Skip to content
CAI
Software that uses CAICheck a score

cjpais/Handy

55.6

Adequate · 14 November 2026

54.8k

lines of production code

Rust

with TypeScript

7

measurements over time

CAI band scale
CAI trend line
CAI lens gauges

What this system is

Features

A new footer component has been introduced to the application, providing users with a persistent bottom bar that displays the current application version and integrates a model selector alongside an update checker. The footer dynamically fetches the app version via Tauri's API, falling back to a default value if retrieval fails, and presents this information along with model selection controls in a styled layout.

src/components/footer · high confidence

Initial internationalization support with 26 languages and RTL handling

The application now supports 26 languages, including English, Simplified and Traditional Chinese, Spanish, French, German, Japanese, Korean, Vietnamese, Polish, Italian, Russian, Ukrainian, Portuguese, Czech, Turkish, Arabic, Hebrew, Swedish, Bulgarian, Dutch, Nepali, Hindi, Danish, Catalan, and Indonesian. This change introduces a new i18n infrastructure that auto-discovers translation files, syncs the user's preferred language from settings or system locale, and correctly handles Right-to-Left (RTL) text direction for Arabic and Hebrew.

src/i18n · high confidence

Introduce settings and OS detection hooks

New React hooks have been added to centralize application state and environment detection. The useSettings hook provides a unified interface for managing application settings, including audio device selection, post-processing configuration, and binding updates, while the useOsType hook detects the current operating system to enable platform-specific keyboard handling.

src/hooks · high confidence

Introduces Apple Intelligence integration on macOS

The application now supports Apple Intelligence as a post-processing provider on macOS. A new \apple\_intelligence.rs\ module bridges to Swift system APIs to check availability and process text with a system prompt, and the main library (\lib.rs\) conditionally loads this module on \aarch64\ Macs. This enables users to leverage the device's local AI capabilities for transcription refinement.

src-tauri/src · high confidence

Introduction of new model and settings state management stores

The application now uses dedicated Zustand stores (\modelStore.ts\ and \settingsStore.ts\) to manage core application state. The model store handles the lifecycle of AI models, including downloading, verifying, extracting, and selecting active models, while syncing progress and status with the backend. The settings store centralizes configuration management, providing a unified interface to load, update, and reset user preferences such as audio device selection, paste methods, theme, and post-processing options, ensuring platform-specific defaults are correctly applied.

src/stores · high confidence

New General Settings page with model and audio configuration

A new General Settings page has been introduced, consolidating configuration options into distinct groups. Users can now manage model-specific settings, including language selection and English translation toggles, which are dynamically shown based on the capabilities of the currently selected model. The audio section provides controls for microphone and output device selection, input channel configuration, volume adjustment, and audio feedback options, alongside shortcut management for transcription and cancellation actions.

src/components/settings/general · high confidence

New LLM-based post-processing configuration interface

The application now includes a dedicated settings page for configuring LLM-based post-processing. Users can select an API provider (including Apple Intelligence or custom endpoints), manage API keys and base URLs, and choose or create specific models. The interface also allows users to manage custom post-processing prompts, enabling them to create, select, and edit text templates that are sent to the selected model for processing.

src/components/settings/post-processing · high confidence

New UI components for configuring post-processing providers

This change introduces a new set of React components in the PostProcessingSettingsApi directory to manage post-processing configuration. It adds an ApiKeyField for secure API key input, a BaseUrlField for custom provider endpoints, a ProviderSelect to switch between available providers (including Apple Intelligence), and a ModelSelect that supports both selecting from a list and creating custom model names. These components are backed by a new hook, usePostProcessProviderState, which handles the state logic for fetching model lists, updating settings, and managing provider-specific constraints like Apple Intelligence availability checks.

src/components/settings/PostProcessingSettingsApi · high confidence

New Voice Activity Detection (VAD) toolkit with multiple backend support

The audio toolkit now includes a modular VAD system in \src-tauri/src/audio\_toolkit/vad\ that supports multiple detection backends, specifically an experimental Earshot VAD (pure Rust, 256-sample frames) and a Silero VAD (30ms frames). A \SmoothedVad\ wrapper provides hysteresis logic to prevent audio tail loss by managing pre-roll buffers, onset detection, and hangover tails, ensuring continuous speech capture. The module defines a common \VoiceActivityDetector\ trait and timing constants (prefill, hangover, onset) to standardize how audio frames are processed and emitted. Also in this area: 1 commit (0 fixes) modifying src-tauri; Long-form audio is now chunked at pauses to preserve transcription accuracy.

