Update app.py

app.py

@@ -9,9 +9,12 @@ print("Loading sentence-transformer model for retrieval...")
 retriever_model = SentenceTransformer('all-MiniLM-L6-v2')
 print("Retriever model loaded.")
 
+# --- THIS IS THE UPDATED LINE ---
 print("Loading generative model for answering (google/flan-t5-base)...")
+# Using the balanced 'base' model for better performance and reliability.
 generator_pipe = pipeline("text2text-generation", model="google/flan-t5-base", device=-1)
 print("Generative model loaded.")
+# --- END OF UPDATE ---
 
 
 # --- 2. Setup ChromaDB ---
@@ -68,14 +71,12 @@ except ValueError:
     print("ChromaDB collection loaded.")
 
 
-# --- 3. Define Chatbot Logic (UPDATED SECTION) ---
+# --- 3. Define Chatbot Logic ---
 def chatbot_response(message, history):
     query_embedding = retriever_model.encode([message]).tolist()
-    
-    # --- 1. INCREASED CONTEXT ---
     results = collection.query(
         query_embeddings=query_embedding,
-        n_results=10 # Retrieve more context
+        n_results=5 # Using 5 results is a good balance for the base model
     )
     retrieved_documents = results['documents'][0]
     
@@ -83,12 +84,8 @@ def chatbot_response(message, history):
         return "I'm sorry, I couldn't find any relevant information in the chat history. 🤔"
     
     context = "\n- ".join(retrieved_documents)
-    
-    # --- 2. ENHANCED PROMPT ---
     prompt = f"""
-    Based ONLY on the following excerpts from a WhatsApp chat, provide a detailed and comprehensive answer to the user's question. 
-    Synthesize the information from all relevant excerpts into a single, well-structured response.
-    Do not make up information.
+    Based on the following excerpts from a WhatsApp chat, provide a helpful and accurate answer to the user's question.
 
     Chat Context:
     - {context}
@@ -96,11 +93,10 @@ def chatbot_response(message, history):
     Question:
     {message}
 
-    Detailed Answer:
+    Answer:
     """
     
-    # --- 3. INCREASED MAX LENGTH ---
-    generated_text = generator_pipe(prompt, max_length=400, num_beams=5, early_stopping=True)
+    generated_text = generator_pipe(prompt, max_length=150, num_beams=5, early_stopping=True)
     response = generated_text[0]['generated_text']
     
     return response
@@ -108,10 +104,10 @@ def chatbot_response(message, history):
 # --- 4. Create the Gradio Interface ---
 iface = gr.ChatInterface(
     fn=chatbot_response,
-    title="WhatsApp Chat Bot ✍️",
-    description="Ask me anything about this WhatsApp chat history. I will provide a detailed response.",
+    title="WhatsApp Chat Bot ⚡️",
+    description="Ask me anything about this WhatsApp chat history. (Powered by flan-t5-base)",
     theme="soft", 
-    examples=["Summarize the conversation about the event in detail.", "What were all the points raised about the project deadline?"],
+    examples=["What was the final decision on the project deadline?", "Summarize the conversation about the event."],
     cache_examples=False
 )