Mnemo The Mnemo Authors
Install Mnemo with WinGet
winget install --id=SnifferLP.Mnemo -e Mnemo is a local-first study platform with flashcards, spaced repetition, adaptive learning, tests, and configurable AI assistance.
Loading...
README
Mnemo
A quieter way to remember.
A local-first desktop study app for flashcards, spaced repetition and deliberate practice. Windows and Linux · optional AI · no account required
Mnemo keeps your study library on your own computer. It combines a native Tauri shell with a focused React interface, stores material in a local SQLite database, and treats AI assistance as an optional extra rather than a dependency.
There is no account system, no telemetry and no vendor-hosted service. AI features call only the provider you configure, and everything else works offline.
Preview
Mnemo uses a calm, card-based dashboard: a soft neutral canvas, white surfaces with rounded corners and soft elevation, a cobalt accent and a full-height navigation rail that collapses to an icon strip. Modern Sans is the default typeface, with the editorial serif still available in Settings. Light and dark themes share the same structure, and the interface adapts from smaller laptop windows to large desktop screens. The token system, sidebar and startup rules are documented in docs/DESIGN.md.
<img src="docs/screenshots/lightmode-set-overview.png" alt="Study set library with search, favourites and folder filters" width="400" />
<img src="docs/screenshots/darkmode-ai-overview.png" alt="AI study assistant in dark mode with per-set context controls" width="400" />
<img src="docs/screenshots/darkmode-settings-overview.png" alt="Appearance settings in dark mode with display font, language, theme and accent colour" width="820" />
Screenshots use neutral sample content from a development profile and show both themes. Requirements for new captures are documented in docs/screenshots/README.md. Please do not add screenshots containing real study content, personal names, API keys or machine-specific paths.
Features
- Library management — folders, tags, favourites, search, sorting, duplication, archiving and progress reset.
- Rich cards — front and back text, explanations, hints, images and audio attachments.
- Starred terms and flashcard options — star individual terms, start a session with the front or back side first, limit it to starred terms, and preview a real card directly on the set page.
- Study modes — flashcards, adaptive learning, writing practice, SM-2 spaced repetition, practice tests and a matching game.
- Insights — due counts, daily goals, timezone-stable current and longest streaks, activity ranges, a 30-week heatmap and per-set progress.
- Import and export — CSV, TSV, JSON and Quizlet-style text import; CSV, TSV, JSON, Markdown and PDF export.
- Optional AI — card generation, document summaries, study guides and tutor chat through OpenAI-compatible APIs, Gemini or local Ollama. Imported documents can be selected or removed individually.
- Application info — the installed version and a repository link are available at the bottom of Settings.
- Automatic updates — signed in-app updates for Windows x64 and Linux x86_64 AppImage builds, with stable and pre-release channels.
- Privacy by default — local SQLite storage, encrypted provider keys and no analytics.
- English and German UI — a language dropdown in Settings applies immediately and persists across restarts, with locale-aware dates, plural forms and bundled offline fonts.
Supported platforms
| Platform | Status | Distribution | Automatic updates |
|---|---|---|---|
| Windows 10/11 x64 | Supported | NSIS installer (.exe) | Yes |
| Linux x86_64 (Arch) | Supported | .pkg.tar.zst | Package manager only |
| Linux x86_64 (general) | Supported | AppImage | Yes |
| Debian/Ubuntu x86_64 | Supported | .deb | Package manager only |
| Linux aarch64 | Not officially released | Source build may work; no package is published | No |
| macOS | Not supported | — | No |
Windows builds are produced on windows-latest in GitHub Actions. Linux checks and package builds run on Linux runners. No macOS artifact is published.
Download and installation
Download the latest artifacts from the repository's Releases page. Generated packages are never committed to the repository.
Windows 10 and 11
- Download
Mnemo__x64-setup.exe. - Run the installer. It installs for the current user, so administrator rights are not normally required.
- Launch Mnemo from the Start Menu or the installer's finish page.
The installer adds a Start Menu entry, uses the Mnemo application icon and registers an uninstaller. Uninstall from Settings → Apps → Installed apps → Mnemo, or from the Start Menu shortcut's context menu.
The installer embeds the Microsoft WebView2 bootstrapper. Windows 11 and current Windows 10 installations normally already have WebView2. If it is missing, the installer downloads it from Microsoft during installation; that step requires an internet connection. No Rust, Node.js, pnpm or other development tools are required to install or run Mnemo.
> Windows code signing: the current installer is not code-signed. Windows SmartScreen may show an "Unknown publisher" warning. Verify that you downloaded the file from this repository's Releases page. Signing may be added in a future release; unsigned artifacts are never presented as signed.