_src-tauri/src/audio\toolkit/vad · medium confidence · unverified

New app shell, theme system, and release notes viewer

The application now uses a new main entry point (App.tsx) that manages onboarding, settings navigation, and error handling via toast notifications. A new theme system (theme.css) introduces Light, Dark, and System appearance modes with shared color tokens, while custom scrollbars are styled for macOS and other platforms. A new "What's New" feature (WhatsNewGate, MarkdownContent) displays release notes in a modal after updates, using a safe, sanitized Markdown renderer. The app also initializes i18n, applies the stored theme synchronously to prevent flash, and sets up platform-specific data attributes for CSS scoping.

src · high confidence

New audio, model, and history management infrastructure

This change introduces a new set of backend managers in src-tauri/src/managers to handle core application logic. The AudioManager now implements cross-platform system mute/unmute capabilities (using Windows Core Audio, Linux PipeWire/PulseAudio/ALSA, and macOS osascript) and integrates Voice Activity Detection (VAD) policies. The ModelManager provides a unified interface for discovering, downloading, and verifying GGUF-based speech models, including SHA-256 integrity checks and resumable HTTP downloads. Additionally, a new HistoryManager manages transcription records using SQLite with rusqlite\_migration for schema versioning, and a GGUF metadata reader allows pre-download capability probing.

src-tauri/src/managers · high confidence

New command layer for audio, history, models, and transcription management

The application introduces a structured Tauri command interface in \src-tauri/src/commands\ to expose core capabilities to the frontend. This includes audio device enumeration and Windows microphone permission checks, a persistent history system with pagination, deletion, and retry logic, and a model management system that supports downloading, switching, and unloading models with configurable timeouts. Additionally, it adds controls for recording retention periods and custom audio feedback sounds.

src-tauri/src/commands · high confidence

New debug settings panel with live logging and diagnostic tools

A new debug settings section has been added to the application, providing users with advanced configuration and troubleshooting capabilities. This includes a live log viewer that streams application logs in real-time with configurable log levels (Error through Trace), and a keyboard diagnostic tool for macOS that detects Secure Input issues and stuck modifiers. Users can also adjust the word correction threshold, paste delay, recording buffer size, and hold threshold for push-to-talk behavior. Additional features include a toggle for reliable paste (macOS/Windows), a selector for keyboard implementation (Tauri vs. Handy Keys), and preview tools for onboarding steps and release notes.

src/components/settings/debug · high confidence

New first-run onboarding flow

Users are now guided through a structured, multi-step onboarding experience that persists across app relaunches. The flow covers language selection, permission requests, and model setup, including a migration screen for users upgrading from ONNX models to GGUF. It features interactive demo and 'Try it' screens to teach shortcut usage, a summary panel for quick settings adjustments, and handles interrupted downloads gracefully. Also in this area: 2 commits (1 fix) modifying src/components.

src/components/onboarding · medium confidence · unverified

New icon components for transcription and cancellation actions

The application now includes dedicated SVG icon components for key user actions: MicrophoneIcon, TranscriptionIcon, CancelIcon, and ResetIcon. These components are exported via the central index file, making them available for use in the UI to represent recording, transcription, cancellation, and reset functionalities respectively. Also in this area: Background file transcription with UI feedback and model locking.

