Repository navigation
Improve recording HUD and protect dictation recovery - #21
Merged
Merged
Conversation
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Summary
Verification
Previously run on the exact local main commit
c1cae6adaa990e2e8429717d86e30b05e9b5dd9a: 529 Release tests passed (excluding actual RealModel and QualityCorpus tests), Release solution build passed without warnings, self-contained win-x64 publish andTest-ReleasePackage.ps1passed.Previously: deterministic fakes cover offline and streaming hangs, including a completed reader with blocked final decode, delayed disposal, cancellation, audio limits, and follow-up sessions. Avalonia checks are headless.
User reports the latest robustness build tested positively. This does not establish coverage of every hardware, animation, or failure case.
Fresh for this PR: fetched
origin/mainatc4c383eba73eb4e506cf4ebab409c658958deff3, checked the clean source/head and duplicate PRs, and verified the published EXE SHA-256686EF5EE0A96D3597C18F17684712C4FE66B1BD0F3B3D26F767BAA6DDD4955D2. No source changes or conflict resolution were needed for publication.Follow-up after the first CI run: one HUD test depended on the runner animation setting and a 60 ms wall-clock window. Test-only commit
2bc64524d46c4052426d7a3573b91bb819e5bc05fixes the test setup and explicitly checks the reveal timer; fresh focused tests 2/2 and the full App suite 137/137 passed locally. That test-only commit retained thec1cae6aproduct sources.AO review follow-up
62cad99: batch Nemotron transcription now feeds recordings longer than nine seconds in bounded chunks, waiting for consumption without changing non-blocking live capture. Fresh Release suite 532/532 and build passed; a new separate self-contained win-x64 package was published from this commit and passed the release package check. The earlierc1cae6aEXE predates this product change.CI follow-up
5463dca: serialize the blocking native fake test classes to avoid thread-pool starvation in the CI runner; Speech tests 123/123 passed locally. Production sources remain identical to the62cad99package.Real GGUF/native model behavior and the full native HUD/reduced-motion matrix remain to be checked on hardware. Native cancellation does not interrupt an already running decode; timeouts quarantine the instance until actual cleanup and require restart. EOS does not prove semantic completeness.