Fix encoding issues, enhance ChromaDB reader with ID query and embedding model selection

Setup.py

@@ -946,13 +946,18 @@ class WolfChatSetup(tk.Tk):
             logger.info(f"Starting bot: {sys.executable} {bot_script_name}")
             # Ensure CWD is script's directory if main.py relies on relative paths
             script_dir = os.path.dirname(os.path.abspath(__file__)) 
+            current_env = os.environ.copy()
+            current_env["PYTHONIOENCODING"] = "utf-8"
             self.bot_process_instance = subprocess.Popen(
                 [sys.executable, bot_script_name],
                 cwd=script_dir, # Run main.py from its directory
                 stdout=subprocess.PIPE, # Capture output
                 stderr=subprocess.STDOUT, # Redirect stderr to stdout
                 text=True,
-                bufsize=1 # Line buffered
+                encoding='utf-8', # Specify UTF-8 encoding
+                errors='replace', # Handle potential encoding errors
+                bufsize=1, # Line buffered
+                env=current_env # Set PYTHONIOENCODING
             )
             bot_process_instance = self.bot_process_instance # Update global
 
@@ -1852,7 +1857,21 @@ class WolfChatSetup(tk.Tk):
                  messagebox.showwarning("Already Running", "Another process is already running. Please stop it first.")
                  return
 
-            self.running_process = subprocess.Popen([sys.executable, "main.py"])
+            # Run main.py, capturing output with UTF-8 encoding and setting PYTHONIOENCODING
+            current_env = os.environ.copy()
+            current_env["PYTHONIOENCODING"] = "utf-8"
+            self.running_process = subprocess.Popen(
+                [sys.executable, "main.py"],
+                stdout=subprocess.PIPE,
+                stderr=subprocess.STDOUT,
+                text=True,
+                encoding='utf-8',
+                errors='replace',
+                bufsize=1,
+                env=current_env # Set PYTHONIOENCODING
+            )
+            # Start a thread to log bot's output for this independent run as well
+            threading.Thread(target=self._log_subprocess_output, args=(self.running_process, "ChatBot"), daemon=True).start()
             print("Attempting to start main.py...")
             self.update_run_button_states(False) # Disable run buttons, enable stop
         except Exception as e:

tools/chroma_view.py

@@ -3,6 +3,7 @@ import tkinter as tk
 from tkinter import filedialog, messagebox
 import json
 import chromadb
+from chromadb.utils import embedding_functions # 新增導入
 import datetime
 import pandas as pd
 import threading
@@ -15,6 +16,8 @@ from ttkbootstrap.scrolled import ScrolledFrame
 import numpy as np
 import logging
 from typing import List, Dict, Any, Optional, Union, Tuple
+import inspect # 用於檢查函數簽名，判斷是否支持混合搜索
+import re # 新增導入 for ID parsing in UI
 
 class ChromaDBReader:
     """ChromaDB備份讀取器的主數據模型"""
@@ -28,6 +31,9 @@ class ChromaDBReader:
         self.query_results = []  # 當前查詢結果
         self.chroma_client = None  # ChromaDB客戶端
         
