whisper.cpp's encoder always runs over a full, padded 30s mel window
(1500 encoder positions) unless audio_ctx is explicitly reduced. The
~4s live-transcription streaming windows were never setting it, so
every window paid the full 30s-equivalent encode cost on one serial
worker thread -- the actual cause of the reported 30-65s lag between
speaking and a segment appearing, not "CPU is just slow".
Scales audio_ctx proportionally to the window's real sample count
instead (audio_ctx_for_window), floored and capped to whisper.cpp's
own valid range. Verified with a controlled A/B against the real
base.en-q5_1 model: ~30% faster even under heavy concurrent CPU
contention on this machine, which is masking most of the improvement
--the fix targets the encoder's O(n^2)-ish attention cost specifically
(1500 -> ~201 positions for a 4s window), so the gain on an otherwise-
idle machine should be substantially larger.
Same speech-bubble-with-sparkles mark, now on its own gradient purple
background instead of a transparent one — cropped from the refined
logo source (mark + gradient, no wordmark text) rather than isolated
via color-keying this time, since the new source is already composed
as a clean standalone icon tile.
Cropped the speech-bubble-with-sparkles mark from the new
WhispAssist logo/promo graphic (not the full graphic — text and
illustration detail don't read at 16-32px), color-keyed its background
to transparent, and regenerated the whole icon set via `tauri icon`
(title bar, taskbar, system tray, installer). Removed the iOS/Android/
Appx/macOS outputs tauri icon generates by default — this project only
targets Windows.
html/body sit above .app in the DOM tree and were left with the
browser default (white background + 8px UA-stylesheet margin) — a gap
invisible against the light theme but a bright white outline around
the whole window in dark mode. Mirrors data-theme onto <html> (so the
dark token overrides actually apply up there) and zeroes the
html/body margin.
Replaces a hardcoded blue (#2563eb) primary button with the semantic
accent token, adds a ShieldAlert icon + heading, and gives it proper
elevation (shadow + radius) to read as a deliberate dialog rather than
a plain bordered box.
Recording/Hardware/Storage/Calendar/Sync/Privacy nav tabs get matching
icons; the active tab now uses the accent color instead of an inverted
fg/bg swap. Check/X/AlertTriangle icons replace ✓/✗/✕/⚠ throughout
(connection test result, close button, third-party-cloud warning
banner, no-egress confirmation). The modal itself gets proper
elevation (shadow + larger radius) instead of a flat bordered box.
Tags/Summary/Action items/Calendar event/Participants/Speakers each
get a matching icon and a consistent uppercase section-header style;
"Generate summary" becomes a proper primary (accent-filled) button;
the reminder bell emoji becomes a Lucide Bell icon that highlights in
the warning color once checked.
Bold/Italic/Heading1/Heading2/List/ListChecks Lucide icons replace the
letter/glyph toolbar buttons (bullet "•" and checkbox "☑" in
particular); export buttons get matching icons. Section labels for
Transcript/Notes get icons too, and the reprocess spinner animates
(respecting prefers-reduced-motion).
Search icon inside the input, Trash2/Download/RotateCcw replacing
emoji, tag chips and the selected-item state now use the semantic
accent tokens instead of a hardcoded blue hex, and list rows get a
hover state + an accent-colored left border when selected.
- Full semantic color token system (surfaces, text, borders, one
accent color) replacing the previous 4-variable set — light/dark
designed together, every pairing checked to >=4.5:1 (WCAG AA).
- Global custom scrollbar styling (thin, theme-aware) — the default
OS scrollbar looked out of place against the app's own chrome.
- ThemeToggle replaces the plain theme <select>.
- Record/Stop buttons and the settings/warning icons are now Lucide
SVG icons instead of Unicode glyphs (●■⚙⚠); the recording indicator
is an animated pulsing dot (paused under prefers-reduced-motion)
instead of a static red dot.
- Elevation (shadow tokens) and consistent radius scale.
Replaces the plain <select> theme dropdown with three icon buttons.
Defaults to "system" (Monitor), matching the existing behavior of
following the OS preference until the user picks an explicit override.
Replaces ad-hoc emoji/Unicode-glyph icons (record/stop dots, trash,
checkbox, bell, etc.) throughout the UI — explicitly recommended by
the ui-ux-pro-max design skill over font-dependent emoji icons, which
render inconsistently across systems and can't be themed/sized via CSS.
The due-date input is new too — action items previously had no way to
set one, which would have made the reminder checkbox unusable (a
reminder needs a due date to schedule against).
reminder_set was accepted on the model but hard-coded to 0 on insert
and never included in the UPDATE. save_action_items's return type
changes from () to Vec<ActionItem> so the caller (which needs the row
id to key reminder scheduling) doesn't need a second round-trip for
newly-inserted items. Also adds list_pending_reminders for startup
reconciliation.
Reschedules anything Windows' own toast schedule lost track of
(uninstall/reinstall, a changed AppUserModelID) — safe to run every
startup since schedule() is idempotent (cancels any existing entry for
the same action item id first).
Chosen over an app-side polling timer specifically to honor NFR-RES-1
("no polling timers running when not recording") — Windows itself
delivers the toast at due_at, and a scheduled toast is cleared by the
OS after it fires, so there's no "already fired" bookkeeping needed in
this app either. Windows-only, matching hardware::dxgi's pattern: the
real WinRT implementation lives in a #[cfg(windows)] submodule behind
plain top-level functions.
Verified against the live Windows toast API (not just compiled): a
throwaway example scheduled a reminder, confirmed
GetScheduledToastNotifications() reported it back with the correct id,
then confirmed cancel() removed it.
Extracts export_meeting_to as a shared helper (was duplicated inline
in export_meeting) so single and bulk export share identical
fetch+render logic. bulk_export_stem sanitizes+suffixes filenames with
an id prefix — meeting titles are very often duplicates ("Untitled
meeting") and would otherwise silently overwrite each other in a
batch. A meeting that fails to export is skipped rather than aborting
the whole run; also fixes export_meeting's doc comment, stale since
T8.4 added pdf/docx.
Adds StartRecordingArgs.template_id + list_note_templates command;
renames built_in_template to summary_prompt_bias to disambiguate it
from the new NoteTemplate concept (same meeting-type identifiers,
different purpose). Also fixes refresh_notes_and_notify (the
rename/merge-speaker refresh path) to route through
Store::update_notes instead of a raw std::fs::write — it was bypassing
T8.2's FTS reindexing, a gap missed there because this call site uses
meeting_dir(meeting_id) without the & that my earlier grep matched on.
NewMeeting/Meeting gain template_id; finalize_meeting now returns it
(fetched via one cheap scalar SELECT) so callers that only have
segments/speakers in memory (not a fetched Meeting) can resolve the
template without a separate round-trip.
4 built-in templates (Standup, 1:1, Sales Call, Retro), keyed by the
same meeting-type ids commands::summary_prompt_bias already uses for
LLM-prompt biasing — one identifier space shared by both, but two
independent lookups (structuring notes.md vs. biasing a summary
prompt). Section headers are prepended as an empty scaffold above the
auto-rendered speaker dialogue.
Templates themselves are a built-in Rust catalog, not a DB table —
this just remembers which one a meeting picked, so re-rendering
notes.md on reprocess/resume keeps the same section structure.
iamdoubz
iamdoubz