324 lines
12 KiB
Markdown
Executable File
324 lines
12 KiB
Markdown
Executable File
# Changelog
|
||
|
||
## 0.8.4
|
||
|
||
### Patch Changes
|
||
|
||
- 8739929: Overhaul audio engine lifecycle: recover recordings across device/route changes instead of silently capturing nothing (#251, #252, #218, #226), rebuild the warm capture engine immediately when devices change while idle instead of leaving a stale engine running (#209), suspend the warm microphone while the screen is locked or asleep and rearm on wake, stop the sound-effects engine when idle (#209), and fix a crash on quit caused by racy audio teardown (#245)
|
||
- 2c5d8ae: Add Whisper Large v3 Turbo (632MB) to the model library: near large-v3 accuracy at several times the speed (#185). Whisper models are now labeled by their real sizes (Tiny, Base) instead of Small/Medium.
|
||
- 9c597cb: Prevent Hex from crashing on first launch before hotkey permissions are granted (#254)
|
||
|
||
## 0.8.3
|
||
|
||
### Patch Changes
|
||
|
||
- 224822d: Fix hotkeys dying after sleep or when macOS permission checks go stale: key events arriving at the tap now self-heal the Input Monitoring state instead of being dropped, the tap survives stale permission denials, and it is recreated on wake from sleep (#250)
|
||
|
||
## 0.8.2
|
||
|
||
### Patch Changes
|
||
|
||
- eae358b: Never silently clear the selected model when a scan misfires, and route hotkey presses to model setup instead of transcribing with no model
|
||
- eae358b: Redesign the model library: visible Download buttons, one-click model switching, and a menu to remove downloaded models or show them in Finder
|
||
|
||
## 0.8.1
|
||
|
||
### Patch Changes
|
||
|
||
- cd971a3: Reuse existing Parakeet downloads after upgrading FluidAudio
|
||
|
||
## 0.8.0
|
||
|
||
### Minor Changes
|
||
|
||
- 276da69: Add lowercase and punctuation-removal paste transforms (#253)
|
||
|
||
### Patch Changes
|
||
|
||
- 10e75b9: Simplify output transforms with clearer word rules, preview, and formatting controls.
|
||
- 8e6e041: Improve update and changelog controls in Settings.
|
||
- 71878b7: Defer Fast Mode capture rebuilds until audio route changes settle
|
||
- db31b0a: Move transcription model selection into a focused model library with automatic downloads on selection.
|
||
- c00a91d: Refresh microphones reliably, harden recording cleanup, and polish the settings overlay behavior.
|
||
- db31b0a: Handle interrupted recording stops explicitly and ignore stale Fast Mode callbacks.
|
||
- 2b5fb15: Recognize downloaded Parakeet models immediately after restarting Hex.
|
||
- 8e6e041: Preserve literal dollar signs in word-remapping replacements.
|
||
|
||
## 0.7.6
|
||
|
||
### Patch Changes
|
||
|
||
- f4cecbb: Show in Finder reveals Parakeet caches at the correct path (#205)
|
||
- f4cecbb: Stay in the menu bar when launched at login (#222)
|
||
|
||
## 0.7.5
|
||
|
||
### Patch Changes
|
||
|
||
- b5a19b4: Fix word remapping deletion crashes (#207)
|
||
- b5a19b4: Enable Super Fast Mode by default for new users
|
||
- b5a19b4: Restore macOS Sonoma compatibility (#215)
|
||
|
||
## 0.7.4
|
||
|
||
### Patch Changes
|
||
|
||
- 5a4af9b: Fix silent recordings from multichannel input devices during calls (#204)
|
||
- b78f049: Stop priming the sound-effects audio engine when sound effects are disabled so Hex avoids unnecessary background audio activity and sleep assertions (#200).
|
||
|
||
## 0.7.3
|
||
|
||
### Patch Changes
|
||
|
||
- 7340d1e: Restore double-tap lock audio capture (#193)
|
||
|
||
## 0.7.2
|
||
|
||
### Patch Changes
|
||
|
||
- 55249a6: Keep the ends of recordings from getting clipped in super fast mode.
