fix(recipes): allow recipes storage path on a different drive (Windows)

os.path.commonpath raises ValueError for paths on different Windows
drives. Treat that as no common root so cross-drive recipes migrations
succeed instead of failing with 'Invalid recipes path change'.

locales/en.json

@@ -773,6 +773,8 @@
       "deleteAll": "Delete Selected",
       "downloadMissingLoras": "Download Missing LoRAs",
       "downloadExamples": "Download Example Images",
+      "downloadMissingExamples": "Download Missing",
+      "reprocessExamples": "Re-process All",
       "clear": "Clear Selection",
       "skipMetadataRefreshCount": "Skip ({count} models)",
       "resumeMetadataRefreshCount": "Resume ({count} models)",
@@ -808,6 +810,8 @@
       "sendToWorkflowReplace": "Send to Workflow (Replace)",
       "openExamples": "Open Examples Folder",
       "downloadExamples": "Download Example Images",
+      "downloadMissingExamples": "Download Missing",
+      "reprocessExamples": "Re-process All",
       "replacePreview": "Replace Preview",
       "setContentRating": "Set Content Rating",
       "moveToFolder": "Move to Folder",

py/nodes/save_image.py

@@ -252,6 +252,13 @@ class SaveImageLM:
                         "tooltip": "When enabled, embeds generation parameters into the saved image metadata. Disable to skip writing generation metadata.",
                     },
                 ),
+                "add_loras_to_prompt": (
+                    "BOOLEAN",
+                    {
+                        "default": False,
+                        "tooltip": "When enabled, appends the LoRA syntax line (e.g. <lora:name:strength>) after the positive prompt in the saved metadata.",
+                    },
+                ),
                 "add_counter_to_filename": (
                     "BOOLEAN",
                     {
@@ -348,7 +355,7 @@ class SaveImageLM:
         type_lower = model_type.lower() if model_type else "other"
         return f"urn:air:{slug}:{type_lower}:civitai:{model_id}@{version_id}"
 
-    def format_metadata(self, metadata_dict: dict) -> str:
+    def format_metadata(self, metadata_dict: dict, add_loras_to_prompt: bool = False) -> str:
         """Format metadata as A1111-compatible parameters string with Hashes JSON and Civitai resources."""
         if not metadata_dict: return ""
 
@@ -458,7 +465,10 @@ class SaveImageLM:
         scheduler_name = scheduler_mapping.get(scheduler, scheduler) if scheduler else None
 
         # Build output lines
-        lines = [prompt] if prompt else [""]
+        prompt_line = prompt if prompt else ""
+        if add_loras_to_prompt and loras_text:
+            prompt_line = f"{prompt_line}\n{loras_text}" if prompt_line else loras_text
+        lines = [prompt_line] if prompt_line else [""]
         if negative_prompt:
             lines.append(f"Negative prompt: {negative_prompt}")
 
@@ -793,6 +803,7 @@ class SaveImageLM:
         save_with_metadata=True,
         add_counter_to_filename=True,
         save_as_recipe=False,
+        add_loras_to_prompt=False,
     ):
         """Save images with metadata"""
         results = []
@@ -801,7 +812,7 @@ class SaveImageLM:
         raw_metadata = get_metadata()
         metadata_dict = MetadataProcessor.to_dict(raw_metadata, id)
 
-        metadata = self.format_metadata(metadata_dict)
+        metadata = self.format_metadata(metadata_dict, add_loras_to_prompt)
 
         # Process filename_prefix with pattern substitution
         filename_prefix = self.format_filename(filename_prefix, metadata_dict)
@@ -943,6 +954,7 @@ class SaveImageLM:
         save_with_metadata=True,
         add_counter_to_filename=True,
         save_as_recipe=False,
+        add_loras_to_prompt=False,
     ):
         """Process and save image with metadata"""
         # Make sure the output directory exists
@@ -974,6 +986,7 @@ class SaveImageLM:
             save_with_metadata,
             add_counter_to_filename,
             save_as_recipe,
+            add_loras_to_prompt,
         )
 
         return {

py/services/settings_manager.py

@@ -1473,10 +1473,12 @@ class SettingsManager:
 
         try:
             common_root = os.path.commonpath([source, target])
-        except ValueError as exc:
-            raise ValueError("Invalid recipes path change") from exc
+        except ValueError:
+            # Windows: paths on different drives share no common root.
+            # A cross-drive move is valid, so treat it as no common root.
+            common_root = None
 
-        if common_root == source:
+        if common_root is not None and common_root == source:
             raise ValueError("Recipes path cannot be moved into a nested directory")
 
         planned_recipe_updates: Dict[str, Dict[str, Any]] = {}

py/utils/example_images_download_manager.py

@@ -172,6 +172,7 @@ class DownloadManager:
         model_types = data.get("model_types", ["lora", "checkpoint"])
         delay = float(data.get("delay", 0.2))
         force = data.get("force", False)
+        model_hashes = data.get("model_hashes", [])
 
         # Step 2: Validate configuration (fast lookup)
         settings_manager = get_settings_manager()
@@ -241,6 +242,7 @@ class DownloadManager:
                         delay,
                         active_library,
                         force,
+                        model_hashes,
                     )
                 )
 
@@ -577,8 +579,9 @@ class DownloadManager:
         delay,
         library_name,
         force: bool = False,
+        model_hashes: list[str] | None = None,
     ):
-        """Download example images for all models."""
+        """Download example images for all models (or only the given hashes)."""
 
         downloader = await get_downloader()
 
@@ -606,6 +609,18 @@ class DownloadManager:
                         if model.get("sha256"):
                             all_models.append((scanner_type, model, scanner))
 
+            # Restrict to the requested hashes when provided (empty = all models).
+            # Explicit targets are a directed user request, so previously failed
+            # models are retried instead of skipped.
+            explicit_targets = bool(model_hashes)
+            if model_hashes:
+                hash_set = {h.lower() for h in model_hashes}
+                all_models = [
+                    (scanner_type, model, scanner)
+                    for scanner_type, model, scanner in all_models
+                    if model.get("sha256", "").lower() in hash_set
+                ]
+
             # Update total count
             self._progress["total"] = len(all_models)
             logger.debug(f"Found {self._progress['total']} models to process")
@@ -629,6 +644,7 @@ class DownloadManager:
                     downloader,
                     library_name,
                     force,
+                    explicit_targets,
                 )
 
                 # Update progress
@@ -725,6 +741,7 @@ class DownloadManager:
         downloader,
         library_name,
         force: bool = False,
+        explicit_targets: bool = False,
     ):
         """Process a single model download."""
 
@@ -747,8 +764,9 @@ class DownloadManager:
             self._progress["current_model"] = f"{model_name} ({model_hash[:8]})"
             await self._broadcast_progress(status="running")
 
-            # Skip if already in failed models (unless force mode is enabled)
-            if not force and model_hash in self._progress["failed_models"]:
+            # Skip if already in failed models (unless force mode is enabled or
+            # the model was explicitly targeted by hash)
+            if not force and not explicit_targets and model_hash in self._progress["failed_models"]:
                 logger.debug(f"Skipping known failed model: {model_name}")
                 return False
 
@@ -757,30 +775,34 @@ class DownloadManager:
             )
             existing_files = _model_directory_has_files(model_dir)
 
-            # Skip if already processed AND directory exists with files
-            if model_hash in self._progress["processed_models"]:
-                if existing_files:
-                    logger.debug(f"Skipping already processed model: {model_name}")
+            # Model-level guard: a populated folder counts as done. Explicitly
+            # targeted models bypass it so the per-image existence pre-check can
+            # fill individual gaps without re-fetching existing files.
+            if not explicit_targets:
+                # Skip if already processed AND directory exists with files
+                if model_hash in self._progress["processed_models"]:
+                    if existing_files:
+                        logger.debug(f"Skipping already processed model: {model_name}")
+                        return False
+
+                    logger.debug(
+                        "Model %s (%s) marked as processed but folder empty or missing, reprocessing triggered",
+                        model_name,
+                        model_hash,
+                    )
+                    # Track that we are reprocessing this model for summary logging
+                    self._progress["reprocessed_models"].add(model_hash)
+                    # Remove from processed models since we need to reprocess
+                    self._progress["processed_models"].discard(model_hash)
+
+                if existing_files and model_hash not in self._progress["processed_models"]:
+                    logger.debug(
+                        "Model folder already populated for %s, marking as processed without download",
+                        model_name,
+                    )
+                    self._progress["processed_models"].add(model_hash)
                     return False
 