src/components/icons · high confidence

New models management page with filtering and status tracking

A new Models Settings page has been added to the application, allowing users to browse, search, and manage available AI models. The interface includes filters for streaming capability, translation support, and specific languages, along with real-time status indicators for downloading, verifying, and extracting models. Users can now select, download, cancel, or delete models directly from this settings view, with safeguards to prevent deleting the currently active model. Also in this area: Background file transcription with UI feedback and model locking.

src/components/settings/models · high confidence

New recording overlay with live transcription and theme support

A new recording overlay has been introduced, providing a persistent UI for live audio capture and transcription. The overlay features a compact pill that expands into a live text panel, displaying real-time transcribed text with auto-scrolling and history support. It includes a waveform visualizer for microphone levels and supports both top and bottom screen placements. The component fully respects the application's light/dark theme settings and handles language direction (RTL/LTR) for proper text alignment.

src/overlay · high confidence

New settings components for transcription acceleration, audio input, and shortcut management

The settings interface now includes dedicated UI components for configuring hardware acceleration (CPU, CUDA, DirectML, ROCm) and selecting specific GPU devices for transcription, as well as managing microphone input including channel selection for multi-channel interfaces and fallback microphones for clamshell mode. Users can also customize global and handy-keys shortcuts with a new recording interface, manage custom words, and adjust audio feedback and auto-submit behaviors.

src/components/settings · high confidence

New settings sidebar and macOS-specific UI components

The settings interface has been restructured into a new Sidebar component that organizes configuration into distinct sections including General, History, Models, Advanced, Post-Processing, Debug, and About, with Post-Processing and Debug sections conditionally enabled based on user settings. This change introduces several new components: a SecureInputWarning banner that detects and alerts users when macOS Secure Input interferes with keyboard shortcuts, an AccessibilityPermissions component for managing macOS accessibility permissions, and an ErrorBoundary component to prevent UI crashes from blanking the entire application. Also in this area: Background file transcription with UI feedback and model locking.

src/components · high confidence

New unified UI component library

The application now includes a comprehensive set of reusable UI components in src/components/ui, introducing a consistent design language across the interface. This update adds foundational elements such as Alert, Badge, Button, Dialog, Dropdown, Input, Textarea, and Select, alongside specialized controls like Slider, ToggleSwitch, and ResetButton. It also introduces layout and utility components including SettingContainer, SettingsGroup, TextDisplay, PathDisplay, and Tooltip, along with an AudioPlayer component for media handling. These components provide standardized styling, accessibility features, and interaction patterns for settings and general UI usage.

src/components/ui · high confidence

New utility modules for formatting, keyboard handling, theming, and RTL support

This change introduces several new utility modules in src/lib/utils. The format.ts module provides a function to display model sizes in MB or GB with appropriate formatting. The keyboard.ts module (along with its test file) implements cross-platform keyboard event handling, including OS-specific modifier key naming (e.g., Command vs Super) and consistent display names for compound keys like Caps Lock. The theme.ts module adds user-selectable light/dark/system appearance options, persisting the choice to localStorage and AppSettings to prevent theme flashing on boot. The rtl.ts module handles right-to-left language support by updating the document's dir and lang attributes based on the selected language. Finally, modelTranslation.ts provides utilities to retrieve translated model names and descriptions from the i18n system.

src/lib/utils · high confidence

New validation scripts and CI tooling for model catalog, translations, and Nix dependencies

Added several new scripts to improve development workflow and data integrity: \check-model-language-coverage.ts\ ensures model language codes map to exactly one frontend intent (including a specific check for Norwegian variants); \check-nix-deps.ts\ replaces manual hash management with \bun2nix\ to keep Nix dependencies in sync with \bun.lock\; \check-translations.ts\ validates that all translation files match the English reference keys; \stage-transcribe-libs.sh\ handles staging of transcribe-cpp dynamic backend libraries for CI; and \gen\_catalog.py\ and \mirror\_models.py\ manage the model catalog generation and mirroring to S3-compatible storage with hash verification. Also in this area: Onboarding overhaul with model catalog, shortcut capture, and theme unification; Model catalog sourced from @handy-computer/models package.

