Add product context and graph connectivity features

Introduces product_info to session state and injects product context into response and greeting generation for more tailored sales interactions. Refactors UI for product/service setup, streamlines code comments, and adds a check_connectivity method to Graph for verifying graph connectivity in Math Lab. Minor UI and logic improvements throughout for clarity and maintainability.

app.py

@@ -21,17 +21,17 @@ if "page" not in st.session_state: st.session_state.page = "dashboard"
 if "messages" not in st.session_state: st.session_state.messages = []
 if "current_node" not in st.session_state: st.session_state.current_node = "start"
 if "lead_info" not in st.session_state: st.session_state.lead_info = {}
-if "visited_history" not in st.session_state: st.session_state.visited_history = []  # Track visited nodes
+if "product_info" not in st.session_state: st.session_state.product_info = {} # NEW: Product Context
+if "visited_history" not in st.session_state: st.session_state.visited_history = []
 if "current_archetype" not in st.session_state: st.session_state.current_archetype = "UNKNOWN"
 if "reasoning" not in st.session_state: st.session_state.reasoning = ""
 if "current_sentiment" not in st.session_state: st.session_state.current_sentiment = 0.0
-# Checklist status based on your screenshot
 if "checklist" not in st.session_state:
     st.session_state.checklist = {
         "Identify Customer": False,
         "Determine Objectives": False,
         "Outline Advantages": False,
-        "Keep it Brief": True, # Always try to be brief
+        "Keep it Brief": True,
         "Experiment/Revise": False
     }
 
@@ -46,7 +46,6 @@ def init_db():
 
 def save_lead_to_db(lead_info, chat_history, outcome):
     init_db()
-    # Ask AI to extract structured data from chat
     model = genai.GenerativeModel(MODEL_NAME)
     chat_text = "\n".join([f"{m['role']}: {m['content']}" for m in chat_history])
     
@@ -61,7 +60,6 @@ def save_lead_to_db(lead_info, chat_history, outcome):
     """
     try:
         response = model.generate_content(prompt)
-        # Simple parsing (in production use structured output)
         ai_data = response.text
     except:
         ai_data = "AI Extraction Failed"
@@ -72,7 +70,7 @@ def save_lead_to_db(lead_info, chat_history, outcome):
         "Company": lead_info.get("company"),
         "Type": lead_info.get("type"),
         "Context": lead_info.get("context"),
-        "Pain Point": "AI Analysis Pending", # Placeholder for simplicity
+        "Pain Point": "AI Analysis Pending",
         "Budget": "Unknown",
         "Outcome": outcome,
         "Summary": f"Call with {len(chat_history)} messages. {outcome}"
@@ -103,11 +101,9 @@ def load_graph_data():
     return graph, node_to_id, id_to_node, nodes, edges
 
 def get_predicted_path(graph, start_id, target_id, id_to_node, node_to_id):
-    # Get visited node IDs and client type for smart pathfinding
     visited_ids = [node_to_id[n] for n in st.session_state.get('visited_history', []) if n in node_to_id]
     client_type = st.session_state.lead_info.get('type', 'B2B')
     
-    # Use enhanced Bellman-Ford with penalties
     dist = bellman_ford_list(graph, start_id, visited_nodes=visited_ids, client_type=client_type)
     if dist[target_id] == float('inf'): return []
     path = [target_id]
@@ -125,9 +121,6 @@ def get_predicted_path(graph, start_id, target_id, id_to_node, node_to_id):
     return [id_to_node[i] for i in reversed(path)]
 
 def analyze_full_context(model, user_input, current_node, chat_history):
-    """
-    Аналізує Інтент з налаштуванням "NEVER GIVE UP".
-    """
     history_text = "\n".join([f"{m['role']}: {m['content']}" for m in chat_history[-4:]])
     
     prompt = f"""
@@ -141,12 +134,7 @@ def analyze_full_context(model, user_input, current_node, chat_history):
     
     CRITICAL RULES FOR INTENT:
     1. **EXIT** triggers ONLY if user is HOSTILE or EXPLICITLY ends the call.
-       - Examples: "Stop calling me", "Fuck off", "Put me on blacklist", "Bye", "Hang up".
-    
     2. **STAY** (Objection Handling) triggers for ANY resistance.
-       - Examples: "Not interested", "No time", "We have a vendor", "Too expensive", "Send info to mail".
-       - Even if they say "No" to the first question -> It is NOT an exit. It is an objection to handle!
-    
     3. **MOVE** triggers only if user agrees or answers a question positively.
     
     OUTPUT JSON format:
@@ -161,31 +149,40 @@ def analyze_full_context(model, user_input, current_node, chat_history):
         clean_text = response.text.replace("```json", "").replace("```", "").strip()
         return json.loads(clean_text)
     except:
-        # За замовчуванням STAY! Краще зайвий раз перепитати, ніж кинути слухавку.
         return {"archetype": "UNKNOWN", "intent": "STAY", "reasoning": "Fallback safety"}
 
-def generate_response(model, instruction_text, user_input, intent, lead_info, archetype):
+def generate_response(model, instruction_text, user_input, intent, lead_info, archetype, product_info={}):
     """
-    Генерує ЖИВУ відповідь на основі інструкції.
+    Generates a generic or product-specific response.
     """
-    # 1. Формуємо портрет клієнта
     bot_name = lead_info.get('bot_name', 'Олексій')
     client_name = lead_info.get('name', 'Клієнт')
     company = lead_info.get('company', 'Компанія')
     context = lead_info.get('context', 'Cold')
     
-    # 2. Налаштування стилю (Tone of Voice)
     tone = "Professional, confident."
     if archetype == "DRIVER": tone = "Direct, concise, results-oriented (Time is money)."
     elif archetype == "ANALYST": tone = "Logical, factual, detailed."
     elif archetype == "EXPRESSIVE": tone = "Energetic, inspiring, emotional."
     elif archetype == "CONSERVATIVE": tone = "Calm, supportive, reassuring."
     
-    # 3. Налаштування довжини (Brevity)
     length_instruction = "Keep it concise."
     if "Cold" in context: length_instruction = "Extremely short and punchy (Elevator Pitch)."
     
-    # 4. Формуємо Промпт
+    # NEW: Product Context Injection
+    product_context = ""
+    if product_info:
+        product_context = f"""
+        PRODUCT CONTEXT:
+        You are selling: {product_info.get('product_name', 'Our Solution')}
+        Value Proposition: {product_info.get('product_value', 'High Value')}
+        Pricing: {product_info.get('product_price', 'Custom Pricing')}
+        Competitive Edge: {product_info.get('competitor_diff', 'Best in class')}
+        
+        CRITICAL INSTRUCTION:
+        Whenever the script graph says "Pitch", "Price", or "Objection", use the PRODUCT CONTEXT above. Do NOT invent fake features.
+        """
+
     prompt = f"""
     ROLE: You are {bot_name}, a top-tier sales representative at SellMe AI.
     CLIENT: {client_name} from {company}.
@@ -194,6 +191,8 @@ def generate_response(model, instruction_text, user_input, intent, lead_info, ar
     INTENT DETECTED: {intent}
     ARCHETYPE: {archetype}
     
+    {product_context}
+    
     TASK: Generate the spoken response in Ukrainian.
     
     CRITICAL RULES:
@@ -211,17 +210,27 @@ def generate_response(model, instruction_text, user_input, intent, lead_info, ar
     except Exception as e:
         return f"[System Error: {e}]"
 
-def generate_greeting(model, start_instruction, lead_info):
+def generate_greeting(model, start_instruction, lead_info, product_info={}):
     bot_name = lead_info.get('bot_name', 'Manager')
     client_name = lead_info.get('name', 'Client')
     context = lead_info.get('context', 'Cold')
     
+    # NEW: Product Context Injection
+    product_context = ""
+    if product_info:
+        product_context = f"""
+        PRODUCT CONTEXT:
+        You are selling: {product_info.get('product_name', 'Our Solution')}
+        """
+
     prompt = f"""
     ROLE: Sales Rep {bot_name}.
     CLIENT: {client_name}.
     CONTEXT: {context} call.
     INSTRUCTION: "{start_instruction}"
     
+    {product_context}
+    
     TASK: Generate the opening line.
     - If Cold Call: Be brief, aggressive (pattern interrupt).
     - If Warm Call: Be welcoming, reference the application.
@@ -234,30 +243,17 @@ def generate_greeting(model, start_instruction, lead_info):
 
 
 def train_brain():
-    """
-    RL MODULE: Аналізує минулі діалоги і оновлює ваги графа.
-    """
-    # 1. Завантажуємо статистику
     df, _ = get_analytics()
     if df is None or df.empty or "Transcript" not in df.columns:
         return "Недостатньо даних для навчання."
 
-    # 2. Завантажуємо поточний граф
     graph, node_to_id, id_to_node, nodes, edges = load_graph_data()
-    
-    # 3. Аналіз успішних/провальних шляхів
-    # (Це спрощена логіка: ми шукаємо ключові слова кроків у транскрипті)
-    success_bonuses = {} # node_name -> bonus
+    success_bonuses = {}
     
     for index, row in df.iterrows():
         is_success = (row["Outcome"] == "Success")
         transcript = str(row["Transcript"])
-        
-        # Проходимо по всіх вузлах і шукаємо, чи були вони в діалозі
-        # (Це грубий метод, в ідеалі треба зберігати шлях ID у базу)
         for node_name, node_text in nodes.items():
-            # Шукаємо унікальні шматки тексту інструкції в транскрипті, щоб зрозуміти, чи були ми там
-            # Або просто перевіряємо, чи згадується цей етап в логах
             snippet = node_text[:20] 
             if snippet in transcript:
                 if is_success:
@@ -265,31 +261,23 @@ def train_brain():
                 else:
                     success_bonuses[node_name] = success_bonuses.get(node_name, 0) - 1
 