+        self.selected_embedding_model_name = "default"  # 用於查詢的嵌入模型
+        self.query_embedding_function = None  # 實例化的查詢嵌入函數, None 表示使用集合內部預設
+
         # 設置日誌
         logging.basicConfig(
             level=logging.INFO,
@@ -118,6 +124,32 @@ class ChromaDBReader:
             self.chroma_client = None
             self.collection_names = []
             return False
+
+    def set_query_embedding_model(self, model_name: str):
+        """設置查詢時使用的嵌入模型"""
+        self.selected_embedding_model_name = model_name
+        if model_name == "default":
+            self.query_embedding_function = None  # 表示使用集合的內部嵌入函數
+            self.logger.info("查詢將使用集合內部嵌入模型。")
+        elif model_name == "all-MiniLM-L6-v2":
+            try:
+                # 注意: sentence-transformers 庫需要安裝
+                self.query_embedding_function = embedding_functions.SentenceTransformerEmbeddingFunction(model_name="all-MiniLM-L6-v2")
+                self.logger.info(f"查詢將使用外部嵌入模型: {model_name}")
+            except Exception as e:
+                self.logger.error(f"無法加載 SentenceTransformer all-MiniLM-L6-v2: {e}。將使用集合內部模型。")
+                self.query_embedding_function = None
+        elif model_name == "paraphrase-multilingual-MiniLM-L12-v2":
+            try:
+                # 注意: sentence-transformers 庫需要安裝
+                self.query_embedding_function = embedding_functions.SentenceTransformerEmbeddingFunction(model_name="paraphrase-multilingual-MiniLM-L12-v2")
+                self.logger.info(f"查詢將使用外部嵌入模型: {model_name}")
+            except Exception as e:
+                self.logger.error(f"無法加載 SentenceTransformer paraphrase-multilingual-MiniLM-L12-v2: {e}。將使用集合內部模型。")
+                self.query_embedding_function = None
+        else:
+            self.logger.warning(f"未知的查詢嵌入模型: {model_name}, 將使用集合內部模型。")
+            self.query_embedding_function = None
     
     def load_collection(self, collection_name: str) -> bool:
         """加載指定的集合"""
@@ -125,6 +157,9 @@ class ChromaDBReader:
             return False
         
         try:
+            # 獲取集合時，如果需要指定 embedding_function (通常在創建時指定)
+            # 此處是讀取，所以集合的 embedding_function 已經固定
+            # 我們將在查詢時使用 self.query_embedding_function 來生成 query_embeddings
             self.current_collection = self.chroma_client.get_collection(collection_name)
             self.logger.info(f"已加載集合: {collection_name}")
             return True
@@ -133,46 +168,220 @@ class ChromaDBReader:
             self.current_collection = None
             return False
     
-    def execute_query(self, query_text: str, n_results: int = 5) -> List[Dict]:
+    def execute_query(self, query_text: str, n_results: int = 5, 
-        """執行查詢並返回結果"""
+                  query_type: str = "basic", 
+                  where: Dict = None, 
+                  where_document: Dict = None,
+                  include: List[str] = None,
+                  metadata_filter: Dict = None,
+                  hybrid_alpha: float = None) -> List[Dict]:
+        """執行查詢並返回結果
+    
+        參數:
+            query_text: 查詢文本
+            n_results: 返回結果數量
+            query_type: 查詢類型 (basic, metadata, hybrid, multi_vector)
+            where: where 過濾條件
+            where_document: 文檔內容過濾條件
+            include: 指定包含的文檔 ID
+            metadata_filter: 元數據過濾條件
+            hybrid_alpha: 混合搜索的權重參數（0-1之間，越大越傾向關鍵詞搜索）
+        """
         if not self.current_collection or not query_text:
             return []
-        
+    
         try:
-            results = self.current_collection.query(
+            query_params = {
-                query_texts=[query_text],
+                "n_results": n_results
-                n_results=n_results
+            }
-            )
             
