fix(summary): await onSave to persist editor edits by arvanus · Pull Request #416 · Zackriya-Solutions/meetily

arvanus · 2026-03-30T18:29:18Z

Summary

Adds missing await on onSave() call in BlockNoteSummaryView.handleSave
Without await, the save handler resolved instantly — dirty flag reset before Tauri command completed, errors were silently swallowed, and spinner/toast feedback was imperceptible
Now the save properly awaits the Tauri api_save_meeting_summary command before clearing dirty state

Fixes

Closes #394

Test plan

Edit a meeting summary in the BlockNote editor
Click Save — confirm spinner appears during save
Confirm success toast shows after save completes
Reopen the meeting and verify edits persisted

🤖 Generated with Claude Code

Adds a `get_build_info` Tauri command that exposes compile-time info (GPU backend, build profile, OS/arch) and displays it in the About page. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Emits `model-loading-status` events from Whisper and Parakeet validation so the frontend can show a loading indicator while the transcription model initializes. Displays an animated banner above recording controls. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Stereo recording (mic=L, system=R) with mono toggle for compatibility. Addresses issue Zackriya-Solutions#241. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Address AudioChunk channel awareness, checkpoint frame math, RecordingMode plumbing path, serde defaults for backward compat, file size notes, and expanded testing section. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

- Default changed to mono (stereo is opt-in advanced option) - Pre-allocated interleave buffer to avoid per-window allocation - debug_assert_eq! instead of assert_eq! in hot path - AAC 256kbps for stereo mode - Retranscribe edge case covered - Future iterations section (downmix playback, export mono) - Design decisions log with rationale Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

10-task plan covering AudioChunk channels field, RecordingMode enum, IncrementalAudioSaver multi-channel math, pipeline stereo interleave, preference wiring, AAC bitrate, frontend toggle, and tests. Fixes from plan review: load_preferences via recording_commands.rs (has AppHandle), no broken commits, split metadata into own task, stereo duration test, explicit RecordingSaver struct changes. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Frame-based checkpoint math ensures stereo checkpoints trigger at correct 30s intervals. Includes tests for stereo checkpoint timing and duration calculation. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

STEP 4 now produces interleaved stereo or mixed mono based on RecordingMode. Uses pre-allocated buffer for zero-alloc hot path. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Serde defaults ensure old metadata.json files without these fields deserialize correctly as mono/1-channel. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

…ng path RecordingMode flows: preferences -> recording_commands -> recording_manager -> pipeline + saver -> incremental_saver -> encoder. Replaces temporary Mono default with actual user preference. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Advanced option to record mic and system audio on separate stereo channels. Default: OFF (mono). Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Vulkan was hardcoded in the Windows target dependencies, causing a dual-backend crash (CUDA + Vulkan) on NVIDIA systems. The error was: "pre-allocated tensor in a backend that cannot run the operation" followed by STATUS_STACK_BUFFER_OVERRUN. GPU backends are now opt-in on all platforms via Cargo features: - pnpm run tauri:dev (CPU, works everywhere) - pnpm run tauri:dev:cuda (NVIDIA GPUs) - pnpm run tauri:dev:vulkan (AMD/Intel GPUs) macOS retains hardcoded Metal+CoreML as those are the only option. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

- window_ms: 600 -> 50 (was a debug value left behind, caused massive latency before any audio reached VAD/Whisper) - can_mix: || -> && (only mix when BOTH mic and system buffers have enough data, preventing silence-padded chunks that confused VAD) Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Partial transcriptions during continuous speech for immediate feedback, replaced by final transcription when speaker pauses. Same UX pattern as YouTube Live Captions / Google Meet. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

When retranscribing a multitrack stereo recording (mic=L, system=R), each channel is now extracted, VAD-processed, and transcribed separately. Results are labeled with [Mic] or [System] prefix and merged by timestamp. - Add extract_channel() and is_multitrack_stereo() to DecodedAudio - Retranscription detects stereo and forks into per-channel processing - Mono recordings continue to work as before (no label prefix) - Progress reporting adapted for dual-channel processing Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Add optional `source` field ("mic", "system", or null) to TranscriptSegment instead of prefixing text with [Mic]/[System]. This allows the frontend to format/filter by source without parsing. - TranscriptSegment gains `source: Option<String>` with serde defaults - Retranscription passes source per channel instead of prefixing text - JSON (transcripts.json) includes source field - DB schema unchanged (source stored in JSON only for now) - Test for source field added Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

When source field is present (multitrack stereo), the transcript text sent to the LLM includes [Mic] or [System] prefix for speaker context: [00:15] [Mic] Testing 1, 2, 3 [00:18] [System] Hello everyone Mono recordings continue without prefix. Also adds source field to Transcript TypeScript interface. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

The write_transcripts_json function was building JSON manually and missing the source field. Now includes "source": "mic"/"system" when present (omitted for mono recordings). Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