-    # 4. Оновлення ваг (Reinforcement)
     new_edges = []
     changes_log = []
-    
     for edge in edges:
         u_name, v_name = edge["from"], edge["to"]
         old_weight = edge["weight"]
         new_weight = old_weight
-        
-        # Якщо вузол 'to' часто зустрічається в успішних діалогах -> зменшуємо вагу вхідних ребер
         score = success_bonuses.get(v_name, 0)
         
-        if score > 0: # Успішний вузол
-            new_weight *= 0.9 # Знижка 10%
-        elif score < 0: # Провальний вузол
-            new_weight *= 1.1 # Штраф 10%
+        if score > 0: new_weight *= 0.9
+        elif score < 0: new_weight *= 1.1
             
-        # Обмеження щоб ваги не зламались
         new_weight = max(1, min(new_weight, 100))
         new_edges.append({"from": u_name, "to": v_name, "weight": round(new_weight, 2)})
         
         if old_weight != new_weight:
             changes_log.append(f"{u_name}->{v_name}: {old_weight} -> {new_weight}")
 
-    # 5. Збереження "Розумного" файлу
     learned_data = {"nodes": nodes, "edges": new_edges}
     with open("sales_script_learned.json", "w", encoding="utf-8") as f:
         json.dump(learned_data, f, ensure_ascii=False, indent=2)
@@ -300,62 +288,27 @@ def train_brain():
 def draw_graph(graph_data, current_node, predicted_path):
     nodes = graph_data[3]
     edges = graph_data[4]
-    
     dot = graphviz.Digraph()
-    
-    # --- НАЛАШТУВАННЯ ГЕОМЕТРІЇ (Compact Mode) ---
-    dot.attr(
-        rankdir='TB',        # Зверху вниз
-        splines='ortho',     # Ламані лінії (прямі кути)
-        nodesep='0.3',       # Мінімальний відступ збоку
-        ranksep='0.4',       # Мінімальний відступ знизу (робить граф коротшим)
-        bgcolor='transparent' # Прозорий фон, щоб зливався з темою
-    )
-    
-    # --- СТИЛЬ БЛОКІВ (Wide & Slim) ---
-    # shape='note' виглядає як документ, або 'box' для суворості
-    # fixedsize='false' дозволяє блоку розтягуватись під текст, але ми задаємо мінімальну ширину
-    dot.attr('node', 
-             shape='box', 
-             style='rounded,filled', 
-             fontname='Arial', 
-             fontsize='11', 
-             width='2.5',      # Робимо їх широкими
-             height='0.5',     # Робимо їх низькими
-             margin='0.1'      # Менше полів всередині блоку
-    )
-    
-    # --- СТИЛЬ ЛІНІЙ ---
+    dot.attr(rankdir='TB', splines='ortho', nodesep='0.3', ranksep='0.4', bgcolor='transparent')
+    dot.attr('node', shape='box', style='rounded,filled', fontname='Arial', fontsize='11', width='2.5', height='0.5', margin='0.1')
     dot.attr('edge', fontname='Arial', fontsize='9', arrowsize='0.6')
 
     for n in nodes:
-        # Базовий стиль (Світло-сірий, непомітний)
         fill = '#F7F9F9'; color = '#BDC3C7'; pen = '1'; font = '#424949'
-        
-        # Поточний крок (Червоний акцент)
         if n == current_node: 
             fill = '#FF4B4B'; color = '#922B21'; pen = '2'; font = 'white'
-            
-        # Золотий шлях (Жовтий підсвіт)
         elif n in predicted_path: 
             fill = '#FEF9E7'; color = '#F1C40F'; pen = '1'; font = 'black'
-            
-        # Малюємо вузол
         dot.node(n, label=n, fillcolor=fill, color=color, penwidth=pen, fontcolor=font)
         
     for e in edges:
-        color = '#D5D8DC'; pen = '1' # Дуже світлі лінії за замовчуванням
-        
-        # Підсвітка шляху
+        color = '#D5D8DC'; pen = '1'
         if e["from"] in predicted_path and e["to"] in predicted_path:
              try:
-                 # Перевіряємо послідовність
                  if predicted_path.index(e["to"]) == predicted_path.index(e["from"]) + 1:
-                    color = '#F1C40F'; pen = '2.5' # Жирна золота лінія
+                    color = '#F1C40F'; pen = '2.5'
              except: pass
-             
         dot.edge(e["from"], e["to"], color=color, penwidth=pen)
-        
     return dot
 