-                logger.debug(
-                    "Model %s (%s) marked as processed but folder empty or missing, reprocessing triggered",
-                    model_name,
-                    model_hash,
-                )
-                # Track that we are reprocessing this model for summary logging
-                self._progress["reprocessed_models"].add(model_hash)
-                # Remove from processed models since we need to reprocess
-                self._progress["processed_models"].discard(model_hash)
-
-            if existing_files and model_hash not in self._progress["processed_models"]:
-                logger.debug(
-                    "Model folder already populated for %s, marking as processed without download",
-                    model_name,
-                )
-                self._progress["processed_models"].add(model_hash)
-                return False
-
             if not model_dir:
                 logger.warning(
                     "Unable to resolve example images folder for model %s (%s)",
@@ -884,7 +906,7 @@ class DownloadManager:
                         model_name,
                     )
                     # Clear failed_models so non-force runs can retry
-                    if force and model_hash in self._progress["failed_models"]:
+                    if (force or explicit_targets) and model_hash in self._progress["failed_models"]:
                         self._progress["failed_models"].discard(model_hash)
                         logger.info(
                             f"Removed {model_name} from failed_models after force retry with rate-limited images"
@@ -904,7 +926,7 @@ class DownloadManager:
                     )
                 elif success:
                     self._progress["processed_models"].add(model_hash)
-                    if force and model_hash in self._progress["failed_models"]:
+                    if (force or explicit_targets) and model_hash in self._progress["failed_models"]:
                         self._progress["failed_models"].discard(model_hash)
                         logger.info(
                             f"Removed {model_name} from failed_models after successful force retry"

py/utils/example_images_processor.py

@@ -113,6 +113,26 @@ class ExampleImagesProcessor:
         message = str(error).lower()
         return '404' in message or 'file not found' in message
 
+    @staticmethod
+    def _example_image_file_exists(model_dir: str, index: int, media_type_hint: str | None = None) -> bool:
+        """Return True when the file that would be written for a media index already exists.
+
+        The final filename (``image_{index}{extension}``) depends on the downloaded
+        content, so the extension cannot be known ahead of time. The post-download
+        check skips the write when the exact target file exists; this pre-check
+        approximates that with the candidate extensions for the media type (videos
+        only when the metadata hints at a video) so the network request is avoided
+        for files that already exist on disk.
+        """
+        if media_type_hint == "video":
+            extensions = SUPPORTED_MEDIA_EXTENSIONS['videos']
+        else:
+            extensions = SUPPORTED_MEDIA_EXTENSIONS['images']
+        return any(
+            os.path.exists(os.path.join(model_dir, f"image_{index}{ext}"))
+            for ext in extensions
+        )
+
     @staticmethod
     async def download_model_images(model_hash, model_name, model_images, model_dir, optimize, downloader):
         """Download images for a single model
@@ -139,7 +159,12 @@ class ExampleImagesProcessor:
             original_url = image_url
             if optimize and 'civitai.com' in image_url:
                 image_url = ExampleImagesProcessor.get_civitai_optimized_url(image_url)
-            
+
+            # Skip the download when the file already exists on disk
+            if ExampleImagesProcessor._example_image_file_exists(model_dir, i, image.get("type")):
+                logger.debug("File already exists, skipping download for %s", image_url)
+                continue
+
             # Download the file first to determine the actual file type
             try:
                 logger.debug(f"Downloading media file {i} for {model_name}")
@@ -229,6 +254,11 @@ class ExampleImagesProcessor:
             if optimize and 'civitai.com' in image_url:
                 image_url = ExampleImagesProcessor.get_civitai_optimized_url(image_url)
 
+            # Skip the download when the file already exists on disk
+            if ExampleImagesProcessor._example_image_file_exists(model_dir, i, image.get("type")):
+                logger.debug("File already exists, skipping download for %s", image_url)
+                continue
+
             async def _attempt_download() -> tuple:
                 logger.debug("Downloading media file %s for %s", i, model_name)
                 return await downloader.download_to_memory(

static/js/api/apiConfig.js

@@ -184,7 +184,8 @@ export const DOWNLOAD_ENDPOINTS = {
     downloadGet: '/api/lm/download-model-get',
     cancelGet: '/api/lm/cancel-download-get',
     progress: '/api/lm/download-progress',
-    exampleImages: '/api/lm/force-download-example-images' // New endpoint for downloading example images
+    exampleImages: '/api/lm/force-download-example-images', // Re-process example images ignoring previous status
+    exampleImagesMissing: '/api/lm/download-example-images' // Download only missing example images
 };
 
 // Hugging Face API endpoints

static/js/api/baseModelApi.js

@@ -1641,7 +1641,7 @@ export class BaseModelApiClient {
         }
     }
 
-    async downloadExampleImages(modelHashes, modelTypes = null) {
+    async downloadExampleImages(modelHashes, modelTypes = null, { force = true } = {}) {
         let ws = null;
 
         await state.loadingManager.showWithProgress(async (loading) => {
@@ -1700,8 +1700,13 @@ export class BaseModelApiClient {
                 // Determine optimize setting
                 const optimize = state.global?.settings?.optimize_example_images ?? true;
 
+                // force=false routes to the regular endpoint, which skips already-processed models
+                const endpoint = force
+                    ? DOWNLOAD_ENDPOINTS.exampleImages
+                    : DOWNLOAD_ENDPOINTS.exampleImagesMissing;
+
                 // Make the API request to start the download process
-                const response = await fetch(DOWNLOAD_ENDPOINTS.exampleImages, {
+                const response = await fetch(endpoint, {
                     method: 'POST',
                     headers: {
                         'Content-Type': 'application/json'
@@ -1710,6 +1715,7 @@ export class BaseModelApiClient {
                         model_hashes: modelHashes,
                         output_dir: outputDir,
                         optimize: optimize,
+                        force: force,
                         model_types: modelTypes || [this.apiConfig.config.singularName]
                     })
                 });

static/js/components/ContextMenu/BulkContextMenu.js

@@ -137,11 +137,10 @@ export class BulkContextMenu extends BaseContextMenu {
             downloadMissingLorasItem.style.display = currentModelType === 'recipes' ? 'flex' : 'none';
         }
 
-        const downloadExampleImagesItem = this.menu.querySelector('[data-action="download-example-images"]');
-        if (downloadExampleImagesItem) {
+        const downloadExampleImagesSubmenu = this.menu.querySelector('[data-has-submenu="download-example-images"]');
+        if (downloadExampleImagesSubmenu) {
             // Show on model pages (loras, checkpoints, embeddings), hide on recipes
-            const modelPages = ['loras', 'checkpoints', 'embeddings'];
-            downloadExampleImagesItem.style.display = modelPages.includes(currentModelType) ? 'flex' : 'none';
+            downloadExampleImagesSubmenu.style.display = ['loras', 'checkpoints', 'embeddings'].includes(currentModelType) ? 'flex' : 'none';
         }
 
         const skipMetadataRefreshItem = this.menu.querySelector('[data-action="skip-metadata-refresh"]');
@@ -294,8 +293,11 @@ export class BulkContextMenu extends BaseContextMenu {
             case 'download-missing-loras':
                 this.handleDownloadMissingLoras();
                 break;
+            case 'download-missing-example-images':
+                this.handleDownloadExampleImages({ force: false });
+                break;
             case 'download-example-images':
-                this.handleDownloadExampleImages();
+                this.handleDownloadExampleImages({ force: true });
                 break;
             case 'clear':
                 bulkManager.clearSelection();
@@ -340,7 +342,7 @@ export class BulkContextMenu extends BaseContextMenu {
         await bulkMissingLoraDownloadManager.downloadMissingLoras(selectedRecipes);
     }
 
