Mityu v1.1.0

What's Changed

• feat(release): versioned download name + release DX (no per-release manual step) by @aydogandagidir in #6
• feat(learning): land ADR-0030 local learning system (through C2 glue) by @aydogandagidir in #7
• feat(landing): learning-loop hero animation + drop redundant static mock by @aydogandagidir in #8
• feat(landing): add 'learning loop' animation before pricing by @aydogandagidir in #9
• feat(landing): scroll-grow entrance for the learning-loop section by @aydogandagidir in #10
• feat(eval-harness): prepare the A5 gate to measure two domains safely by @aydogandagidir in #11
• test(transparency): make the Art. 50 AI marking a gate instead of a convention by @aydogandagidir in #12
• feat(inference): add the Tier-0 capability probe and local-backend seam by @aydogandagidir in #13
• chore: remove stale build logs and correct two ADRs the repo had outgrown by @aydogandagidir in #14
• feat(legal): A5 KVKK analysis, draft notices, and a recordable withdrawal path by @aydogandagidir in #15
• feat(ask): answer questions about one meeting, from that meeting only by @aydogandagidir in #16
• feat(diarization): the design gate, and the schema step it allows by @aydogandagidir in #17
• feat(eval): DER harness — the instrument that must exist before a diarization engine by @aydogandagidir in #18
• fix(privacy): meeting deletion left the human's rewritten text behind by @aydogandagidir in #19
• feat(eval): validate the DER scorer against NIST md-eval on 216 real recordings by @aydogandagidir in #20
• docs(adr): ADR-0035 — diarization engine, decided by measurement not prediction by @aydogandagidir in #21
• fix: audit round 2 — ADR-0035's central claim was wrong, and the ledger was 10 PRs stale by @aydogandagidir in #22
• feat(ci): verify the sherpa-onnx native archive before cargo can download it (H2) by @aydogandagidir in #23
• feat(diarize): the diarization engine, out of process (H3a) by @aydogandagidir in #24
• feat(diarize): pinned, verified model resolution for the sidecar (H3b-1) by @aydogandagidir in #25
• feat(diarize): wire the sidecar into the app (H3b-2) by @aydogandagidir in #26
• chore: a UI screenshot tool that refuses to pass on a blank page by @aydogandagidir in #27
• feat(diarize): speaker labels and talk time in the UI (H6) by @aydogandagidir in #28
• chore(ui): replace the phantom type-scale classes with Tailwind's real ones by @aydogandagidir in #29
• feat(diarize): model notices, and a GPL-3.0 finding that blocks the other half by @aydogandagidir in #30
• fix(diarize): stop shipping GPL-3.0 espeak-ng in the sidecar (H10) by @aydogandagidir in #31
• feat(diarize): third-party notice for the sidecar, now that it is GPL-free (H7) by @aydogandagidir in #32
• chore(release): 1.0.4 → 1.1.0 by @aydogandagidir in #34

Full Changelog: v1.0.4...v1.1.0

Mityu v1.0.4

What's Changed

• Mityu v1.0.4 release candidate by @aydogandagidir in #4
• fix(release): release-summary needs contents:write to read the unpublished draft by @aydogandagidir in #5

Full Changelog: v1.0.3...v1.0.4

Mityu 1.0.3

Mityu 1.0.3

Onboarding tour fix. The first-run product tour now works correctly end to end:

• Step 2 (the source-linked, human-approved summary) is shown reliably — previously it could be skipped when the summary panel wasn't immediately visible.
• Back now works throughout the tour.
• If a step's target isn't on screen, the step is shown centered instead of being skipped, so the walkthrough is never interrupted.

Everything stays local-first: capture, transcription, and summarization run on your machine.

Install

• Windows: Mityu_1.0.3_x64-setup.exe (recommended) or the .msi.
• Existing 1.0.x installs update automatically.

The installer isn't code-signed with an Authenticode certificate yet, so Windows SmartScreen may warn on first run — click "More info → Run anyway."