 # --- MAIN APP ---
@@ -363,21 +316,12 @@ st.sidebar.title("🛠️ SellMe Control")
 mode = st.sidebar.radio("Mode", ["🤖 Sales Bot CRM", "🧪 Math Lab"])
 
 if mode == "🤖 Sales Bot CRM":
-    # --- SALES BOT DEMO (Moved & Indented) ---
-
-    # --- API KEY SETUP (Robust) ---
     api_key = None
     try:
-        # Try to get key from secrets (Cloud)
-        if "GOOGLE_API_KEY" in st.secrets:
-            api_key = st.secrets["GOOGLE_API_KEY"]
-    except:
-        # If secrets file is missing (Local run), just ignore and pass
-        pass
+        if "GOOGLE_API_KEY" in st.secrets: api_key = st.secrets["GOOGLE_API_KEY"]
+    except: pass
 
-    # Fallback to manual input if no key found yet
-    if not api_key:
-        api_key = st.sidebar.text_input("Google API Key", type="password")
+    if not api_key: api_key = st.sidebar.text_input("Google API Key", type="password")
 
     if st.sidebar.button("📊 Dashboard"): st.session_state.page = "dashboard"; st.rerun()
     if st.sidebar.button("📞 New Call"): st.session_state.page = "setup"; st.rerun()
@@ -389,103 +333,84 @@ if mode == "🤖 Sales Bot CRM":
     configure_genai(api_key)
     model = genai.GenerativeModel(MODEL_NAME)
     graph_data = load_graph_data()
-    graph, node_to_id, id_to_node, nodes, edges = graph_data
+    graph, nodes = graph_data[0], graph_data[3]
+    node_to_id, id_to_node = graph_data[1], graph_data[2]
 
-    # --- PAGE: DASHBOARD ---
+    # --- DASHBOARD ---
     if st.session_state.page == "dashboard":
         st.title("📊 CRM & Analytics Hub")
-        
-        # Кнопка для запуску навчання
         if st.button("🧠 Train AI on History (RL)"):
-            with st.spinner("Analyzing patterns... Updating weights..."):
-                msg = train_brain()
+            with st.spinner("Analyzing patterns... Updating weights..."): msg = train_brain()
             st.success(msg)
         
         data, stats = get_analytics()
-        
         if data is not None and not data.empty:
-            # Метрики
             c1, c2, c3 = st.columns(3)
             c1.metric("Total Calls", stats["total"])
             c2.metric("Success Rate", f"{stats['success_rate']}%")
-            c3.metric("AI Learning Iterations", "v1.2") # Фейкова метрика для краси
-            
+            c3.metric("AI Learning Iterations", "v1.2")
             st.divider()
             
-            # Вибір дзвінка для детального аналізу
             st.subheader("🕵️ Call Inspector")
-            
-            # Створюємо список для селектора: "Дата - Ім'я - Результат"
             options = data.apply(lambda x: f"{x['Date']} | {x['Name']} ({x['Outcome']})", axis=1).tolist()
             selected_option = st.selectbox("Select a call to review:", options)
-            
             if selected_option:
-                # Знаходимо вибраний рядок
                 selected_row = data.iloc[options.index(selected_option)]
-                
                 with st.expander("📝 Full Transcript & Insights", expanded=True):
                     st.markdown(f"**Client:** {selected_row['Name']} ({selected_row['Type']})")
                     st.markdown(f"**Result:** {selected_row['Outcome']}")
                     st.text_area("Transcript", str(selected_row.get("Transcript", "No transcript available")), height=300)
-                    
                     if "AI Insights" in selected_row and selected_row["AI Insights"]:
                         st.info(f"💡 **AI Insight:** {selected_row['AI Insights']}")
-                    else:
-                        st.warning("No insights generated for this call.")
-                        
-        else:
-            st.info("Database is empty. Make some calls!")
+                    else: st.warning("No insights generated for this call.")
+        else: st.info("Database is empty. Make some calls!")
 
-    # --- PAGE: SETUP ---
+    # --- SETUP WITH PRODUCT INFO ---
     elif st.session_state.page == "setup":
         st.title("👤 Налаштування Дзвінка")
-        
         with st.form("lead_form"):
-            st.markdown("### 👨‍💼 Хто дзвонить?")
-            bot_name = st.text_input("Ваше ім'я (Менеджера)", "Олексій")
-            
-            st.markdown("### 📞 Кому дзвонимо?")
             c1, c2 = st.columns(2)
+            # Lead Info
+            c1.markdown("### 👨‍💼 Lead Info")
+            bot_name = c1.text_input("Ваше ім'я (Менеджера)", "Олексій")
             name = c1.text_input("Ім'я Клієнта", "Олександр")
-            company = c2.text_input("Компанія (для B2B)", "SoftServe")
-            
+            company = c1.text_input("Компанія", "SoftServe")
             type_ = c1.selectbox("Тип бізнесу", ["B2B", "B2C"])
-            context = c2.selectbox("Контекст", ["Холодний дзвінок", "Теплий лід (заявка)", "Повторний дзвінок"])
+            context = c1.selectbox("Контекст", ["Холодний дзвінок", "Теплий лід", "Повторний дзвінок"])
+
+            # NEW: Product Info
+            c2.markdown("### 📦 Product / Service Info")
+            p_name = c2.text_input("Product Name", "AI Sales Engine")
+            p_value = c2.text_input("Main Benefit (Value)", "Increases sales by 300%")
+            p_price = c2.text_input("Price / Pricing Model", "$100/month")
+            p_diff = c2.text_input("Competitive Edge (Why us?)", "Learns from every call")
             
