Changelog
Every Shruvi release, written in plain English. New entries land with each desktop tag.
-
HUD now draws over fullscreen apps on macOS Tahoe
- macOS 26 (Tahoe) fullscreen fix. The recording HUD uses an NSPanel
with
type: 'panel'so it draws above fullscreen apps. This regressed in the macOS 26 beta and is now resolved. - HUD follows active display. If you have multiple monitors, the HUD appears on the display that currently has focus.
- Admin dashboard polish. Several internal-only fixes for the ops dashboard so we can monitor backend health without SSH.
- macOS 26 (Tahoe) fullscreen fix. The recording HUD uses an NSPanel
with
-
Faster transcription, Deepgram learned vocabulary, Electron 39
- Faster on short clips. We now skip the AI cleanup pass on very short utterances where it adds latency without changing the output.
- One less resample. A redundant resampling step in the audio path is gone — about 340ms shaved off the Deepgram path.
- Deepgram learned vocabulary. When you teach Shruvi a custom word, we now pass it as a biasing keyterm to Deepgram automatically. (Other providers were already supported.)
- Electron 39 runtime. Bumped from Electron 30 to 39 to track the active LTS support window.
- Connection-per-IP limit. Backend now caps WebSocket connections per
source IP via
SHRUVI_WS_CONNECT_PER_IP_LIMIT.
-
Public status page and cache-busted download links
- status.shruvi.com. A public status page (Better Stack on free tier, Cloudflare DNS-only CNAME). Users can now check whether the backend is healthy at a glance.
- Download cache fix.
/download/macand/download/windowsnow include a?v=<version>cache-buster. If you were a Windows user downloading "the latest" only to find an older version installed, this is the fix — the oldCache-Control: immutableheaders on the binary URLs were poisoning browser caches for up to a year. - Cleanup model upgrade. AI cleanup now uses
llama-3.3-70b-versatilefor sharper output on the same latency budget.
-
Full visual redesign across every page
A pre-launch UX overhaul. Every page now speaks the same visual language:
- Ambient hero with the Aurora wallpaper
- Audiowide eyebrows on every section header
- Nunito display numerals
- Bare rows with hairline dividers — no SaaS card density
- Single, consistent grid across Settings, History, Account, and Home
If you've been on Shruvi since v1.3.x, this is the biggest visual change the app has ever shipped.
-
xAI Aurora transcription
- xAI Aurora support. Real-time streaming and batch transcription through xAI's new Aurora model. Speaker diarization built in. Twenty-five languages supported.
- xAI handshake fix. Audio is now held until xAI's
transcript.createdevent, fixing silent failures on cold sessions. - Provider system. xAI joins Groq (default), OpenAI Realtime, Google Chirp 2, and Deepgram in the pluggable provider registry. Switch via the admin API at runtime — no restart.