feat(discovery): AI tagging, semantic search, similarity, duplicate finder, folder management

Adds a full discovery and organisation layer to the gallery.

## AI Tagger
- WD tagger (ONNX via ort) with DirectML/CPU acceleration
- Per-image and per-folder tag queuing; configurable batch size and
  confidence threshold; model downloaded on first use via ureq/zip
- Tags stored with source ('ai' | 'user'); user tags are never
  overwritten by AI; AI tags protected from accidental promotion
- Tagger pause/resume per folder; in-flight batches discarded cleanly
  on pause or cancellation without leaving jobs stuck in processing

## Semantic & Tag Search
- CLIP text-query embedding via candle (HuggingFace hub model)
- Progressive candidate doubling with filter post-processing so
  folder/rating/media-kind filters do not silently under-return results
- Tag search with DB-backed pagination (offset + COUNT total)
- Filename search unchanged; prefix syntax: s:, t:, f: in search bar

## Similarity Search
- HNSW in-memory index (hnsw_rs) with monotonic embedding_revision
  counter for cache invalidation; build_index retries if a concurrent
  write advances the revision during construction
- Folder-scoped similar search uses brute-force cosine scan (no k limit)
- Region-based similarity: crop an arbitrary rectangle in the lightbox,
  embed it with CLIP, find the nearest images globally or within folder
- Pagination for both similar and region results; scope toggle
  (all media / current folder) re-runs the query without reopening image

## Duplicate Finder
- Three-phase detection: stat (size) → sample hash (4×16 KB windows)
  → full-file hash (xxh3, large files only) to eliminate false positives
- Filesystem-first deletion: only removes DB rows for files successfully
  deleted from disk; failed files remain visible for retry
- Persisted scan cache (SQLite) with invalidation on reindex and deletion;
  both global and per-folder scopes invalidated after any deletion

## Tag Cloud & Explore
- Visual cluster explore: k-means over CLIP embeddings, configurable k,
  representative thumbnail per cluster; cache keyed on xxh3 of embedding set
- Tag explore: ranked tag list with image counts and representative
  thumbnail per tag; invalidated when AI tagging completes or folder removed
- Tag autocomplete for search bar

## Folder Management
- Inline rename (display name only, not OS rename)
- Relocate: file-picker to update folder path; all image paths rewritten
  using SUBSTR prefix replacement to avoid corrupting paths that contain
  the folder name as a substring
- Missing-folder recovery banner: Locate or Remove, never silent deletion
- Right-click context menu on sidebar items: Reindex, Rename, Locate,
  Remove; hover buttons preserved alongside context menu

## Background Tasks
- Per-folder progress panel with embedding, tagging, caption, and
  thumbnail stage breakdown
- Retry button per task for failed embeddings and tagging jobs
- Completed-task notifications (system tray via tauri-plugin-notification)
- Scan errors shown inline; WalkDir permission errors protect existing
  records from deletion rather than cascading data loss

## Backend correctness
- sqlite-vec DML performed outside transactions throughout (virtual-table
  limitation); embedding revision incremented on delete as well as insert
- Tagging job discard checks status = 'processing' so paused jobs reset
  to 'pending' are also skipped, not just cancelled ones
- Stale async responses in Lightbox guarded with cancellation flags and
  currentImageIdRef for both tag-load and tag-add callbacks
- Duplicate cache cleared on reindex; folder-removal clears vector rows
  outside transaction before deleting the folder row

README.md

@@ -1,116 +1,58 @@
 # Phokus
 
-## Overview
+A local-first desktop media library for browsing, filtering, and curating image and video folders.
 