-            # --- ПЕРЕВІРКА ІСТОРІЇ ---
-            # Спробуємо знайти клієнта в базі (якщо підключено Google Sheets)
-            if st.checkbox("🔍 Перевірити в базі (за ім'ям)"):
+            if c1.checkbox("🔍 Перевірити в базі"):
                 try:
-                    # Це працює, якщо у нас є leads_manager з Google Sheets
                     from leads_manager import connect_to_gsheet
                     sheet = connect_to_gsheet()
                     if sheet:
                         records = sheet.get_all_records()
                         found = [r for r in records if str(r['Name']).lower() == name.lower()]
-                        
                         if found:
-                            last_interaction = found[-1] # Останній запис
-                            st.info(f"📜 Знайдено історію! Останній контакт: {last_interaction['Date']}")
-                            st.warning(f"Результат минулого разу: {last_interaction['Outcome']}")
-                            st.caption(f"Нотатки: {last_interaction.get('Summary', '')}")
-                            
-                            # Можна автоматично змінити контекст на "Повторний"
+                            last = found[-1]
+                            st.info(f"📜 Contact Found: {last['Date']}")
                             context = "Повторний дзвінок" 
-                        else:
-                            st.success("✨ Новий клієнт")
-                except Exception as e:
-                    st.error("База даних недоступна.")
-            
-            submitted = st.form_submit_button("🚀 Почати розмову")
+                        else: st.success("✨ New Client")
+                except: st.error("Database unavailable.")
             
+            submitted = st.form_submit_button("🚀 Start Call")
             if submitted:
-                # Зберігаємо все, включаючи ім'я бота
                 st.session_state.lead_info = {
-                    "bot_name": bot_name,
-                    "name": name,         # <--- ВИПРАВИЛИ НА "name"
-                    "company": company, 
-                    "type": type_, 
-                    "context": context
+                    "bot_name": bot_name, "name": name, 
+                    "company": company, "type": type_, "context": context
+                }
+                # Store Product Info
+                st.session_state.product_info = {
+                    "product_name": p_name,
+                    "product_value": p_value,
+                    "product_price": p_price,
+                    "competitor_diff": p_diff
                 }
                 st.session_state.messages = []
                 st.session_state.current_node = "start"
@@ -494,170 +419,94 @@ if mode == "🤖 Sales Bot CRM":
                 st.session_state.visited_history = []
                 st.rerun()
 
-    # --- PAGE: CHAT ---
+    # --- CHAT ---
     elif st.session_state.page == "chat":
         st.markdown(f"### Call with {st.session_state.lead_info['name']}")
-        
         col_chat, col_tools = st.columns([1.5, 1])
         
         with col_tools:
-            st.markdown("#### 🎯 Call Objectives")
-            # Logic to auto-update checklist based on node
+            st.markdown("#### 🎯 Objectives")
             if "qualification" in st.session_state.current_node: st.session_state.checklist["Identify Customer"] = True
-            if "pain" in st.session_state.current_node or "shame" in st.session_state.current_node: st.session_state.checklist["Determine Objectives"] = True
+            if "pain" in st.session_state.current_node: st.session_state.checklist["Determine Objectives"] = True
             if "pitch" in st.session_state.current_node: st.session_state.checklist["Outline Advantages"] = True
+            for goal, done in st.session_state.checklist.items(): st.write(f"{'✅' if done else '⬜'} {goal}")
             
-            for goal, done in st.session_state.checklist.items():
-                icon = "✅" if done else "⬜"
-                st.write(f"{icon} {goal}")
-            
-            # Display Client Profile (Real-time)
-            st.markdown("#### 🧠 Client Profile (Real-time)")
-            
-            # Get current archetype from session
+            st.markdown("#### 🧠 Profile")
             current_archetype = st.session_state.get("current_archetype", "Analyzing...")
-            
-            # Visual cards for archetypes
             cols = st.columns(4)
-            
-            # Styles for highlighting
-            def get_opacity(target): return "1.0" if current_archetype == target else "0.3"
-            
-            cols[0].markdown(f"<div style='opacity:{get_opacity('DRIVER')}; font-size:20px; text-align:center'>🔴<br>Boss</div>", unsafe_allow_html=True)
-            cols[1].markdown(f"<div style='opacity:{get_opacity('ANALYST')}; font-size:20px; text-align:center'>🔵<br>Analyst</div>", unsafe_allow_html=True)
-            cols[2].markdown(f"<div style='opacity:{get_opacity('EXPRESSIVE')}; font-size:20px; text-align:center'>🟡<br>Fan</div>", unsafe_allow_html=True)
-            cols[3].markdown(f"<div style='opacity:{get_opacity('CONSERVATIVE')}; font-size:20px; text-align:center'>🟢<br>Safe</div>", unsafe_allow_html=True)
-            
-            if st.session_state.reasoning:
-                st.caption(f"🤖 AI Insight: {st.session_state.reasoning}")
+            def op(t): return "1.0" if current_archetype == t else "0.3"
+            cols[0].markdown(f"<div style='opacity:{op('DRIVER')};text-align:center'>🔴<br>Boss</div>", unsafe_allow_html=True)
+            cols[1].markdown(f"<div style='opacity:{op('ANALYST')};text-align:center'>🔵<br>Analyst</div>", unsafe_allow_html=True)
+            cols[2].markdown(f"<div style='opacity:{op('EXPRESSIVE')};text-align:center'>🟡<br>Fan</div>", unsafe_allow_html=True)
+            cols[3].markdown(f"<div style='opacity:{op('CONSERVATIVE')};text-align:center'>🟢<br>Safe</div>", unsafe_allow_html=True)
+            if st.session_state.reasoning: st.caption(f"🤖 {st.session_state.reasoning}")
                 
-            st.markdown("#### 📊 AI Strategy")
+            st.markdown("#### 📊 Strategy")
             curr_id = node_to_id[st.session_state.current_node]
             target_id = node_to_id["close_standard"]
             path = get_predicted_path(graph, curr_id, target_id, id_to_node, node_to_id)
-            st.graphviz_chart(
-                draw_graph(graph_data, st.session_state.current_node, path),
-                use_container_width=True  # Розтягує граф на всю ширину колонки
-            )
+            st.graphviz_chart(draw_graph(graph_data, st.session_state.current_node, path), use_container_width=True)
             
-            # --- BELLMAN-FORD ALGORITHM LOGS ---
-            st.markdown("---")
-            with st.expander("🧮 Алгоритм Беллмана-Форда (Live Logs)", expanded=False):
-                st.markdown("""
-                **Математика прийняття рішень:**
-                Алгоритм шукає шлях $P$, де сума ваг $W$ є мінімальною:
-                $$ D[v] = \\min(D[v], D[u] + W(u, v)) $$
-                """)
-                
-                # 1. Calculate real data
+            with st.expander("🧮 Bellman-Ford Logs"):
                 visited_ids = [node_to_id[n] for n in st.session_state.get('visited_history', []) if n in node_to_id]
                 client_type = st.session_state.lead_info.get('type', 'B2B')
                 current_sentiment = st.session_state.get("current_sentiment", 0.0)
+                raw_dist = bellman_ford_list(graph, curr_id, visited_nodes=visited_ids, client_type=client_type, sentiment_score=current_sentiment)
                 
-                # Call algorithm to get distance array
-                raw_dist = bellman_ford_list(
-                    graph, 
-                    curr_id, 
-                    visited_nodes=visited_ids, 
-                    client_type=client_type, 
-                    sentiment_score=current_sentiment
-                )
-                
-                # 2. Build beautiful table for humans
                 debug_data = []
-                target_path_set = set(path)  # Path we already found for graph
-                
+                target_path_set = set(path)
                 for i, d in enumerate(raw_dist):
                     node_name = id_to_node[i]
-                    
-                    # Format infinity
-                    cost_display = "∞" if d == float('inf') else round(d, 2)
-                    
-                    # Node status
                     status = "⬜"
                     if node_name == st.session_state.current_node: status = "📍 Start"
                     elif node_name in target_path_set: status = "✨ Path"
                     elif d == float('inf'): status = "🚫 Unreachable"
-                    
-                    debug_data.append({
-                        "Node": node_name,
-                        "Cost (Weight)": cost_display,
-                        "Status": status
-                    })
+                    debug_data.append({"Node": node_name, "Cost": "∞" if d==float('inf') else round(d,2), "Status": status})
                 
-                # Convert to DataFrame
                 df_log = pd.DataFrame(debug_data)
-                
-                # Sort: path first, then cheap, then expensive
-                df_log["sort_key"] = df_log["Cost (Weight)"].apply(lambda x: 9999 if x == "∞" else float(x))
-                df_log = df_log.sort_values(by="sort_key").drop(columns=["sort_key"])
-                
-                # Display
-                st.dataframe(
-                    df_log, 
-                    use_container_width=True,
-                    hide_index=True
-                )
-                
-                # 3. Explanation "Why?"
-                st.info(f"""
-                **Фактори впливу:**
-                - 🎭 **Емоція:** {current_sentiment} (впливає на вартість агресивних кроків)
-                - 🏢 **Тип:** {client_type} (змінює пріоритет швидкості)
-                - 🔄 **Повтори:** Вузли, де ми вже були, мають штраф x50.
-                """)
+                df_log["sort"] = df_log["Cost"].apply(lambda x: 9999 if x=="∞" else float(x))
+                st.dataframe(df_log.sort_values("sort").drop(columns=["sort"]), hide_index=True)
 
         with col_chat:
             for msg in st.session_state.messages:
                 with st.chat_message(msg["role"]): st.write(msg["content"])
                 
