Automated deployment via API

app.py

@@ -359,69 +359,94 @@ def upload_data():
 
 from huggingface_hub import InferenceClient
 
-# Initialize Inference Client (uses HF_TOKEN from environment)
-# Switch to Mistral-7B-Instruct for better reliability on free tier
-client = InferenceClient("mistralai/Mistral-7B-Instruct-v0.2")
+# Initialize Inference Client
+# Using Phi-3 Mini for high reliability and low latency on public API
+client = InferenceClient("microsoft/Phi-3-mini-4k-instruct")
 
 def generate_ai_response(query):
-    """Generate AI response using Mistral-7B LLM with RAG-lite context"""
+    """Generate AI response using Phi-3 LLM with strict fallback"""
+    # Default Fallback (pre-calculated to always be available)
+    fallback_response = "⚠️ **Neural Link Unstable**. Falling back to local protocols."
     try:
-        # 1. Gather Context
         current_temp = app_state.get_temperature()
         optimal_low = CONFIG['TEMP_OPTIMAL_LOW']
         optimal_high = CONFIG['TEMP_OPTIMAL_HIGH']
-        current_energy = app_state.get_energy()
+        
+        # Smart Fallback Logic
+        if 'pour' in query.lower():
+            if 1410 <= current_temp <= 1430:
+                 fallback_response = f"✅ **POUR READY** (Offline Mode). Current Temp {current_temp:.1f}°C is optimal."
+            else:
+                 fallback_response = f"⚠️ **HOLD POUR** (Offline Mode). Current Temp {current_temp:.1f}°C is out of range ({optimal_low}-{optimal_high}°C)."
+        elif 'temp' in query.lower():
+            fallback_response = f"🌡️ **Offline Status**: {current_temp:.1f}°C. (Optimal: {optimal_low}-{optimal_high}°C)"
+        elif 'energy' in query.lower():
+            fallback_response = f"⚡ **Energy Status**: {app_state.get_energy():.1f} kWh."
+        
+        # 1. Gather Context for Prompt
         is_anomaly = app_state.is_anomaly
         risk_score = app_state.anomaly_risk * 100
         
-        # 2. Construct System Prompt
         system_prompt = f"""
-        SYSTEM: You are the IronGuard Foundry AI, an expert metallurgist assistant. 
-        REAL-TIME STATUS:
-        - Temperature: {current_temp:.1f}°C (Optimal: {optimal_low}-{optimal_high}°C)
-        - Energy Usage: {current_energy:.1f} kWh (Optimal: ~450 kWh)
-        - Anomaly Detected: {is_anomaly} (Risk Score: {risk_score:.1f}%)
+        You are the IronGuard Foundry AI.
+        LIVE SENSOR DATA:
+        - Temp: {current_temp:.1f}°C (Target: {optimal_low}-{optimal_high})
+        - Energy: {app_state.get_energy():.1f} kWh
+        - Anomaly Risk: {risk_score:.1f}%
         
         INSTRUCTIONS:
-        - Answer the user's question mostly based on the REAL-TIME STATUS.
-        - If the temperature is > {CONFIG['TEMP_MAX']}, warn the user immediately.
-        - If asked about "pouring", only say YES if temp is under {optimal_high} and above {optimal_low}.
-        - Be concise, professional, and helpful. Do not mention you are an AI model.
-        - Use formatting like **bold** for key metrics.
+        Answer the user's question using the LIVE SENSOR DATA.
+        If Temp > {CONFIG['TEMP_MAX']}, warn immediately.
+        Keep answers short and professional.
         """
         
-        # 3. Call Inference API with fallback
-        messages = [
-            {"role": "user", "content": system_prompt + "\n\nUSER QUESTION: " + query}
-        ]
-        
+        # 2. Call API (Nested Try to catch API specific errors)
         try:
-            response_text = ""
-            # Using basic text_generation if chat_completion is flaky for this model
-            response_text = client.text_generation(
-                system_prompt + "\n\nUSER: " + query + "\n\nASSISTANT:",
-                max_new_tokens=150,
-                temperature=0.7
-            )
-            return response_text
+            messages = [
+                {"role": "user", "content": system_prompt + "\nUSER: " + query}
+            ]
+            response = ""
+            for message in client.chat_completion(messages, max_tokens=150, stream=True):
+                if message.choices and message.choices[0].delta.content:
+                    response += message.choices[0].delta.content
+            return response if response else fallback_response
+
         except Exception as api_err:
-            logger.warning(f"⚠️ Inference API failed: {api_err}. Using fallback.")
-            raise api_err # Trigger fallback
+            logger.warning(f"⚠️ API Error: {api_err}")
+            return fallback_response
 
     except Exception as e:
-        logger.error(f"❌ AI Gen Error: {e}")
-        # ROBUST FALLBACK - specific responses if API fails
-        fallback_temp = app_state.get_temperature()
-        if 'pour' in query.lower():
-            if 1410 <= fallback_temp <= 1430:
-                 return f"✅ **POUR READY** (Offline Mode). Temp {fallback_temp:.1f}°C is optimal."
-            return f"⚠️ **HOLD POUR** (Offline Mode). Temp {fallback_temp:.1f}°C is out of range."
-        return f"🤖 **System Status**: Temp {fallback_temp:.1f}°C, Energy {app_state.get_energy():.1f}kWh. (Neural Network Unreachable, verifying safety protocols...)"
-
-# ============================================================================
-# 6. SOCKET.IO EVENTS
-# ============================================================================
+        logger.error(f"❌ Critical Logic Error: {e}")
+        return fallback_response
 
+@app.route('/api/chat', methods=['POST'])
+def chat():
+    """AI Chat interface with fail-safe return"""
+    try:
+        data = request.json
+        query = data.get('query', '').lower()
+        
+        # Guaranteed to return a string, never raises
+        response = generate_ai_response(query)
+        
+        app_state.chat_history.append({
+            'user': query,
+            'bot': response,
+            'timestamp': datetime.now().isoformat()
+        })
+        
+        return jsonify({
+            'response': response,
+            'confidence': 1.0, # Artificial confidence for UX
+            'timestamp': datetime.now().isoformat()
+        })
+    except Exception as e:
+        logger.error(f"❌ Critical Route Error: {e}")
+        # Absolute last resort JSON to prevent frontend 'System Error'
+        return jsonify({
+            'response': "⚠️ **System Critical**: Local fallback active. Please refresh console.",
+            'confidence': 0.0
+        })
 @socketio.on('connect')
 def handle_connect():
     """Handle client connection"""