VibeFuse v1.1.1 updates Code Studio, local speech, and the canvas.
Code Studio
- Clicking a file in the explorer opens that file in the center text editor.
- Scrolling with the mouse wheel over Code Studio moves the editor, the file tree, and the side panels. It no longer zooms the canvas.
- Markdown files have a Preview button that renders the file like a GitHub README: headings, lists, tables, code blocks, and images. Code switches back to the text editor.
- Diff still compares the open file with its saved copy.
Voice
- Whisper speech-to-text and Piper speech run locally, so recognition and spoken replies stay on this PC.
- After the wake word, listening ends shortly after you stop talking instead of staying open.
- When LM Studio is the selected model provider, VibeFuse starts it at launch if it is not already running.
- A voice or model server that cannot be reached fails within seconds instead of waiting several minutes.
Canvas and browser
- Widget header font size and page zoom update immediately instead of stalling the app.
- Browser audio keeps playing when you switch canvases.
- A YouTube widget resumes the playback time you left.
- The browser can confirm a saved login, fill it in, and submit when you return to that site.
- Canvas zoom and the window position are restored the next time VibeFuse starts.
- Right-click copy and paste works in Jarvis fields and in browser pages.
Also in this version
- Vocal notes can be taken from the orb and reviewed in a summary.
- Groq Code is in the agent catalog.
- Settings includes a decision-model choice, including JEV, next to the local LM Studio server.
- Voice chat keeps a history of the conversation.
- Startup waits for catalogs to load instead of opening on a timeout.
Get it
Re-download VibeFuseInstaller.exe from fuseintelligence.org/downloads/VibeFuse/ or open Settings → About → Update.