Commit 669a661 had a stale staged deletion of cc-design-system.css and
reverted three other CSS-variable files. This commit restores them to
the state from cd37771 (adopt --cc-* CSS variable system).
navigationStore: rewritten off Neo4j db names — Supabase whiteboard_rooms table,
setAuthInfo(token, userId) pattern, auto-creates default room per context on first use
snapshotService: rewritten to Supabase Storage REST (/storage/v1/object/authenticated/cc.users/…),
setAccessToken() instance method, static methods take accessToken not dbName
AuthContext/NeoUserContext: auth injected into nav store, no Neo4j db names required
singlePlayerPage: loadNodeData no longer calls Neo4j; snapshot wired via accessToken
navigation types: NeoGraphNode updated for Supabase-backed tree structure
transcriptionStore/Service: getSession() removed, accessToken via AuthContext
LLMConfigModal: auth context wiring fixes
GraphNavigator/GraphSidebar: updated nav components
Co-Authored-By: Claude Sonnet 4.6 <[email protected]>
Root causes of disconnection and slow transcription:
- AudioWorklet was firing every 128 native samples (~48 kHz), sending
~375 tiny WebSocket messages/sec. Server flooded with tiny frames
during silence → keepalive ping timed out → connection dropped.
- JS resampling 48 kHz → 16 kHz added CPU overhead on every chunk.
- Audio started on ws.onopen before server sent SERVER_READY, so early
frames were dropped.
Fixes:
- audioWorklet.js: accumulate 4096 samples before posting (256 ms/chunk
at 16 kHz, ~4 messages/sec), transfer ArrayBuffer zero-copy.
- transcriptionService: AudioContext({ sampleRate: 16000 }) — browser
handles native resampling, no JS resampling needed. Remove
resampleTo16kHZ entirely.
- Wait for SERVER_READY message before calling setupAudioProcessing().
- Send 'END_OF_AUDIO' string on stop so server can finalise last segment.
Co-Authored-By: Claude Sonnet 4.6 <[email protected]>
- transcriptionService: track finalizedSegmentCount so only newly-final
segments are emitted per WS message (was re-processing full array each
time, causing the live segment to freeze in the completed list)
- transcriptionStore: saveSegment isFinal branch now appends the passed
text directly instead of currentSegment (currentSegment was stale
relative to the incoming final)
- CCTranscriptionPanel: record button colour changed from var(--color-text)
to explicit #2563eb so it is visible in dark mode; completed segment
backgrounds changed from hardcoded #fff to var(--color-muted) so text
is readable in both themes; keyword Add button gets same blue fix
Co-Authored-By: Claude Sonnet 4.6 <[email protected]>
The old startTranscription() called enumerateDevices() first to find a
device ID, then getUserMedia. Without microphone permission, enumerateDevices()
returns devices with deviceId="" (empty string, falsy). This caused the
!this.selectedDeviceId check to bail out early, never calling getUserMedia,
never prompting the user for mic permission, and never creating the WebSocket.
Result: user clicks Start Recording → isRecording=true → button changes to
Stop Recording → but no mic prompt, no WebSocket, no transcription.
Fix: call getUserMedia directly (with optional echoCancellation/noiseSuppression
if no device pre-selected). getUserMedia triggers the browser permission prompt
automatically. Device selection is still honoured via exact deviceId constraint
when one has been explicitly chosen.
Co-Authored-By: Claude Sonnet 4.6 <[email protected]>
CC Worker
kcarandClaude Sonnet 4.6