Mityu 1.0.2

Mityu 1.0.2

First-run onboarding tour. New users now get a short guided walkthrough on a pre-loaded sample meeting — see how Mityu turns a recording into a searchable transcript and a source-linked, human-approved summary, all on your device. Replay it anytime from Settings → Replay product tour.

Everything stays local-first: capture, transcription, and summarization run on your machine.

Install

• Windows: Mityu_1.0.2_x64-setup.exe (recommended) or the .msi.
• Existing 1.0.x installs update automatically.

The installer isn't code-signed with an Authenticode certificate yet, so Windows SmartScreen may warn on first run — click "More info → Run anyway."

Mityu 1.0.1

Mityu 1.0.1

Small but meaningful polish on the 1.0 launch.

Fixed / new

• Correct version everywhere — the in-app version now reads the real app version (the sidebar previously showed an old number).
• A sample meeting on first launch — new installs open with one high-quality, source-linked example (transcript → summary → action items) so you can see what Mityu produces before recording your own. It's clearly marked as a sample and you can delete it anytime.

Existing 1.0.0 installs update automatically. Windows may show an "unknown publisher" SmartScreen prompt — More info → Run anyway (a code-signing certificate is coming shortly).

Mityu 1.0.0

Mityu 1.0.0 — first public release

On-device meeting intelligence for Windows. Mityu records, transcribes, and reports on your meetings entirely on your machine — no bots join your calls, nothing is uploaded, no account required.

Highlights

• On-device transcription — Whisper large-v3 or Parakeet, fully offline
• Source-linked summaries — every AI-generated line ties back to its exact transcript moment
• Human-approved — every AI draft is labeled "AI-generated · review required" and reviewed by you before it's final
• Encrypted at rest (SQLCipher)
• Any meeting app — captures your mic + system audio, so it works with Zoom, Google Meet, or Teams without a bot joining
• BYOK for cloud summaries (Ollama, OpenAI, Anthropic, Groq, OpenRouter) — optional; local by default

Install (Windows 10/11, 64-bit)

Download Mityu_1.0.0_x64-setup.exe and run it. Windows may show an "unknown publisher" SmartScreen prompt — click More info → Run anyway. (A code-signing certificate is coming shortly.)

License

14-day free trial (no card). Mityu Pro — one-time $79, activate on up to 2 devices, 1 year of updates. macOS support is in development.

FFmpeg dependency binaries (LGPL, v8.1)

Third-party FFmpeg sidecar binaries bundled by the Mityu build (frontend/src-tauri/build/ffmpeg.rs), self-hosted here instead of depending on a third party's servers at every build/install.

Why LGPL instead of the previous GPL build: the binary Mityu previously bundled (mirrored from Zackriya-Solutions/ffmpeg-binaries, itself from gyan.dev) was compiled with --enable-gpl --enable-version3 (includes libx264/libx265). Mityu only ever invokes FFmpeg for raw-audio -> AAC/MP4 transcoding (-c:a aac, no -c:v) and never uses any GPL-only codec, so these are LGPL-only static builds instead — same functionality for Mityu's actual use, without the GPL source-redistribution obligation that would otherwise fall on this project as a commercial distributor.

Source: unmodified re-uploads of the pinned n8.1 static LGPL builds from BtbN/FFmpeg-Builds (a widely-used, actively maintained community CI build of upstream FFmpeg). Original release: https://github.com/BtbN/FFmpeg-Builds/releases/tag/latest — mirrored here so ffmpeg.rs doesn't depend on a :latest tag that can change out from under a pinned build.

Coverage: Windows x64 and Linux x64/arm64 only. macOS (Intel + Apple Silicon) is intentionally NOT included yet — no equivalently-licensed (LGPL-only) static macOS build was identified during this pass; the macOS targets still use the previous source until that's resolved separately.

Verification note: archive layout and file types were checked (valid PE32+ / correct bin/ffmpeg[.exe] paths matching what ffmpeg.rs already expects), but the binaries were not execution-tested end-to-end as part of this release — run one real cargo build (after clearing any cached frontend/src-tauri/binaries/ffmpeg-* file) before shipping a signed release.

Docs: see docs/RELEASE_CHECKLIST.md #3 and docs/GA_READINESS.md.