|
||
- d9e40cc: Use the capture engine for normal recordings to reduce startup drift
|
||
- d9e40cc: Keep the microphone picker visible and refresh it when audio devices change
|
||
|
||
## 0.7.1
|
||
|
||
### Patch Changes
|
||
|
||
- ed69836: Suppress startup windows when Hex launches as a hidden login item (#146)
|
||
|
||
## 0.7.0
|
||
|
||
### Minor Changes
|
||
|
||
- c5d5162: Add Super Fast mode to keep the mic warm and prepend a short in-memory buffer
|
||
|
||
## 0.6.10
|
||
|
||
### Patch Changes
|
||
|
||
- c018c40: Add setting to disable double-tap lock for hands-free recording
|
||
- 7af7cd9: Update dependencies: TCA 1.23, Sparkle 2.8, swift-dependencies 1.11
|
||
|
||
## 0.6.9
|
||
|
||
### Patch Changes
|
||
|
||
- 74893ab: Support escape sequences (\n, \t, \\) in word remappings for newlines, tabs, and literal backslashes (#140)
|
||
|
||
## 0.6.8
|
||
|
||
### Patch Changes
|
||
|
||
- e2000d8: Fix Icon Composer app icon not displaying (#148)
|
||
- 75bc323: Update macOS Tahoe app icon (#145)
|
||
|
||
## 0.6.7
|
||
|
||
### Patch Changes
|
||
|
||
- cc99650: Prepare release metadata for 0.6.6
|
||
|
||
## 0.6.6
|
||
|
||
### Patch Changes
|
||
|
||
- 3b6c966: Improve transcript modifications layout and remove log export settings
|
||
- 3b6c966: Add opt-in regex word removals for transcripts (#121)
|
||
|
||
## 0.6.5
|
||
|
||
### Patch Changes
|
||
|
||
- 140c205: Fix Sparkle auto-update for sandboxed app by adding required XPC entitlements and SUEnableInstallerLauncherService. Users on 0.6.3 will need to manually download this update.
|
||
|
||
## 0.6.4
|
||
|
||
### Patch Changes
|
||
|
||
- c00f79e: Reduce code duplication: add ModelPatternMatcher, FileManager helpers, settingsCaption style, notification constants, and Core Audio helper
|
||
- 658a755: Fix silent recordings caused by device-level microphone mute - automatically detects and fixes muted input devices before recording
|
||
|
||
## 0.6.3
|
||
|
||
### Patch Changes
|
||
|
||
- b4c54ce: Fix microphone priming and media pause races
|
||
- 5217d3f: Add word remappings and remove LLM UI (#000)
|
||
- 4d38708: Add persistent MCP config editing for Claude Code modes
|
||
- bbd0b80: Show system default mic name in picker
|
||
- bbd0b80: Fix Parakeet polling cleanup and organize paste flow
|
||
- 3413d68: Rename Transformations tab to Modes
|
||
- 4d38708: Fix microphone freezing and speech cutoff when using custom microphone. Only switch input device when actually needed, re-prime recorder after device changes, and add cleanup on app termination.
|
||
|
||
## 0.6.2
|
||
|
||
### Patch Changes
|
||
|
||
- 7e325ad: Fix Sequoia hotkey deadlock by removing Input Monitoring guard that prevented CGEventTap creation. Tap creation triggers permission prompt naturally. Re-add 'force quit Hex now' voice escape hatch from v0.5.8 (#122 #124)
|
||
- 7e325ad: Add missing-model callout and focus settings when transcription starts without a model
|
||
|
||
## 0.6.0
|
||
|
||
### Patch Changes
|
||
|
||
- 3bf2fb0: Fix voice prefix matching with punctuation - now strips punctuation (.,;:!?) when matching prefixes
|
||
|
||
## 0.5.13
|
||
|
||
### Patch Changes
|
||
|
||
- 083513c: Add comprehensive documentation to HotKeyProcessor and extract magic numbers into named constants (HexCoreConstants)
|
||
|
||
## 0.5.12
|
||
|
||
### Patch Changes
|
||
|
||
- 471310c: Fix Input Monitoring permission enforcement for hotkey reliability
|
||
|
||
## 0.5.11
|
||
|
||
### Patch Changes
|
||
|
||
- 1deda2a: Route Advanced → Export Logs through the new swift-log diagnostics file so Sequoia permission bugs (#122 #124) can be diagnosed locally without relying on macOS unified logs.
