Update app.py

app.py

@@ -78,6 +78,30 @@ def find_similar_products(query_embedding, top_k=8):
     
     return matching_products
 
+json_schema = {
+    "name": "CookingIngredientsSchema",
+    "description": "Extract cooking ingredients and analysis from user query.",
+    "strict": True,
+    "schema": {
+        "type": "object",
+        "properties": {
+            "analysis": {
+                "type": "string",
+                "description": "完整的需求分析，解釋用戶的目標和需要的物品。"
+            },
+            "ingredients": {
+                "type": "array",
+                "description": "提取的食材或關鍵物品清單。",
+                "items": {
+                    "type": "string",
+                    "description": "單一食材或關鍵物品名稱。"
+                }
+            }
+        },
+        "required": ["analysis", "ingredients"],
+        "additionalProperties": False
+    }
+}
 
 # Analyze query and find products
 def analyze_query_and_find_products(query: str) -> str:
@@ -100,15 +124,22 @@ def analyze_query_and_find_products(query: str) -> str:
             model="gpt-4o",
             messages=analysis_messages,
             temperature=0.7,
-            max_tokens=500
+            response_format={
+            "type": "json_schema",
+            "json_schema": json_schema
+            }
         )
         
         analysis = analysis_response.choices[0].message.content
+        analysis_json = json.loads(analysis)
+
+        description = analysis_json["analysis"]  # 取得分析結果
+        ingredients = analysis_json["ingredients"]  # 取得提取的食材清單
 
-        print("AAA:" + analysis)
+        print("AAA:" + ingredients)
         
         # Generate embedding for the query
-        query_embedding = get_embedding(query + " " + analysis)
+        query_embedding = get_embedding(query + " " + description)
         
         # Find similar products
         matching_products = find_similar_products(query_embedding)