VAD now emits intermediate partial segments every ~1.5s during ongoing speech, providing real-time visual feedback before the speaker pauses. When the speaker pauses (400ms silence), a final segment replaces all partials for that speech segment. Backend: - VAD emits partial segments (confidence=-1.0 marker) every 1.5s - Pipeline passes is_partial flag on AudioChunk - Worker uses chunk-level partial flag instead of Whisper's - Recording saver skips partials (only finals are persisted) Frontend: - TranscriptContext replaces partials with finals by audio_start_time - Partials displayed in gray italic, finals in normal black - TranscriptSegmentData gains is_partial field Same UX pattern as YouTube Live Captions / Google Meet. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Resolved conflict in pipeline.rs: added is_partial field to recording chunk (multitrack stereo STEP 4). Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Only one speech segment is active at a time, so simply remove all existing partials when a new partial or final arrives. Fixes issue where partials accumulated instead of being replaced due to float precision in audio_start_time comparison. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

When the worker picks up a partial chunk, it now drains any newer partials from the queue and only processes the latest one. This prevents partial accumulation when the speaker talks faster than Whisper can transcribe (~1s per chunk). If a final chunk is found while draining, it processes the final instead (finals are never skipped). Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Partial transcription works at the backend level but the frontend display needs a separate data path. Currently partials leak into the transcript list and recording saver despite filters. Next step: brainstorm clean separation of partial vs final data flow. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Partials now use 'transcript-partial' event (text only), finals use 'transcript-update' (unchanged). This completely isolates partial data from the recording saver and transcript list. Backend: - Worker emits transcript-partial for partials, transcript-update for finals - recording_commands listener no longer needs is_partial filter Frontend: - transcriptService.onTranscriptPartial() listens for partial event - TranscriptContext sets partialText on partial, clears on final - TranscriptPanel shows partialText as live preview box Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

The VAD now emits partial segments (confidence < 0) for live preview, but get_speech_chunks_with_progress (used by retranscription and import) should only return final segments. Partials were leaking into retranscription results, causing duplicate/incremental text. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Partial text now renders inside VirtualizedTranscriptView, directly above the "Listening..." indicator, instead of outside the component. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

The save handler was calling onSave() without await, causing the dirty flag to reset before the Tauri command completed. This made save errors silent and gave no visual feedback to the user. Fixes Zackriya-Solutions#394. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

arvanus and others added 30 commits March 12, 2026 20:25

feat: add build info (GPU backend, platform) to About page

51677a4

Adds a `get_build_info` Tauri command that exposes compile-time info (GPU backend, build profile, OS/arch) and displays it in the About page. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

docs: add multitrack audio design spec

a76262c

Stereo recording (mic=L, system=R) with mono toggle for compatibility. Addresses issue Zackriya-Solutions#241. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

feat(audio): add channels field to AudioChunk struct

9f9b390

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

feat(audio): add RecordingMode enum to recording preferences

6d4013f

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

feat(audio): add stereo interleave and recording mode fork in pipeline

f09f4b1

STEP 4 now produces interleaved stereo or mixed mono based on RecordingMode. Uses pre-allocated buffer for zero-alloc hot path. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

feat(audio): add channels and recording_mode to MeetingMetadata

45a0c7d

Serde defaults ensure old metadata.json files without these fields deserialize correctly as mono/1-channel. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

feat(audio): increase AAC bitrate to 256kbps for stereo recording

fce4ec2

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

feat(ui): add stereo recording toggle in settings

8d38be6

Advanced option to record mic and system audio on separate stereo channels. Default: OFF (mono). Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

test(audio): add unit tests for stereo interleave function

a5997f0

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Merge branch 'enhance/about-build-info' into test/all-features

0287068

Merge branch 'fix/whisper-gpu-backend' into test/all-features

983cc7a

Merge branch 'feat/multitrack-audio' into test/all-features

a0ffdbf

Merge branch 'fix/whisper-gpu-backend' into feat/multitrack-audio

06a9f4d

Merge branch 'feat/multitrack-audio' into test/all-features

211f2d7

merge: feat/streaming-partial-transcription into test/all-features

c2474c4

Resolved conflict in pipeline.rs: added is_partial field to recording chunk (multitrack stereo STEP 4). Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

fix(ui): destructure isPartial prop in TranscriptSegment component

b8aa65e

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

arvanus and others added 7 commits March 20, 2026 22:19

fix(ui): move partial preview above Listening indicator

4a3c0fe

Partial text now renders inside VirtualizedTranscriptView, directly above the "Listening..." indicator, instead of outside the component. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

fix(summary): await onSave to persist editor edits#416

fix(summary): await onSave to persist editor edits#416
arvanus wants to merge 37 commits intoZackriya-Solutions:mainfrom
arvanus:fix/save-summary-await

arvanus commented Mar 30, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Conversation

arvanus commented Mar 30, 2026

Summary

Fixes

Test plan

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant