Update app.py

app.py

@@ -1,5 +1,5 @@
 # app.py
-# @title Beer Game Final Version (v4 - English UI & Introduction)
+# @title Beer Game Final Version (v4.2 - Input State Bug Fixed)
 
 # -----------------------------------------------------------------------------
 # 1. Import Libraries
@@ -56,15 +56,13 @@ else:
 
 
 # -----------------------------------------------------------------------------
-# 3. Core Game Logic Functions
+# 3. Core Game Logic Functions (No changes in this section)
 # -----------------------------------------------------------------------------
 
 def get_customer_demand(week: int) -> int:
-    """Defines the end-customer demand pattern."""
     return 4 if week <= 4 else 8
 
 def init_game_state(llm_personality: str, info_sharing: str):
-    """Initializes or resets the game state in st.session_state."""
     roles = ["Retailer", "Wholesaler", "Distributor", "Factory"]
     human_role = random.choice(roles)
     participant_id = str(uuid.uuid4())[:8]
@@ -74,8 +72,8 @@ def init_game_state(llm_personality: str, info_sharing: str):
         'human_role': human_role, 'llm_personality': llm_personality,
         'info_sharing': info_sharing, 'logs': [], 'echelons': {},
         'factory_production_pipeline': deque([0] * FACTORY_LEAD_TIME, maxlen=FACTORY_LEAD_TIME),
-        'decision_step': 'initial_order', # Controls the two-step decision process
-        'human_initial_order': None,      # Stores the player's initial order
+        'decision_step': 'initial_order',
+        'human_initial_order': None,
     }
 
     for i, name in enumerate(roles):
@@ -95,7 +93,6 @@ def init_game_state(llm_personality: str, info_sharing: str):
     st.info(f"New game started! AI Mode: **{llm_personality} / {info_sharing}**. You have been randomly assigned the role of: **{human_role}**.")
 
 def get_llm_order_decision(prompt: str, echelon_name: str) -> (int, str):
-    """Calls the OpenAI API to get a decision and returns the integer order and raw text."""
     if not client: return 8, "NO_API_KEY_DEFAULT"
     with st.spinner(f"Getting AI decision for {echelon_name}..."):
         try:
@@ -117,7 +114,6 @@ def get_llm_order_decision(prompt: str, echelon_name: str) -> (int, str):
             return 8, f"API_ERROR: {e}"
 
 def get_llm_prompt(echelon_state: dict, week: int, llm_personality: str, info_sharing: str, all_echelons_state: dict) -> str:
-    """Generates the prompt for the LLM based on the game scenario."""
     # This function's logic is complex and correct, so it remains unchanged.
     base_info = f"Your Current Status at the **{echelon_state['name']}** for **Week {week}**:\n- On-hand inventory: {echelon_state['inventory']} units.\n- Backlog (unfilled orders): {echelon_state['backlog']} units.\n- Incoming order this week (from your customer): {echelon_state['incoming_order']} units.\n- Shipments on the way to you: {list(echelon_state['incoming_shipments'])}\n- Orders you have placed being processed by your supplier: {list(echelon_state['order_pipeline'])}"
     if llm_personality == 'perfect_rational' and info_sharing == 'full':
@@ -132,7 +128,7 @@ def get_llm_prompt(echelon_state: dict, week: int, llm_personality: str, info_sh
         supply_line = sum(echelon_state['incoming_shipments']) + sum(echelon_state['order_pipeline'])
         calculated_order = anchor_demand + inventory_correction - supply_line
         rational_local_order = max(0, int(calculated_order))
-        return f"**You are a perfectly rational supply chain AI with ONLY LOCAL information.**\nYou must use a logical heuristic to make a stable decision. A proven method is \"Anchoring and Adjustment\".\n\n{base_info}\n\n**Rational Calculation (Anchoring & Adjustment):**\n1.  **Anchor on Demand:** Your best guess for future demand is your last incoming order: **{anchor_demand} units**.\n2.  **Adjust for Inventory:** You want to hold a safety stock of {safety_stock} units. Your current stock is {echelon_state['inventory'] - echelon_state['backlog']}. You need to order an extra **{inventory_correction} units** to correct this.\n3.  **Account for Supply Line:** You already have **{supply_line} units** in transit or being processed. These should be subtracted from your new order.\n\n**Final Calculation:**\n* Order = (Anchor Demand) + (Inventory Adjustment) - (Supply Line)\n* Order = {anchor_demand} + {inventory_correction} - {supply_line} = **{rational_local_order} units**.\n\n**Your Task:** Confirm this locally rational quantity. Respond with a single integer."
+        return f"**You are a perfectly rational supply chain AI with ONLY LOCAL information.**\nYou must use a logical heuristic to make a stable decision. A proven method is \"Anchoring and Adjustment\".\n\n{base_info}\n\n**Rational Calculation (Anchoring & Adjustment):**\n1.  **Anchor on Demand:** Your best guess for future demand is your last incoming order: **{anchor_demand} units**.\n2.  **Adjust for Inventory:** You want to hold a safety stock of {safety_stock} units. Your current stock is {echelon_state['inventory'] - echelon_state['backlog']}. You need to order an extra **{inventory_correction} units** to correct this.\n3.  **Account for Supply Line:** You already have **{supply_line} units** in transit or being processed. These should be subtracted from your new order.\n\n**Final Calculation:**\n* Order = (Anchor Demand) + (Inventory Adjustment) - (Supply Line)\n* Order = {anchor_demand} + {inventory_correction} - {supply_line} = **{rational_local_order} units**.\n**Your Task:** Confirm this locally rational quantity. Respond with a single integer."
     elif llm_personality == 'human_like' and info_sharing == 'full':
         full_info_str = f"\n**Full Supply Chain Information:**\n- End-Customer Demand this week: {get_customer_demand(week)} units.\n"
         for name, e_state in all_echelons_state.items():
@@ -142,14 +138,12 @@ def get_llm_prompt(echelon_state: dict, week: int, llm_personality: str, info_sh
         return f"**You are a reactive supply chain manager for the {echelon_state['name']}.** You have a limited view and tend to over-correct based on fear.\n\n**Your Mindset: **Your top priority is try to not have a backlog.\n\n{base_info}\n\n**Your Task:** You just saw your own inventory and a new order coming. Your gut instinct is to panic and order enough to ensure you are never caught with a backlog again.\n\n**React emotionally.** What is your knee-jerk order quantity? Respond with a single integer."
 
