TeXterity is a native, offline-first desktop game for learning to type LaTeX used in graduate mathematics and physics. It is written in Rust with eframe/egui and uses RaTeX for production parsing, layout, and display-list generation.

The defining interaction is source-linked formula teaching: every pedagogical region in a rendered prompt retains its canonical byte span. Hovering or keyboard-focusing a region reveals the exact source fragment that produced it. Grading uses a tutor-owned syntax tree, so safe syntax variants such as x^2 and x^{2} can be equivalent without confusing mathematical equivalence with LaTeX fluency.

Run

cargo run -p tutor_app --bin texterity

tutor_app is the Cargo workspace package; texterity is the executable and installed application name.

The application stores progress in the platform data directory. Set TEXTERITY_DATA_DIR to use an explicit directory for testing or portable operation. No account, network connection, browser, JavaScript runtime, webview, or system TeX installation is required.

Learn

The dark-first desktop interface opens on one clear action: Choose a lesson. The collapsible left rail provides Learn, Practice, Progress, Reference, Achievements, and Settings; its manual state is remembered, while compact windows use a temporary navigation drawer. Foundation is the first authored curriculum track, and every locked lesson names the prerequisite that unlocks it.

Tutorials keep the authored learning gates but present them as one continuous workbench: read the combined lesson brief, restore the required source fragment, then type complete LaTeX beneath the large rendered formula. The formula inspector, fixed feedback area, and multiline IBM Plex Mono editor remain in stable positions. Press Ctrl+Enter to check, Ctrl+Shift+Enter to check and continue, or plain Enter to add a line. See the user guide for the complete journey and keyboard behavior.

Verify

cargo fmt --check
cargo clippy --workspace --all-targets --all-features -- -D warnings
cargo test --workspace --all-features
cargo doc --workspace --no-deps
python3 tools/generate_builtin_content.py --check
python3 scripts/audit-runtime-boundaries.py
cargo xtask verify --release

CI executes every benchmark budget and compiles all six libFuzzer targets. See performance verification, fuzzing, and release packaging for the corresponding local commands. The acceptance matrix traces every normative AT-001 through AT-026 contract to its automated and packaged-host evidence.

See the architecture walkthrough to trace a learner answer from the native input widget through scanning, structural grading, the state reducer, SQLite persistence, and UI feedback. Content authors should start with the content guide.

Architecture boundaries

• tutor_core: deterministic game and learning-domain state.
• tutor_latex: source spans, tutor syntax, macros, grading, diagnostics.
• tutor_renderer_ratex: the only crate allowed to import RaTeX; its optional themed paint path changes visible default ink without changing layout or mapping.
• tutor_content: immutable built-in/custom content indexes and validation.
• tutor_storage: synchronous SQLite repositories and atomic JSON interchange.
• tutor_app: native presentation, embedded visual assets, and effect execution; it may streamline navigation but cannot redefine domain transitions.
• xtask: content, corpus, architecture, packaging, and quality gates.

The complete normative product specification is preserved in latex_typing_game_spec.

License

MIT. The executable embeds KaTeX math fonts, IBM Plex Sans/Mono interface fonts, and Lucide icons for offline operation. Their upstream licenses and pinned payload details are recorded in THIRD_PARTY_NOTICES.md.