Hours
Weeks
Commits
Role
Pastry
Native-Grade Clipboard Intelligence with Rust-Powered OCR, Color Analysis, and Licensing
Pastry is a privacy-first desktop clipboard manager that turns everything you copy—text, screenshots, and colors—into a searchable, organized history.
pastry-app/desktop
Pastry is a privacy-first desktop clipboard manager that turns everything you copy—text, screenshots, and colors—into a searchable, organized history. It exists for power users who live in their keyboard and hate losing context: developers, designers, and writers who constantly switch apps and need instant access to past snippets, captured screens, and UI palettes. Instead of being “just another history list,” Pastry treats the clipboard as a real data stream, layering on full-text search, screenshot OCR, and color extraction so users can rediscover information they didn’t even realize they needed to save. Under the hood, Pastry uses a Tauri architecture with a Rust core and React/TypeScript frontend so performance-critical paths stay native-fast while the UI remains easy to evolve. The Rust backend runs a watchdog-driven clipboard monitor that survives flaky OS APIs and ties directly into a hardened licensing system backed by SQLite and machine-bound HMAC state. Native macOS integrations (via Objective‑C FFI) handle app activation, paste simulation, and OCR through Apple’s Vision framework, while a filesystem watcher turns new screenshots into OCR’d, searchable items. A custom color pipeline converts Display P3 to sRGB and runs Lab-space k-means clustering to surface accurate dominant colors for designers. To support real-world usage, Pastry includes file-backed, category-aware logging with runtime toggles for targeted diagnostics, plus an always-on Rust audio engine for low-latency feedback during rapid keyboard-driven workflows. These systems make Pastry behave like a polished desktop product rather than a demo app, and demonstrate end-to-end ownership from UX to OS-level integration and licensing.
Engineered a resilient Rust clipboard monitor with health watchdog, periodic handle refresh, and license-aware gating, delivering stable real-time history even under flaky OS clipboard behavior.
Built a macOS-native window and focus controller that coordinates animated overlay toggling with atomic race protection, restoring the previous app instantly during rapid keyboard-driven switching.
Implemented a screenshot ingestion pipeline that watches macOS screenshot folders, debounces noisy events, deduplicates via SHA256, and runs OCR to turn images into searchable clipboard entries.
Developed a macOS color extraction engine that converts Display P3 images to sRGB and applies Lab-space k-means clustering to surface visually accurate dominant colors for UI tools.
Integrated Apple’s Vision framework via Objective‑C FFI into an async Rust OCR service, supporting multiple scripts and feeding recognized text directly into the clipboard and screenshot flows.
Designed a hardened licensing and trial system combining SQLite state, machine-bound HMAC signatures, and OS-specific files to resist tampering and reinstall-based trial resets.
Created a file-backed, category-aware logging subsystem with runtime toggling and log rotation, enabling targeted production diagnostics without impacting performance in normal usage.
Built an asynchronous sound engine with an always-on output stream, worker threads, and fade effects, providing low-latency, non-blocking audio feedback for copy and paste actions.
Provides a fast, privacy-preserving desktop clipboard manager that unifies text, images, screenshots, and rich context into a searchable, organized history for power users.