v0.4.0

Cassini is now called Cassini talk. Your Library stays in Music/Cassini.

• Recording starts reliably when the computer is silent: the other people's
  side waits for the first sound instead of failing the start
• When a microphone or the computer's sound cannot open, Cassini says why and
  keeps recording what did open; if nothing opened, no empty meeting is left
  behind. Quit no longer waits for a microphone that never delivers audio
• Silent sources are saved as exact silence, so writing down and speaker
  separation skip them instead of reading noise
• Every part of Cassini places each track at the same start, so playback,
  transcripts and speakers line up; words stay inside the time a person was
  present, and a turn never spans a gap. Cassini for Mac passes the same check
• Browser meetings: set up the companion from Cassini, connect in the
  background, keep recording through page reloads, record automatically only
  once a call has really started, and record with no window open. Browser and
  computer audio share one clock, even when the system clock is adjusted
  during a meeting. The companion is now version 0.2.13: a long silent stretch
  no longer slows a browser recording's clock
• GNOME and Mac pass the same end-to-end tests: a recording followed live
  while it is in progress (shared test 2), and the same three outcome files
  (shared test 8)
• While recording: a live spectrogram for each side, Recording details that
  show what is being captured, a warning when muted output would silence the
  computer's sound, and a test recording to check your setup
• Library: move meetings to Trash with five seconds to undo, rename with F2,
  and keep each saved meeting as a portable Cassini file
• Writing down shows its progress before the first words, reads a track again
  without voice detection when it finds too few words there, and names the app
  version when it downloads models
• Translated into Italian, French, Spanish and Portuguese (Brazil and Portugal)
• Many capture fixes: interrupted and re-anchored inputs, frozen browser
  connections, and a live spectrum that no longer stops after the window
  stalls

What's Changed

• Move saved meetings to Trash with five-second Undo by @silviot in #98
• Wait for inbound call evidence before automatic recording by @silviot in #97
• Pin reviewed call evidence and gate shared auto-start vectors by @silviot in #99
• Keep row menu actions connected until activation finishes by @silviot in #100
• Prepare saved voices for older meetings and explain missing models by @silviot in #101
• Remove timestamped notes from the Recording page by @silviot in #102
• Apply the recording title caret fix to main by @silviot in #106
• Make custom recording and playback buttons accessible by @silviot in #105
• Connect to the browser in the background by @silviot in #107
• Show writing progress before the first transcript words by @silviot in #109
• Ask before enabling the global recording shortcut by @silviot in #108
• Balance the empty Library heading by @silviot in #111
• Keep the Flatpak Library when XDG Music is unavailable by @silviot in #104
• Restore Keyboard Shortcuts in the Library menu by @silviot in #110
• Keep a new recording visible when an older capture finishes by @silviot in #112
• Make the microphone tone gate tolerate clock phase corrections by @silviot in #113
• Keep exports outside Flatpak’s temporary home and show success by @silviot in #114
• D35: Rename saved meetings from either menu or F2 by @silviot in #115
• D35: Cancel on Escape and explicit background recording permission by @silviot in #116
• D37: align GNOME consent, browser states and saved meeting journeys by @silviot in #117
• Prevent the delayed GTK warning during Rename saves by @silviot in #119
• Show Recording details with live capture and writing-down facts by @silviot in #96
• Fix playback speed and harden companion setup responsiveness by @silviot in #122
• Add the browser companion from Cassini (D39) by @silviot in #123
• Restore automatic names and clean displayed quotation marks by @silviot in #118
• Restart installed updates safely and heal dead browser registrations by @silviot in #127
• Keep browser service tests out of the user's library by @silviot in #126
• Library v2 step 2: every track reader accepts WAV and compacted Opus spans (S-12 §5) by @silviot in #128
• Library v2 step 3a: names, journal, durable publish, mixdown and finalize (S-12) by @silviot in #130
• Library v2: direct regression tests for #130's commit guards by @silviot in #131
• Keep the browser control link open and keep stopped calls stopped by @silviot in #125
• D45: rename Cassini talk and migrate existing installations by @silviot in #129
• Finish Quit when a microphone never produces audio by @silviot in #132
• Keep browser recording batches inside the app request timeout by @silviot in #134
• Background Apps for windowless automatic recording (D41) by @silviot in #124
• Keep browser recordings through reloads and save how they ended by @silviot in #121
• Keep browser meeting identities and queued work across reconnects by @silviot in #133
• Preserve capture identity and stopping in tab-scoped browser status by @silviot in #135
• Bump GitHub Actions to current majors by @silviot in #136
• Warn when muted output silences a system recording by @silviot in #138
• Publish stopped captures as Library Cassini files by @silviot in #137
• Complete the clean Cassini talk rename (D45.2 / D51) by @silviot in #140
• ci: align rolling openSUSE container snapshots by @silviot in #141
• Draft: recording spectrogram (D36.1) by @silviot in #139
• test: synchronize broker meter delivery and fixture clocks by @silviot in #142
• Adopt D52 multistream capture accounting and lifecycle by @silviot in #143
• Draft source diagnostics in portable meeting files by @silviot in #144
• D50: Localize Cassini talk in six languages by @silviot in #145
• Bound recognition timestamps to each source audio extent by @silviot in #146
• Add producer drop counts to the recorded span instead of replacing them by @silviot in #147
• Never expect an announced page microphone when Me is recorded natively by @silviot in #150
• Record gap ranges and derive presence without reading PCM (D52.2, G6) by @silviot in #148
• Raise the pytest suite deadline to 360 s by @silviot in #152
• Bump actions/setup-node from 6 to 7 by @dependabot[bot] in #154
• Pin the test-Library isolation protocol (#30) by @silviot in #153
• Close frozen browser transports on deadlines instead of waiting for EOF (G7, part A) by @silviot in #151
• Decode a track without VAD when its VAD pass finds too few words (G10) by @silviot in #149
• Record when a write-down assumed a track's start at 0 (D55) by @silviot in #158
• Compute ActivityMeter.inside() once per frame count by @silviot in #159
• Name the app and its version in every model download (D54) by @silviot in #157
• G12: tie voice names to the diarization input (voiceSource) by @silviot in #160
• Wait for the picker paste before Open in companion-instal...