scripts · high confidence

NixOS and Home Manager integration for Handy

Handy is now available as a declarative NixOS module and a Home Manager module. The NixOS module (nix/module.nix) installs the package and configures a udev rule to grant the 'input' group access to /dev/uinput, which is required for virtual input device creation. The Home Manager module (nix/hm-module.nix) provides a systemd user service to automatically start Handy upon graphical session login.

nix · high confidence

Offline model catalog with mirror fallbacks for reliable downloads

The application now includes a bundled, offline model catalog (\catalog.json\) that is compiled directly into the binary, ensuring users can discover and select models without requiring network access. This catalog lists specific models (such as Parakeet, Nemotron, and Whisper variants) with their capabilities, scores, and file hashes. To improve download reliability, the system supports a mirror fallback mechanism: if the primary Hugging Face download fails, the application automatically attempts to fetch the model files from the configured mirror (\https://blob.handy.computer\), verifying integrity using the SHA256 hashes stored in the catalog. Also in this area: Transcribe files into History and track dictation outcomes; Model catalog sourced from @handy-computer/models package.

src-tauri/src/catalog · high confidence

Portable installation mode added to Windows installer

The Windows installer now supports a portable installation mode, allowing users to choose between a standard installation and a portable setup that does not write to system directories or the registry. This feature is implemented via a new installer script that includes a dedicated UI page for selecting the install type, while preserving legacy portable markers during updates and requiring a specific magic string to prevent false positives on Scoop installs.

src-tauri/nsis · high confidence

Behavioural changes

1 commit (0 fixes) modifying src-tauri/resources/models

A change to existing behaviour in src-tauri/resources/models — 1 commit, 1 file: “transcribe.cpp vad (\#2267)” (refs \#2267).

src-tauri/resources/models · low confidence · unverified

2 commits (0 fixes) modifying src-tauri/src/language\_detection

A change to existing behaviour in src-tauri/src/language\_detection — 2 commits, 4 files: “bump transcribe.cpp (\#2489)”; “add language detection (\#2331)” (refs \#2489, \#2331).

_src-tauri/src/language\detection · medium confidence · unverified

3 commits (1 fix) modifying src/components/shared/modelTable

A change to existing behaviour in src/components/shared/modelTable — 3 commits (1 fix), 8 files: “fix: ignore IME confirmation Enter in text inputs (\#2248)”; “new onboarding (\#2303)”; “feat: transcribe files into history (\#2315)” (refs \#2248, \#2303, \#2315).

src/components/shared/modelTable · medium confidence · unverified

5 commits (0 fixes) modifying src-tauri/icons

A change to existing behaviour in src-tauri/icons — 5 commits, 51 files: “icon update + color tray when recording”; “who knows”; “working? ability to package/run on macos”; and 2 more.

src-tauri/icons · medium confidence · unverified

Add scripts to normalize Bun's isolated install output for reproducible builds

New TypeScript scripts in \.nix/scripts\ ensure that \bun install --linker=isolated\ produces a bit-reproducible \node\_modules\ tree, which is required for Nix fixed-output derivations. The \normalize-install.ts\ entry point runs three sub-scripts: \canonicalize-node-modules.ts\ rebuilds the \.bun/node\_modules\ symlink tree in lexicographic order to fix non-deterministic creation order; \normalize-bun-binaries.ts\ similarly sorts \.bin\ symlinks within each package's private directory; and \heal-peer-dep-bins.ts\ fixes a timing race in Bun's installer that causes missing \.bin\ symlinks for circular peer dependencies by recreating the intended links. Together, these changes eliminate host-specific drift in the dependency tree.

.nix/scripts · high confidence

Audio toolkit restructure with language-aware filler removal and custom word fixes

The audio toolkit module has been reorganized into dedicated source files, introducing language-aware filler word removal that uses high-confidence text-based language detection to apply language-specific rules. Custom word correction now supports multi-word phrases via n-gram matching and correctly preserves ampersands in user-defined terms. Additionally, the audio subsystem defaults to the ALSA host on Linux for microphone access.

_src-tauri/src/audio\toolkit · high confidence

Audio toolkit restructured with multi-channel support and improved recording reliability

The audio toolkit has been reorganized into a new module structure, introducing a dedicated device listing system that identifies default input and output devices. A key behavioral change is the addition of input channel selection for multi-channel audio interfaces, allowing users to choose specific channels rather than always averaging them. Recording reliability is improved through a robust ring-buffer architecture that prevents audio loss during pauses and ensures recordings are saved completely before transcription. The system now includes a frame resampler that resets its state between recordings to prevent audio crosstalk, and it handles microphone initialization failures gracefully without aborting the entire application. Also in this area: Transcribe files into History and track dictation outcomes.

_src-tauri/src/audio\toolkit/audio · high confidence

CI script added to stage transcribe-cpp dynamic backend libraries

A new CI script, \scripts/ci/stage-transcribe-libs.sh\, has been introduced to handle the packaging of transcribe-cpp's dynamic backend libraries. This script copies the core \libtranscribe\ shared library and the required \libggml-cpu\ backend modules from the build output into the final distribution directory, ensuring they are co-located for runtime loading by the \handy\ binary. The script includes validation to fail the build if these critical components are missing, guaranteeing that the necessary compute backends are available for the application.

scripts/ci · high confidence

Long-form audio is now chunked at pauses to preserve transcription accuracy

For models that do not natively handle long audio in a single pass, recordings are now automatically split into \~30-second chunks at detected pauses before decoding, preventing the accuracy degradation that previously occurred with longer clips. This behavior is controlled by a new 'Long-form chunking' setting (enabled by default) and applies to both live recordings and batch transcriptions, including headless CLI usage. The implementation introduces a VAD-based chunker and a unified decoding policy that selects between streaming, chunked, or single-run modes based on model capabilities. Also in this area: Isolate transcribe.cpp backend in a separate worker process. Also in this area: 1 commit (0 fixes) modifying src-tauri; 15 commits (13 fixes) fixing src-tauri.

_src-tauri/src/engine\supervisor · medium confidence · unverified

More reliable clipboard paste on macOS and Windows

The clipboard paste mechanism has been replaced with a receipt-sequenced approach that waits for the target application to actually read the pasted content before restoring the previous clipboard. On macOS, this uses a promise-based pasteboard owner, and on Windows, delayed-render clipboard formats. This prevents the previous clipboard from being restored prematurely (e.g., if the target app is slow to read), which previously caused the old content to be pasted instead. The change also ensures that if the user copies something else during the process, their new content is preserved.

_src-tauri/src/paste\tx · high confidence

Native library bundling and macOS microphone entitlements

The macOS build now includes an Entitlements.plist granting microphone access and an Info.plist with the required usage description, ensuring the app can capture audio without permission prompts. The build process (build.rs) and Tauri configuration have been updated to bundle native shared libraries (transcribe-cpp, ONNX Runtime, and VC++ runtime DLLs) into a transcribe-libs directory, which is then packaged alongside the executable on Windows and installed into app-private paths on Linux, resolving dependency issues on systems without these libraries pre-installed.

src-tauri · high confidence

New History Settings UI with Linux audio playback and clipboard fixes

The History tab now features a dedicated settings component that supports infinite scrolling for browsing transcription history and real-time updates as new entries are added. A key behavioral change addresses audio playback on Linux by reading files directly into blobs, ensuring recordings play correctly in that environment. The interface also includes a button to open the recordings folder and ensures that copy-to-clipboard actions only show success feedback after the write operation completes, improving reliability.

src/components/settings/history · high confidence

New model selector UI with download progress and status indicators

The model selection interface has been replaced with a new component suite that provides real-time feedback during model lifecycle events. Users now see a status button indicating whether a model is ready, loading, downloading, verifying, extracting, or in error, along with a dropdown to select from available models. The UI includes a progress bar that displays download speed and percentage, and automatically switches to the newly downloaded model once the process completes, provided no transcription is active. This change also introduces internationalization support for model names and descriptions within the selector.

