monolabsdev/openbench-ai

v0.19.0 Feature

This release adds 3 notable features for engineering teams evaluating rollout.

Published 16d LLM Frameworks

View tool

✓ No known CVEs patched

Read the diff → Tool health → What is this tool? →

✓ No known CVEs patched in this version

Topics

local-llm multi-model offline-ai ollama privacy-first tauri

Summary

AI summary

Updates Highlights, Bug Fixes, and speech-to-text across a mixed release.

Changes in this release

Type	Severity	Summary	CVE
Feature
Feature	Medium	Introduces mobile web chat interface with Wi‑Fi QR pairing and optimized input component. Introduces mobile web chat interface with Wi‑Fi QR pairing and optimized input component. Source: llm_adapter@2026-07-15 Confidence: high	—
Feature	Medium	Integrates Supertonic TTS plugin with engine selector in Settings → Speech. Integrates Supertonic TTS plugin with engine selector in Settings → Speech. Source: llm_adapter@2026-07-15 Confidence: high	—
Feature	Medium	Adds experimental voice mode with local Whisper transcription and UI controls. Adds experimental voice mode with local Whisper transcription and UI controls. Source: llm_adapter@2026-07-15 Confidence: low	—
Feature	Low	Adds experimental Voice Mode with local Whisper transcription, VAD, sleep/wake commands, quiet mic normalization, and duplicate sentence deduplication. Adds experimental Voice Mode with local Whisper transcription, VAD, sleep/wake commands, quiet mic normalization, and duplicate sentence deduplication. Source: granite4.1:30b@2026-07-15-audit Confidence: low	—
Performance	Medium	Improves TTS streaming using chunked state machine with word‑boundary events. Improves TTS streaming using chunked state machine with word‑boundary events. Source: llm_adapter@2026-07-15 Confidence: high	—
Bugfix	Medium	Fixes dictation robustness by exposing partial transcripts, audio level, error state, and auto‑restart guard. Fixes dictation robustness by exposing partial transcripts, audio level, error state, and auto‑restart guard. Source: llm_adapter@2026-07-15 Confidence: high	—
Bugfix	Medium	Switches tauri-plugin-supertonic from local path dependency to published crate v0.2 for reliable builds. Switches tauri-plugin-supertonic from local path dependency to published crate v0.2 for reliable builds. Source: llm_adapter@2026-07-15 Confidence: low	—

Full changelog

Highlights

Experimental Voice Mode

Hands-free voice interaction powered by local Whisper transcription. Tap the mic in the sidebar to enter a full-screen overlay with a live breathing orb, real-time partial transcripts, and audio level metering.

Voice Activity Detection — adaptive silence/speech thresholds with energy tracking and hangover window. The mic activates on speech and pauses on silence automatically.
Sleep & Wake Commands — say "go to sleep" to pause dictation and "wake up" to resume, without touching the UI. Keywords are configurable in Settings → Speech.
Quiet Mic Normalization — audio from low-gain mics is scaled up before Whisper transcription for better accuracy on quiet input.
Duplicate Sentence Dedup — Whisper's greedy decoder sometimes emits the same sentence twice; these duplicates are now filtered automatically.

Mobile Web Chat

A dedicated mobile chat interface accessible over your local network.

Wi-Fi Pairing — scan a QR code on desktop to instantly open a chat session on your phone. No app install required.
Mobile Chat Input — a mobile-optimized input component with send button, Dictation (speech-to-text), and TTS playback controls.
Settings Tab — a new Mobile settings tab for managing paired devices and connection preferences.

Supertonic TTS Integration

Text-to-speech is now powered by the Supertonic plugin with a new engine selector in Settings → Speech. Choose between available TTS engines and preview voices directly in settings.

Improvements

TTS Streaming — the TTS store now uses a chunked streaming state machine with word-boundary events for smoother playback and real-time status tracking.
Dictation Robustness — the dictation hook now exposes partial transcripts, audio level, error state, and an auto-restart guard to prevent infinite retry loops.
Sidebar Voice Indicator — conversation items in the sidebar show a voice indicator when a conversation was initiated via voice mode.

Bug Fixes

Switched tauri-plugin-supertonic from local path dependency to published crate (v0.2) for reliable builds.

What's Changed

Sync dev into main: supertonic TTS, engine selector, mobile pairing guards by @TheoSlater in https://github.com/monolabsdev/poly-ui/pull/109
feat: experimental voice mode by @TheoSlater in https://github.com/monolabsdev/poly-ui/pull/110

Full Changelog: https://github.com/monolabsdev/poly-ui/compare/v0.18.0...v0.19.0

View diff on GitHub

Weekly OSS security release digest.

The CVE patches and breaking changes that affected production tools this week. One email, every Sunday.

No spam, unsubscribe anytime.

Share this release

Share on X Share on Bluesky

Track monolabsdev/openbench-ai

Get notified when new releases ship.

About monolabsdev/openbench-ai

All releases →