- Added LiveMonitor class to scan text areas in the frontmost window for grammar and spelling issues.
- Introduced DetectedField struct to represent text areas with issues, including suggestions for corrections.
- Implemented functionality to apply corrections and dismiss suggestions.
- Added Permissions class to manage Accessibility permissions required for monitoring.
- Created SnippetStore for managing text-expansion snippets, including persistence to UserDefaults.
- Developed TextCapture class to read and write text from the frontmost application using Accessibility API and clipboard fallback.
- Implemented TextFieldScanner to traverse the Accessibility tree and collect editable text elements.
- Added TextReplacementCoordinator to orchestrate the text correction process.
- Integrated UpdaterManager for handling application updates using Sparkle.
- Created comprehensive unit tests for various components, including grammar corrections and entitlement checks.
Replace the MLXEngine stub with real Qwen2.5-3B-4bit inference via
mlx-swift-examples (2.25.9): loads the local model snapshot on disk (no
network), chat-templated generation per AIAction, and tone classification.
Wired behind the existing AIEngine seam; AppState resolves the model dir
through ModelManager.
Build notes:
- arm64-only now (MLX is Metal / Apple-Silicon only; matches the spec).
- swift build (CLI) cannot compile mlx-swift's .metal kernels, so the
metallib is generated once via scripts/build-metallib.sh (xcodebuild) and
vendored at AppSupport/mlx.metallib; scripts/build.sh ships it as
Contents/Resources/mlx-swift_Cmlx.bundle/Contents/Resources/default.metallib
(the path mlx resolves, and it codesigns cleanly).
- A hidden --mlx-selftest entrypoint runs one headless generation to verify
the runtime end-to-end.
Verified: the signed app's self-test produces a real rewrite; full test
suite green (the MLX integration test is gated behind GRAMMAGEM_MLX_TEST).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Implements the highest-leverage items from docs/grammarly-value-plan.md:
- Live Check: on-device 0-100 writing score, Correctness/Clarity/Polish
count chips, and per-suggestion Dismiss / Add-to-dictionary (a cheap,
always-available 'no' is what makes 'Apply' feel safe).
- Value loop: day-streak, correction milestones and a 'subscription you
didn't pay' figure on the dashboard + a streak line in the menu bar
(all local, self-vs-self, no telemetry).
- First-run demo: a planted sample the real engine cleans in one click,
before any permission ask.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Native SwiftUI menu-bar app: global hotkey → capture selected text (AX API /
clipboard fallback) → grammar (Harper) / local AI (MLX) → replace in place.
Free/paid entitlement gate + Lemon Squeezy device-cap licensing, $0 marginal
cost per user. Heavyweight engines behind protocol stubs so it builds out of the box.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