-            # --- ГЕНЕРАЦІЯ ПЕРШОГО ПОВІДОМЛЕННЯ ---
             if not st.session_state.messages:
-                with st.spinner("AI готується до дзвінка..."):
-                    # Викликаємо AI для генерації живого привітання
-                    greeting = generate_greeting(model, nodes["start"], st.session_state.lead_info)
-                    
+                with st.spinner("AI warming up..."):
+                    # Pass Product Info
+                    greeting = generate_greeting(model, nodes["start"], st.session_state.lead_info, st.session_state.product_info)
                 st.session_state.messages.append({"role": "assistant", "content": greeting})
                 st.rerun()
 
             if user_input := st.chat_input("Reply..."):
                 st.session_state.messages.append({"role": "user", "content": user_input})
-                
-                # Logic - Analyze with full context including archetype detection
                 current_text = nodes[st.session_state.current_node]
                 analysis = analyze_full_context(model, user_input, st.session_state.current_node, st.session_state.messages)
-                intent = analysis.get("intent", "STAY")
-                archetype = analysis.get("archetype", "UNKNOWN")
-                reasoning = analysis.get("reasoning", "")
-                
-                # Store archetype and reasoning for display
+                intent, archetype = analysis.get("intent", "STAY"), analysis.get("archetype", "UNKNOWN")
                 st.session_state.current_archetype = archetype
-                st.session_state.reasoning = reasoning
+                st.session_state.reasoning = analysis.get("reasoning", "")
                 
                 if "EXIT" in intent:
                     outcome = "Success" if "close" in st.session_state.current_node else "Fail"
                     save_lead_to_db(st.session_state.lead_info, st.session_state.messages, outcome)
                     st.success("Call Saved!")
                     st.session_state.page = "dashboard"; st.rerun()
-                
                 elif "STAY" in intent:
-                    resp = generate_response(model, current_text, user_input, "STAY", st.session_state.lead_info, archetype)
-                
+                    # Pass Product Info
+                    resp = generate_response(model, current_text, user_input, "STAY", st.session_state.lead_info, archetype, st.session_state.product_info)
                 else: # MOVE
-                    # Track current node in visited history before moving
-                    if st.session_state.current_node not in st.session_state.visited_history:
+                     if st.session_state.current_node not in st.session_state.visited_history:
                         st.session_state.visited_history.append(st.session_state.current_node)
-                    
-                    curr_id = node_to_id[st.session_state.current_node]
-                    best_next = None; min_w = float('inf')
-                    for n, w in graph.adj_list[curr_id]:
+                     curr_id = node_to_id[st.session_state.current_node]
+                     best_next = None; min_w = float('inf')
+                     for n, w in graph.adj_list[curr_id]:
                         if w < min_w: min_w = w; best_next = n
-                    
-                    if best_next is not None:
+                     if best_next is not None:
                         st.session_state.current_node = id_to_node[best_next]
                         new_text = nodes[st.session_state.current_node]
-                        resp = generate_response(model, new_text, user_input, "MOVE", st.session_state.lead_info, archetype)
-                    else:
+                        # Pass Product Info
+                        resp = generate_response(model, new_text, user_input, "MOVE", st.session_state.lead_info, archetype, st.session_state.product_info)
+                     else:
                         resp = "Call finished."
                         save_lead_to_db(st.session_state.lead_info, st.session_state.messages, "End of Script")
 
@@ -666,95 +515,61 @@ if mode == "🤖 Sales Bot CRM":
 
 elif mode == "🧪 Math Lab":
     st.title("🧪 Computational Math Lab")
-    
     st.markdown("### Section A: Graph Inspector")
     col1, col2 = st.columns(2)
     n_nodes = col1.slider("N (Vertices)", 5, 15, 10)
     density = col2.slider("Density", 0.1, 1.0, 0.5)
     
     if st.button("Generate Graph"):
-         # Generate graph using experiments module
          graph = experiments.generate_erdos_renyi(n_nodes, density)
          st.session_state.lab_graph = graph
     
