RecorderDesk v0.2.0

What's new in v0.2.0

• feat(licensing): version entitlement schema, close anon RPC bypass, add dev:prod split; go live on Stripe
• feat(licensing): version entitlement schema, close anon RPC bypass, add dev:prod split; go live on Stripe
• test(visual): tolerate unconfigured-licensing fail-open in assist console guard
• ja rename
• docs: drop config and commands that no longer exist
• build: fail fast on missing build config, and sync .env to the Windows release box
• fix(licensing): move the publishable-key allowlist off the reserved SUPABASE_ prefix
• fix(dev): re-sign Electron after install to survive its revoked signature
• test(onboarding): extract canTestProvider and cover per-family Test gating
• test(call-badge): cover call-status derivation via extracted resolveCallStatus
• fix(ui): keep the transcript toolbar on one row, and free space at the minimum window size (#83)
• fix(import): keep the ASR timestamps instead of re-deriving them from line index
• fix(asr): feed the VAD frame by frame — imports were losing most of the transcript
• feat(import): choose the spoken language when importing a file
• fix: don't claim "Processed locally" for a cloud summary; title imports on any provider
• fix(asr): import/re-transcribe English with parakeet-en, not the removed Whisper
• fix(asr): stop Settings offering Whisper models the registry no longer has
• fix(asr): download the model English actually runs on (parakeet-en, not zipformer)
• fix(cli): don't let a stale ANTHROPIC_BASE_URL hang the subscription CLI
• fix(win): stop CLI providers hanging when the app is launched from Explorer
• fix(cli): make a CLI timeout say what actually happened
• fix(app): single instance, and show elapsed time while a CLI test runs
• fix(cli): find the CLI binary on Windows (agy/claude "not found" despite being installed)
• refactor(onboarding): make the auto-download decision pure and tested
• fix(onboarding): match the welcome mark to the real app icon proportions
• fix(onboarding): stop the ASR model re-downloading in an endless loop
• fix(onboarding): end the model download, round progress, use the app icon
• fix(win): show the window even when the first paint never arrives
• fix(win): make the window actually appear when launched from the installer
• fix(test): make joinWithin expectations platform-agnostic
• test: cover the mobile-companion flag, ICS provider, and Ask-AI prompt assembly
• refactor(ui): generalized call badge; drop redundant dashboard pill and a vacuous test
• feat(onboarding): AI model step with Local/Cloud/CLI selector and a real Test
• fix(summarizer): bound on-device prompt and chunk long transcripts (fixes machine OOM)
• fix(settings): make decodingMethod/hotwordsScore round-trip; denylist trial/pairing state; add completeness guard
• feat(onboarding): local-only wizard, Full default, brighter UI, fix English Continue
• fix: sidebar Shortcuts contrast on light theme + Teams auto-record false positives (#67)
• fix(call-detection): match COMMAND column and raise UDP threshold off idle churn
• test: cover accent/text-size/clock + fix visual suite serving a stale bundle
• fix(shortcuts): render and match platform-native keys on Windows
• fix(teams): implement Windows call detection via tasklist + netstat UDP
• refactor(ui): migrate residual violet to cobalt tokens; solidify mind-map Re-generate button
• refactor(ui): migrate residual violet literals to cobalt accent tokens
• feat(ui): shift app icon from violet to cobalt
• feat(ui): brand accent migration + Accent/Text-Size settings + zoom-responsive fixes
• test(export,i18n): cover the mind-map HTML export and data-held language keys
• fix(teams): stop auto-record firing on wake from sleep
• fix(teams): await the Teams window lookup when choosing which screen to record
• fix(teams): detect calls by the process names Teams actually uses
• fix(live-transcript): make speaker merges stick and let a bad turn be corrected
• fix(summary): scale summary depth with meeting length
• feat(meeting-detail): open the automation runs from the badge
• fix(diarization): run each window in its own process so memory is reclaimed
• fix(meeting-detail): stop truncating chapter labels, compact the header
• feat(screenshots): show recording offsets and seek the player from them
• fix(ui): make btn-danger opaque so it can't show the page through it
• fix(asr): only claim a first-time MLX download when one is happening
• perf(diarization): diarize long recordings in bounded windows
• fix(chapters): sample the whole recording instead of only its opening
• fix(transcript): roll timestamps over into hours past the first hour
• test(ipc): assert the IPC contract against the real preload surface
• test(asr): assert supervision statuses as structured payloads
• feat(recording): checkpoint the live transcript so quitting mid-recording keeps the words
• fix(recording): journal the session at stop so an interrupted save can't lose the meeting
• fix(assist): bound the Meeting Progress prompt and stop hard-failing on model JSON
• fix(asr): supervise the dictation worker and make Restart ASR actually restart
• feat(audio): add "start recordings muted" setting (#60)
• fix(i18n): structured dictation-status payload instead of hardcoded English (#59)
• feat(asr): lead model pickers with tier, not model name
• chore(i18n): remove dead settings.asr keys from en/ja locales
• fix(security): move Google OAuth secret server-side; sign entitlement responses

Changes since v0.1.10.