-    async handleDownloadExampleImages() {
+    async handleDownloadExampleImages({ force = true } = {}) {
         if (state.selectedModels.size === 0) {
             return;
         }
@@ -361,7 +363,7 @@ export class BulkContextMenu extends BaseContextMenu {
 
         try {
             const apiClient = getModelApiClient();
-            await apiClient.downloadExampleImages([...hashes]);
+            await apiClient.downloadExampleImages([...hashes], null, { force });
         } catch (error) {
             console.error('Bulk download example images failed:', error);
         }

static/js/components/ContextMenu/ModelContextMenuMixin.js

@@ -347,7 +347,10 @@ export const ModelContextMenuMixin = {
                 openExampleImagesFolder(this.currentCard.dataset.sha256);
                 return true;
             case 'download-examples':
-                this.downloadExampleImages();
+                this.downloadExampleImages(false);
+                return true;
+            case 'download-examples-force':
+                this.downloadExampleImages(true);
                 return true;
             case 'civitai':
                 if (this.currentCard.dataset.from_civitai === 'true') {
@@ -378,7 +381,7 @@ export const ModelContextMenuMixin = {
     },
 
     // Download example images method
-    async downloadExampleImages() {
+    async downloadExampleImages(force = false) {
         const modelHash = this.currentCard.dataset.sha256;
         if (!modelHash) {
             showToast('toast.contextMenu.missingHash', {}, 'error');
@@ -387,7 +390,7 @@ export const ModelContextMenuMixin = {
 
         try { 
             const apiClient = getModelApiClient();
-            await apiClient.downloadExampleImages([modelHash]);
+            await apiClient.downloadExampleImages([modelHash], null, { force });
         } catch (error) {
             console.error('Error downloading example images:', error);
         }

templates/checkpoints.html

@@ -32,7 +32,12 @@
     <div class="context-menu-separator menu-section-break"></div>
     <!-- Media / Preview -->
     <div class="context-menu-item" data-action="preview"><i class="fas fa-folder-open"></i> {{ t('loras.contextMenu.openExamples') }}</div>
-    <div class="context-menu-item" data-action="download-examples"><i class="fas fa-download"></i> {{ t('loras.contextMenu.downloadExamples') }}</div>
+    <div class="context-menu-item has-submenu" data-has-submenu="download-examples"><i class="fas fa-download"></i> {{ t('loras.contextMenu.downloadExamples') }} <i class="fas fa-chevron-right submenu-arrow"></i>
+        <div class="context-submenu">
+            <div class="context-menu-item" data-action="download-examples"><i class="fas fa-download"></i> {{ t('loras.contextMenu.downloadMissingExamples') }}</div>
+            <div class="context-menu-item" data-action="download-examples-force"><i class="fas fa-redo-alt"></i> {{ t('loras.contextMenu.reprocessExamples') }}</div>
+        </div>
+    </div>
     <div class="context-menu-item" data-action="replace-preview"><i class="fas fa-image"></i> {{ t('loras.contextMenu.replacePreview') }}</div>
     <div class="context-menu-separator menu-section-break"></div>
     <!-- Attributes -->

templates/components/context_menu.html

@@ -44,8 +44,18 @@
     <div class="context-menu-item" data-action="preview">
         <i class="fas fa-folder-open"></i> <span>{{ t('loras.contextMenu.openExamples') }}</span>
     </div>
-    <div class="context-menu-item" data-action="download-examples">
-        <i class="fas fa-download"></i> <span>{{ t('loras.contextMenu.downloadExamples') }}</span>
+    <div class="context-menu-item has-submenu" data-has-submenu="download-examples">
+        <i class="fas fa-download"></i>
+        <span>{{ t('loras.contextMenu.downloadExamples') }}</span>
+        <i class="fas fa-chevron-right submenu-arrow"></i>
+        <div class="context-submenu">
+            <div class="context-menu-item" data-action="download-examples">
+                <i class="fas fa-download"></i> <span>{{ t('loras.contextMenu.downloadMissingExamples') }}</span>
+            </div>
+            <div class="context-menu-item" data-action="download-examples-force">
+                <i class="fas fa-redo-alt"></i> <span>{{ t('loras.contextMenu.reprocessExamples') }}</span>
+            </div>
+        </div>
     </div>
     <div class="context-menu-item" data-action="replace-preview">
         <i class="fas fa-image"></i> <span>{{ t('loras.contextMenu.replacePreview') }}</span>
@@ -136,8 +146,18 @@
     </div>
     <div class="context-menu-section" data-section="download">
         <div class="context-menu-section-header">{{ t('loras.bulkOperations.sections.download') }}</div>
-        <div class="context-menu-item" data-action="download-example-images">
-            <i class="fas fa-download"></i> <span>{{ t('loras.bulkOperations.downloadExamples') }}</span>
+        <div class="context-menu-item has-submenu" data-has-submenu="download-example-images">
+            <i class="fas fa-download"></i>
+            <span>{{ t('loras.bulkOperations.downloadExamples') }}</span>
+            <i class="fas fa-chevron-right submenu-arrow"></i>
+            <div class="context-submenu">
+                <div class="context-menu-item" data-action="download-missing-example-images">
+                    <i class="fas fa-download"></i> <span>{{ t('loras.bulkOperations.downloadMissingExamples') }}</span>
+                </div>
+                <div class="context-menu-item" data-action="download-example-images">
+                    <i class="fas fa-redo-alt"></i> <span>{{ t('loras.bulkOperations.reprocessExamples') }}</span>
+                </div>
+            </div>
         </div>
         <div class="context-menu-item" data-action="download-missing-loras">
             <i class="fas fa-download"></i> <span>{{ t('loras.bulkOperations.downloadMissingLoras') }}</span>

templates/embeddings.html

@@ -32,7 +32,12 @@
     <div class="context-menu-separator menu-section-break"></div>
     <!-- Media / Preview -->
     <div class="context-menu-item" data-action="preview"><i class="fas fa-folder-open"></i> {{ t('loras.contextMenu.openExamples') }}</div>
-    <div class="context-menu-item" data-action="download-examples"><i class="fas fa-download"></i> {{ t('loras.contextMenu.downloadExamples') }}</div>
+    <div class="context-menu-item has-submenu" data-has-submenu="download-examples"><i class="fas fa-download"></i> {{ t('loras.contextMenu.downloadExamples') }} <i class="fas fa-chevron-right submenu-arrow"></i>
+        <div class="context-submenu">
+            <div class="context-menu-item" data-action="download-examples"><i class="fas fa-download"></i> {{ t('loras.contextMenu.downloadMissingExamples') }}</div>
+            <div class="context-menu-item" data-action="download-examples-force"><i class="fas fa-redo-alt"></i> {{ t('loras.contextMenu.reprocessExamples') }}</div>
+        </div>
+    </div>
     <div class="context-menu-item" data-action="replace-preview"><i class="fas fa-image"></i> {{ t('loras.contextMenu.replacePreview') }}</div>
     <div class="context-menu-separator menu-section-break"></div>
     <!-- Attributes -->

