The headline of this release is a rewritten AI chat window.
- Attachments live in the conversation. A screenshot, a picture from the clipboard or an attached file no longer sit in a strip beneath the window. They appear in the flow where you added them, outlined in a dashed border so "staged" never reads as "sent" — and once you send, they stay in exactly that place. Nothing jumps.
- There is a preview. Clicking an attachment opens it in full: a picture at full size with a caption saying where it came from, a text file with its contents, verbatim. You can check what the model will read before it is sent rather than after.
- Answers are formatted. Lists, headings and code blocks look like lists, headings and code blocks, and the text arrives as it is written instead of in one lump at the end.
- The window is a sensible size. It now fits itself to your monitor. It used to be a fixed box that looked tiny on a large screen.
Also in this release:
- The floating badges were redrawn. The recording indicator with its level bars, the assistant badge and the Translit Mode badge — the circles and bars no longer have stepped edges.
- We stopped writing "⚠ LLM" at you. When dictation correction fails it now says so: "AI correction failed". That matters more than it sounds: the text still goes into your document, just uncorrected, and a silent badge looked exactly like success.
- The AI assistant tab in the console is translated. Some labels there stayed English whatever language you had chosen. Dictation correction moved to the Voice tab, next to dictation, where people look for it.
To update: the console will offer the update itself, or download the installer from the site. It installs over your current copy — settings, dictionaries and licence are kept.