 def step_game(human_final_order: int, human_initial_order: int, ai_suggestion: int):
-    """Processes one week of the game and records detailed logs, including two-step decision data."""
     state = st.session_state.game_state
     week, echelons, human_role = state['week'], state['echelons'], state['human_role']
     llm_personality, info_sharing = state['llm_personality'], state['info_sharing']
     echelon_order = ["Retailer", "Wholesaler", "Distributor", "Factory"]
     llm_raw_responses = {}
 
-    # Core game simulation steps (unchanged)
     factory_state = echelons["Factory"]
     if state['factory_production_pipeline']: factory_state['inventory'] += state['factory_production_pipeline'].popleft()
     for name in ["Retailer", "Wholesaler", "Distributor"]:
@@ -167,7 +161,6 @@ def step_game(human_final_order: int, human_initial_order: int, ai_suggestion: i
         receiver = echelons[sender]['downstream_name']
         if receiver: echelons[receiver]['incoming_shipments'].append(echelons[sender]['shipment_sent'])
 
-    # Agents place orders
     for name in echelon_order:
         e = echelons[name]
         if name == human_role:
@@ -181,9 +174,8 @@ def step_game(human_final_order: int, human_initial_order: int, ai_suggestion: i
 
     state['factory_production_pipeline'].append(echelons["Factory"]['order_placed'])
 
-    # Update costs and record logs
     log_entry = {'timestamp': datetime.utcnow().isoformat() + "Z", 'week': week, **state}
-    del log_entry['echelons'], log_entry['factory_production_pipeline'], log_entry['logs'] # Clean up complex objects
+    del log_entry['echelons'], log_entry['factory_production_pipeline'], log_entry['logs']
     for name in echelon_order:
         e = echelons[name]
         e['weekly_cost'] = (e['inventory'] * HOLDING_COST) + (e['backlog'] * BACKLOG_COST); e['total_cost'] += e['weekly_cost']
@@ -191,20 +183,16 @@ def step_game(human_final_order: int, human_initial_order: int, ai_suggestion: i
             log_entry[f'{name}.{key}'] = e[key]
         log_entry[f'{name}.llm_raw_response'] = llm_raw_responses.get(name, "")
     
-    # Record two-step decision data
     log_entry[f'{human_role}.initial_order'] = human_initial_order
     log_entry[f'{human_role}.ai_suggestion'] = ai_suggestion
     
     state['logs'].append(log_entry)
 
-    # Advance week and reset decision step
     state['week'] += 1
     state['decision_step'] = 'initial_order'
     if state['week'] > WEEKS: state['game_running'] = False
 
