rocky/OSGKeyboard
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

refactor: drop Qwen3 CoreML ASR, add local-engine cloud polish toggle

Browse Source 
Rolls back the v0.2.0 Qwen3 CoreML on-device ASR stack and replaces the
'local engine' UX with iOS 26 SpeechAnalyzer + DictationTranscriber only.

The 'Cloud polish after ASR' toggle (ProviderConfig.localModeCloudPolishEnabled)
lets users opt into a post-ASR DeepSeek round-trip from the local engine.
Defaults to off so the local engine stays genuinely local. New PolishError.missingAPIError
surfaces an inline 'fill in your key' warning when the toggle is on but the
Keychain is empty. DeepSeek preset default model bumped to deepseek-v4-flash.

Deleted:
  - OSGKeyboard/ThirdParty/Qwen3Speech/ (74 files, ~16k LoC)
  - OSGKeyboard/Services/ModelManager.swift (492)
  - OSGKeyboard/Services/OnDeviceModelWarmup.swift (197)
  - OSGKeyboard/Services/Qwen3ASRService.swift (257)
  - OSGKeyboard/Services/ModelDownloadSourcePicker.swift (126)
  - OSGKeyboard/Views/OnDeviceModelsView.swift (184)
  - OSGKeyboard/Views/DownloadConfirmSheet.swift (96)
  - OSGKeyboardShared/Models/OnDeviceModel.swift (140)
  - OSGKeyboardShared/Services/OnDeviceModelStatus.swift (104)
  - Qwen3ASRServiceProvider registration in OSGKeyboardApp
  - Qwen3Speech package declaration in project.yml
  - 5 .qwen3ASR enum / branch reference sites in HomeView, OnboardingView,
    LocalEngineSettingsRows, FlowSessionManager, ASRService, EngineServiceLabel
  - Two pre-existing Swift 6 strict-concurrency errors in
    LiveDictationController + FlowSessionManager (the weak [weak self] in
    detached-task MainActor.run blocks) that were blocking clean builds

Added:
  - LocalModelsGroup: 'Built-in iOS SpeechAnalyzer' badge + 'Cloud polish
    after ASR' Switch toggle
  - PolishingService: honour localModeCloudPolishEnabled; new .missingAPIKey
    error case with localised warning
  - AppGroupStore.localModeCloudPolishEnabled (mirrored into App Group
    so the keyboard extension honours the toggle during live dictation)
  - SettingsView: show provider/api sections when local-mode cloud polish
    is on so the user can paste a DeepSeek key
  - FlowSessionManager: route through PolishingService for local + polish-on
    flow; translate missingAPIKey into a polished warning
  - KeyboardViewController: handle PolishingService.PolishError.missingAPIKey
    in the keyboard-side live polish path
  - CHANGELOG v0.2.1: documents the rollback + new toggle
  - README.md / README.zh.md: engine matrix section, data flow note

Verified: xcodebuild -scheme OSGKeyboard -destination 'generic/platform=iOS Simulator'
build succeeds under SWIFT_STRICT_CONCURRENCY=complete.
This commit is contained in:
Rocky
2026-06-24 01:51:34 +08:00
parent 39690c0a93
commit c07cf4db9f
119 changed files with 401 additions and 18858 deletions
Download Patch File Download Diff File Expand all files Collapse all files
README.md
+13
View File
@@ -29,6 +29,7 @@ The audio stays on-device (transcribed by Apple's on-device `SpeechAnalyzer` + `
- 🎙 **Push-to-talk** with a Typeless-style circular mic button
- 🧠 **On-device ASR** (`SpeechAnalyzer` + `DictationTranscriber`, iOS 26+)
- ✍️ **AI polishing** — adds structure, punctuation, fixes grammar, optionally produces lists
- 🧩 **Local + cloud polish toggle** — local engine is ASR-only by default; opt into a post-ASR cloud polish step (DeepSeek by default) when the iOS speech recognition isn't strong enough for your environment (noisy far-field audio, strong accents, etc.)
- 🔌 **Bring-your-own API** — works with any OpenAI-compatible endpoint (OpenAI, DeepSeek, Qwen DashScope, your own self-hosted server, …)
- 🔒 **Privacy first** — audio never leaves your device; transcripts only sent to the LLM you choose
- 🎨 **Native SwiftUI** — dark theme, frosted glass, ~2000 lines of Swift
@@ -108,6 +109,18 @@ OSGKeyboard/
textDocumentProxy.insertText(polished)
```
**Engine modes:**
- `cloud` (default): ASR runs on-device via `SpeechAnalyzer`, the transcript is
sent to your configured LLM for polish.
- `local`: ASR runs on-device via `SpeechAnalyzer`. The transcript is inserted
as-is — no cloud round-trip.
- `local` + "Cloud polish after ASR" toggle (Settings → On-device models):
same as `cloud` in spirit, but routed via the local-engine pipeline so the
keyboard extension can still use the on-device ASR; the transcript is sent
to the configured LLM before insertion. Requires a DeepSeek (or other
OpenAI-compatible) API key in the Keychain.
---
## Adding a new LLM provider