src/components/model-selector · high confidence

New model selector UI with download progress and status indicators

The model selection interface has been replaced with a new component suite that provides real-time feedback during model lifecycle events. Users now see a status button indicating whether a model is ready, loading, downloading, verifying, extracting, or in error, along with a dropdown to select from available models. The UI includes a progress bar that displays download speed and percentage, and automatically switches to the newly downloaded model once the process completes, provided no transcription is active. This change also introduces internationalization support for model names and descriptions within the selector.

src/components/shared · high confidence

The update checker component now detects when the application is running in portable mode and, instead of attempting an in-place self-update (which fails on Windows), it displays a dialog that deep-links directly to the correct NSIS installer for the user's specific platform and architecture. This logic is implemented in the new \portableInstaller.ts\ module, which resolves the installer URL from the update manifest, and includes a test suite to verify correct URL resolution and fallback behavior.

src/components/update-checker · high confidence

Shortcut handling now supports a pluggable backend architecture

The shortcut management system has been refactored to support multiple keyboard input backends, allowing users to switch between Tauri's built-in global-shortcut plugin and the more granular handy-keys library via a new settings option. This change introduces a unified handler that routes events to the active backend, ensuring consistent behavior for actions like transcription and cancel bindings regardless of the underlying implementation. If the handy-keys backend fails to initialize, the system automatically falls back to the Tauri implementation and updates the settings to prevent repeated failures on subsequent launches.

src-tauri/src/shortcut · high confidence

Standardized language selection and recognition logic

The application now uses a centralized language constant file to manage a comprehensive list of supported languages, including explicit options for Simplified and Traditional Chinese. This change refines the user experience by hiding the generic 'Chinese' option from the selection picker to avoid ambiguity, while ensuring that language codes are correctly normalized (e.g., mapping Norwegian Bokmål 'nb' to 'no' and Filipino 'fil' to 'tl') for accurate backend recognition.

src/lib/constants · high confidence

Fixes

3 commits (3 fixes) fixing src

A fix in src — 3 commits (3 fixs), 2 files: “fix: say so when the microphone sends only silence (\#2486)”; “fix: show a No Model Selected notice instead of a blank model-not-found error (\#2418)”; “fix(macos): clear stale permission grants before asking again (\#2057)” (refs \#2486, \#2418, \#2057). Also in this area: 1 commit (0 fixes) modifying src, src/lib; Fix compatibility with older Safari versions by shimming Object.hasOwn.

src/lib · low confidence · unverified

Automated Nix dependency hashing via bun2nix

The Nix build system now uses \bun2nix\ to automatically generate and manage SHA-512 hashes for Bun package dependencies. This replaces the previous manual hash management process, introducing new \bun-lock-hash\ and \bun.nix\ files to ensure reproducible builds without requiring developers to manually calculate or update hashes when dependencies change. Also in this area: Onboarding overhaul with model catalog, shortcut capture, and theme unification; Model catalog sourced from @handy-computer/models package.

.nix · high confidence

Test coverage

1 commit adding/updating tests in src-tauri/tests; Added Playwright E2E tests for app startup and structure.

Dependencies

Initial dependency manifests for Handy v0.9.7

This change introduces the foundational dependency configuration for the Handy application, establishing the project at version 0.9.7. It adds the frontend \package.json\ defining the React, Vite, and Tauri plugin ecosystem, alongside the backend \Cargo.toml\ and \Cargo.lock\ which specify the Rust dependencies for audio processing, speech recognition (transcribe-rs/cpp), and system integration. This commit sets the baseline library versions for the application's core functionality.

