a group of their own), Copy conversation ID (see recall across chats) and Delete.
- The header shows Back / Forward arrows over the pages you visited (⌘[ and ⌘]), the chat's title, the model chip (see Providers & models), the context ring (below), Chat options (rename, favourite, export as Markdown or PDF, delete) and the Files, tools and sources panel.
- ⌘W closes the chat on screen and takes you home. Nothing is deleted.
How full the context is
Once a chat has answered, a small ring in the header shows how much of the model's context the chat fills — hover it for the numbers. It turns amber when the conversation reaches the point where older messages get compacted.
Click it to open the panel on Context: one bar split into what the last request carried — the system prompt, skills, MCP, tools, memory, summaries and messages — with the free space beside it, the time it was measured and where compaction starts. The panel's other tabs list the chat's Files, Tools and Sources; a row takes you to the answer it came from.