-
 def plot_results(df: pd.DataFrame, title: str, human_role: str):
-    """Generates and returns the end-of-game plots."""
     fig, axes = plt.subplots(4, 1, figsize=(12, 22))
     fig.suptitle(title, fontsize=16)
     echelons = ['Retailer', 'Wholesaler', 'Distributor', 'Factory']
@@ -217,19 +205,15 @@ def plot_results(df: pd.DataFrame, title: str, human_role: str):
                 'total_cost': row[f'{e}.total_cost']})
     plot_df = pd.DataFrame(plot_data)
 
-    # Plot 1: Inventory
     inventory_pivot = plot_df.pivot(index='week', columns='echelon', values='inventory').reindex(columns=echelons)
     inventory_pivot.plot(ax=axes[0], kind='line', marker='o', markersize=4); axes[0].set_title('Inventory Levels'); axes[0].grid(True, linestyle='--'); axes[0].set_ylabel('Stock (Units)')
 
-    # Plot 2: Orders (Bullwhip Effect)
     order_pivot = plot_df.pivot(index='week', columns='echelon', values='order_placed').reindex(columns=echelons)
     order_pivot.plot(ax=axes[1], style='--'); axes[1].plot(range(1, WEEKS + 1), [get_customer_demand(w) for w in range(1, WEEKS + 1)], label='Customer Demand', color='black', lw=2.5); axes[1].set_title('Order Quantities (The Bullwhip Effect)'); axes[1].grid(True, linestyle='--'); axes[1].legend(); axes[1].set_ylabel('Ordered (Units)')
 
-    # Plot 3: Costs
     total_costs = plot_df.groupby('echelon')['total_cost'].max().reindex(echelons)
     total_costs.plot(kind='bar', ax=axes[2], rot=0); axes[2].set_title('Total Cumulative Cost'); axes[2].set_ylabel('Cost ($)')
 
-    # Plot 4: Human Decision Analysis
     human_df = df[['week', f'{human_role}.initial_order', f'{human_role}.ai_suggestion', f'{human_role}.order_placed']].copy()
     human_df.rename(columns={
         f'{human_role}.initial_order': 'Your Initial Order', f'{human_role}.ai_suggestion': 'AI Suggestion', f'{human_role}.order_placed': 'Your Final Order'
@@ -239,7 +223,6 @@ def plot_results(df: pd.DataFrame, title: str, human_role: str):
     plt.tight_layout(rect=[0, 0, 1, 0.96]); return fig
 
 def save_logs_and_upload(state: dict):
-    """Saves logs locally and uploads to Hugging Face Hub at the end of the game."""
     if not state.get('logs'): return
     participant_id = state['participant_id']
     df = pd.json_normalize(state['logs'])
@@ -266,21 +249,17 @@ if st.session_state.get('initialization_error'):
 else:
     # --- Game Setup & Instructions ---
     if 'game_state' not in st.session_state or not st.session_state.game_state.get('game_running', False):
-        
-        # --- NEW: Introduction Section ---
         st.markdown("---")
         st.header("📖 Welcome to the Beer Game!")
         st.markdown("""
         The Beer Game is a classic supply chain simulation that demonstrates a phenomenon called the **"Bullwhip Effect."** Even with stable customer demand, small variations in orders can amplify as they move up the supply chain, causing massive shortages and overstocks.
         """)
-
         st.subheader("Your Goal")
         st.markdown("""
         Minimize the total cost for your position in the supply chain. Costs are incurred for:
         - **Holding Inventory:** **$0.50 per unit per week**
         - **Backlog (Unfilled Orders):** **$1.00 per unit per week**
         """)
-
         col1, col2 = st.columns(2)
         with col1:
             st.subheader("🔗 The Supply Chain")
@@ -296,7 +275,6 @@ else:
             st.markdown("""
             The key challenge is managing delays. There is a **communication delay** for orders to reach your supplier and a **shipping delay** for goods to arrive. You must order enough to meet future demand without creating a huge pile of expensive inventory.
             """)
-        
         st.subheader("🎮 How to Play This Version")
         st.markdown("""
         1.  **Configure the Game:** Choose the AI's behavior and the level of information sharing.
@@ -307,15 +285,12 @@ else:
         4.  **Advance:** Once you submit your final order, the week advances, and all AI agents make their moves.
         """)
         st.markdown("---")
-        
-        # --- Game Configuration ---
         st.header("⚙️ Game Configuration")
         c1, c2 = st.columns(2)
         with c1:
              llm_personality = st.selectbox("AI Agent 'Personality'", ('human_like', 'perfect_rational'), format_func=lambda x: x.replace('_', ' ').title(), help="**Human-like:** Tends to react emotionally, potentially over-ordering. **Perfect Rational:** Uses a mathematical heuristic to make stable, logical decisions.")
         with c2:
             info_sharing = st.selectbox("Information Sharing Level", ('local', 'full'), format_func=lambda x: x.title(), help="**Local:** You and the AI agents can only see your own inventory and incoming orders. **Full:** Everyone can see the entire supply chain's status and the true end-customer demand.")
-        
         if st.button("🚀 Start Game", type="primary", disabled=(client is None)):
             init_game_state(llm_personality, info_sharing)
             st.rerun()
@@ -327,7 +302,6 @@ else:
         
