Update app.py

app.py

@@ -60,28 +60,17 @@ def init_embedding_model():
     
 @st.cache_resource
 def init_vector_search() -> MongoDBAtlasVectorSearch:
-    from langchain_community.embeddings import HuggingFaceInferenceAPIEmbeddings
-    from langchain_community.vectorstores import MongoDBAtlasVectorSearch
-
-    HF_TOKEN = os.getenv("HF_TOKEN", "").strip()
-    # model_name = "thenlper/gte-small"
+    # Load local embedding model
+    embedding_model = init_embedding_model()
 
     try:
-        st.write(f"🔌 Connecting to Hugging Face model: `{model_name}`")
-        # embedding_model = HuggingFaceInferenceAPIEmbeddings(
-        #     api_key=HF_TOKEN,
-        #     model_name=model_name
-        # )
-
-        embedding_model=init_embedding_model()
-        # Test if embedding works
+        # Test embedding
         test_vector = embedding_model.embed_query("Test query for Grant Buddy")
-        st.success(f"✅ HF embedding model connected. Vector length: {len(test_vector)}")
-
+        st.success(f"✅ Local embedding model loaded. Vector length: {len(test_vector)}")
     except Exception as e:
-        st.error("❌ Failed to connect to Hugging Face Embedding API")
+        st.error("❌ Failed to compute embedding locally")
         st.error(f"Error: {e}")
-        raise e  # Stop app here if embedding fails
+        raise e
 
     # MongoDB setup
     user = quote_plus(os.getenv("MONGO_USERNAME", "").strip())
@@ -93,7 +82,6 @@ def init_vector_search() -> MongoDBAtlasVectorSearch:
 
     MONGO_URI = f"mongodb+srv://{user}:{password}@{cluster}/{db_name}?retryWrites=true&w=majority"
 
-    # Connect to vector search
     try:
         vector_store = MongoDBAtlasVectorSearch.from_connection_string(
             connection_string=MONGO_URI,
@@ -103,12 +91,21 @@ def init_vector_search() -> MongoDBAtlasVectorSearch:
         )
         st.success("✅ Connected to MongoDB Vector Search")
         return vector_store
-
     except Exception as e:
         st.error("❌ Failed to connect to MongoDB Atlas Vector Search")
         st.error(f"Error: {e}")
         raise e
+✅ Optional: Clean Up init_embedding_model() Too
+Just to be safe, you can explicitly name the model inside init_embedding_model():
 
+python
+Copy
+Edit
+@st.cache_resource
+def init_embedding_model():
+    model_name = "sentence-transformers/all-mpnet-base-v2"
+    st.write(f"📦 Loading local embedding model: `{model_name}`")
+    return HuggingFaceEmbeddings(model_name=model_name)
 
 # =================== Format Retrieved Chunks ===================
 def format_docs(docs: List[Document]) -> str: