Release 2.5.0#33
Merged
Merged
Conversation
Add OpenAI-style LAN STT support with optional Keychain bearer token, UI configuration, history filters, and onboarding. Pause hosted AI polish while offline so local transcription can keep pasting raw text.
Add public synthetic WAV clips and a source transcript for 1, 2, 3, and 6 minute STT checks. Add a benchmark script that prints timing and transcript preview without saving private results.
Document the Local AI Server engine, public fixture exception, security handling, and release notes. Keep public AGENTS notes compact and free of private runtime details.
Save active recordings cancelled with Esc as recoverable history audio. Clear cancelled failure metadata after successful history re-transcription. Refresh public docs for recovery behavior and repo hygiene.
Apply conservative vocabulary recognition corrections before AI polish so canonical technical terms survive with polish disabled. Add public English and Spanish fixtures plus Local AI, Deepgram, and ElevenLabs batch benchmark tooling.
Warm the input node with the queried hardware format so launch preflight does not trip on stale AVAudioEngine defaults.
Group longform and technical transcription fixtures into subfolders. Update benchmark examples and docs to use the new fixture paths.
Replaces synthetic Spanish STT fixtures with a natural public Spanish benchmark. Expands conservative vocabulary corrections for technical dictation. Keeps cloud/local benchmark scripts aligned with the app correction layer.
Prevents stale ElevenLabs realtime audio sender work from mutating a new session. Corrects git commit and git push dictation when users save split vocabulary terms.
Stores pending realtime audio bytes outside Data to avoid removeAll traps during final commit. Keeps history replay stable for ElevenLabs realtime retries.
Add Local Server polish support with optional API keys and adaptive correction memory. Surface learned suggestions in vocabulary settings and cover the flow with replay tooling/tests. Keep local replay defaults public-safe and exclude ignored private files from tree secret scans.
Add local AI polish memory for ranked terms and reviewable correction suggestions. Support local polish providers with longer timeout budgets and safer fidelity checks. Mark AI-suggested corrections clearly in the Vocabulary UI.
Scope AI polish settings per provider and keep local-server auth optional. Redact provider setup failures and document correction-only memory rules.
Remove CLAUDE.md from the ignored local docs list. Track the lightweight pointer to AGENTS.md for Claude users.
Treat invalid mixed-separator numeric fragments as unreliable fidelity anchors. Keep valid formatted numbers protected and align the history replay guard.
Detect reviewable correction suggestions from accepted polish output. Mark accepted AI suggestions in the vocabulary UI and extend replay coverage.
Allow minor numeric pruning in long same-language narrative dictations while preserving hard drift checks. Keep short numeric, URL, email, vocabulary, and identifier safeguards strict. Update replay tooling, docs, and regression coverage.
Limit fidelity checks to protected tokens and retry polish up to three attempts. Update replay logic, tests, localized copy, and agent notes for the relaxed guard.
Move personal workflow detail out of public agent notes. Ignore repo-local skills while keeping AGENTS.md concise and public-safe.
Treat polish transcripts as inert text and reject assistant-style responses. Use the shared output-language picker for translation profiles without forcing English.
Move Unreleased changes into the 2.5.0 changelog section. Bump app version/build and ignore the local polish benchmark helper.
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.
Summary
Checks