Update app.py

app.py

@@ -1,14 +1,14 @@
 import streamlit as st
 from transformers import pipeline
 
-# Load models
+# Load pipelines
 emotion_classifier = pipeline(
     "text-classification",
     model="j-hartmann/emotion-english-distilroberta-base",
     return_all_scores=True
 )
 intent_classifier = pipeline("zero-shot-classification", model="facebook/bart-large-mnli")
-text_generator = pipeline("text2text-generation", model="declare-lab/flan-alpaca-base")
+text_generator = pipeline("text2text-generation", model="google/flan-t5-large")
 
 candidate_tasks = [
     "change mobile plan", "top up balance", "report service outage",
@@ -48,17 +48,22 @@ def get_emotion_score(emotion):
 def generate_response(intent, human=True):
     if human:
         prompt = (
-            f"You are a telecom customer service agent. For the customer intent '{intent}', generate a professional 3-part response:\n"
-            "1. Greeting (e.g., Thank you for contacting us...)\n"
-            "2. Current plan summary (e.g., You're currently on Plan X at ¥X/month. We suggest Plan Y with XXGB for ¥Y/month.)\n"
-            "3. Close with a question (e.g., Would you like to switch?)"
+            f"You are a professional telecom support agent. The customer intends to '{intent}'. "
+            "Please generate a structured 3-part response:\n\n"
+            "1. Start with a short, polite greeting.\n"
+            "2. Mention customer is currently on a fictional plan (e.g., Plan X, ¥68/month), then recommend a better plan (e.g., Plan Y, ¥88/month, 20GB).\n"
+            "3. Ask if they'd like to proceed with the switch or need more details."
         )
     else:
         prompt = (
-            f"As a telecom assistant, generate a full 3-part structured response for customer intent: '{intent}'.\n"
-            "Include greeting, summary of their current plan and a better offer (fictional), and end with a follow-up question."
+            f"The user intends to '{intent}' in a telecom support context. "
+            "Generate a complete structured reply with three parts:\n\n"
+            "1. Greet the user.\n"
+            "2. Provide a fictional current plan and recommend an upgrade.\n"
+            "3. End with a follow-up question to confirm.\n\n"
+            "Do not ask what plan the user has — assume and invent fictional details clearly."
         )
-    result = text_generator(prompt, max_new_tokens=120, do_sample=False)
+    result = text_generator(prompt, max_new_tokens=150, do_sample=False)
     return result[0]['generated_text'].strip()
 
 # Streamlit UI
@@ -81,14 +86,14 @@ if selected_customer not in st.session_state.chat_sessions:
 session = st.session_state.chat_sessions[selected_customer]
 st.title("Smart Customer Support Assistant (for Agents Only)")
 
-# Conversation History
+# Show conversation
 st.markdown("### Conversation")
 for msg in session["chat"]:
     avatar = "👤" if msg['role'] == 'user' else ("🤖" if msg.get("auto") else "👨‍💼")
     with st.chat_message(msg['role'], avatar=avatar):
         st.markdown(msg['content'])
 
-# Customer Input
+# Input + Analyze
 col1, col2 = st.columns([6, 1])
 with col1:
     user_input = st.text_input("Enter customer message:", key="customer_input")
@@ -127,11 +132,10 @@ with col2:
                 session["agent_reply"] = ""
             st.rerun()
 
-# Status Summary
+# Priority & Agent reply
 if session["support_required"]:
     st.markdown(f"### {session['support_required']}")
 
-# Agent Reply
 st.subheader("Agent Response Console")
 session["agent_reply"] = st.text_area("Compose your reply:", value=session["agent_reply"], key="agent_reply_box")
 if st.button("Send Reply"):
@@ -142,7 +146,7 @@ if st.button("Send Reply"):
         session["support_required"] = ""
         st.rerun()
 
-# Human-level Recommendation
+# Detected needs
 if session["system_result"] is not None:
     st.markdown("#### Customer Status")
     st.markdown(f"- **Emotion:** {session['system_result']['emotion'].capitalize()}")
@@ -153,7 +157,7 @@ if session["system_result"] is not None:
         st.markdown(f"**• {intent.capitalize()}**")
         st.code(suggestion)
 
-# End
+# End conversation
 if st.button("End Conversation"):
     session["chat"] = []
     session["system_result"] = None