Jake
3770815fbf
Closes the code-side v0.1 ship gate. All quality gates green: cargo fmt/clippy/test (~327 tests), npm check (0/0), vitest 13/13, scripts/dogfood-rebrand-drill.sh 8/8. Phase F — first-run onboarding promoted to v0.1 - FirstRunPage with skip-to-main + failure recovery + event recording - Six onboarding commands (record/list/has-completed + lumotia_events) - Storage migration v17 (onboarding_events + lumotia_events tables) UI hardening (in-scope items from v0.1-ui-hardening.md) - StatusPill + PostCaptureCard components, 21st preview entry - Sidebar recording-as-sacred-state (opacity + aria-disabled, reduced-motion) - Settings 6-section regroup + Help section + Activation log + Privacy toggle - Error-state copy sweep (DictationPage + SettingsPage, plain-language) - Global :focus-visible rule, textarea outlines restored - Ctrl+K / Ctrl+, / Escape bindings in +layout LLM resilience - rule_based_extract_tasks (regex-free imperative-verb extractor) + extract_tasks_with_fallback wrapper — task extraction never returns zero - tokio::time::timeout(120s) wraps cleanup/tags/tasks commands Release artefacts - LICENSE (canonical AGPL-3.0), CHANGELOG (Keep-a-Changelog format) - v0.1-release-notes, privacy-and-ai-use, install-warnings, tester-onboarding-kit, tester-acceptance-runbook, code-signing-setup, apple-silicon-rb08-runbook, virtual-audio-setup, v0.1-contrast-audit - Workspace versioning + AGPL spdx; npm exact-pin (10 ranges removed) - AppImage SHA-256 sidecar in build.yml - README v0.1 section + Reporting-issues; canonical repo slug Closure pass — items moved from human-required to code-complete - KI-02 Linux idle inhibit: zbus 5 → org.freedesktop.login1.Manager.Inhibit - KI-03 Windows sleep prevention: SetThreadExecutionState(ES_CONTINUOUS|...) - acquire/release_idle_inhibit Tauri commands, wired in DictationPage - Diagnostic-bundle frontend wire-up (Settings → Help button) - WCAG-AA contrast fix via .btn-filled-text utility (no token changes) - 8 destructive-action sites wrapped in plain-language confirm() guards - KNOWN-ISSUES.md + v0.1-known-limitations.md updated (KI-02/03 fixed) Scripts - pre-tag-verify.sh, tag-day.sh, smoke-linux + driver - parse-diagnostic-bundle.sh, parse-activation-log.py Per-item audit trail: docs/release/v0.1-completion-status.md Remaining: W-01…W-08 (signing certs, hardware probes, smoke matrix, tester recruitment) — see docs/release/v0.1-known-limitations.md.
4.8 KiB
4.8 KiB
Changelog
All notable user-facing changes to Lumotia are documented here. Format: Keep a Changelog. Versioning: SemVer.
Unreleased
0.1.0 - 2026-MM-DD
Added
- Dictation with on-device speech recognition. Speak naturally; Lumotia transcribes locally using Whisper or Parakeet. Six Whisper variants (Tiny through Distil-Large v3) and Parakeet for lower-latency English transcription. The first-run hardware probe selects the fastest-accurate pair for your machine. Models download once and run entirely offline.
- Live streaming transcription. Audio is processed as you speak rather than in one batch at the end, so you see results as they arrive. Speech-gated chunking and a duplicate-boundary filter keep the output clean.
- Transcript cleanup without touching your original. A rule-based pass removes filler words, collapses repetition, and applies consistent spelling. If you have a local LLM model downloaded, it runs a second pass for natural-language polish. Your raw Whisper transcript is always preserved — cleanup is additive, never destructive.
- Automatic task extraction with a safe fallback. Lumotia identifies action items from your transcript using either the local LLM or a rule-based verb-list extractor. If the LLM fails, the fallback runs silently and tasks still appear.
- MicroSteps and a 5-minute focus timer. Any extracted task can be broken into 3–7 concrete sub-steps. Select a MicroStep to start a 5-minute timer scoped to that one step, so you can work on one thing at a time.
- History with full-text search. Every transcript is indexed and searchable. Filter by date, tag, or keyword. Open any past transcript in the editor to review or correct it; edits autosave.
- Markdown export with YAML frontmatter. Export any transcript as a Markdown file ready for Obsidian or any plain-text workflow. One button, native save dialog.
- Read-only MCP server. An optional
lumotia-mcpbinary lets Claude Desktop, Cline, Cursor, or any MCP-compatible client read your transcripts and tasks over a local stdio connection. It cannot write, edit, or delete anything. - First-run onboarding flow. A short guided setup covers microphone selection, model download, and a practice recording with a pre-supplied prompt, so you know what to say.
- Per-profile custom vocabulary. Add domain-specific terms to a profile and Lumotia feeds them to the transcription engine as hints, reducing misrecognition of names and jargon.
- Content tags and topic suggestions. Lumotia suggests tags from your transcript. Promote them to your manual tag list with one click, or ignore them.
- Keyboard-navigable throughout. The full capture flow — record, review, extract tasks, start a timer, export — is completable without a mouse. Focus rings are visible on every interactive element.
prefers-reduced-motionis respected app-wide.
Privacy
- All transcription, cleanup, and task extraction runs on your device. No audio, transcript, or task data is sent anywhere.
- No telemetry, no analytics, no crash reports leave the machine unless you explicitly bundle one for a support request.
- An optional local activation log records when the hotkey fires. It never leaves your machine and can be disabled in Settings → Privacy.
- The optional MCP server is read-only and communicates over stdio only — no network listener, no remote access. Enabling it gives your MCP client read access to your full transcript history; treat it accordingly.
- Full disclosure:
docs/release/privacy-and-ai-use.md(link added when that page lands).
Known limitations
See docs/release/v0.1-known-limitations.md for the full list. Brief summary:
- macOS App Nap may pause Lumotia when its window loses focus during long sessions. The protection code is in place but not yet verified on Apple Silicon hardware.
- Linux idle inhibit is not wired. The compositor may lock or suspend during a long dictation session. Workaround: launch with
systemd-inhibit --what=idle:sleep lumotiaor raise your system screen-lock timeout. - Windows sleep prevention is not yet implemented. Set your active power plan's sleep timeout to "Never" while dictating.
- If the local LLM hangs mid-generation, the status indicator may stay on "Cleaning up" until you restart the app. Your transcript is preserved and the rest of the app continues to work.
- Cloud transcription (OpenAI Whisper API and equivalents) is not available in this release.
- The MCP server grants read access to your entire transcript history with no per-row permission boundary. A finer-grained permission system is planned for a later release.