(dependencies) · high confidence

Housekeeping

Initial project scaffolding and documentation

The repository has been initialized with core configuration files including \.prettierignore\ and \.prettierrc\ to enforce code formatting standards, and a \bun.lock\ file to manage frontend dependencies. Comprehensive documentation has been added, including \README.md\ for installation and usage, \BUILD.md\ for platform-specific build instructions, \CONTRIBUTING.md\ for development guidelines, and \AGENTS.md\ to provide context for AI coding assistants. A \LICENSE\ file (MIT) and \CLAUDE.md\ (pointing to AGENTS.md) are also included to establish the project's open-source terms and development workflow.

(repo-wide) · high confidence

Written by cca.canine.dev from the codebase's own history, inside the signed delivery this page is composed from.

How this codebase got here

Score

  • CAI 49 → 56 (+6.9)
  • Rubric changed (rubric-2026.10.5 → rubric-2026.11.4) — scores are not directly comparable.

Lenses

  • Code Health 63 → 74 (+11.2)
  • Architecture 92 → 91 (-0.6)
  • Maturity 66 → 65 (-0.2)
  • Readiness 44 → 60 (+15.4)
  • Security 56 → 60 (+3.2)
  • Event Sourcing 100 → 100 (+0.0)
  • Accessibility 41 → 45 (+4.7)
  • Performance 100 → 100 (+0.0)

Resolved (21)

  • AudioPlayer.AudioPlayer (cognitive 45) (src/components/ui/AudioPlayer.tsx)
  • Duplicated block (10 lines × 2) (src-tauri/src/clipboard.rs)
  • Duplicated block (11 lines × 2) (src-tauri/src/llm_client.rs)
  • Duplicated block (14 lines × 2) (src-tauri/swift/apple_intelligence.swift)
  • High CVE: [GHSA redacted] (bun.lock)
  • High CVE: [GHSA redacted] (bun.lock)
  • High CVE: [GHSA redacted] (src-tauri/Cargo.lock)
  • High CVE: [GHSA redacted] (src-tauri/Cargo.lock)
  • Hotspot: src/components/settings/models/ModelsSettings.tsx (src/components/settings/models/ModelsSettings.tsx)
  • Medium vulnerability: [GHSA redacted] (src-tauri/Cargo.lock)
  • Medium vulnerability: RUSTSEC-2026-0367 (src-tauri/Cargo.lock)
  • MethodTooLong: ModelManager.download_hf_model (src-tauri/src/managers/model.rs)
  • ModelManager::download_hf_model (cognitive 42) (src-tauri/src/managers/model.rs)
  • ModelManager::download_hf_model (cyclomatic 24) (src-tauri/src/managers/model.rs)
  • No ADRs found
  • Off-boarding risk: anonymized user #1
  • Outdated: transcribe-cpp
  • Repeated repair: src-tauri/src/commands/audio.rs (src-tauri/src/commands/audio.rs)
  • check-translations.validateTranslations (cognitive 28) (scripts/check-translations.ts)
  • mirror_models.main (cognitive 49) (scripts/mirror_models.py)
  • …and 1 more

New (60)

  • AudioPlayer.AudioPlayer::togglePlay (cognitive 18) (src/components/ui/AudioPlayer.tsx)
  • Critical CVE: [GHSA redacted] (src-tauri/Cargo.lock)
  • Dropdown.Dropdown (cognitive 16) (src/components/ui/Dropdown.tsx)
  • Duplicated block (10 lines × 2) (src-tauri/src/managers/audio.rs)
  • Duplicated block (11 lines × 2) (src-tauri/src/llm_client.rs)
  • Duplicated block (14 lines × 2) (src-tauri/swift/apple_intelligence.swift)
  • Duplicated block (5 lines × 2) (src-tauri/src/commands/audio.rs)
  • Duplicated block (9 lines × 2) (src-tauri/src/managers/gguf_meta.rs)
  • Edited copies of one member (3 members, 11+ corresponding lines) (src-tauri/src/commands/mod.rs)
  • Edited copies of one member (3 members, 14+ corresponding lines) (src-tauri/src/clipboard.rs)
  • Edited copies of one member (3 members, 14+ corresponding lines) (src-tauri/src/commands/audio.rs)
  • Edited copies of one member (4 members, 15+ corresponding lines) (src-tauri/src/shortcut/mod.rs)
  • Edited copies of one member (6 members, 15+ corresponding lines) (src-tauri/src/shortcut/mod.rs)
  • Edited copy of a member (14 corresponding lines) (src-tauri/src/commands/models.rs)
  • Edited copy of a member (14 corresponding lines) (src-tauri/src/shortcut/handy_keys.rs)
  • Edited copy of a member (16 corresponding lines) (src-tauri/src/clipboard.rs)
  • Edited copy of a member (18 corresponding lines) (src-tauri/src/commands/audio.rs)
  • Edited copy of a member (19 corresponding lines) (src-tauri/src/managers/transcription.rs)
  • Edited copy of a member (23 corresponding lines) (src-tauri/src/commands/audio.rs)
  • Edited copy of a member (24 corresponding lines) (src-tauri/src/shortcut/handy_keys.rs)
  • …and 40 more

