fix: address all 6 P2 review issues from PR #8 round 4

- BackgroundTasks: per-task Retry now dispatches to the correct backend —
  retryFailedEmbeddings for embedding failures, queueTaggingJobs for tagging
  failures, or both; previously always called embedding retry only

- commands: tag-cloud cache key now hashes both image IDs and embedding bytes
  (xxh3) so re-embedding a file after reindex correctly invalidates the cache;
  removed now-dead fnv_hash_ids helper

- db: update_folder_path uses SUBSTR instead of replace() to rewrite only the
  leading path prefix, preventing corruption when the old folder name also
  appears in a subdirectory name

- db: add_user_tag promotes an existing AI-owned tag to user-owned on conflict
  instead of DO NOTHING, preventing the tagger from deleting a manually added
  tag that coincides with an AI tag

- db/commands: add clear_duplicate_scan_cache / invalidate_duplicate_scan_cache
  so deletion removes the stale persisted cache entry; without this a restart
  would reload groups containing deleted images

- store: setSimilarScope re-runs loadSimilarByRegion (with saved crop) when the
  active collection is Region Search Results, instead of silently switching to
  whole-image similarity and discarding the crop

src-tauri/src/commands.rs

@@ -753,15 +753,6 @@ pub struct TagCloudEntry {
     pub image_ids: Vec<i64>,
 }
 
-fn fnv_hash_ids(ids: &[i64]) -> u64 {
-    let mut h: u64 = 0xcbf29ce484222325;
-    for &id in ids {
-        for b in id.to_le_bytes() {
-            h = h.wrapping_mul(0x100000001b3) ^ (b as u64);
-        }
-    }
-    h
-}
 
 /// Clusters the library's image embeddings with k-means and returns one representative
 /// image per cluster — the member whose embedding is closest to its cluster centroid.
@@ -782,11 +773,22 @@ pub async fn get_tag_cloud(
         return Ok(vec![]);
     }
 
-    // Compute a hash of the current embedded image IDs (sorted for stability)
+    // Sort by ID for stable ordering; hash both IDs and embedding bytes so that
-    let mut sorted_ids: Vec<i64> = embeddings_with_ids.iter().map(|(id, _)| *id).collect();
+    // replacing a file and regenerating its embedding invalidates the cache even
-    sorted_ids.sort_unstable();
+    // when the set of image IDs hasn't changed.
-    let current_hash = fnv_hash_ids(&sorted_ids);
+    let mut sorted_pairs: Vec<_> = embeddings_with_ids.iter().collect();
-
+    sorted_pairs.sort_unstable_by_key(|(id, _)| *id);
+    let current_hash = {
+        use xxhash_rust::xxh3::Xxh3;
+        let mut hasher = Xxh3::new();
+        for (id, embedding) in &sorted_pairs {
+            hasher.update(&id.to_le_bytes());
+            for val in embedding.iter() {
+                hasher.update(&val.to_le_bytes());
+            }
+        }
+        hasher.digest()
+    };
     let folder_scope = match folder_id {
         Some(id) => format!("folder_{}", id),
         None => "all".to_string(),
@@ -1097,6 +1099,19 @@ pub async fn load_duplicate_scan_cache(
     }
 }
 
+#[tauri::command]
+pub async fn invalidate_duplicate_scan_cache(
+    db: State<'_, DbState>,
+    folder_id: Option<i64>,
+) -> Result<(), String> {
+    let folder_scope = match folder_id {
+        Some(id) => format!("folder:{}", id),
+        None => "all".to_string(),
+    };
+    let conn = db.get().map_err(|e| e.to_string())?;
+    db::clear_duplicate_scan_cache(&conn, &folder_scope).map_err(|e| e.to_string())
+}
+
 #[tauri::command]
 pub async fn delete_images_from_disk(
     db: State<'_, DbState>,

src-tauri/src/db.rs

@@ -1371,9 +1371,10 @@ pub fn update_folder_path(conn: &Connection, folder_id: i64, old_path: &str, new
     )?;
     // Rewrite image paths so the indexer can match them by path and skip
     // re-generating thumbnails and embeddings for unchanged files.
-    // SQLite's replace() does a literal prefix substitution on each path.
+    // Use SUBSTR to replace only the leading prefix; SQLite's replace()
+    // would corrupt paths where the old folder name also appears deeper in the tree.
     conn.execute(
-        "UPDATE images SET path = replace(path, ?1, ?2) WHERE folder_id = ?3",
+        "UPDATE images SET path = ?2 || SUBSTR(path, LENGTH(?1) + 1) WHERE folder_id = ?3",
         params![old_path, new_path, folder_id],
     )?;
     Ok(())
@@ -1887,7 +1888,11 @@ pub fn add_user_tag(conn: &Connection, image_id: i64, tag: &str) -> Result<Image
     conn.execute(
         "INSERT INTO image_tags (image_id, tag, source, ai_model, confidence, created_at)
          VALUES (?1, ?2, 'user', NULL, NULL, datetime('now'))
-         ON CONFLICT(image_id, tag) DO NOTHING",
+         ON CONFLICT(image_id, tag) DO UPDATE SET
+             source = 'user',
+             ai_model = NULL,
+             confidence = NULL
+         WHERE source = 'ai'",
         params![image_id, tag],
     )?;
     let row = conn.query_row(
@@ -2130,6 +2135,15 @@ pub fn get_duplicate_scan_cache(
     }
 }
 
+/// Deletes the duplicate scan cache for the given scope.
+pub fn clear_duplicate_scan_cache(conn: &Connection, folder_scope: &str) -> Result<()> {
+    conn.execute(
+        "DELETE FROM duplicate_scan_cache WHERE folder_scope = ?1",
+        params![folder_scope],
+    )?;
+    Ok(())
+}
+
 /// Upserts the duplicate scan cache for the given scope.
 pub fn set_duplicate_scan_cache(
     conn: &Connection,

src-tauri/src/lib.rs

@@ -132,6 +132,7 @@ pub fn run() {
             commands::search_tags_autocomplete,
             commands::find_duplicates,
             commands::load_duplicate_scan_cache,
+            commands::invalidate_duplicate_scan_cache,
             commands::delete_images_from_disk,
             commands::rename_folder,
             commands::update_folder_path,

src/components/BackgroundTasks.tsx

@@ -57,6 +57,7 @@ export function BackgroundTasks() {
   const indexingProgress = useGalleryStore((state) => state.indexingProgress);
   const mediaJobProgress = useGalleryStore((state) => state.mediaJobProgress);
   const retryFailedEmbeddings = useGalleryStore((state) => state.retryFailedEmbeddings);
+  const queueTaggingJobs = useGalleryStore((state) => state.queueTaggingJobs);
   const duplicateScanning = useGalleryStore((state) => state.duplicateScanning);
   const duplicateScanProgress = useGalleryStore((state) => state.duplicateScanProgress);
   const [expanded, setExpanded] = useState(false);
@@ -475,7 +476,10 @@ export function BackgroundTasks() {
                   {taskHasFailed && (
                     <button
                       className="rounded-lg border border-amber-500/20 bg-amber-500/10 px-2 py-0.5 text-[11px] text-amber-300 hover:bg-amber-500/20 transition-colors shrink-0"
-                      onClick={() => void retryFailedEmbeddings(task.id)}
+                      onClick={() => {
+                        if (task.hasFailedEmbeddings) void retryFailedEmbeddings(task.id);
+                        if (task.hasFailedTagging) void queueTaggingJobs(task.id);
+                      }}
                     >
                       Retry
                     </button>

src/store.ts

@@ -1126,10 +1126,14 @@ export const useGalleryStore = create<GalleryState>((set, get) => ({
 
   setSimilarScope: (similarScope) => {
     set({ similarScope });
-    const { similarSourceImageId, similarSourceFolderId, selectedFolderId } = get();
+    const { similarSourceImageId, similarSourceFolderId, selectedFolderId, collectionTitle, similarCrop } = get();
     if (similarSourceImageId === null) return;
     const folderId = similarScope === "current_folder" ? (similarSourceFolderId ?? selectedFolderId) : null;
+    if (collectionTitle === "Region Search Results" && similarCrop !== null) {
+      void get().loadSimilarByRegion(similarSourceImageId, similarCrop, folderId, similarSourceFolderId);
+    } else {
       void get().loadSimilarImages(similarSourceImageId, folderId, true, similarSourceFolderId);
+    }
   },
 
   suggestImageTags: async (imageId) => {
@@ -1498,7 +1502,7 @@ export const useGalleryStore = create<GalleryState>((set, get) => ({
   clearDuplicateSelection: () => set({ duplicateSelectedIds: new Set() }),
 
   deleteSelectedDuplicates: async () => {
-    const { duplicateSelectedIds } = get();
+    const { duplicateSelectedIds, duplicateScanFolderId } = get();
     const ids = Array.from(duplicateSelectedIds);
     if (ids.length === 0) return 0;
     const deleted = await invoke<number>("delete_images_from_disk", { params: { image_ids: ids } });
@@ -1509,6 +1513,8 @@ export const useGalleryStore = create<GalleryState>((set, get) => ({
         .map((g) => ({ ...g, images: g.images.filter((img) => !duplicateSelectedIds.has(img.id)) }))
         .filter((g) => g.images.length > 1),
     }));
+    // Invalidate the persisted cache so a restart doesn't reload stale entries
+    await invoke("invalidate_duplicate_scan_cache", { folderId: duplicateScanFolderId ?? null });
     return deleted;
   },