Arch Linux
sudo pacman -U mnemo--1-x86_64.pkg.tar.zst
The package installs the binary, desktop entry, hicolor icons and license notices.
AppImage
chmod +x Mnemo__amd64.AppImage
./Mnemo__amd64.AppImage
The AppImage needs FUSE 2. If FUSE is unavailable, run it with:
./Mnemo__amd64.AppImage --appimage-extract-and-run
Automatic updates
Mnemo can update itself in place on Windows x64 NSIS installations and Linux x86_64 AppImages downloaded from this repository's GitHub Releases. The .deb and Arch packages are deliberately left to apt/pacman and are not modified by the in-app updater.
Open Settings → Updates to choose a channel:
- Stable Releases (default) only considers GitHub releases that are not marked as pre-releases.
- Pre-Releases also considers beta and test releases. It never switches the channel automatically.
Checks run in the background at startup without blocking the app. If a compatible update is found, a dismissible notification appears in the bottom-right corner. Mnemo downloads and installs only after Download and install is selected, verifies the Tauri signature, and then restarts on Linux or lets the NSIS installer relaunch the Windows app. A failed check or installation leaves the current installation usable.
Update signing uses Tauri's minisign key. The public key is embedded in src-tauri/tauri.conf.json; the encrypted private key and its password are GitHub Actions secrets named TAURI_SIGNING_PRIVATE_KEY and TAURI_SIGNING_PRIVATE_KEY_PASSWORD. Unsigned or tampered update payloads are rejected. Do not publish a release whose updater artifacts are missing signatures.
Debian and Ubuntu
sudo apt install ./Mnemo__amd64.deb
The Debian package depends on the standard WebKitGTK 4.1 runtime libraries provided by your distribution.
Getting started
- Launch Mnemo. A new, empty library is created on first start; no demo data is inserted.
- Create a study set, or import one from CSV, TSV, JSON or plain text.
- Add cards with front/back text, an explanation, a hint and optional media.
- Choose a study mode and start learning. Flashcards and keyboard shortcuts work without any AI configuration.
- Open Settings → Data to see the exact database, media and backup locations on your machine.
Backups created from Settings are ordinary SQLite copies. They are not encrypted; store them somewhere you trust.
AI configuration
AI is optional. Without a provider, the library, import/export, flashcards, spaced repetition, tests, statistics and backups all work normally.
- Open Settings → AI providers → Add provider.
- Select a preset or Custom.
- Enter the base URL and model.
- Paste your own API key. For local Ollama, leave the key empty.
- Test the connection and optionally mark the provider as default.
| Provider type | Example base URL |
|---|---|
| OpenAI | https://api.openai.com/v1 |
| OpenRouter | https://openrouter.ai/api/v1 |
| Mistral | https://api.mistral.ai/v1 |
| Gemini | Google Gemini API endpoint |
| Ollama (local) | http://localhost:11434/v1 |
| Custom OpenAI-compatible | Your provider's /v1 endpoint |
Provider keys are encrypted at rest with AES-256-GCM. The master key stays on the local machine: on Linux it is written with mode 0600; on Windows it is protected with the current user's DPAPI credentials. Keys are never sent to the frontend or written to logs.
AI requests are sent directly to the provider you choose, under that provider's privacy terms. Mnemo does not proxy them through a project-operated server.
Privacy and local data
Mnemo has no accounts, telemetry, analytics, cloud sync or bundled API keys. The application creates its own data directory on first launch.
| Data | Default location |
|---|---|
| Linux data directory | $XDG_DATA_HOME/com.mnemo.study/ or ~/.local/share/com.mnemo.study/ |
| Windows data directory | %APPDATA%\com.mnemo.study\ |
| Database | /mnemo.db (SQLite, WAL mode) |
| Media attachments | /media/ |
| API-key vault | /secure.key |
| Development builds | /dev/ |
On Linux, secure.key is created with 0600 permissions. On Windows, the file is wrapped with DPAPI and can only be decrypted by the same Windows user profile. Moving the data directory to another Windows account requires removing secure.key and re-entering provider keys; the database and study content are unaffected.
To remove all local data, quit Mnemo and delete its data directory. Uninstalling the application does not delete the data directory automatically.
System requirements
- Windows: Windows 10 or Windows 11, x64, Microsoft WebView2 Runtime (installed automatically if missing and an internet connection is available).
- Linux: x86_64 distribution with WebKitGTK 4.1, GTK 3 and libsoup 3; AppImage users also need FUSE 2.
- A local SQLite-capable filesystem. Network shares are not recommended for the live database.
- AI features require network access to the configured provider. Ollama can run entirely locally.