         st.header(f"Week {week} / {WEEKS}")
         st.subheader(f"Your Role: **{human_role}** | AI Mode: **{state['llm_personality'].replace('_', ' ')}** | Information: **{state['info_sharing']}**")
-        
         st.markdown("---")
         st.subheader("Supply Chain Status")
         if info_sharing == 'full':
@@ -339,7 +313,7 @@ else:
                     st.metric("Inventory", e['inventory']); st.metric("Backlog", e['backlog'])
                     st.write(f"Incoming Order: **{e['incoming_order']}**")
                     st.write(f"Arriving Next Week: **{list(e['incoming_shipments'])[0] if e['incoming_shipments'] else 0}**")
-        else: # local information
+        else:
             st.info("In Local Information mode, you can only see your own status dashboard.")
             e = echelons[human_role]
             st.markdown(f"### 👤 {human_role} (Your Dashboard)")
@@ -349,8 +323,6 @@ else:
             col3.write(f"**Incoming Order (This Week):**\n# {e['incoming_order']}")
             col4.write(f"**Shipment Arriving (Next Week):**\n# {list(e['incoming_shipments'])[0] if e['incoming_shipments'] else 0}")
         st.markdown("---")
-
-        # --- Two-Step Decision UI ---
         st.header("Your Decision")
         human_echelon_state = echelons[human_role]
 
@@ -368,18 +340,49 @@ else:
             prompt_sugg = get_llm_prompt(human_echelon_state, week, state['llm_personality'], state['info_sharing'], echelons)
             ai_suggestion, _ = get_llm_order_decision(prompt_sugg, f"{human_role} (Suggestion)")
             
+            # =============== BUG FIX STARTS HERE ===============
+            # Initialize the session_state key for the input box only once per step.
+            if 'final_order_input' not in st.session_state:
+                st.session_state.final_order_input = ai_suggestion
+            # =============== BUG FIX ENDS HERE =================
+
             with st.form(key="final_order_form"):
                 st.markdown(f"#### **Step 2:** The AI suggests ordering **{ai_suggestion}** units.")
                 st.markdown("Considering the AI's advice, submit your **final** order to end the week.")
-                final_order = st.number_input("Your Final Order Quantity:", min_value=0, step=1, value=ai_suggestion)
+                
+                # =============== BUG FIX STARTS HERE ===============
+                # Use the 'key' parameter to bind the input to session state.
+                # Remove the problematic 'value' parameter.
+                st.number_input(
+                    "Your Final Order Quantity:",
+                    min_value=0,
+                    step=1,
+                    key='final_order_input' 
+                )
+                # =============== BUG FIX ENDS HERE =================
+
                 if st.form_submit_button("Submit Final Order & Advance to Next Week"):
-                    step_game(int(final_order), state['human_initial_order'], ai_suggestion)
+                    # =============== BUG FIX STARTS HERE ===============
+                    # Read the final order value directly from session state.
+                    final_order_value = st.session_state.final_order_input
+                    
+                    # Call the game logic
+                    step_game(final_order_value, state['human_initial_order'], ai_suggestion)
+                    
+                    # Clean up the session state key to prepare for the next week.
+                    del st.session_state.final_order_input
+                    # =============== BUG FIX ENDS HERE =================
+                    
                     st.rerun()
 
         st.sidebar.header("Game Info")
         st.sidebar.markdown(f"**Game ID**: `{state['participant_id']}`\n\n**Current Week**: {week}")
         if st.sidebar.button("🔄 Reset Game"):
-            del st.session_state.game_state; st.rerun()
+            # Clean up state if the user resets
+            if 'final_order_input' in st.session_state:
+                del st.session_state.final_order_input
+            del st.session_state.game_state
+            st.rerun()
 
     # --- Game Over Interface ---
     if 'game_state' in st.session_state and not st.session_state.game_state.get('game_running', False) and st.session_state.game_state['week'] > WEEKS:
@@ -390,4 +393,9 @@ else:
         st.pyplot(fig)
         save_logs_and_upload(state)
         if st.button("✨ Start a New Game"):
-            del st.session_state.game_state; st.rerun()
+            del st.session_state.game_state
+            st.rerun()
+
+
+  
+