-    # Display if graph exists in session state
     if 'lab_graph' in st.session_state:
         graph = st.session_state.lab_graph
-        
         tab1, tab2, tab3 = st.tabs(["Visual Graph", "Adjacency Matrix", "Adjacency List"])
         
         with tab1:
             st.subheader("Graphviz Visualization")
+            is_connected, unreachable = graph.check_connectivity(0)
+            if is_connected: st.success("✅ Graph is Fully Connected (from Node 0)")
+            else: st.error(f"⚠️ Warning: Unreachable nodes: {unreachable}")
+            
             dot = graphviz.Digraph()
-            # Basic Graphviz from graph.adj_list
             for u, neighbors in graph.adj_list.items():
                 dot.node(str(u), label=str(u))
-                for v, w in neighbors:
-                    dot.edge(str(u), str(v), label=str(w))
+                for v, w in neighbors: dot.edge(str(u), str(v), label=str(w))
             st.graphviz_chart(dot)
             
         with tab2:
-            st.subheader("Adjacency Matrix")
-            # Reuse logic to show infinity as string
+            st.subheader("Adjacency Matrix (Heatmap)")
             matrix = graph.to_adjacency_matrix()
-            display_matrix = [[("∞" if x == float('inf') else x) for x in row] for row in matrix]
-            st.dataframe(pd.DataFrame(display_matrix))
+            df_matrix = pd.DataFrame(matrix)
+            df_heatmap = df_matrix.replace(float('inf'), None)
+            st.dataframe(df_heatmap.style.background_gradient(cmap="Blues", axis=None).format(formatter=lambda x: f"{x:.0f}" if pd.notnull(x) else "∞"))
             
         with tab3:
             st.subheader("Adjacency List")
             st.write(graph.adj_list)
             
     st.divider()
-    
     st.markdown("### Section B: Scientific Experiments")
     st.markdown("Comparing Bellman-Ford implementations: **Adjacency List vs Adjacency Matrix**.")
-    
-    # Preset parameters
-    sizes_preset = list(range(20, 220, 20)) # [20, 40, ..., 200]
+    sizes_preset = list(range(20, 220, 20)) 
     densities_preset = [0.1, 0.3, 0.5, 0.7, 0.9]
     
     if st.button("🚀 Run Scientific Benchmark"):
-        with st.spinner("Running rigorous scientific benchmarks (20 runs/config)... This may take a moment."):
-            # Execute run_scientific_benchmark from experiments.py
+        with st.spinner("Running benchmarks..."):
             results = experiments.run_scientific_benchmark(sizes_preset, densities_preset, num_runs=20)
-            
-            # Show Results Table
             df_results = pd.DataFrame(results)
-            st.subheader("Raw Experimental Data")
+            st.subheader("Raw Data")
             st.dataframe(df_results)
-            
-            # --- Visualization Logic ---
             st.divider()
-            st.subheader("Performance Comparison Chart")
-            
-            # Use columns for layout
-            col_chart, col_filter = st.columns([3, 1])
-            
-            with col_filter:
-                # Filter by Density
-                selected_density = st.selectbox(
-                    "Select Density to Visualize:", 
-                    densities_preset,
-                    index=2 # Default to 0.5
-                )
-                
-                # Analysis
-                st.info(f"""
-                **Analysis for Density {selected_density}:**
-                
-                - 🟦 **Time_List**: Uses Bellman-Ford on Adjacency List. Complexity $O(V \\cdot E)$.
-                - 🟥 **Time_Matrix**: Uses Bellman-Ford on Adjacency Matrix. Complexity $O(V^3)$.
-                """)
             
-            with col_chart:
-                # Filter data
-                filtered_df = df_results[df_results["Density"] == selected_density].copy()
-                filtered_df = filtered_df.sort_values(by="Vertices (N)")
-                
-                # Prepare for Chart (Index=N, Columns=[Time_List, Time_Matrix])
-                chart_data = filtered_df.set_index("Vertices (N)")[["Time_List", "Time_Matrix"]]
-                
-                # Plot
-                st.line_chart(chart_data)
-                
-            st.success(f"Benchmarking complete! Notice how Matrix implementation performance stays constant or redundant for sparse graphs, while List implementation scales with edges.")
+            c_chart, c_filter = st.columns([3, 1])
+            with c_filter:
+                sel_density = st.selectbox("Density:", densities_preset, index=2)
+                st.info(f"**Analysis:** List O(E) vs Matrix O(V^3).")
+            with c_chart:
+                filtered_df = df_results[df_results["Density"] == sel_density].sort_values("Vertices (N)")
+                st.line_chart(filtered_df.set_index("Vertices (N)")[["Time_List", "Time_Matrix"]])
+            st.success("Benchmarking complete!")

graph_module.py

@@ -59,3 +59,28 @@ class Graph:
 
     def get_list(self):
         return self.adj_list
+
+    def check_connectivity(self, start_node_id):
+        """
+        Checks if all nodes are reachable from start_node_id using BFS.
+        Returns: (is_connected: bool, unreachable_nodes: list)
+        """
+        if start_node_id < 0 or start_node_id >= self.num_vertices:
+            return False, []
+
+        visited = set()
+        queue = [start_node_id]
+        visited.add(start_node_id)
+        
+        while queue:
+            u = queue.pop(0)
+            for v, weight in self.adj_list[u]:
+                if v not in visited:
+                    visited.add(v)
+                    queue.append(v)
+        
+        all_nodes = set(range(self.num_vertices))
+        unreachable = list(all_nodes - visited)
+        unreachable.sort()
+        
+        return (len(unreachable) == 0), unreachable