new update

app.py

@@ -45,24 +45,8 @@ model = AutoModelForCausalLM.from_pretrained(model_name, device_map="balanced",
 
 # Step 4: Define the Retrieval Function
 def retrieve_documents(query, top_k=3):
-    # Find embeddings matching the query
-    matched_embeddings = [embeddings[i] for i in range(len(metadata)) if query.lower() in metadata[i].lower()]
-    
-    # If no matches found, set a default query embedding
-    if matched_embeddings:
-        query_embedding = np.mean(matched_embeddings, axis=0)
-    else:
-        # Fallback: use the mean of all embeddings as a default embedding
-        query_embedding = np.mean(embeddings, axis=0)
-        print("No exact matches found for query. Using default query embedding.")
-    
-    # Reshape query_embedding to match FAISS expected shape (1, d)
-    query_embedding = query_embedding.reshape(1, -1)
-
-    # Perform the search
-    distances, indices = index.search(query_embedding, top_k)
-    
-    # Retrieve document metadata based on indices
+    query_embedding = np.mean([embeddings[i] for i in range(len(metadata)) if query.lower() in metadata[i].lower()], axis=0)
+    distances, indices = index.search(np.array([query_embedding]), top_k)
     retrieved_docs = [metadata[idx] for idx in indices[0]]
     return retrieved_docs
 
@@ -89,5 +73,5 @@ iface = gr.Interface(
     description="Enter a query to search for relevant courses using Retrieval Augmented Generation."
 )
 
-if __name__ == "__main__":
-    iface.launch()
+if _name_ == "_main_":
+    iface.launch()