Development and build
Requirements
- Rust stable (MSVC toolchain on Windows)
- Node.js 22 LTS or newer
- pnpm 12.18.0 (also declared in
package.json) - Windows: Visual Studio Build Tools with the Desktop development with C++ workload and WebView2
- Linux: WebKitGTK 4.1, GTK 3, libsoup 3 and the usual Tauri build dependencies
Install the Linux dependencies on Debian/Ubuntu:
sudo apt install libwebkit2gtk-4.1-dev libgtk-3-dev libsoup-3.0-dev \
libayatana-appindicator3-dev librsvg2-dev patchelf build-essential \
curl file pkg-config
On Arch Linux:
sudo pacman -S webkit2gtk-4.1 gtk3 libsoup3 librsvg patchelf base-devel
On Windows, install Rust, Node.js 22 and pnpm, then make sure the MSVC Rust toolchain is active:
rustup default stable-msvc
corepack enable
corepack prepare pnpm@12.18.0 --activate
Run locally
git clone https://github.com/SnifferLP/Mnemo.git
cd Mnemo
pnpm install --frozen-lockfile
pnpm tauri dev
Vite serves the UI on http://127.0.0.1:1420; Tauri opens the native window. Development builds use a separate dev data directory so they cannot modify an installed library.
Build installers
pnpm tauri build
On Linux this produces .deb and .AppImage bundles under src-tauri/target/release/bundle/. On Windows, src-tauri/tauri.windows.conf.json selects NSIS and produces an installer under src-tauri/target/release/bundle/nsis/.
These ordinary builds do not produce updater signatures. To build signed updater artifacts locally, set TAURI_SIGNING_PRIVATE_KEY and TAURI_SIGNING_PRIVATE_KEY_PASSWORD, then merge the release config:
pnpm tauri build --config src-tauri/tauri.release.conf.json
On Linux this adds Mnemo__amd64.AppImage.sig; on Windows it adds Mnemo__x64-setup.exe.sig. Release automation generates the platform manifest as latest.json and attaches it to the GitHub Release.
Arch Linux packaging
Build the release binary without Tauri's own Linux bundles, then create the Arch package:
pnpm tauri build --no-bundle
./packaging/arch/build-binary-package.sh
See packaging/arch/README.md and docs/RELEASING.md for source-package and release details.
Tests
Run the same checks locally that CI runs:
pnpm typecheck
pnpm test
pnpm build
cd src-tauri
cargo fmt --all --check
cargo clippy --all-targets --all-features -- -D warnings
cargo test --all-targets
The current local suite contains 33 Rust unit tests, 15 Rust integration tests and 56 frontend tests. The frontend suite covers learning-round state, activity-chart axis helpers, flashcard option helpers, card-duplicate matching, settings merging, automatic-update defaults, motion preferences, update-channel persistence and translation catalogues. Rust tests cover SM-2 scheduling, answer normalization, import/export, AI request construction, settings sanitization, database migrations, CRUD, document removal, updater channel selection, statistics, persistence, starred terms, archive round trips and upgrade behavior.
GitHub Actions also runs the Rust suite on Windows, builds the NSIS installer, builds Linux AppImage/.deb bundles and builds the Arch package. A successful compile is not a substitute for manual testing of Windows installation, display scaling, audio playback and other interactive behavior; see docs/RELEASING.md for the manual QA checklist.
License and third-party licenses
Mnemo's source code is released under the MIT License; see LICENSE.
Bundled fonts are licensed separately under the SIL Open Font License 1.1. Third-party components, including windows-sys, Tauri, React and cryptography crates, are documented in THIRD_PARTY_NOTICES.md. The MIT license does not replace those terms.
Reporting bugs and getting support
- Search existing issues before opening a new one.
- Use the bug report template and include the Mnemo version, platform, steps to reproduce and relevant logs with secrets and personal data removed.
- For suspected security vulnerabilities, follow
SECURITY.mdand use the repository's private Report a vulnerability flow. Do not open a public issue for a vulnerability. - Feature requests and general questions are welcome through the repository's issue tracker.
When reporting a Windows problem, please state whether it occurs with the installed NSIS build or a local development build, and whether WebView2 is installed.
Known limitations
- No cloud sync, sharing links, mobile app or multi-user accounts.
- The Windows installer is currently unsigned. Updater artifacts are separately signed with Tauri's update key and are not affected by Windows Authenticode status.
- Automatic updates are not implemented for
.deb, Arch or other package-manager installations. Those packages must be updated through their package manager. - AI quality and availability depend on the configured provider; AI grading is off by default and consumes provider tokens.
- Scanned PDFs require external OCR.
- PDF export uses a limited built-in font encoding; use JSON or Markdown for full Unicode.
- Tests and study material are stored locally without a separate answer-key permission model; Mnemo is a single-user desktop application.
- macOS is not supported.