v0.3.0

What's Changed

• Fix Library contrast, transcript routing and playback lifecycle by @silviot in #14
• UX-play: mix and seek recording-folder audio from the Library by @silviot in #10
• Keep live recordings and Library previews responsive by @silviot in #15
• Spike: bundle pinned Whistle and prove offline Mac parity by @silviot in #16
• UX3: read meetings in the Library with timed playback by @silviot in #17
• UX5: record native audio straight into the Library by @silviot in #18
• UX13: write meetings down with the shared offline pipeline by @silviot in #19
• Fix resumable shutdown and stop-and-save exit by @silviot in #20
• Stream long received audio through bounded child PCM blocks by @silviot in #21
• Protect Library write boundaries and retain validated speaker identity by @silviot in #22
• Bound transcript publication and keep live reading stable by @silviot in #23
• Finish close when capture ends behind its confirmation by @silviot in #24
• Ship offline writing down (Whistle) in every package by @silviot in #25
• Preserve Meeting playback across background Library refreshes by @silviot in #27
• Fix capture close holds, finalization and saved queue intent by @silviot in #26
• Keep Stop responsive while independent audio sources finish by @silviot in #29
• Keep partial transcript intact while Try again catches up by @silviot in #28
• Add the Recording page and controls for hidden capture by @silviot in #30
• Ship a private, pinned ONNX Runtime instead of the distribution's by @silviot in #31
• Import received meetings and add Cassini/Markdown export by @silviot in #32
• Search every meeting with phrase matches and D14 filters by @silviot in #33
• Add verified resumable model download core by @DrMeyers in #34
• Keep model removal inside the configured store by @DrMeyers in #36
• Bring the README up to date with what Cassini does today by @silviot in #37
• Complete native Preferences and durable Library relocation by @DrMeyers in #35
• Preserve capture failure evidence before pipeline teardown by @DrMeyers in #39
• Document GPL licence for vendored browser files by @DrMeyers in #38
• Fix hidden capture completion and window lifetime by @DrMeyers in #40
• Protect model storage and HTTP resume checkpoints by @DrMeyers in #43
• Fail export on streaming read errors at EOS by @DrMeyers in #42
• Preserve detailed failures in the native GTK capture probe by @DrMeyers in #45
• Keep Search cache transactions and playback consistent by @DrMeyers in #44
• Preserve recording notes and pending title edits by @DrMeyers in #46
• Validate received imports before Library publication by @DrMeyers in #41
• Keep the last browser capture observer owned until save by @DrMeyers in #48
• Use current meeting metadata in Export without GTK file IO by @DrMeyers in #47
• Use a monotonic clock for independent Pulse capture sources by @DrMeyers in #49
• Add pinned Parakeet Ultra worker and offline golden proofs by @DrMeyers in #50
• Download Parakeet in Preferences and choose models per job by @DrMeyers in #51
• Export recovered recordings with honest dates and offsets by @DrMeyers in #52
• Reuse green main release packages and retain full CI coverage by @DrMeyers in #53
• Orchestrate CI areas and share verified build profiles by @DrMeyers in #54
• Ship the pinned Parakeet runtime in Linux packages by @silviot in #55
• Run engine and Library gates for imported source changes by @silviot in #56
• Keep pinned note times intact through undo and line edits by @silviot in #57