|
||
|
||
## 0.5.10
|
||
|
||
### Patch Changes
|
||
|
||
- 3560bdb: Keep hotkeys alive on Sequoia and add voice force-quit plus Advanced log export (#122 #124)
|
||
|
||
## 0.5.9
|
||
|
||
### Patch Changes
|
||
|
||
- 6c2f1bd: Add comprehensive permissions logging for improved debugging and log export support
|
||
|
||
## 0.5.8
|
||
|
||
### Patch Changes
|
||
|
||
- 03b81c7: Let the hotkey tap start even when Input Monitoring is missing so Sequoia users get prompts again, while keeping the accessibility watchdog (#122 #124). Add a spoken “force quit Hex now” escape hatch in case permissions clobber input.
|
||
|
||
## 0.5.7
|
||
|
||
### Patch Changes
|
||
|
||
- 539b0a4: Pad sub-1.5s Parakeet recordings so FluidAudio accepts them
|
||
|
||
## 0.5.6
|
||
|
||
### Patch Changes
|
||
|
||
- a1eb1d0: Restore hotkeys when Input Monitoring permission is missing (#122, #124)
|
||
- 1ee452a: Add non-interactive changeset creation for AI agents
|
||
- 68475f5: Fix clipboard restore timing for slow apps – increased delay from 100ms to 500ms to prevent paste failures in apps that read clipboard asynchronously
|
||
|
||
## 0.5.5
|
||
|
||
### Patch Changes
|
||
|
||
- 0045f28: Fix recording chime latency by switching to AVAudioEngine with pre-loaded buffers
|
||
- 7f6c5db: Actually request macOS Input Monitoring permission when installing the key event tap so Sequoia users can record hotkeys again (#122, #124).
|
||
|
||
## 0.5.4
|
||
|
||
### Patch Changes
|
||
|
||
- Fix hotkey monitoring on macOS Sequoia 15.7.1 by properly handling Input Monitoring permissions (#122, #124)
|
||
|
||
## 0.5.3
|
||
|
||
### Patch Changes
|
||
|
||
- Fix Sparkle update delivery by regenerating appcast with correct bundle versions and updating release tooling to prevent duplicate CFBundleVersion issues
|
||
|
||
## 0.5.2
|
||
|
||
### Patch Changes
|
||
|
||
- Fix Sparkle update delivery by regenerating appcast with correct bundle versions and updating release tooling to prevent duplicate CFBundleVersion issues
|
||
|
||
## 0.5.1
|
||
|
||
### Patch Changes
|
||
|
||
- Fix Sparkle appcast generation by cleaning duplicate bundle versions and updating release pipeline to preserve last 3 DMGs for delta generation
|
||
|
||
## 0.5.0
|
||
|
||
### Minor Changes
|
||
|
||
- 049592c: Add support for multiple Parakeet model variants: choose between English-only (v2) or multilingual (v3) based on your transcription needs.
|
||
|
||
### Patch Changes
|
||
|
||
- aca9ad5: Fix microphone access retained when recording canceled with ESC (#117)
|
||
- 049592c: Polish paste-last-transcript hotkey UI with improved layout and clearer instructions.
|
||
- 049592c: Improve hotkey reliability with accessibility trust monitoring and automatic recovery from tap disabled events (#89, #81, #87).
|
||
- 049592c: Improve media pausing reliability by using MediaRemote API instead of simulated keyboard events.
