Update app.py

app.py

@@ -98,6 +98,40 @@ index = initialize_faiss()
 # Save FAISS index after modifications
 save_faiss_index(index)
 
+# Load document store and populate FAISS index
+knowledgebase_file = = os.path.join(UPLOAD_DIR, "knowledge_text.txt")  # Ensure this path is correct
+
+def load_document_store():
+    """Loads knowledgebase.txt into a dictionary where FAISS IDs map to text and embeddings"""
+    global document_store
+    document_store = {}  # Reset document store
+    all_texts = []
+    
+    if os.path.exists(knowledgebase_file):
+        with open(knowledgebase_file, "r", encoding="utf-8") as f:
+            lines = f.readlines()
+        
+        for i, line in enumerate(lines):
+            text = line.strip()
+            if text:
+                document_store[i] = {"text": text}  # Store text mapped to FAISS ID
+                all_texts.append(text)  # Collect all texts for embedding
+
+        print(f"Loaded {len(document_store)} documents into document_store.")
+    else:
+        print("Error: knowledgebase.txt not found!")
+    
+    # Generate embeddings for all documents
+    embeddings = bertmodel.encode(all_texts)
+    embeddings = embeddings.astype("float32")
+    
+    # Add embeddings to FAISS index
+    index.add_with_ids(embeddings, np.array(list(document_store.keys()), dtype=np.int64))
+    print(f"Added {len(all_texts)} document embeddings to FAISS index.")
+
+# Load document store and index the documents
+load_document_store()
+
 # Function to upload document
 def upload_document(file_path, embed_model):
     try:
@@ -279,11 +313,16 @@ def generate_response(user_input, model_id):
        
         # Append chat history
         func_caller = []
-        
+         
         query_vector = bertmodel.encode(user_input).reshape(1, -1).astype("float32")
         D, I = index.search(query_vector, 1)
-        
-        retrieved_knowledge = document_store.get(I[0][0], {}).get("text", "No relevant information found.")
+
+        # Retrieve document
+        retrieved_id = I[0][0]
+        retrieved_knowledge = (
+            document_store.get(retrieved_id, {}).get("text", "No relevant information found.")
+            if retrieved_id != -1 else "No relevant information found."
+        )
         
         # Construct the knowledge prompt
         prompt = f"Use the following knowledge:\n{retrieved_knowledge}"