Make Every File Searchable On Your Desktop
让电脑没有难搜的文件
English | 简体中文 | Latest Release
SoEasy is a native, privacy-first, ultra-fast desktop search engine. Built with modern C++17 and Qt5, it indexes and searches source code repositories, Office suites, spreadsheets, presentations, PDFs, and scanned documents entirely offline with sub-millisecond query response and instant content preview.
Zero Electron bloat, zero cloud telemetry, and complete crash resilience through multi-process sandbox isolation.
- Instant Full-Text & Filename Search: Powered by SQLite FTS5 with Double-Array Trie Jieba tokenization and BM25 ranking. Features live debounced in-document keyword preview.
-
Crash-Resilient Sandbox Isolation: All document parsing runs within an isolated subprocess (
soeasy-worker) governed by modernreproc++. Malformed or looping files are killed and self-healed in milliseconds without freezing the UI. -
Double-Array Trie (DAT) + mmap Tokenization: Precompiles 350,000 dictionary words into contiguous DAT binary blocks, mapped via kernel
mmap. Memory footprint plunges from 120 MB down to 7.2 MB (94% RAM reduction). -
Lightweight Deep Learning OCR: Bundles PP-OCRv6 with zero-allocation Header Probe nanosecond sniffing (
$< 1,\mu\text{s}$ ) to filter micro-icons and borders before decoding. Cuts RAM by 34% and accelerates inference by >40%. - Hardware-Bound Local Encryption: Zero cloud dependence. Index databases are protected with page-level AES-128 encryption (SQLite3MultipleCiphers) and Zstandard Level 3 compression, bound to host hardware to prevent offline theft.
- Modern Component Architecture: Features composite filename and path presentation, a clean single-line search bar, full keyboard navigation (Enter, Ctrl+O, Ctrl+P, F3), and real-time incremental filesystem watching.
💡 Deep Architectural Specifications: For in-depth implementation details on process supervision, DAT Trie derivation, nanosecond Header Probe heuristics, and machine-bound envelope encryption, see 📖 Deep Architectural Principles (docs/features.md).
flowchart TD
subgraph UI ["Presentation Layer (soeasy-gui / Qt5)"]
MainWindow["MainWindow (Single-Line Bar & Dual-Pane)"]
ResultView["ResultView (Virtual Viewport & Hover Preview)"]
PreviewPanel["PreviewPanel & DocumentViewerWidget"]
MainWindow --> ResultView
MainWindow --> PreviewPanel
end
subgraph Core ["Core Service Layer (soeasy-service / CQRS)"]
IndexPipeline["IndexPipeline (Service Facade)"]
subgraph QuerySide ["Query Side (Read Path)"]
SearchCoord["ISearchCoordinator / SearchCoordinator"]
NameIndex["NameIndex (In-Memory Instant Index)"]
SearchCoord --> NameIndex
end
subgraph CommandSide ["Command Side (Write Path)"]
Orchestrator["IndexOrchestrator (Concurrent Scheduler)"]
WatchEngine["WatchEngine & EventDebouncer (efsw)"]
WorkerGuard["WorkerGuard (reproc++ Sandbox Watchdog)"]
WriteQueue["AsyncWriteQueue (Batched Writer)"]
Orchestrator --> WatchEngine
Orchestrator --> WorkerGuard
Orchestrator --> WriteQueue
end
IndexPipeline --> SearchCoord
IndexPipeline --> Orchestrator
end
subgraph Sandbox ["Isolated Parser Process (soeasy-worker)"]
ParserRouter["Tiered Parser Router (L1 Text / L2 Office & PDF / L3 PP-OCRv6)"]
end
subgraph Storage ["Storage & Infrastructure Layer (soeasy-storage / common)"]
IStorage["«interface» IStorageEngine (DIP Boundary)"]
StorageEngine["StorageEngine (SQLite3MC AES-128 + Zstd)"]
Subsystems["ReaderConnectionPool | Fts5JiebaTokenizer (DAT+mmap) | MachineKeyProvider"]
StorageEngine -.->|"implements"| IStorage
StorageEngine --> Subsystems
end
MainWindow --> IndexPipeline
ResultView --> SearchCoord
PreviewPanel --> SearchCoord
SearchCoord --> IStorage
WriteQueue --> IStorage
WorkerGuard ==>|"Cap'n Proto + Shared Memory IPC (Process Isolation)"| ParserRouter
| Category | File Extensions | Extraction Engine |
|---|---|---|
| Plain Text & Source Code | .txt, .md, .log, .csv, .json, .xml, .html, .c, .cpp, .py, .go, .rs, .java, .ts, .sh, etc. |
L1 Native parser + uchardet auto-sniffing |
| Microsoft Office (OOXML) | .docx, .xlsx, .pptx |
L2 DocToText / Fast stream extraction |
| Microsoft Office Binary | .doc, .xls, .ppt, .xlsb |
L2 DocToText binary extractors |
| WPS Office Suite | .wps, .et, .dps |
L2 DocToText |
| OpenDocument (ODF) | .odt, .ods, .odp, .odg, .fodt, .fods, .fodp |
L2 DocToText |
| Portable Documents | .pdf (text layer + automatic scan queueing) |
L2 Poppler natural flow + OCR fallback |
| Apple iWork | .pages, .numbers, .key, .keynote |
L2 DocToText |
| Rich Text & Emails | .rtf, .eml |
L2 DocToText |
| Images & Vector Graphics | .png, .jpg, .jpeg, .webp, .bmp, .tiff, .tif, .jfif, .pnm, .svg |
L3 PP-OCRv6 deep learning OCR + SVG Parser |
📌 Roadmap: Recursive extraction and streaming search for compressed archives (
.zip,.tar,.gz,.7z,.rar,.bz2,.xz) are planned for future releases (currently greyed out in settings).
- Clone: Use
git clone --recursiveto initialize all submodules - Compiler: GCC 8+ or Clang 9+ supporting C++17 (MSYS2 MinGW-w64 recommended on Windows)
- CMake: 3.16 or higher
- Dependencies: Qt5 (
Widgets,Gui,Core), Cap'n Proto, uchardet, zlib, libxml2, libiconv, OpenCV, Boost
# Generate Release configuration
cmake -B build -DCMAKE_BUILD_TYPE=Release
# Build in parallel
cmake --build build -j$(nproc)
# Run full test suite
ctest --test-dir build --output-on-failure
# Launch SoEasy
./build/soeasy-gui💡 Aligned with official GitHub Actions CI: MSYS2 MINGW64 environment is recommended:
# 1. Open MSYS2 MINGW64 terminal and install toolchain & dependencies
pacman -S --noconfirm \
git base-devel \
mingw-w64-x86_64-toolchain \
mingw-w64-x86_64-cmake \
mingw-w64-x86_64-ninja \
mingw-w64-x86_64-qt5-base \
mingw-w64-x86_64-capnproto \
mingw-w64-x86_64-opencv \
mingw-w64-x86_64-libxml2 \
mingw-w64-x86_64-libiconv \
mingw-w64-x86_64-zlib \
mingw-w64-x86_64-uchardet \
mingw-w64-x86_64-boost
# 2. Configure and build with Ninja
cmake -B build -G "Ninja" -DCMAKE_BUILD_TYPE=Release
cmake --build build -j$(nproc)
# 3. Run full automated tests
ctest --test-dir build --output-on-failure
# 4. Launch executable
./build/soeasy-gui.exeSoEasy is distributed under the Apache-2.0 License.
