Use exact same GPT-5 prompt from working BF space

- Copy identical prompt structure from BF space
- Include GOBIK brand rules
- Include product type understanding
- Keep same examples and format

smart_warehouse_complete.py

@@ -188,21 +188,32 @@ def get_warehouse_stock_smart_complete(user_message):
     # GPT-5 prompt with enhanced instructions
     smart_prompt = f"""User is asking: "{user_message}"
 
-CRITICAL: Find products where the NAME contains "{user_message.upper()}".
-For example, if user asks "madone sl 6", find products where name contains "MADONE SL 6".
-DO NOT return products with different names even if they seem similar.
-
+Find ALL products that match this query from the list below.
 If user asks about specific size (S, M, L, XL, XXL, SMALL, MEDIUM, LARGE, X-LARGE), return only that size.
 If user asks generally (without size), return ALL variants of the product.
 {warehouse_filter}
 
+IMPORTANT BRAND AND PRODUCT TYPE RULES:
+- GOBIK: Spanish textile brand we import. When user asks about "gobik", return ALL products with "GOBIK" in the name.
+- Product names contain type information: FORMA (jersey/cycling shirt), TAYT (tights), İÇLİK (base layer), YAĞMURLUK (raincoat), etc.
+- Understand Turkish/English terms: 
+  * "erkek forma" / "men's jersey" -> Find products with FORMA in name
+  * "tayt" / "tights" -> Find products with TAYT in name  
+  * "içlik" / "base layer" -> Find products with İÇLİK in name
+  * "yağmurluk" / "raincoat" -> Find products with YAĞMURLUK in name
+- Gender: UNISEX means for both men and women. If no gender specified, it's typically men's.
+
 Products list (with warehouse availability):
 {json.dumps(products_summary, ensure_ascii=False, indent=2)}
 
-Return ONLY index numbers of products where the name contains "{user_message.upper()}".
-Return as comma-separated list (e.g., "5,8,12,15").
+Return ONLY index numbers of ALL matching products as comma-separated list (e.g., "5,8,12,15").
 If no products found, return ONLY: -1
-DO NOT return empty string or any explanation, ONLY numbers or -1"""
+DO NOT return empty string or any explanation, ONLY numbers or -1
+
+Examples of correct responses:
+- "2,5,8,12,15,20" (multiple products found)
+- "45" (single product found)  
+- "-1" (no products found)"""
 
     headers = {
         "Content-Type": "application/json",