|
||
- 049592c: Fix menu bar rendering issue where items appeared as single embedded view instead of separate clickable menu items.
|
||
- 1b9bd52: Optimize recorder startup by keeping AVAudioRecorder primed between sessions, eliminating ~500ms latency for successive recordings
|
||
- 55fb4f8: Add a sound effects volume slider beneath the toggle so users can fine-tune feedback relative to the existing 20% baseline, keeping 100% at the legacy loudness (#000).
|
||
|
||
## 0.4.0
|
||
|
||
### Minor Changes
|
||
|
||
- e50478d: Add Parakeet TDT v3 plus the first-run model bootstrap, faster recording pipeline, and solid Fn/modifier hotkeys so the next release captures all of the recent feature work (#71, #97, #113, #89, #81, #87).
|
||
|
||
### Patch Changes
|
||
|
||
- ea42b5b: Move `HexSettings` + `RecordingAudioBehavior` into HexCore and add fixtures/tests so we can migrate historic settings blobs safely before shipping new media-ducking options.
|
||
- e50478d: Adopt Changesets for SemVer + changelog management, wire release.ts to fail without pending fragments, and sync the aggregated release notes into the bundled changelog + GitHub releases.
|
||
- 2fbbe7a: Wait for NSPasteboard changeCount to advance before pasting so panel apps always receive the latest transcript (#69, #42).
|
||
|
||
All notable changes to Hex are documented in this file. The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/) and the project follows [Semantic Versioning](https://semver.org/).
|
||
|
||
## Unreleased
|
||
|
||
### Added
|
||
|
||
- Added NVIDIA Parakeet TDT v3 support with a redesigned model manager so you can swap between Parakeet and curated Whisper variants without juggling files (#71).
|
||
- Added first-run model bootstrap: Hex now automatically downloads the recommended model, shows progress/cancel controls, and prevents transcription from starting until a model is ready (#97).
|
||
- Added a global hotkey to paste the last transcript plus contextual actions to cancel or delete model downloads directly from Settings, making recovery workflows faster.
|
||
|
||
### Improved
|
||
|
||
- Model downloads now surface the failing host/domain in their error message so DNS or network issues are easier to debug (#112).
|
||
- Recording starts ~200–700 ms faster: start sounds play immediately, media pausing runs off the main actor, and transcription errors skip the extra cancel chime for less audio clutter (#113).
|
||
- The transcription overlay tracks the active window so UI hints stay anchored to whichever app currently has focus.
|
||
- HexSettings now lives inside HexCore with fixture-based migration tests, giving us a single source of truth for future settings changes.
|
||
|
||
### Fixed
|
||
|
||
- Printable-key hotkeys (for example `⌘+'`) can now trigger short recordings just like modifier-only chords, so quick phrases aren’t discarded anymore (#113).
|
||
- Fn and other modifier-only hotkeys respect left/right side selection, ignore phantom arrow events, and stop firing when combined with other keys, resolving long-standing regressions (#89, #81, #87).
|
||
- Paste reliability: Hex now waits for the clipboard write to commit before firing ⌘V, so panel apps like Alfred, Raycast, and IntelliBar always receive the latest transcript instead of the previous clipboard contents (#69, #42).
|
||
|
||
## 1.4
|
||
|
||
### Patch Changes
|
||
|
||
- Bump version for stable release
|
||
|
||
## 0.1.33
|
||
|
||
### Added
|
||
|
||
- Add copy to clipboard option
|
||
- Add support for complete keyboard shortcuts
|
||
- Add indication for model prewarming
|
||
|
||
### Fixed
|
||
|
||
- Fix issue with Hex showing in Mission Control and Cmd+Tab
|
||
- Improve paste behavior when text input fails
|
||
- Rework audio pausing logic to make it more reliable
|
||
|
||
## 0.1.26
|
||
|
||
### Added
|
||
|
||
- Add changelog
|
||
- Add option to set minimum record time
|