tests/frontend/components/contextMenu.interactions.test.js

@@ -2155,4 +2155,35 @@ describe('Interaction-level regression coverage', () => {
     excludedItem.dispatchEvent(new Event('click', { bubbles: true }));
     expect(window.pageControls.enterExcludedView).toHaveBeenCalledTimes(1);
   });
+
+  it('routes single-model example downloads to missing-only and force paths', async () => {
+    document.body.innerHTML = `
+      <div id="loraContextMenu" class="context-menu">
+        <div class="context-menu-item has-submenu" data-has-submenu="download-examples">
+          <div class="context-submenu">
+            <div class="context-menu-item" data-action="download-examples"></div>
+            <div class="context-menu-item" data-action="download-examples-force"></div>
+          </div>
+        </div>
+      </div>
+    `;
+
+    const { LoraContextMenu } = await import('../../../static/js/components/ContextMenu/LoraContextMenu.js');
+    const contextMenu = new LoraContextMenu();
+
+    const card = document.createElement('div');
+    card.className = 'model-card';
+    card.dataset.filepath = '/models/test.safetensors';
+    card.dataset.sha256 = 'abc123hash';
+    document.body.appendChild(card);
+
+    contextMenu.showMenu(100, 100, card);
+
+    document.querySelector('[data-action="download-examples"]').dispatchEvent(new Event('click', { bubbles: true }));
+    expect(downloadExampleImagesApiMock).toHaveBeenCalledWith(['abc123hash'], null, { force: false });
+
+    contextMenu.showMenu(100, 100, card);
+    document.querySelector('[data-action="download-examples-force"]').dispatchEvent(new Event('click', { bubbles: true }));
+    expect(downloadExampleImagesApiMock).toHaveBeenCalledWith(['abc123hash'], null, { force: true });
+  });
 });

tests/nodes/test_save_image.py

@@ -86,6 +86,41 @@ def test_save_image_skips_png_parameters_when_metadata_disabled_and_keeps_workfl
         assert img.info["workflow"] == json.dumps(workflow)
 
 
+def test_save_image_does_not_append_loras_to_prompt_by_default(monkeypatch, tmp_path):
+    _configure_save_paths(monkeypatch, tmp_path)
+    _configure_metadata(
+        monkeypatch,
+        {"prompt": "prompt text", "seed": 123, "loras": "<lora:foo:0.7>"},
+    )
+
+    node = SaveImageLM()
+    node.save_images([_make_image()], "ComfyUI", "png", id="node-1")
+
+    image_path = tmp_path / "sample_00001_.png"
+    with Image.open(image_path) as img:
+        assert "<lora:" not in img.info["parameters"]
+        assert img.info["parameters"] == "prompt text\nSeed: 123, Version: ComfyUI"
+
+
+def test_save_image_appends_loras_to_prompt_when_enabled(monkeypatch, tmp_path):
+    _configure_save_paths(monkeypatch, tmp_path)
+    _configure_metadata(
+        monkeypatch,
+        {"prompt": "prompt text", "seed": 123, "loras": "<lora:foo:0.7>"},
+    )
+
+    node = SaveImageLM()
+    node.save_images(
+        [_make_image()], "ComfyUI", "png", id="node-1", add_loras_to_prompt=True
+    )
+
+    image_path = tmp_path / "sample_00001_.png"
+    with Image.open(image_path) as img:
+        assert img.info["parameters"] == (
+            "prompt text\n<lora:foo:0.7>\nSeed: 123, Version: ComfyUI"
+        )
+
+
 def test_save_image_skips_jpeg_metadata_when_disabled(monkeypatch, tmp_path):
     _configure_save_paths(monkeypatch, tmp_path)
     _configure_metadata(monkeypatch, {"prompt": "prompt text", "seed": 123})
@@ -451,6 +486,14 @@ class TestParameterDefaultConsistency:
         assert SaveImageLM.save_images.__defaults__[5] == 0
         assert SaveImageLM.process_image.__defaults__[7] == 0
 
+    def test_add_loras_to_prompt_defaults_are_consistent(self):
+        input_types = SaveImageLM.INPUT_TYPES()
+        optional = input_types["optional"]
+
+        assert optional["add_loras_to_prompt"][1]["default"] is False
+        assert SaveImageLM.save_images.__defaults__[-1] is False
+        assert SaveImageLM.process_image.__defaults__[-1] is False
+
 
 def test_png_does_not_pass_webp_method_or_jpeg_subsampling(monkeypatch, tmp_path):
     _configure_save_paths(monkeypatch, tmp_path)

tests/services/test_example_images_download_manager_async.py

@@ -529,7 +529,8 @@ async def test_not_found_example_images_are_cleaned(
 
     model_dir = images_root / model_hash
     model_dir.mkdir(parents=True, exist_ok=True)
-    (model_dir / "image_0.png").write_bytes(b"first")
+    # Pre-existing file collides with the valid image index (1) so the
+    # pre-download existence check must skip it without a network request
     (model_dir / "image_1.png").write_bytes(b"second")
 
     async def fake_process_local_examples(*_args, **_kwargs):
@@ -608,11 +609,188 @@ async def test_not_found_example_images_are_cleaned(
     ]
 
     files = sorted(p.name for p in model_dir.iterdir())
-    assert files == ["image_0.png", "image_1.png"]
-    assert (model_dir / "image_0.png").read_bytes() == b"first"
+    assert files == ["image_1.png"]
     assert (model_dir / "image_1.png").read_bytes() == b"second"
 
 
+async def test_failed_models_retried_when_explicitly_targeted(
+    monkeypatch: pytest.MonkeyPatch,
+    tmp_path,
+    settings_manager,
+):
+    ws_manager = RecordingWebSocketManager()
+    manager = download_module.DownloadManager(ws_manager=ws_manager)
+
+    images_root = tmp_path / "examples"
+    monkeypatch.setitem(settings_manager.settings, "example_images_path", str(images_root))
+
+    model_hash = "a" * 64
+    model_path = tmp_path / "model.safetensors"
+    model_path.write_text("data", encoding="utf-8")
+
+    model_metadata = {
+        "sha256": model_hash,
+        "model_name": "Failed Example",
+        "file_path": str(model_path),
+        "file_name": "model.safetensors",
+        "civitai": {"images": [{"url": "https://example.com/valid.png"}]},
+    }
+
+    scanner = StubScanner([model_metadata.copy()])
+    _patch_scanner(monkeypatch, scanner)
+
+    # Persist a previous failure so the skip path is exercised
+    images_root.mkdir(parents=True, exist_ok=True)
+    (images_root / ".download_progress.json").write_text(
+        json.dumps(
+            {
+                "failed_models": [model_hash],
+                "processed_models": [],
+                "rate_limited_models": [],
+            }
+        ),
+        encoding="utf-8",
+    )
+
+    async def fake_process_local_examples(*_args, **_kwargs):
+        return False
+
+    async def fake_get_updated_model(model_hash_arg, _scanner):
+        return model_metadata
+
+    class DownloaderStub:
+        def __init__(self):
+            self.calls: list[str] = []
+
+        async def download_to_memory(self, url, *_args, **_kwargs):
+            self.calls.append(url)
+            return True, b"\x89PNG\r\n\x1a\n", {"content-type": "image/png"}
+
+    downloader = DownloaderStub()
+
+    async def fake_get_downloader():
+        return downloader
+
+    monkeypatch.setattr(
+        download_module.ExampleImagesProcessor,
+        "process_local_examples",
+        staticmethod(fake_process_local_examples),
+    )
+    monkeypatch.setattr(
+        download_module.MetadataUpdater,
+        "get_updated_model",
+        staticmethod(fake_get_updated_model),
+    )
+    monkeypatch.setattr(download_module, "get_downloader", fake_get_downloader)
+
+    # Without explicit hashes the previously failed model is skipped
+    skipped_manager = download_module.DownloadManager(ws_manager=RecordingWebSocketManager())
+    result = await skipped_manager.start_download({"model_types": ["lora"], "delay": 0})
+    assert result["success"] is True
+    if skipped_manager._download_task is not None:
+        await asyncio.wait_for(skipped_manager._download_task, timeout=1)
+    assert downloader.calls == []
+
+    # With explicit hashes the previously failed model is retried and cleared
+    result = await manager.start_download(
+        {"model_types": ["lora"], "delay": 0, "model_hashes": [model_hash]}
+    )
+    assert result["success"] is True
+    if manager._download_task is not None:
+        await asyncio.wait_for(manager._download_task, timeout=1)
+    assert downloader.calls == ["https://example.com/valid.png"]
+    assert manager._progress["failed_models"] == set()
+    assert model_hash in manager._progress["processed_models"]
+
+
+async def test_explicit_targets_fill_partial_example_gaps(
+    monkeypatch: pytest.MonkeyPatch,
+    tmp_path,
+    settings_manager,
+):
+    ws_manager = RecordingWebSocketManager()
+
+    images_root = tmp_path / "examples"
+    monkeypatch.setitem(settings_manager.settings, "example_images_path", str(images_root))
+
+    model_hash = "b" * 64
+    model_path = tmp_path / "model.safetensors"
+    model_path.write_text("data", encoding="utf-8")
+
+    model_metadata = {
+        "sha256": model_hash,
+        "model_name": "Partial Example",
+        "file_path": str(model_path),
+        "file_name": "model.safetensors",
+        "civitai": {
+            "images": [
+                {"url": "https://example.com/first.png"},
+                {"url": "https://example.com/second.png"},
+            ]
+        },
+    }
+
+    scanner = StubScanner([model_metadata.copy()])
+    _patch_scanner(monkeypatch, scanner)
+
+    # Simulate a partially populated folder: index 0 already downloaded
+    model_dir = images_root / model_hash
+    model_dir.mkdir(parents=True, exist_ok=True)
+    (model_dir / "image_0.png").write_bytes(b"existing")
+
+    async def fake_process_local_examples(*_args, **_kwargs):
+        return False
+
+    async def fake_get_updated_model(model_hash_arg, _scanner):
+        return model_metadata
+
+    class DownloaderStub:
+        def __init__(self):
+            self.calls: list[str] = []
+
+        async def download_to_memory(self, url, *_args, **_kwargs):
+            self.calls.append(url)
+            return True, b"\x89PNG\r\n\x1a\n", {"content-type": "image/png"}
+
+    downloader = DownloaderStub()
+
+    async def fake_get_downloader():
+        return downloader
+
+    monkeypatch.setattr(
+        download_module.ExampleImagesProcessor,
+        "process_local_examples",
+        staticmethod(fake_process_local_examples),
+    )
+    monkeypatch.setattr(
+        download_module.MetadataUpdater,
+        "get_updated_model",
+        staticmethod(fake_get_updated_model),
+    )
+    monkeypatch.setattr(download_module, "get_downloader", fake_get_downloader)
+
+    # Untargeted run treats the populated folder as done
+    untargeted = download_module.DownloadManager(ws_manager=RecordingWebSocketManager())
+    result = await untargeted.start_download({"model_types": ["lora"], "delay": 0})
+    assert result["success"] is True
+    if untargeted._download_task is not None:
+        await asyncio.wait_for(untargeted._download_task, timeout=1)
+    assert downloader.calls == []
+
+    # Explicitly targeted run fills only the missing index, skipping the
+    # existing file without a network request
+    targeted = download_module.DownloadManager(ws_manager=ws_manager)
+    result = await targeted.start_download(
+        {"model_types": ["lora"], "delay": 0, "model_hashes": [model_hash]}
+    )
+    assert result["success"] is True
+    if targeted._download_task is not None:
+        await asyncio.wait_for(targeted._download_task, timeout=1)
+    assert downloader.calls == ["https://example.com/second.png"]
+    assert (model_dir / "image_1.png").exists()
+    assert (model_dir / "image_0.png").read_bytes() == b"existing"
+
+
 @pytest.fixture
 def settings_manager():
     return get_settings_manager()

tests/services/test_settings_manager.py

@@ -860,6 +860,59 @@ def test_set_recipes_path_rewrites_symlinked_recipe_metadata(manager, tmp_path):
     assert not old_json_path.exists()
 
 
+def test_set_recipes_path_allows_cross_drive_migration(manager, tmp_path, monkeypatch):
+    # Windows regression: os.path.commonpath raises ValueError for paths on
+    # different drives (ntpath semantics). Cross-drive moves must succeed.
+    lora_root = tmp_path / "loras"
+    old_recipes_dir = lora_root / "recipes" / "nested"
+    old_recipes_dir.mkdir(parents=True)
+    manager.set("folder_paths", {"loras": [str(lora_root)]})
+
+    recipe_id = "recipe-cross-drive"
+    old_image_path = old_recipes_dir / f"{recipe_id}.webp"
+    old_json_path = old_recipes_dir / f"{recipe_id}.recipe.json"
+    old_image_path.write_bytes(b"image-bytes")
+    old_json_path.write_text(
+        json.dumps(
+            {
+                "id": recipe_id,
+                "file_path": str(old_image_path),
+                "title": "Recipe Cross Drive",
+            }
+        ),
+        encoding="utf-8",
+    )
+
+    new_recipes_dir = tmp_path / "N_drive" / "AI" / "Library" / "Recipes"
+
+    # The effective current recipes dir (source of the migration) is
+    # lora_root/recipes — the nested subdirectory holds the recipe files.
+    source = str(lora_root / "recipes")
+    target = str(new_recipes_dir)
+    real_commonpath = os.path.commonpath
+
+    def fake_commonpath(paths):
+        # Simulate ntpath on Windows: a source/target pair on different
+        # drives shares no common root and raises ValueError.
+        if {source, target} <= set(paths):
+            raise ValueError("Paths don't have the same drive")
+        return real_commonpath(paths)
+
+    monkeypatch.setattr(os.path, "commonpath", fake_commonpath)
+
+    manager.set("recipes_path", str(new_recipes_dir))
+
+    migrated_image_path = new_recipes_dir / "nested" / f"{recipe_id}.webp"
+    migrated_json_path = new_recipes_dir / "nested" / f"{recipe_id}.recipe.json"
+
+    assert manager.get("recipes_path") == str(new_recipes_dir.resolve())
+    assert migrated_image_path.read_bytes() == b"image-bytes"
+    migrated_payload = json.loads(migrated_json_path.read_text(encoding="utf-8"))
+    assert migrated_payload["file_path"] == str(migrated_image_path)
+    assert not old_image_path.exists()
+    assert not old_json_path.exists()
+
+
 def test_set_recipes_path_rejects_file_target(manager, tmp_path):
     lora_root = tmp_path / "loras"
     lora_root.mkdir()

tests/utils/test_example_images_download_manager_unit.py

@@ -63,7 +63,7 @@ async def test_start_download_bootstraps_progress_and_task(
     release = asyncio.Event()
 
     async def fake_download(
-        self, output_dir, optimize, model_types, delay, library_name, force=False
+        self, output_dir, optimize, model_types, delay, library_name, force=False, model_hashes=None
     ):
         started.set()
         await release.wait()
@@ -93,6 +93,44 @@ async def test_start_download_bootstraps_progress_and_task(
     assert manager._progress["status"] == "completed"
 
 
+async def test_start_download_forwards_model_hashes(
+    monkeypatch: pytest.MonkeyPatch, tmp_path
+) -> None:
+    settings_manager = get_settings_manager()
+    settings_manager.settings["example_images_path"] = str(tmp_path)
+    settings_manager.settings["libraries"] = {"default": {}}
+    settings_manager.settings["active_library"] = "default"
+
+    manager = download_module.DownloadManager(ws_manager=RecordingWebSocketManager())
+
+    received: Dict[str, Any] = {}
+
+    async def fake_download(
+        self, output_dir, optimize, model_types, delay, library_name, force=False, model_hashes=None
+    ):
+        received["model_hashes"] = model_hashes
+        async with self._state_lock:
+            self._is_downloading = False
+            self._download_task = None
+        self._progress["status"] = "completed"
+
+    monkeypatch.setattr(
+        download_module.DownloadManager,
+        "_download_all_example_images",
+        fake_download,
+    )
+
+    result = await manager.start_download(
+        {"model_types": ["lora"], "delay": 0, "model_hashes": ["abc123", "def456"]}
+    )
+    assert result["success"] is True
+
+    task = manager._download_task
+    assert task is not None
+    await asyncio.wait_for(task, timeout=1)
+    assert received["model_hashes"] == ["abc123", "def456"]
+
+
 async def test_pause_and_resume_flow(monkeypatch: pytest.MonkeyPatch, tmp_path) -> None:
     settings_manager = get_settings_manager()
     settings_manager.settings["example_images_path"] = str(tmp_path)

tests/utils/test_example_images_processor_unit.py

@@ -100,6 +100,54 @@ def test_get_file_extension_media_type_hint_low_priority() -> None:
     assert ext == ".mp4"
 