-Phokus is a Tauri desktop app for building a fast, local media library from folders on disk. It indexes images and videos, stores metadata in SQLite, and gives you a dense browsing workflow with filtering, favorites, ratings, and a lightbox preview.
+## Features
 
-The current app is optimized for:
-
-- local folders instead of cloud import flows
-- large visual libraries
-- quick review and curation
-- mixed image and video browsing
-
-## Current features
-
-- Add and remove media folders
-- Background indexing with progress updates
-- Browse all media or filter by folder
-- Search by filename
-- Filter by images, videos, or favorites
-- Sort by modified date, name, or file size
-- Grid density controls
-- Lightbox preview with keyboard navigation
-- Favorite and star-rating metadata saved in SQLite
-- Virtualized/local-first architecture built on Tauri + React
+- Add and remove media folders; background indexing with live progress
+- Browse all media or filter by folder, type (image/video), favorites, or star rating
+- **Filename search**, **semantic search** (`/s query`), and **tag search** (`/t tag`)
+- **Similar image search** — find visually similar media by image or selected region
+- **AI tagging** via WD tagger (ONNX, CPU/DirectML) with confidence threshold control
+- **Explore view** — visual cluster map and tag cloud for browsing by theme
+- **Duplicate finder** — scan for exact duplicates by file hash with bulk delete
+- Lightbox preview with keyboard navigation, inline tag editing, and rating controls
+- Sort by date, name, size, rating, or duration
+- Grid density controls (compact / comfortable / detail)
 
 ## Supported formats
 
-Images:
-
-- `jpg`
-- `jpeg`
-- `png`
-- `gif`
-- `bmp`
-- `tiff`
-- `tif`
-- `webp`
-- `avif`
-- `heic`
-- `heif`
-
-Videos:
-
-- `mp4`
-- `mov`
-- `m4v`
-- `webm`
+| Images | Videos |
+|--------|--------|
+| jpg, jpeg, png, gif, bmp | mp4, mov, m4v |
+| tiff, tif, webp, avif, heic, heif | webm |
 
 ## Stack
 
-- Tauri 2
-- React 19
-- TypeScript
-- Zustand
-- Rust
-- SQLite + `sqlite-vec`
-- Vite
-
-## Project structure
-
-- `src/`: React UI, state, and components
-- `src-tauri/src/commands.rs`: Tauri command surface
-- `src-tauri/src/db.rs`: SQLite schema and queries
-- `src-tauri/src/indexer.rs`: folder crawling and batch indexing
-- `src-tauri/src/vector.rs`: vector table setup for future semantic workflows
+- Tauri 2 + Rust backend
+- React 19 + TypeScript + Zustand
+- SQLite + `sqlite-vec` (vector search)
+- ONNX Runtime (`ort`) for AI tagging
+- Candle (Rust ML) for visual embeddings
+- FFmpeg sidecar for video thumbnails and metadata
+- Vite + Tailwind CSS v4
 
 ## Development
 
-### Prerequisites
-
-- Node.js 20+
-- `pnpm`
-- Rust toolchain
-- Tauri system prerequisites for Windows
-
-### Install
+**Prerequisites:** Node.js 20+, pnpm, Rust toolchain, Tauri system prerequisites for Windows.
 
 ```bash
 pnpm install
-```
 
-### Run in development
+# Run with hot-reload (frontend + Rust)
+pnpm dev:app
 
-```bash
-pnpm tauri dev
-```
+# Frontend only
+pnpm dev:vite
 
-### Build
-
-```bash
-pnpm tauri build
+# Production build
+pnpm build:app
 ```
 
 ## How it works
 
-1. Add a folder from the sidebar or Library menu.
-2. The Rust indexer walks the directory recursively.
-3. Supported files are written into SQLite with metadata such as path, size, dimensions, media type, rating, and favorite state.
+1. Add a folder from the sidebar — the Rust indexer walks it recursively.
+2. Supported files are written to SQLite with metadata (path, dimensions, media type, etc.).
+3. Background workers generate thumbnails, compute visual embeddings, and run AI tagging.
 4. Progress events stream back to the UI while the gallery updates incrementally.
-5. The gallery view loads media in pages and opens items in a lightbox for review.
-
-## Notes
-
-- This is currently a local desktop library, not a sync product.
-- Search is filename-based right now.
-- The vector table and embedding fields exist, but semantic search is not wired into the UI yet.
-- Some visible UI copy may still use the old working name until the frontend text is updated.
-
-## Positioning
-
-The clearest product description today is:
-
-> A local-first desktop media library for browsing, filtering, and curating image and video folders.
-
-That description is more accurate than "gallery" alone and gives you a better base for future branding, onboarding copy, and a landing page.
+5. Embeddings power semantic search and the similar images feature via an HNSW index.