Changes since last survey

  • 59 commits — 20 feature/other, 39 fixes

By area

  • src-tauri/src — 27 commits
  • src/i18n — 9 commits
  • src/components — 7 commits
  • src-tauri/Cargo.lock — 4 commits
  • (root) — 3 commits
  • (repo) — 2 commits
  • src-tauri/Cargo.toml — 2 commits
  • .github/workflows — 1 commit
  • docs/troubleshooting — 1 commit
  • public/release-notes — 1 commit
  • src-tauri/tests — 1 commit
  • src/overlay — 1 commit

Notable commits

  • fix: fix(deps): address TLS dependency security advisories (#2432)
  • fix: fix(linux): add 'None' keyboard implementation for Wayland compositors (#1300)
  • fix: fix(linux): keep transparent overlay painting on WebKitGTK 2.54 (#2360)
  • fix: fix(linux): type multi-line text correctly with dotool (#2199)
  • fix: fix(macos): clear stale permission grants before asking again (#2057)
  • fix: fix(macos): prevent silent AirPods recordings when also playing audio from iPhone (#1991)
  • fix: fix(overlay): disable WebView2 context menu on the recording overlay (#2403)
  • fix: fix(overlay): don't steal focus when the overlay falls back to X11 (#2488)
  • fix: fix(shortcut): classify tap vs hold by key event time, not dequeue time (#2097)
  • fix: fix(shortcuts): avoid the ibus hotkey on Linux defaults (#2367)
  • fix: fix: bound post-processing with a 60 s deadline
  • fix: fix: build libopus on windows arm
  • fix: fix: build libopus without simd on windows arm
  • fix: fix: build the async runtime before streaming logs to the webview (#2478)
  • fix: fix: bump hf-hub so downloads survive proxies that strip Range (#2318)
  • fix: fix: convert Chinese script based on the text, not the detected language
  • fix: fix: declare OpenSSL 3 runtime dependency in Debian packages (#2336)
  • fix: fix: disable ggml Metal on Intel macOS builds (#2340)
  • fix: fix: dismiss focused setting tooltips with Escape (#2251)
  • fix: fix: find the ydotoold socket when YDOTOOL_SOCKET is not set (#2427)
  • …and 39 more

Written by cca.canine.dev from the codebase's own history, inside the signed delivery this page is composed from.

Survey your own repository

cjpais/Handy was measured the same way every project in this corpus was: the same rubric, at a pinned commit, with the result published in full. Point a surveyor at a repository you know and see whether you agree with it.

About this page

  • The score is its most recent published measurement, taken on 14 November 2026 at a pinned commit. It is not a live figure and does not change until the project is measured again.
  • Measured at commit d2ee87178c04d27a247bf1e7cd30a4049da1702c — the exact code this score is about.
  • Scored under rubric-2026.11.4 — the same rubric and the same method as every other entry in this index.
  • Measured by Canine Code Analyzer (cca.canine.dev) using cca measure-2026.11.65+preprod-36d4cd83012f.