• Keep capture ownership when auxiliary windows close by @silviot in #59
• Recover damaged model installs and completed download state by @silviot in #58
• Render Settings boards with installed speech runtimes by @silviot in #61
• Clean review docs and omit caches from test image context by @silviot in #62
• Keep Library row timers on the native capture clock by @silviot in #63
• Keep received source symlinks visible to no-follow opens by @silviot in #60
• Record browser meetings through the Library and companion setup by @silviot in #64
• Declare libzstd for model downloads in every native package by @silviot in #66
• Add What is recorded controls and automatic system-sound fallback by @silviot in #65
• Enable the browser companion in Flatpak with isolated host relays by @silviot in #67
• Keep native recording intact when a source change fails by @silviot in #70
• Keep Flatpak startup responsive and retry the browser bridge by @silviot in #71
• Constrain relay networking and cover board test dependencies by @silviot in #72
• Stop ended browser calls and bound fallback recovery by @silviot in #68
• Keep browser recording intact when a source change fails by @silviot in #69
• UX15: tell speakers apart with optional Nemotron by @silviot in #73
• UX16: name speakers, merge voices and undo merges by @silviot in #74
• Bring the README up to date with Waves 5 to 7 by @silviot in #75
• Keep native recording clocks active without the browser bridge by @silviot in #76
• Keep cancelled diarization resumable without late transcript writes by @silviot in #77
• Bound fallback lifetime when a browser peer disappears before status by @silviot in #79
• Keep browser PCM flowing after rejected source changes by @silviot in #80
• Cancel pending native source changes when Stop is requested by @silviot in #81
• Isolate Flatpak host probes and stop retrying missing permission by @silviot in #82
• Keep cleared source speaker names through writing down by @silviot in #78
• Make model-control GUI readiness checks deterministic by @silviot in #83
• UX17: recognise and manage saved voices locally by @silviot in #84
• Preserve native input gaps without turning clock drift into silence by @silviot in #85
• Keep confirmed saved voices through Merge and Undo by @silviot in #87
• Preserve saved voice meeting links while the Library is unavailable by @silviot in #88
• Preserve voice associations and respect explicit naming intent by @silviot in #89
• Finish voice jobs atomically and recompute Undo after navigation by @silviot in #90
• Keep saved-voice dialogs, Search badges and sample cleanup accurate by @silviot in #91
• Describe saved voices in the README by @silviot in #92
• D...

v0.2.0

What's Changed

• W1-G3: isolated screenshot harness spike for the current GNOME window by @silviot in #5
• Wait for real audio readiness and sample counts in recording tests by @silviot in #7
• UX1a: derive a read-only Library model from capture and Received folders by @silviot in #6
• Complete multi-distro packaging and release 0.2.0 by @silviot in #8
• UX1b: Library shell, meeting previews and first run by @silviot in #9
• Polish Library live header and row text to shared R11 by @silviot in #11
• Fix post-merge review findings in packaging and the player by @silviot in #12
• Rewrite the README for people who use Cassini by @silviot in #13

New Contributors

• @silviot made their first contribution in #5

Full Changelog: v0.1.1...v0.2.0

v0.1.1

Full Changelog: https://github.com/codemyriad/cassini-gnome/commits/v0.1.1