Update app.py

app.py

@@ -5,30 +5,16 @@ import pandas as pd
 import numpy as np
 import os
 from openai import OpenAI
-from typing import List, Dict
 import pickle
-import time
-from sklearn.metrics.pairwise import cosine_similarity
-from huggingface_hub import HfApi, hf_hub_download, upload_file
-from pathlib import Path
 import faiss
 
 # Initialize OpenAI client
 client = OpenAI(api_key=os.environ.get("OPENAI_API_KEY"))
 
-def get_embedding(text: str, model="text-embedding-ada-002"):
-    """Get embeddings for a text using OpenAI's API"""
-    try:
-        text = text.replace("\n", " ")
-        response = client.embeddings.create(
-            input=[text],
-            model=model
-        )
-        return response.data[0].embedding
-    except Exception as e:
-        print(f"Error getting embedding: {e}")
-        return None
-        
+# Load Excel data
+df = pd.read_excel("your_file.xlsx")  # 修改為你的 xlsx 文件路徑
+
+# Initialize FAISS Index
 def initialize_embeddings_from_pkl(pkl_path: str, faiss_path: str):
     """從 PKL 檔案載入嵌入，並加載 FAISS 索引"""
     if not os.path.exists(pkl_path):
@@ -47,21 +33,34 @@ def initialize_embeddings_from_pkl(pkl_path: str, faiss_path: str):
     print(f"FAISS index loaded with {index.ntotal} embeddings.")
     return index, embeddings
 
-# Load CSV data
-#df = pd.read_csv("item_new.csv", encoding='utf-8')
-
+# Paths to FAISS index and PKL file
+faiss_path = "product_index.faiss"  # 修改為你的 FAISS 索引文件路徑
+pkl_path = "product_embeddings.pkl"  # 修改為你的嵌入文件路徑
 
-# Load embeddings from FAISS
+# Initialize embeddings
 print("Initializing embeddings...")
-faiss_path = "product_index.faiss"  # Path to FAISS index file
-pkl_path = "product_embeddings.pkl"  # Path to embeddings file
-
-# 初始化嵌入
 faiss_index, product_embeddings_array = initialize_embeddings_from_pkl(pkl_path, faiss_path)
+assert faiss_index.ntotal == len(df), "FAISS 索引與 xlsx 文件的行數不一致！"
 print(f"Loaded embeddings with shape: {len(product_embeddings_array)} x {len(product_embeddings_array[0])}")
 print("Embeddings initialized")
 
 
+# Generate embeddings for query
+def get_embedding(text: str, model="text-embedding-ada-002"):
+    """Get embeddings for a text using OpenAI's API"""
+    try:
+        text = text.replace("\n", " ")
+        response = client.embeddings.create(
+            input=[text],
+            model=model
+        )
+        return response.data[0].embedding
+    except Exception as e:
+        print(f"Error getting embedding: {e}")
+        return None
+
+
+# Find similar products
 def find_similar_products(query_embedding, top_k=8):
     """Find most similar products using FAISS index"""
     if faiss_index is None:
@@ -74,18 +73,19 @@ def find_similar_products(query_embedding, top_k=8):
     distances, indices = faiss_index.search(query_embedding, top_k)
     
     # Retrieve metadata for matching products
-    matching_products = [product_metadata_list[idx] for idx in indices[0]]
-    matching_distances = distances[0]
+    matching_products = df.iloc[indices[0]].copy()  # 從 df 中提取對應的行
+    matching_products["similarity"] = distances[0]  # 加入相似度數據
     
-    return matching_products, matching_distances
+    return matching_products
 
-# Rest of the code remains the same...
+
+# Analyze query and find products
 def analyze_query_and_find_products(query: str) -> str:
     if not query.strip():
         return "請輸入您的問題或搜尋需求"
     
     try:
-        # First, analyze the query to understand intent
+        # Analyze the query to understand intent
         analysis_messages = [
             {"role": "system", "content": f"""You are a knowledgeable shopping assistant. 
              When given a query:
@@ -105,40 +105,15 @@ def analyze_query_and_find_products(query: str) -> str:
         
         analysis = analysis_response.choices[0].message.content
 
-        print("現在要比對向量的是＜analysis＞：" + analysis)
+        print("AAA:" + analysis)
         
-        # Get embedding for the query
+        # Generate embedding for the query
         query_embedding = get_embedding(query + " " + analysis)
         
         # Find similar products
-        matching_products, similarities = find_similar_products(query_embedding)
+        matching_products = find_similar_products(query_embedding)
         print(f"Found {len(matching_products)} matching products")
         
