GPU to CPU

app.py

@@ -23,8 +23,8 @@ r = redis.Redis(
     decode_responses=True
 )
 
-# Device configuration
-device = "cuda" if torch.cuda.is_available() else "cpu"
+# Device configuration - always use CPU
+device = "cpu"
 
 # Load CSV from Google Drive
 def load_csv_from_drive():
@@ -105,8 +105,8 @@ def retrieve_relevant_resources(query_vector, embeddings, similarity_threshold=0
     query_embedding = torch.from_numpy(query_vector).to(torch.float32)
     if len(query_embedding.shape) == 1:
         query_embedding = query_embedding.unsqueeze(0)
-    query_embedding = query_embedding.cuda()
-
+    
+    # Removed CUDA-specific code
     if embeddings.shape[1] != query_embedding.shape[1]:
         query_embedding = torch.nn.functional.pad(
             query_embedding,
@@ -190,8 +190,8 @@ def ask_with_history_v3(query: str, conversation_id: str, isFirst):
         embeddings=embeddings
     )
     
-    scores_cpu = [score.cpu() for score in scores]
-    filtered_pairs = [(score, idx) for score, idx in zip(scores_cpu, indices) if score.item() >= threshold]
+    # No need for CPU conversion since we're already on CPU
+    filtered_pairs = [(score.item(), idx) for score, idx in zip(scores, indices) if score.item() >= threshold]
     
     if filtered_pairs:
         filtered_scores, filtered_indices = zip(*filtered_pairs)