 
+def test_example_image_file_exists_checks_plausible_extensions(tmp_path) -> None:
+    proc = processor_module.ExampleImagesProcessor
+    assert proc._example_image_file_exists(str(tmp_path), 0) is False
+    Path(tmp_path, "image_0.webp").write_bytes(b"x")
+    assert proc._example_image_file_exists(str(tmp_path), 0) is True
+    assert proc._example_image_file_exists(str(tmp_path), 1) is False
+
+
+def test_example_image_file_exists_video_hint_only_checks_video_extensions(tmp_path) -> None:
+    proc = processor_module.ExampleImagesProcessor
+    Path(tmp_path, "image_2.jpg").write_bytes(b"x")
+    # An existing image file must not satisfy a video-hinted lookup
+    assert proc._example_image_file_exists(str(tmp_path), 2, "video") is False
+    Path(tmp_path, "image_2.mp4").write_bytes(b"x")
+    assert proc._example_image_file_exists(str(tmp_path), 2, "video") is True
+
+
+async def test_download_model_images_with_tracking_skips_existing_files(tmp_path) -> None:
+    proc = processor_module.ExampleImagesProcessor
+    images = [
+        {"url": "https://image.civitai.com/a/b", "type": "image"},
+        {"url": "https://image.civitai.com/c/d", "type": "image"},
+    ]
+    Path(tmp_path, "image_0.jpg").write_bytes(b"existing")
+
+    class RecordingDownloader:
+        def __init__(self) -> None:
+            self.calls: list[str] = []
+
+        async def download_to_memory(self, url, use_auth=False, return_headers=False):
+            self.calls.append(url)
+            return True, b"\xff\xd8\xff" + b"data", {}
+
+    downloader = RecordingDownloader()
+    success, is_stale, failed, rate_limited = await proc.download_model_images_with_tracking(
+        "hash", "model", images, str(tmp_path), False, downloader
+    )
+
+    assert success is True
+    assert is_stale is False
+    assert failed == []
+    assert rate_limited == []
+    # Only the missing image is requested; the existing one is skipped without a network call
+    assert len(downloader.calls) == 1
+    assert "c/d" in downloader.calls[0]
+    assert Path(tmp_path, "image_1.jpg").exists()
+
+
 class StubScanner:
     def __init__(self, models: list[Dict[str, Any]]) -> None:
         self._cache = SimpleNamespace(raw_data=models)

vue-widgets/src/components/AutocompleteTextWidget.vue

@@ -45,7 +45,7 @@ export interface AutocompleteTextWidgetInterface {
 const props = defineProps<{
   widget: AutocompleteTextWidgetInterface
   node: { id: number }
-  modelType?: 'loras' | 'embeddings' | 'custom_words' | 'prompt'
+  modelType?: 'loras' | 'prompt'
   placeholder?: string
   showPreview?: boolean
   spellcheck?: boolean

vue-widgets/src/components/LoraInfoWidget.vue

@@ -98,7 +98,7 @@ interface LoraInfoWidget {
   onSetValue?: (v: unknown) => void
   callback?: unknown
   options?: {
-    getValue?: () => LoraInfoWidgetValue
+    getValue?: () => unknown
     setValue?: (v: unknown) => void
   }
   node?: { widgets?: Array<{ id?: string }>; widgets_values?: Array<unknown> }
@@ -299,8 +299,12 @@ onMounted(() => {
 
   // ComponentWidgetImpl.value getter/setter delegates to options.getValue/options.setValue.
   // These must be set for workflow JSON persistence (LGraphNode.serialize/configure) to work.
-  props.widget.options.getValue = buildValue
-  props.widget.options.setValue = applyValue
+  if (props.widget.options) {
+    props.widget.options.getValue = buildValue
+    props.widget.options.setValue = applyValue
+  } else {
+    console.warn('[LoraInfoWidget] widget.options missing, value persistence disabled')
+  }
 
   // Also set serializeValue for prompt/API serialization path (executionUtil.ts)
   props.widget.serializeValue = async () => buildValue()

vue-widgets/src/composables/useAutocomplete.ts

@@ -3,7 +3,7 @@ import { ref, onMounted, onUnmounted, type Ref } from 'vue'
 // Dynamic import type for AutoComplete class
 type AutoCompleteClass = new (
   inputElement: HTMLTextAreaElement,
-  modelType: 'loras' | 'embeddings' | 'custom_words' | 'prompt',
+  modelType: 'loras' | 'prompt',
   options?: AutocompleteOptions
 ) => AutoCompleteInstance
 
@@ -29,7 +29,7 @@ export interface UseAutocompleteOptions {
 
 export function useAutocomplete(
   textareaRef: Ref<HTMLTextAreaElement | null>,
-  modelType: 'loras' | 'embeddings' | 'custom_words' | 'prompt' = 'loras',
+  modelType: 'loras' | 'prompt' = 'loras',
   options: UseAutocompleteOptions = {}
 ) {
   const autocompleteInstance = ref<AutoCompleteInstance | null>(null)

vue-widgets/src/main.ts

@@ -36,6 +36,9 @@ const AUTOCOMPLETE_TEXT_MIN_HEIGHT_DEFAULT = 300
 const AUTOCOMPLETE_METADATA_VERSION = 1
 const LORA_MANAGER_WIDGET_IDS_PROPERTY = '__lm_widget_ids'
 
+// Access LiteGraph global for Vue DOM mode detection (matches AutocompleteTextWidget.vue)
+declare const LiteGraph: { vueNodesMode?: boolean } | undefined
+
 // @ts-ignore - ComfyUI external module
 import { app } from '../../../scripts/app.js'
 // @ts-ignore - ComfyUI external module
@@ -718,7 +721,7 @@ function createLoraInfoWidget(node: any) {
 function createAutocompleteTextWidgetFactory(
   node: any,
   widgetName: string,
-  modelType: 'loras' | 'embeddings' | 'prompt',
+  modelType: 'loras' | 'prompt',
   inputOptions: { placeholder?: string } = {}
 ) {
   const metadataWidgetName = `__lm_autocomplete_meta_${widgetName}`
@@ -835,7 +838,7 @@ function createAutocompleteTextWidgetFactory(
     applyAutocompleteTextLayoutFix(
       widget,
       container,
-      typeof LiteGraph !== 'undefined' && LiteGraph.vueNodesMode
+      typeof LiteGraph !== 'undefined' && LiteGraph.vueNodesMode === true
     )
   }
 
@@ -964,13 +967,7 @@ app.registerExtension({
         const options = widgetInputOptions.get(`${node.comfyClass}:text`) || {}
         return createAutocompleteTextWidgetFactory(node, 'text', 'loras', options)
       },
-      // Autocomplete text widget for embeddings (used by Prompt node)
-      // @ts-ignore
-      AUTOCOMPLETE_TEXT_EMBEDDINGS(node) {
-        const options = widgetInputOptions.get(`${node.comfyClass}:text`) || {}
-        return createAutocompleteTextWidgetFactory(node, 'text', 'embeddings', options)
-      },
-      // Autocomplete text widget for prompt (supports both embeddings and custom words)
+      // Autocomplete text widget for prompt (used by Prompt and Text nodes)
       // @ts-ignore
       AUTOCOMPLETE_TEXT_PROMPT(node) {
         const options = widgetInputOptions.get(`${node.comfyClass}:text`) || {}

web/comfyui/vue-widgets/lora-manager-widgets.js

@@ -2118,14 +2118,14 @@ to { transform: rotate(360deg);
   padding: 20px 0;
 }
 
-.autocomplete-text-widget[data-v-3f3d7a1a] {
+.autocomplete-text-widget[data-v-55e3316e] {
   background: transparent;
   height: 100%;
   display: flex;
   flex-direction: column;
   box-sizing: border-box;
 }
-.input-wrapper[data-v-3f3d7a1a] {
+.input-wrapper[data-v-55e3316e] {
   position: relative;
   flex: 1;
   display: flex;
@@ -2133,7 +2133,7 @@ to { transform: rotate(360deg);
 }
 
