Core round 1: whisper CoreML/ANE + GPU eligibility + insights remote-readiness#581
Merged
Merged
Conversation
…ownload - CoreML toggle is now real: use AND download gated on the user setting (default ON) plus a valid encoder asset; OFF forces CPU. Reload-on-change so flipping it takes effect. The only reliable escape on a device where CoreML crashes/garbles (native failures can't be caught in-process; no denylist). - [Whisper][ACCEL] logs state the resolved coreML/gpu/flashAttn config per load and point at the native "Core ML model loaded" line so a log pull confirms what ran. - Harden ensureCoreMLEncoder: stall watchdog (aborts a dropped connection), extract to a temp dir + atomic rename, and integrity-check (coremldata.bin) so a partial/ corrupt encoder is never treated as present. "present" now means valid. - deleteModel removes the .mlmodelc encoder alongside the .bin (no orphaned ANE dir). Co-Authored-By: Dishit Karia <hanmadishit74@gmail.com>
… list Optional coreMLStatus prop on the shared ModelCard (additive; only the Transcription tab passes it, so no drift to Text/Image cards). iOS-only: shows ANE when a valid encoder is present, CPU otherwise. Co-Authored-By: Dishit Karia <hanmadishit74@gmail.com>
Mounts the real LocketRecordingScreen with a recovered clip whose WAV header is damaged (needsRepair). Asserts the Repair banner renders, and tapping it drives the real repairRecording service -> native repairWavHeader(path) -> needsRepair cleared -> banner gone. Fakes only device boundaries (icons/theme/nav/toast + AudioNormalizer/RecordingPlayer). Skips in open-core CI where pro/ is absent. Co-Authored-By: Dishit Karia <hanmadishit74@gmail.com>
…layer smoke The LocketPlayer screen was removed (its transcript + speaker-turn rendering moved to the detail), so locketPlayerSmoke.test.tsx mounted a screen that no longer exists - delete it. Replace with locketSpeakerTurns: mounts the real detail with an analysed, tdrz-segmented recording and asserts the "Speaker 1/2" labels render iff [SPEAKER_TURN] markers are present - guarding the diarization fix (turns now show on the screen the user actually reaches). Both branches falsified. Co-Authored-By: Dishit Karia <hanmadishit74@gmail.com>
The recorder's Analyse guard needs to know whether a summary backend is available WITHOUT a local model download (a connected remote server counts). Expose the summarizer's existing remote check as a public method so the guard reads readiness from the same signal the summarizer routes on, never a second copy that could drift. Co-Authored-By: Dishit Karia <hanmadishit74@gmail.com>
The lib (2.2.0, unmaintained since 2021) calls the deprecated requestAccessToEntityType, which iOS 17+ no longer allows for reads, so fetchAllEvents returns nothing on modern iPhones. patch-package branch: use requestFullAccessToEventsWithCompletion on iOS 17+, old API on iOS 16 and earlier. Applied automatically via the postinstall hook. Co-Authored-By: Dishit Karia <hanmadishit74@gmail.com>
GPU offload is decided by one hardwareService predicate (deviceSupportsGpuOffload / whisperSupportsGpu) and enforced at whisperService.loadModel, so a stored setting can never request the GPU on an ineligible device. iOS -> Metal (real device, >4GB); Android -> the ggml OpenCL backend (Adreno/Mali, OpenCL 3.0). The recorder settings UI reads the same predicate to show/hide the GPU toggle, so UI and enforcement can't drift. Remove the flash-attention option (GPU-only, no benefit with the ANE encoder, and unsupported by the OpenCL backend) - now forced off. Also simplify the iOS CoreML encoder download (drop the 90/10 progress split; fetch the encoder before dropping the download row). Co-Authored-By: Dishit Karia <hanmadishit74@gmail.com>
|
Important Review skippedAuto reviews are disabled on base/target branches other than the default branch. Please check the settings in the CodeRabbit UI or the ⚙️ Run configurationConfiguration used: defaults Review profile: CHILL Plan: Pro Plus Run ID: You can disable this status message by setting the Use the checkbox below for a quick retry:
✨ Finishing Touches🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
|
ⓘ Qodo reviews are paused because your trial has ended. Ask your workspace admin to add credits to resume reviews. Manage billing |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Core-side round 1 of the locket work, stacked on `feat/locket-pro`.
What's in it
Notes