Update changelog and README to reflect new features and improvements in V1.0.8

CHANGELOG.md

@@ -6,6 +6,39 @@ into one file for convenience.
 
 ---
 
+## [V1.0.8] - 2026-07-14
+
+### Features
+- **Lightbox image navigation** — while previewing a matched image, press `←` / `→`
+  to jump to the previous / next match without leaving the lightbox. Floating circular
+  nav arrows sit in the dark screen margins (outside the image, so they never obscure
+  content), and an image counter (e.g. `3 / 12`) is centered at the top.
+- **Image info in preview** — new `GET /api/image-info` endpoint returns the image
+  width, height, and file size. The lightbox meta panel now shows the resolution and a
+  human-readable file size (e.g. `1291 × 800 · 97.7 KB`).
+
+### Performance & Robustness
+- **Cache auto-pruning** — the OCR cache (`ocr_cache/*.json`) and thumbnail cache
+  (`thumb_cache/*.webp`) are now periodically pruned so they can't grow without bound.
+  Two new settings, `max_ocr_cache_files` (default 5000) and `max_thumb_cache_files`
+  (default 3000), cap each cache; the oldest files (by mtime) beyond the limit are
+  deleted during normal cache writes.
+- **`prune_cache_dir` rewrite** — replaced the O(n²) `pop(0)` loop with a single slice,
+  and clarified the `max_files` contract: negative = unlimited (no prune), 0 = remove
+  all matching files, positive = keep at most N.
+
+### Testing & CI
+- **E2E scan-stream integration tests** — `TestScanStreamIntegration` drives the full
+  SSE scan pipeline through `fastapi.testclient.TestClient`: happy-path match + copy,
+  no-match / no-copy, and bad target dir → `400` with no scan-lock leak.
+- **`requirements-test.txt`** — new test-only dependency file pinning `httpx` (required
+  by `TestClient`). It is not bundled into the packaged app; CI installs it separately.
+- **Cross-platform CI** — the workflow now runs the matrix on `ubuntu-latest` **and**
+  `windows-latest` with `fail-fast: false`, so one OS failing no longer cancels the
+  other. Fixes the CI break where `TestClient` raised because `httpx` was absent.
+
+---
+
 ## [V1.0.7] - 2026-07-12
 
 ### Features

README.md

@@ -14,6 +14,8 @@ FocusOCR is a lightweight, fully offline desktop web app that scans images for t
 - **ANY / AND matching** — OR logic copies to per-keyword folders; AND logic creates a single combined folder
 - **Auto-organized output** — matched images are copied into `dest/<keyword>/` folders with conflict resolution
 - **Thumbnail gallery** — cached WebP thumbnails, lightbox preview, highlighted OCR snippets
+- **Lightbox navigation** — browse matches with `←` / `→` arrow keys (or floating side arrows); an image counter shows position (e.g. `3 / 12`)
+- **Image info in preview** — the lightbox shows each image's resolution and file size via `GET /api/image-info`
 - **Export CSV / JSON** — download match results with one click
 - **Scan history** — last 10 scans saved in browser; click to restore parameters and gallery
 - **Folder history** — recent target/dest directories remembered per-field
@@ -24,6 +26,7 @@ FocusOCR is a lightweight, fully offline desktop web app that scans images for t
 - **Scan concurrency protection** — overlapping scans return 409; stale locks from disconnected clients auto-reclaim after 60s
 - **Toast notifications** — auto-dismissing toasts replace alert() for non-blocking feedback
 - **OCR result caching** — `~/.focusocr/ocr_cache/` avoids re-scanning unchanged files; configurable toggle in UI
+- **Bounded caches** — OCR and thumbnail caches are auto-pruned (oldest-first) to `max_ocr_cache_files` / `max_thumb_cache_files` so disk usage stays capped
 - **Scan result persistence** — save/load results to `~/.focusocr/results/`; auto-pruned to 20 files; each deletable from the Load Results modal
 - **Bounding box overlay** — lightbox highlights exactly which text regions matched, mapped by line index (not text content) to avoid duplicate highlights on repeated text
 - **Confidence filter** — slider (0–1) excludes low-quality OCR text from matching
@@ -62,7 +65,9 @@ venv/
 tests/
   test_ocr_engine.py             Unit tests for OCR engine and scan lock
   test_config.py                 Unit tests for settings validation
-  test_app.py                    Unit tests for API endpoints (results, reveal, lock leak)
+  test_app.py                    Unit tests + SSE scan-stream integration tests
+requirements.txt                 Runtime dependencies
+requirements-test.txt            Test-only dependencies (httpx for TestClient)
 ```
 
 ---
@@ -101,6 +106,7 @@ Server starts on port **9000** (falls back to 9001, 9002...).
 
 ### Run tests
 ```bash
+pip install -r requirements-test.txt   # httpx, needed by the SSE integration tests
 python -m unittest discover -s tests -v
 ```
 
@@ -118,6 +124,8 @@ Settings stored in `~/.focusocr/config.json`:
   "max_snippets_per_match": 3,
   "max_history_per_dir": 5,
   "max_saved_results": 20,
+  "max_ocr_cache_files": 5000,
+  "max_thumb_cache_files": 3000,
   "enable_ocr_cache": true
 }
 ```
@@ -130,6 +138,8 @@ Settings stored in `~/.focusocr/config.json`:
 | `max_snippets_per_match` | `3` | Snippets shown per match in gallery |
 | `max_history_per_dir` | `5` | Recent directories remembered |
 | `max_saved_results` | `20` | Max result files kept in `~/.focusocr/results/` |
+| `max_ocr_cache_files` | `5000` | Max OCR cache files kept (oldest pruned first) |
+| `max_thumb_cache_files` | `3000` | Max thumbnail cache files kept (oldest pruned first) |
 | `enable_ocr_cache` | `true` | Skip re-scan of unchanged files |
 
 The cache toggle and confidence slider are also available directly in the UI.