Add Confucius4-R2T2 MLX speech model - #111
Merged
Merged
Conversation
IchenDEV
marked this pull request as ready for review
September 23, 2026 14:57
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Outcome
Adds
mlx-community/Confucius4-R2T2-8bitas a selectable local speech model in Utter. The model uses the existing Qwen-compatible MLX runtime. A model-specific 0.5-second silence tail preserves final words when a recording stops mid-sentence.The model download now uses pinned-revision parallel HTTP ranges through the application's existing
ModelCatalog.downloadASRpath. Each file must match its pinned size and SHA-256 before the generation is published. Other models retain their Hub download path.Verification
deleteASR.bash scripts/sdlc-checks.sh,bash scripts/ci-basic-checks.sh, andswift testpassed (756 XCTest cases, 16 environment skips, no failures; one Swift Testing case passed).bash scripts/build-app.sh --app-onlypassed release-style app and Metal artifact verification.review-evidence/2026-09-23-confucius4-r2t2-mlx/.Review and release
The user confirmed the model-license decision. The app links the Chinese and English license text and conversion notice; model weights are downloaded, not bundled. The live test invoked the same catalog method as the UI download button. Native button clicks could not be observed because the computer-use native pipe failed to start; this is recorded in the SDLC verification artifact. This PR does not publish a signed release.
SDLC bundle:
docs/sdlc/changes/2026-09-23-confucius4-r2t2-mlx/(medium risk). Reviewer focus: pinned model manifest and hashes, range retry/cancellation, generation publication, model selection, and license links.