Slice 2: Region Capture → Reading Window with Dummy Data

[emcifuentesc]

Parent

• PRD: NihongOCR MVP #1 (PRD: NihongOCR MVP)

What to build

Add the first complete user-facing path: menu bar "Capture Region" triggers macOS screencapture -i, the selected region is saved to a temp file, and the path is sent to the Python sidecar (still returning hardcoded Reading JSON). The Reading Engine renders the response fully: horizontal layout, furigana (ruby text) above Kanji, Romaji below each line, alternating background colors per line, copy buttons, and always-on-top centered window behavior. Temp files are cleaned up after processing.

This slice builds the entire UI shell and input plumbing using fake data, so the product is demoable even before real OCR is wired in.

Acceptance criteria

• Menu bar icon offers "Capture Region" action
• screencapture -i integration captures region and saves to temp file
• Capture Service passes temp path to Sidecar Manager
• Sidecar returns hardcoded Reading JSON (horizontal, 2+ lines, mixed Kanji/Kana)
• Reading window opens centered, always-on-top, closeable
• Reading Engine renders horizontal layout with ruby furigana above Kanji
• Romaji row appears directly below each Japanese line
• Consecutive lines have alternating background colors
• Two copy buttons: "Copy Japanese" (surface forms) and "Copy Romaji"
• Temp image file deleted immediately after processing
• New capture replaces the current Reading in the same window
• Closing the window returns app to menu-bar-only state
• Reading Engine component tests verify DOM structure and copy-button behavior with mock Reading data

Blocked by

• Slice 1: Project Scaffold + Sidecar Lifecycle #2 (Slice 1: Project Scaffold + Sidecar Lifecycle)

Testing notes

Write Reading Engine component tests using mock Reading data. Verify horizontal layout DOM structure, furigana placement, alternating colors, and that copy buttons extract expected plain text.

[emcifuentesc]

This was generated by AI during triage.

Agent Brief

Goal: Build the first complete user-facing path — menu bar "Capture Region" triggers screencapture -i, the image is saved to a temp file, and the Reading Engine renders a hardcoded Reading JSON with full UI: horizontal layout, furigana (ruby text), Romaji, alternating line colors, copy buttons, always-on-top centered window.

Context:

• Slice 1 (Slice 1: Project Scaffold + Sidecar Lifecycle #2) is complete: Tauri v2 scaffold, menu bar, splash screen, Python sidecar stub with NDJSON protocol
• Domain glossary and ADRs in repo root
• PRD: PRD: NihongOCR MVP #1

Current state:

• src-tauri/src/lib.rs — sets up tray, splash window, spawns sidecar
• src-tauri/src/sidecar.rs — SidecarManager with NDJSON stdin/stdout
• src/splash.html + src/splash.ts — Tailwind splash screen
• src/index.html + src/main.ts — placeholder main window
• python-worker/main.py — stub that returns dummy Reading JSON

What to build:

1. Capture Service (Rust) — Tauri command capture_region() that runs screencapture -i to a temp file, sends the path to the sidecar via SidecarManager::send(), and returns the Reading JSON to the frontend.
2. Menu bar action — Add "Capture Region" to the tray menu. On click, trigger the capture flow.
3. Reading Engine (TypeScript) — Render a Reading with:
   • Horizontal layout: Japanese line with ruby furigana above Kanji, Romaji row below
   • Alternating background colors per line pair
   • Two copy buttons: "Copy Japanese" (surface forms) and "Copy Romaji"
   • Always-on-top, centered, closeable window
4. Temp file cleanup — Delete the temp image after processing.
5. Tests — Reading Engine component tests with mock Reading data.

Key acceptance criteria:

• Menu bar "Capture Region" works
• screencapture -i saves to temp file
• Sidecar returns hardcoded Reading JSON
• Reading window renders correctly with all UI elements
• Copy buttons work
• Temp file cleaned up
• Component tests pass

Blockers: None — Slice 1 is complete.