-        # Get recommendations based on found products
-        product_descriptions = "\n".join([
-            f"- {row['item_id']} ({row['item_name']})"
-            for _, row in matching_products.iterrows()
-        ])
-        
-        recommendation_messages = [
-            {"role": "system", "content": f"""Based on the query and available products,
-             provide helpful recommendations and tips. Consider:
-             1. How the products can be used
-             2. What to look for when choosing
-             3. Alternative options if available
-             Respond in Traditional Chinese."""},
-            {"role": "user", "content": f"""Query: {query}
-            Analysis: {analysis}
-            Available products: {product_descriptions}"""}
-        ]
-        
-        recommendation_response = client.chat.completions.create(
-            model="gpt-4o",
-            messages=recommendation_messages,
-            temperature=0.7,
-            max_tokens=250
-        )
-        
         # Format the response
         response_parts = [
             "🔍 需求分析：",
@@ -146,54 +121,40 @@ def analyze_query_and_find_products(query: str) -> str:
             "\n📦 相關商品推薦：\n"
         ]
         
-        for (_, product), similarity in zip(matching_products.iterrows(), similarities):
-            confidence = similarity * 100
+        for _, row in matching_products.iterrows():
             product_info = f"""
-• {product['item_name']}
-  分類: {product['tags']}
-  規格: {product['description']}
-  相關度: {confidence:.1f}%"""
+• {row['item_name']}
+  描述: {row['description']}
+  分類: {row['tags']}
+  規格: {row['spec']}
+  相似度: {row['similarity']:.2f}"""
             response_parts.append(product_info)
         
-        response_parts.extend([
-            "\n💡 購物建議：",
-            recommendation_response.choices[0].message.content
-        ])
-        
+        response_parts.append("\n💡 購物建議：")
+        response_parts.append("根據您的需求，以上是推薦的商品！")
+
         return "\n".join(response_parts)
     
     except Exception as e:
         print(f"Error in search: {str(e)}")
         return f"搜尋發生錯誤: {str(e)}"
 
-# Add system status message
+
+# Get system status
 def get_system_status():
     """Get system initialization status"""
-    # 判斷是否加載了 FAISS 索引
     embeddings_loaded = faiss_index is not None
     embedding_count = faiss_index.ntotal if embeddings_loaded else 0
-    
-    # 假設 count 是從 .pkl 文件中提取的筆數
-    pkl_path = "product_embeddings.pkl"
-    try:
-        with open(pkl_path, "rb") as f:
-            embeddings = pickle.load(f)
-            if isinstance(embeddings, list):
-                count = len(embeddings)
-            elif isinstance(embeddings, np.ndarray):
-                count = embeddings.shape[0]
-            else:
-                count = 0  # 資料格式不明時設為 0
-    except FileNotFoundError:
-        count = 0  # 如果文件不存在，設為 0
-    
+    product_count = len(df)
+
     return {
         "embeddings_loaded": embeddings_loaded,
         "embedding_count": embedding_count,
-        "product_count": count  # 新增 count
+        "product_count": product_count
     }
 
-# Modified interface with status
+
+# Gradio Interface
 with gr.Blocks(theme=gr.themes.Soft()) as demo:
     gr.Markdown(
         """
@@ -218,68 +179,24 @@ with gr.Blocks(theme=gr.themes.Soft()) as demo:
     
     # Main interface
     with gr.Column():
-        # Input area
         input_text = gr.Textbox(
             label="請輸入您的問題或需求",
-            placeholder="您可以詢問任何商品相關的問題，例如：\n- 想找一些適合做便當的食材\n- 需要營養均衡的食材\n- 想買一些新鮮的海鮮\n- 有什麼適合老人家的食物",
+            placeholder="例如：需要適合便當的食材",
             lines=3
         )
-        
-        # Buttons
-        with gr.Row():
-            submit_btn = gr.Button("搜尋", variant="primary")
-            clear_btn = gr.Button("清除")
-        
-        # Output area
         output_text = gr.Textbox(
             label="分析結果與建議",
-            lines=25
+            lines=20
         )
+        submit_btn = gr.Button("搜尋")
         
-        # Clear function
-        def clear_inputs():
-            return {"input_text": "", "output_text": ""}
-        
-        # Setup button actions
         submit_btn.click(
             fn=analyze_query_and_find_products,
             inputs=input_text,
-            outputs=output_text,
-            api_name="search"  # This enables API access
-        )
-        
-        clear_btn.click(
-            fn=clear_inputs,
-            inputs=[],
-            outputs=[input_text, output_text],
-            api_name="clear"
+            outputs=output_text
         )
     
-    # Examples section
-    gr.Markdown("### 搜尋範例")
-    with gr.Row():
-        examples = gr.Examples(
-            examples=[
-                ["想找一些適合做便當的食材"],
-                ["需要一些營養均衡的食物"],
-                ["有沒有適合老人家吃的食物？"],
-                ["想買一些新鮮的海鮮，有什麼推薦？"],
-                ["最近感冒了，有什麼食材可以幫助恢復？"],
-            ],
-            inputs=input_text,
-            outputs=output_text,
-            fn=analyze_query_and_find_products,
-            cache_examples=True
-        )
-    
-    # Footer
-    gr.Markdown(
-        """
-        ---
-        系統���用AI語意分析技術，能更好地理解您的需求並提供相關建議。
-        如有任何問題或建議，歡迎反饋。
-        """
-    )
+    gr.Markdown("--- 系統使用 AI 分析需求並推薦商品。")
 
 # Launch the app
-demo.launch()
+demo.launch()