trying more ai awareness

ai_chat.py

@@ -113,7 +113,12 @@ Key guidelines:
 - If someone asks for clarification about property terms, explain clearly and helpfully
 - Only ask about their property needs when they show interest in searching or viewing properties
 - Keep responses conversational and not robotic
-- If user says "hi" or "hello", respond with a greeting like "Hi [Name]! How can I help you with industrial properties today?" """
+- If user says "hi" or "hello", respond with a greeting like "Hi [Name]! How can I help you with industrial properties today?"
+- If someone mentions they need space for a specific business type (like "manufacturing plant"), understand the context and suggest appropriate sizes
+- Don't keep asking the same question if the user has already given context about their needs
+- Be understanding and helpful, not pushy or repetitive
+- If someone says "big enough for manufacturing plant" or similar, understand they want a large industrial space and suggest appropriate sizes
+- Don't be robotic - respond like a real human would in a conversation"""
     
     if user_info.get("name") and user_info["name"] != "Unknown":
         system_message += f"\nThe user's name is {user_info['name']} - use their name to make it personal."

persona_manager.py

@@ -197,19 +197,19 @@ async def fallback_parse_response(user_message: str, field: str) -> Tuple[bool,
     if any(word in user_message_lower for word in affirmative_words):
         # For affirmative responses, we need to get the value from context
         # This will be handled by the LLM parsing, but we can provide a fallback
-        return True, "confirmed", "Got it! Thanks for confirming."
+        return True, "confirmed", "Got it! Thanks for confirming.", False, False
     
     # Check for skip requests
     skip_words = ["not sure", "skip", "later", "don't know", "maybe later", "pass", "no idea"]
     if any(word in user_message_lower for word in skip_words):
-        return True, None, PERSONA_FIELDS[field]["skip_response"]
+        return True, None, PERSONA_FIELDS[field]["skip_response"], False, True
     
     # Basic field-specific parsing
     if field == "intent":
         if any(word in user_message_lower for word in ["buy", "purchase", "own"]):
-            return True, "buy", "Got it, you're looking to buy. That's great!"
+            return True, "buy", "Got it, you're looking to buy. That's great!", False, False
         elif any(word in user_message_lower for word in ["lease", "rent"]):
-            return True, "lease", "Perfect, leasing gives you flexibility."
+            return True, "lease", "Perfect, leasing gives you flexibility.", False, False
     
     elif field == "budget":
         import re
@@ -235,7 +235,7 @@ async def fallback_parse_response(user_message: str, field: str) -> Tuple[bool,
                 elif "m" in pattern or "million" in pattern:
                     budget *= 1000000
                 
-                return True, budget, f"Thanks! I'll look for properties around ${budget:,}."
+                return True, budget, f"Thanks! I'll look for properties around ${budget:,}.", False, False
         
         # Fallback: just look for numbers
         numbers = re.findall(r'\d+', user_message)
@@ -245,7 +245,7 @@ async def fallback_parse_response(user_message: str, field: str) -> Tuple[bool,
                 budget *= 1000
             elif "m" in user_message_lower or "million" in user_message_lower:
                 budget *= 1000000
-            return True, budget, f"Thanks! I'll look for properties around ${budget:,}."
+            return True, budget, f"Thanks! I'll look for properties around ${budget:,}.", False, False
     
     elif field == "size_preference_sqm":
         import re
@@ -266,13 +266,20 @@ async def fallback_parse_response(user_message: str, field: str) -> Tuple[bool,
                 # Convert sq ft to sqm if needed
                 if 'ft' in pattern and 'sq' in pattern:
                     size = int(size * 0.0929)  # Convert sq ft to sqm
-                return True, size, f"Perfect! {size} sqm should give you good options."
+                return True, size, f"Perfect! {size} sqm should give you good options.", False, False
+        
+        # Check for context-based size suggestions
+        if any(word in user_message_lower for word in ["manufacturing", "factory", "plant", "production"]):
+            if "large" in user_message_lower or "big" in user_message_lower:
+                return True, 5000, "Perfect! For a large manufacturing plant, I'd recommend around 5000 sqm. This gives you plenty of space for production lines, storage, and office areas.", False, False
+            else:
+                return True, 3000, "Great! For a manufacturing plant, I'd suggest around 3000 sqm. This should accommodate your production needs well.", False, False
         
         # Fallback: just look for numbers
         numbers = re.findall(r'\d+', user_message)
         if numbers:
             size = max(int(n) for n in numbers)
-            return True, size, f"Perfect! {size} sqm should give you good options."
+            return True, size, f"Perfect! {size} sqm should give you good options.", False, False
     
     elif field == "location_preference":
         # Extract location names (basic approach)
@@ -284,10 +291,10 @@ async def fallback_parse_response(user_message: str, field: str) -> Tuple[bool,
                 break
         
         if found_location:
-            return True, found_location, f"Great! {found_location.title()} has good options."
+            return True, found_location, f"Great! {found_location.title()} has good options.", False, False
         else:
             # Assume the whole message is a location
-            return True, user_message.strip(), f"Got it! I'll look in {user_message.strip()}."
+            return True, user_message.strip(), f"Got it! I'll look in {user_message.strip()}.", False, False
     
     elif field == "must_have":
         features = []
@@ -304,9 +311,9 @@ async def fallback_parse_response(user_message: str, field: str) -> Tuple[bool,
                 features.append(feature)
         
         if features:
-            return True, features, f"Perfect! I'll make sure to find places with {', '.join(features)}."
+            return True, features, f"Perfect! I'll make sure to find places with {', '.join(features)}.", False, False
     
-    return False, None, "I didn't quite understand. Could you try again?"
+    return False, None, "I didn't quite understand. Could you try again?", False, False
 
 async def should_ask_persona_question(persona: Dict, conversation_context: str = "") -> Tuple[bool, Optional[str]]:
     """
@@ -352,6 +359,8 @@ Rules:
 - If the user is asking about properties, locations, business needs, or showing interest in finding space → return "yes"
 - If the user mentions locations (including abbreviations like cpt, pta, jhb) → return "yes"
 - If the user mentions business types (clothing, manufacturing, etc.) → return "yes"
+- If the user has already provided context about their needs (like "big enough for manufacturing plant") → return "no" (they've given enough context)
+- If the user is frustrated or asking you to stop asking questions → return "no"
 
 Return ONLY "yes" or "no"."""
 
@@ -415,17 +424,17 @@ Available persona fields:
 - size_preference_sqm: Extract size in square meters (convert from sq ft if needed)
 - must_have: Extract features as array (e.g., ["truck access", "office space"])
 
-Return ONLY a JSON object with the extracted fields. Only include fields that are clearly mentioned or implied.
-If no relevant information is found for a field, don't include it in the response.
-
-IMPORTANT: For location_preference, be very generous in extraction. If the user mentions ANY location (including "pretoria", "pta", "johannesburg", "jhb", "cape town", "cpt", etc.), extract it.
+IMPORTANT: You MUST return a valid JSON object. If no information is found, return {} (empty object).
+For location_preference, be very generous in extraction. If the user mentions ANY location (including "pretoria", "pta", "johannesburg", "jhb", "cape town", "cpt", etc.), extract it.
 
 Examples:
 - "I want a property in cpt" → {"location_preference": "cape town"}
 - "Looking for warehouse in pta around 500k" → {"location_preference": "pretoria", "budget": 500000}
 - "Looking for a property in pretoria for my clothing business" → {"location_preference": "pretoria"}
 - "Need space for clothing business with office" → {"must_have": ["office"]}
-"""
+- "I want one big enough for a phara manufacturing plant" → {"size_preference_sqm": 5000, "must_have": ["manufacturing"]}
+
+Return ONLY the JSON object, no other text."""
 
         messages = [
             {"role": "system", "content": system_prompt},
@@ -455,7 +464,22 @@ Examples:
             
         except json.JSONDecodeError:
             print(f"Failed to parse AI extraction response: {response.content}")
-            return {}
+            print(f"Response type: {type(response.content)}")
+            print(f"Response length: {len(response.content)}")
+            # Try to clean the response
+            cleaned_response = response.content.strip()
+            if cleaned_response.startswith('```json'):
+                cleaned_response = cleaned_response[7:]
+            if cleaned_response.endswith('```'):
+                cleaned_response = cleaned_response[:-3]
+            cleaned_response = cleaned_response.strip()
+            try:
+                extracted = json.loads(cleaned_response)
+                print(f"Successfully parsed after cleaning: {extracted}")
+                return extracted
+            except:
+                print(f"Still failed after cleaning: {cleaned_response}")
+                return {}
             
     except Exception as e:
         print(f"Error in AI persona extraction: {e}")