Architecture Overview

nque is split into three cooperating layers.

┌─────────────────────────────────────────────────────────────┐
│  React 19 frontend (src/)                                   │
│  components · Zustand slices · lib/tauri.ts · lib/events.ts │
└───────────────▲───────────────────────────────┬─────────────┘
                │ invoke(command)               │ emit/listen(event)
┌───────────────┴───────────────────────────────▼─────────────┐
│  Tauri 2 shell (src-tauri/)                                 │
│  lib.rs wiring · commands.rs · window management            │
└───────────────▲─────────────────────────────────────────────┘
                │ plain Rust calls
┌───────────────┴─────────────────────────────────────────────┐
│  Cargo workspace crates (crates/)                           │
│  shared · ocr (ingest) · classify · show (construct/filter) │
└─────────────────────────────────────────────────────────────┘

Frontend

A React SPA built with Vite, styled with Tailwind CSS 4 and shadcn/ui, with all global state in Zustand. It never talks to the filesystem directly; it calls Tauri commands through thin wrappers in src/lib/tauri.ts and subscribes to events via src/lib/events.ts. See Frontend and State management.

Tauri shell

src-tauri/ owns the app lifecycle, the window definitions, and the invoke_handler that registers every command. It contains almost no business logic: it resolves native dialogs, marshals arguments, moves blocking work onto a blocking thread, and emits progress events. See Backend.

Domain crates

The real logic lives in a Cargo workspace under crates/:

Crate	Responsibility
shared	Small cross-crate pure utilities.
ocr	Ingest — PDF → text (native text layer, else render + OCR).
classify	Classify — line-level heading, stage-direction, and dialog rules.
show	Construct + filter — build the show tree, clean it, save/load it.

Keeping the pipeline in plain Rust crates (not src-tauri) makes it testable without a Tauri runtime and reusable outside the desktop app.

Data flow of a PDF import

1. Frontend calls import_pdf_show.
2. Tauri opens a file dialog and reads the PDF bytes.
3. ocr::pipeline::extract_text probes the text layer and extracts or OCRs.
4. show::pipeline::build_show_from_text classifies, constructs, and filters.
5. Progress is emitted as ocr-import-progress events throughout.
6. The resulting ShowFile is returned and loaded into the store.

Continue with Processing pipelines.