 /* Canvas mode styles (default) - matches built-in comfy-multiline-input */
-.text-input[data-v-3f3d7a1a] {
+.text-input[data-v-55e3316e] {
   flex: 1;
   width: 100%;
   background-color: var(--comfy-input-bg, #222);
@@ -2152,7 +2152,7 @@ to { transform: rotate(360deg);
 }
 
 /* Vue DOM mode styles - matches built-in p-textarea in Vue DOM mode */
-.text-input.vue-dom-mode[data-v-3f3d7a1a] {
+.text-input.vue-dom-mode[data-v-55e3316e] {
   background-color: var(--color-charcoal-400, #313235);
   color: #fff;
   padding: 8px 12px 30px 12px;  /* Reserve bottom space for clear button */
@@ -2161,12 +2161,12 @@ to { transform: rotate(360deg);
   font-size: 12px;
   font-family: inherit;
 }
-.text-input[data-v-3f3d7a1a]:focus {
+.text-input[data-v-55e3316e]:focus {
   outline: none;
 }
 
 /* Clear button styles */
-.clear-button[data-v-3f3d7a1a] {
+.clear-button[data-v-55e3316e] {
   position: absolute;
   right: 6px;
   bottom: 6px;  /* Changed from top to bottom */
@@ -2189,31 +2189,31 @@ to { transform: rotate(360deg);
 }
 
 /* Show clear button when hovering over input wrapper */
-.input-wrapper:hover .clear-button[data-v-3f3d7a1a] {
+.input-wrapper:hover .clear-button[data-v-55e3316e] {
   opacity: 0.7;
   pointer-events: auto;
 }
-.clear-button[data-v-3f3d7a1a]:hover {
+.clear-button[data-v-55e3316e]:hover {
   opacity: 1;
   background: rgba(255, 100, 100, 0.8);
 }
-.clear-button svg[data-v-3f3d7a1a] {
+.clear-button svg[data-v-55e3316e] {
   width: 12px;
   height: 12px;
 }
 
 /* Vue DOM mode adjustments for clear button */
-.text-input.vue-dom-mode ~ .clear-button[data-v-3f3d7a1a] {
+.text-input.vue-dom-mode ~ .clear-button[data-v-55e3316e] {
   right: 8px;
   bottom: 10px;  /* Changed from top to bottom, adjusted for Vue DOM padding */
   width: 20px;
   height: 20px;
   background: rgba(107, 114, 128, 0.6);
 }
-.text-input.vue-dom-mode ~ .clear-button[data-v-3f3d7a1a]:hover {
+.text-input.vue-dom-mode ~ .clear-button[data-v-55e3316e]:hover {
   background: oklch(62% 0.18 25);
 }
-.text-input.vue-dom-mode ~ .clear-button svg[data-v-3f3d7a1a] {
+.text-input.vue-dom-mode ~ .clear-button svg[data-v-55e3316e] {
   width: 14px;
   height: 14px;
 }
@@ -2224,7 +2224,7 @@ to { transform: rotate(360deg);
   resize: vertical !important;
 }
 
-.lora-info-widget[data-v-a99cc1ab] {
+.lora-info-widget[data-v-d7692b6f] {
   padding: 12px;
   background: rgba(40, 44, 52, 0.6);
   border-radius: 4px;
@@ -2240,45 +2240,45 @@ to { transform: rotate(360deg);
    determined solely by CSS — not by descendant content. This breaks the
    feedback loop where content grows → ResizeObserver resizes → content
    reflows → repeat. Same technique used by tags_widget.js + lm_styles.css. */
-.lora-info-widget.lm-vue-node[data-v-a99cc1ab] {
+.lora-info-widget.lm-vue-node[data-v-d7692b6f] {
   contain: layout size;
 }
 
 /* ── Tab bar ── */
-.lora-info-tabs[data-v-a99cc1ab] {
+.lora-info-tabs[data-v-d7692b6f] {
   display: flex;
   gap: 0;
   margin-bottom: 10px;
   border-bottom: 1px solid var(--border-color, #444);
   flex-shrink: 0;
 }
-.lora-info-tab[data-v-a99cc1ab] {
+.lora-info-tab[data-v-d7692b6f] {
   flex: 1;
   text-align: center;
   cursor: pointer;
   padding: 6px 0;
   position: relative;
 }
-.lora-info-tab-input[data-v-a99cc1ab] {
+.lora-info-tab-input[data-v-d7692b6f] {
   position: absolute;
   opacity: 0;
   width: 0;
   height: 0;
 }
-.lora-info-tab-label[data-v-a99cc1ab] {
+.lora-info-tab-label[data-v-d7692b6f] {
   font-size: 12px;
   font-weight: 500;
   color: var(--fg-color, #fff);
   opacity: 0.5;
   transition: opacity 0.15s;
 }
-.lora-info-tab:hover .lora-info-tab-label[data-v-a99cc1ab] {
+.lora-info-tab:hover .lora-info-tab-label[data-v-d7692b6f] {
   opacity: 0.75;
 }
-.lora-info-tab.active .lora-info-tab-label[data-v-a99cc1ab] {
+.lora-info-tab.active .lora-info-tab-label[data-v-d7692b6f] {
   opacity: 1;
 }
-.lora-info-tab.active[data-v-a99cc1ab]::after {
+.lora-info-tab.active[data-v-d7692b6f]::after {
   content: '';
   position: absolute;
   bottom: -1px;
@@ -2290,16 +2290,16 @@ to { transform: rotate(360deg);
 }
 
 /* ── Tab content ── */
-.tab-content[data-v-a99cc1ab] {
+.tab-content[data-v-d7692b6f] {
   flex: 1;
   min-height: 0;
   overflow: hidden;
 }
-.notes-tab[data-v-a99cc1ab] {
+.notes-tab[data-v-d7692b6f] {
   display: flex;
   flex-direction: column;
 }
-.description-tab[data-v-a99cc1ab] {
+.description-tab[data-v-d7692b6f] {
   display: flex;
   flex-direction: column;
   overflow-y: auto;
@@ -2307,12 +2307,12 @@ to { transform: rotate(360deg);
 }
 
 /* ── Info fields (shared) ── */
-.info-field[data-v-a99cc1ab] {
+.info-field[data-v-d7692b6f] {
   display: flex;
   flex-direction: column;
   gap: 4px;
 }
-.info-label[data-v-a99cc1ab] {
+.info-label[data-v-d7692b6f] {
   font-size: 10px;
   font-weight: 600;
   text-transform: uppercase;
@@ -2320,7 +2320,7 @@ to { transform: rotate(360deg);
   color: var(--fg-color, #fff);
   opacity: 0.6;
 }
-.lora-filename[data-v-a99cc1ab] {
+.lora-filename[data-v-d7692b6f] {
   font-size: 13px;
   font-weight: 500;
   color: var(--fg-color, #fff);
@@ -2331,11 +2331,11 @@ to { transform: rotate(360deg);
   user-select: text;
   -webkit-user-select: text;
 }
-.notes-field[data-v-a99cc1ab] {
+.notes-field[data-v-d7692b6f] {
   flex: 1;
   min-height: 0;
 }
-.lora-notes[data-v-a99cc1ab] {
+.lora-notes[data-v-d7692b6f] {
   width: 100%;
   flex: 1;
   min-height: 60px;
@@ -2350,14 +2350,14 @@ to { transform: rotate(360deg);
   font-family: inherit;
   outline: none;
 }
-.lora-notes[data-v-a99cc1ab]:focus {
+.lora-notes[data-v-d7692b6f]:focus {
   border-color: var(--comfy-input-border, #444);
 }
-.lora-notes[data-v-a99cc1ab]:disabled {
+.lora-notes[data-v-d7692b6f]:disabled {
   opacity: 0.6;
   cursor: not-allowed;
 }
-.save-btn[data-v-a99cc1ab] {
+.save-btn[data-v-d7692b6f] {
   width: 100%;
   margin-top: 8px;
   padding: 6px 12px;
@@ -2371,11 +2371,11 @@ to { transform: rotate(360deg);
   box-sizing: border-box;
   flex-shrink: 0;
 }
-.save-btn[data-v-a99cc1ab]:hover:not(:disabled) {
+.save-btn[data-v-d7692b6f]:hover:not(:disabled) {
   background: rgba(66, 153, 225, 0.25);
   border-color: rgba(66, 153, 225, 0.6);
 }
-.save-btn[data-v-a99cc1ab]:disabled {
+.save-btn[data-v-d7692b6f]:disabled {
   opacity: 0.4;
   cursor: not-allowed;
   background: rgba(66, 153, 225, 0.05);
@@ -2383,7 +2383,7 @@ to { transform: rotate(360deg);
 }
 
