Skip to content

feat(core): add opt-in expression formatting learning - #1413

Draft
appergb wants to merge 3 commits into
feat/macos-accessibility-drag-guidefrom
feat/awareness-expression
Draft

appergb wants to merge 3 commits into
feat/macos-accessibility-drag-guidefrom
feat/awareness-expression

Conversation

@appergb

@appergb appergb commented Nov 8, 2026

Copy link
Copy Markdown
Collaborator

Adds an independently opted-in, local expression profile for dictation formatting. Only uniquely anchored formatting edits that preserve content and Latin word boundaries are accepted; sensitive or ambiguous text is excluded. The persisted profile contains bounded session IDs, dates and abstract directional evidence, not observed text. Ten consistent independent samples spanning at least three days are required before fixed form-only instructions can enter future dictation prompts. The user can inspect stable directions, disable learning and clear the profile; consent is excluded from cloud sync.

Stale observation generations are invalidated across shared repository facades, settings contention does not block dictation, and corrupt profiles are never overwritten without an explicit clear. This is a conservative initial slice (line breaks, exclamation marks and terminal punctuation), not the requested full per-app perception, blacklist or correction system. It builds on #1407 and retains the Type4Me MIT notice for the reviewed reference implementation.

Validation: 1429 core tests passed (3 ignored); targeted expression, consent, corruption, stale-observer and sync-boundary tests; macOS Tauri cargo check; frontend build and Chromium/WebKit settings tests. Native end-to-end learning across multiple sessions and days still requires acceptance testing.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant