diff --git a/javascript/tagAutocomplete.js b/javascript/tagAutocomplete.js
index 9ebed8c..ec1a985 100644
--- a/javascript/tagAutocomplete.js
+++ b/javascript/tagAutocomplete.js
@@ -1348,6 +1348,13 @@ async function refreshTacTempFiles(api = false) {
     }
 }
 
+async function refreshEmbeddings() {
+    await postAPI("tacapi/v1/refresh-embeddings", null);
+    embeddings = [];
+    await processQueue(QUEUE_FILE_LOAD, null);
+    console.log("TAC: Refreshed embeddings");
+}
+
 function addAutocompleteToArea(area) {
     // Return if autocomplete is disabled for the current area type in config
     let textAreaId = getTextAreaIdentifier(area);
@@ -1452,6 +1459,7 @@ async function setup() {
                 if (mutation.type === "attributes" && mutation.attributeName === "title") {
                     currentModelHash = mutation.target.title;
                     updateModelName();
+                    refreshEmbeddings();
                 }
             }
         });
diff --git a/scripts/tag_autocomplete_helper.py b/scripts/tag_autocomplete_helper.py
index 07057c5..0a0ae15 100644
--- a/scripts/tag_autocomplete_helper.py
+++ b/scripts/tag_autocomplete_helper.py
@@ -262,20 +262,22 @@ def _get_lyco():
 
 # Attempt to use the build-in Lora.networks Lora/LyCORIS models lists.
 try:
-    import importlib
-    lora_networks = importlib.import_module("extensions-builtin.Lora.networks")
+    import sys
+    from modules import extensions
+    sys.path.append(Path(extensions.extensions_builtin_dir).joinpath("Lora").as_posix())
+    import lora # pyright: ignore [reportMissingImports]
 
     def _get_lora():
         return [
             Path(model.filename).absolute()
-            for model in lora_networks.available_networks.values()
+            for model in lora.available_loras.values()
             if Path(model.filename).absolute().is_relative_to(LORA_PATH)
         ]
 
     def _get_lyco():
         return [
             Path(model.filename).absolute()
-            for model in lora_networks.available_networks.values()
+            for model in lora.available_loras.values()
             if Path(model.filename).absolute().is_relative_to(LYCO_PATH)
         ]
 
@@ -379,12 +381,24 @@ if EMB_PATH.exists():
     # Get embeddings after the model loaded callback
     script_callbacks.on_model_loaded(get_embeddings)
 
+def refresh_embeddings(force: bool, *args, **kwargs):
+    try:
+        # Fix for SD.Next infinite refresh loop due to gradio not updating after model load on demand.
+        # This will just skip embedding loading if no model is loaded yet (or there really are no embeddings).
+        # Try catch is just for safety incase sd_hijack access fails for some reason.
+        loaded = sd_hijack.model_hijack.embedding_db.word_embeddings
+        skipped = sd_hijack.model_hijack.embedding_db.skipped_embeddings
+        if len((loaded | skipped)) > 0:
+            load_textual_inversion_embeddings(force_reload=force)
+            get_embeddings(None)
+    except Exception:
+        pass
+
 def refresh_temp_files(*args, **kwargs):
     global WILDCARD_EXT_PATHS
     WILDCARD_EXT_PATHS = find_ext_wildcard_paths()
-    load_textual_inversion_embeddings(force_reload = True) # Instant embedding reload.
     write_temp_files()
-    get_embeddings(shared.sd_model)
+    refresh_embeddings(force=True)
 
 def write_temp_files():
     # Write wildcards to wc.txt if found
@@ -597,6 +611,10 @@ def api_tac(_: gr.Blocks, app: FastAPI):
     async def api_refresh_temp_files():
         refresh_temp_files()
 
+    @app.post("/tacapi/v1/refresh-embeddings")
+    async def api_refresh_embeddings():
+        refresh_embeddings(force=False)
+
     @app.get("/tacapi/v1/lora-info/{lora_name}")
     async def get_lora_info(lora_name):
         return await get_json_info(LORA_PATH, lora_name)