This release adds 3 notable features for engineering teams evaluating rollout.
✓ No known CVEs patched in this version
Summary
AI summaryLong dictations finish substantially faster (from 4.519 s to 0.952 s) and multi-point feedback is formatted more clearly.
Changes in this release
| Type | Severity | Summary | CVE |
|---|---|---|---|
| Feature | Low |
Dictated issue reports, feedback, tasks, and requirements can automatically become numbered lists without explicit request. Dictated issue reports, feedback, tasks, and requirements can automatically become numbered lists without explicit request. Source: llm_adapter@2026-07-19 Confidence: high |
— |
| Feature | Low |
Literal list markers are preserved rather than rewritten by hidden post‑processing heuristics. Literal list markers are preserved rather than rewritten by hidden post‑processing heuristics. Source: llm_adapter@2026-07-19 Confidence: high |
— |
| Performance | Medium |
Dictation processing speed improves from 4.519s to 0.952s for a 70‑second input. Dictation processing speed improves from 4.519s to 0.952s for a 70‑second input. Source: llm_adapter@2026-07-19 Confidence: high |
— |
| Bugfix | Medium |
Streaming list continuity is now guarded across speech segments; numbering continues and invalid numbering falls back safely. Streaming list continuity is now guarded across speech segments; numbering continues and invalid numbering falls back safely. Source: llm_adapter@2026-07-19 Confidence: high |
— |
| Bugfix | Low |
Romanize no longer disables streaming cleanup for Latin‑script dictation; English remains on the fast during‑recording path while non‑Latin uses whole‑text romanization. Romanize no longer disables streaming cleanup for Latin‑script dictation; English remains on the fast during‑recording path while non‑Latin uses whole‑text romanization. Source: llm_adapter@2026-07-19 Confidence: high |
— |
Full changelog
Long dictations now finish substantially faster, and multi-point feedback is formatted more clearly.
What changed:
- Romanize no longer disables streaming cleanup for Latin-script dictation. English stays on the fast during-recording path, while non-Latin speech still uses the correct whole-text romanization path.
- Dictated issue reports, feedback, tasks, and requirements can become numbered lists even when you do not explicitly ask for a list.
- Streaming list continuity is guarded across speech segments. Items stay on separate lines, numbering continues across segments, and invalid numbering falls back to safe whole-text cleanup instead of publishing a broken list.
- Literal list markers are preserved rather than rewritten by a hidden post-processing heuristic.
Measured on an archived 70-second dictation, server stop-to-final improved from 4.519s to 0.952s.
Verified with 331 engine tests, 581 Swift self-checks, exact-model formatting benchmarks, Apple notarization, stapling, and Gatekeeper assessment.
Weekly OSS security release digest.
The CVE patches and breaking changes that affected production tools this week. One email, every Sunday.
No spam, unsubscribe anytime.
Share this release
About Velora
All releases →Beta — feedback welcome: [email protected]