 /* ── Description states ── */
-.description-state[data-v-a99cc1ab] {
+.description-state[data-v-d7692b6f] {
   display: flex;
   align-items: center;
   justify-content: center;
@@ -2395,22 +2395,22 @@ to { transform: rotate(360deg);
   min-height: 0;
   flex-shrink: 0;
 }
-.description-state.error[data-v-a99cc1ab] {
+.description-state.error[data-v-d7692b6f] {
   opacity: 0.7;
   color: #f87171;
 }
 
 /* ── Description content ── */
-.description-content[data-v-a99cc1ab] {
+.description-content[data-v-d7692b6f] {
   min-height: 0;
 }
-.description-section[data-v-a99cc1ab] {
+.description-section[data-v-d7692b6f] {
   margin-bottom: 14px;
 }
-.description-section[data-v-a99cc1ab]:last-child {
+.description-section[data-v-d7692b6f]:last-child {
   margin-bottom: 0;
 }
-.description-text[data-v-a99cc1ab] {
+.description-text[data-v-d7692b6f] {
   padding: 8px 0;
   font-size: 12px;
   line-height: 1.5;
@@ -2422,41 +2422,41 @@ to { transform: rotate(360deg);
   user-select: text;
   -webkit-user-select: text;
 }
-.description-text[data-v-a99cc1ab] p {
+.description-text[data-v-d7692b6f] p {
   margin: 0 0 8px 0;
 }
-.description-text[data-v-a99cc1ab] p:last-child {
+.description-text[data-v-d7692b6f] p:last-child {
   margin-bottom: 0;
 }
-.description-text[data-v-a99cc1ab] a {
+.description-text[data-v-d7692b6f] a {
   color: rgba(66, 153, 225, 0.9);
 }
-.description-text[data-v-a99cc1ab] ul,
-.description-text[data-v-a99cc1ab] ol {
+.description-text[data-v-d7692b6f] ul,
+.description-text[data-v-d7692b6f] ol {
   padding-left: 20px;
   margin: 4px 0;
 }
-.description-text[data-v-a99cc1ab] h1,
-.description-text[data-v-a99cc1ab] h2,
-.description-text[data-v-a99cc1ab] h3 {
+.description-text[data-v-d7692b6f] h1,
+.description-text[data-v-d7692b6f] h2,
+.description-text[data-v-d7692b6f] h3 {
   font-size: 13px;
   margin: 10px 0 4px 0;
   font-weight: 600;
   opacity: 0.95;
 }
-.description-text[data-v-a99cc1ab] code {
+.description-text[data-v-d7692b6f] code {
   background: rgba(255, 255, 255, 0.08);
   padding: 1px 4px;
   border-radius: 3px;
   font-size: 11px;
 }
-.description-text[data-v-a99cc1ab] img {
+.description-text[data-v-d7692b6f] img {
   max-width: 100%;
   border-radius: 4px;
 }
 
 /* ── Placeholder (shared) ── */
-.placeholder[data-v-a99cc1ab] {
+.placeholder[data-v-d7692b6f] {
   font-style: italic;
   color: rgba(226, 232, 240, 0.5);
   text-align: center;
@@ -2465,10 +2465,10 @@ to { transform: rotate(360deg);
 }
 
 /* ── Spinner (Font Awesome) ── */
-.fa-spinner[data-v-a99cc1ab] {
-  animation: fa-spin-a99cc1ab 1s linear infinite;
+.fa-spinner[data-v-d7692b6f] {
+  animation: fa-spin-d7692b6f 1s linear infinite;
 }
-@keyframes fa-spin-a99cc1ab {
+@keyframes fa-spin-d7692b6f {
 0% { transform: rotate(0deg);
 }
 100% { transform: rotate(360deg);
@@ -15316,7 +15316,7 @@ const _sfc_main$1 = /* @__PURE__ */ defineComponent({
     };
   }
 });
-const AutocompleteTextWidget = /* @__PURE__ */ _export_sfc(_sfc_main$1, [["__scopeId", "data-v-3f3d7a1a"]]);
+const AutocompleteTextWidget = /* @__PURE__ */ _export_sfc(_sfc_main$1, [["__scopeId", "data-v-55e3316e"]]);
 const _hoisted_1 = { class: "lora-info-tabs" };
 const _hoisted_2 = { class: "tab-content notes-tab" };
 const _hoisted_3 = { class: "info-field" };
@@ -15511,8 +15511,12 @@ const _sfc_main = /* @__PURE__ */ defineComponent({
           if (data.filePath !== void 0) filePath.value = data.filePath;
         }
       };
-      props.widget.options.getValue = buildValue;
-      props.widget.options.setValue = applyValue;
+      if (props.widget.options) {
+        props.widget.options.getValue = buildValue;
+        props.widget.options.setValue = applyValue;
+      } else {
+        console.warn("[LoraInfoWidget] widget.options missing, value persistence disabled");
+      }
       props.widget.serializeValue = async () => buildValue();
       props.widget.onSetValue = applyValue;
       const widgetIndex = (_b = (_a2 = props.widget.node) == null ? void 0 : _a2.widgets) == null ? void 0 : _b.findIndex(
@@ -15641,7 +15645,7 @@ const _sfc_main = /* @__PURE__ */ defineComponent({
     };
   }
 });
-const LoraInfoWidget = /* @__PURE__ */ _export_sfc(_sfc_main, [["__scopeId", "data-v-a99cc1ab"]]);
+const LoraInfoWidget = /* @__PURE__ */ _export_sfc(_sfc_main, [["__scopeId", "data-v-d7692b6f"]]);
 function createVueWidgetCleanup(vueApp, onCleanup) {
   let didUnmount = false;
   return () => {
@@ -16637,7 +16641,7 @@ function createAutocompleteTextWidgetFactory(node, widgetName, modelType, inputO
     applyAutocompleteTextLayoutFix(
       widget,
       container,
-      typeof LiteGraph !== "undefined" && LiteGraph.vueNodesMode
+      typeof LiteGraph !== "undefined" && LiteGraph.vueNodesMode === true
     );
   }
   const vueCleanup = createVueWidgetCleanup(vueApp, () => {
@@ -16747,13 +16751,7 @@ app$1.registerExtension({
         const options = widgetInputOptions.get(`${node.comfyClass}:text`) || {};
         return createAutocompleteTextWidgetFactory(node, "text", "loras", options);
       },
-      // Autocomplete text widget for embeddings (used by Prompt node)
-      // @ts-ignore
-      AUTOCOMPLETE_TEXT_EMBEDDINGS(node) {
-        const options = widgetInputOptions.get(`${node.comfyClass}:text`) || {};
-        return createAutocompleteTextWidgetFactory(node, "text", "embeddings", options);
-      },
-      // Autocomplete text widget for prompt (supports both embeddings and custom words)
+      // Autocomplete text widget for prompt (used by Prompt and Text nodes)
       // @ts-ignore
       AUTOCOMPLETE_TEXT_PROMPT(node) {
         const options = widgetInputOptions.get(`${node.comfyClass}:text`) || {};