v1: close product parity gaps for daily dictation #77

BLCVoice v1 must not ship as a functional-but-thin dictation shell. This umbrella issue tracks the user-facing product gaps identified during real KDE/Wayland validation and competitor benchmarking.

Product parity scope

Core daily-use UX

• Compact dictation overlay with real microphone level waveform
• Reliable KDE/Wayland placement strategy (layer-shell or equivalent compositor-native anchoring)
• Preserve target-app focus during dictation
• Customizable shortcut surfaced in-app
• Start/stop audio feedback controls
• Launch at login
• Clipboard fallback UX including restore/side-effect handling

History

• Search
• Today / Yesterday / older grouping
• Copy
• Retry / reprocess
• Delete
• Metadata/details view
• Useful usage stats (words, sessions, recent activity)

Personalization

• Dictionary / custom vocabulary
• Deterministic replacements
• Snippets / reusable spoken expansions
• Profiles / modes for different writing contexts
• Model recommendation labels based on speed vs accuracy, not a blanket Base-Q5 recommendation

App structure

• Move technical backend details behind Advanced/Diagnostics
• Strengthen Home as a daily-use hub
• Expand Settings beyond microphone/language
• Keep local-first/offline-first behavior and avoid paid/backend dependencies

Branding and Linux desktop integration

• Final BLCVoice visual identity / app icon
• Complete platform icon set
• Correct Linux desktop launcher/app-id/taskbar identity
• Package/install behaves as BLCVoice rather than the terminal host/dev process

Acceptance bar

• No item above is considered complete by UI mock alone.
• Each feature must be wired to real local persistence/runtime behavior where applicable.
• Existing dictation, local history, privacy guarantees, portal/EIS behavior and CI must remain intact.
• Real KDE/Wayland validation remains mandatory before release.

This issue is intentionally broader than PR #68. PR #68 should remain focused on current Linux/runtime hardening and first UX pass; follow-up implementation may be split into reviewable PRs, but this umbrella scope must be fully closed before calling BLCVoice v1 product-complete.