-            # 轉換結果為更易用的格式
+            # 基本查詢處理邏輯
-            processed_results = []
+            if query_type == "basic":
-            for i, (doc_id, document, metadata, distance) in enumerate(zip(
+                query_params["query_texts"] = [query_text]
-                results['ids'][0],
+            # 多向量查詢（用於比較多個查詢之間的相似性）
-                results['documents'][0],
+            elif query_type == "multi_vector":
-                results['metadatas'][0] if 'metadatas' in results and results['metadatas'][0] else [{}] * len(results['ids'][0]),
+                # 支持以 "|||" 或換行符分隔的多個查詢文本
-                results['distances'][0] if 'distances' in results else [0] * len(results['ids'][0])
+                if "|||" in query_text:
-            )):
+                    query_texts = [text.strip() for text in query_text.split("|||")]
-                # 計算相似度分數 (將距離轉換為相似度: 1 - 歸一化距離)
+                else:
-                # 注意: 根據ChromaDB使用的距離度量可能需要調整
+                    query_texts = [text.strip() for text in query_text.splitlines() if text.strip()]
-                similarity = 1.0 - min(distance, 1.0)  # 確保值在0-1之間
+                query_params["query_texts"] = query_texts
+            
+            # 添加其他查詢參數
+            if where:
+                query_params["where"] = where
+            if where_document:
+                query_params["where_document"] = where_document
+            if include:
+                query_params["include"] = include
+            if metadata_filter:
+                # 直接將元數據過濾條件轉換為 where 條件
+                if "where" not in query_params:
+                    query_params["where"] = {}
+                query_params["where"].update(metadata_filter)
+            
+            # 混合搜索處理
+            if query_type == "hybrid" and hybrid_alpha is not None:
+                # 檢查 ChromaDB 版本是否支持混合搜索
+                if hasattr(self.current_collection, "query") and "alpha" in inspect.signature(self.current_collection.query).parameters:
+                    query_params["alpha"] = hybrid_alpha
+                    # 混合搜索通常需要 query_texts
+                    if "query_texts" not in query_params:
+                         query_params["query_texts"] = [query_text]
+                else:
+                    self.logger.warning("當前 ChromaDB 版本不支持混合搜索，將使用基本查詢")
+                    query_type = "basic" # 降級為基本查詢
+                    query_params["query_texts"] = [query_text]
+            elif query_type == "hybrid" and hybrid_alpha is None:
+                # 如果是混合搜索但未提供 alpha，則默認為基本搜索
+                self.logger.warning("混合搜索未提供 Alpha 值，將使用基本查詢")
+                query_type = "basic"
+                query_params["query_texts"] = [query_text]
+
+
+            # 如果 query_type 不是 multi_vector 且 query_texts 未設置，則設置
+            if query_type not in ["multi_vector", "hybrid"] and "query_texts" not in query_params:
+                 query_params["query_texts"] = [query_text]
+
+            # 如果選擇了外部嵌入模型且不是混合查詢，則生成查詢嵌入
+            if query_type != "hybrid" and \
+               "query_texts" in query_params and \
+               self.query_embedding_function:
                 
-                processed_results.append({
+                texts_to_embed = query_params["query_texts"]
-                    "rank": i + 1,
+                try:
-                    "id": doc_id,
+                    # self.query_embedding_function 接受 List[str] 返回 List[List[float]]
-                    "document": document,
+                    generated_embeddings = self.query_embedding_function(texts_to_embed)
-                    "metadata": metadata,
+
-                    "similarity": similarity,
+                    if generated_embeddings and all(isinstance(emb, list) for emb in generated_embeddings):
-                    "distance": distance
+                        query_params["query_embeddings"] = generated_embeddings
-                })
+                        if "query_texts" in query_params: # 確保它存在才刪除
+                            del query_params["query_texts"]
+                        self.logger.info(f"使用 {self.selected_embedding_model_name} 生成了 {len(generated_embeddings)} 個查詢嵌入。")
+                    else:
+                        self.logger.warning(f"未能使用 {self.selected_embedding_model_name} 為所有查詢文本生成有效嵌入。將回退到使用集合預設嵌入函數進行文本查詢。嵌入結果: {generated_embeddings}")
+                except Exception as e:
+                    self.logger.error(f"使用 {self.selected_embedding_model_name} 生成查詢嵌入時出錯: {e}。將回退到使用集合預設嵌入函數進行文本查詢。")
+
+            # 執行查詢
+            results = self.current_collection.query(**query_params)
+            
+            # 處理結果
+            processed_results = []
+            
+            # 獲取查詢返回的所有結果列表
+            ids_list = results.get('ids', [[]])
+            documents_list = results.get('documents', [[]])
+            metadatas_list = results.get('metadatas', [[]])
+            distances_list = results.get('distances', [[]])
+            
+            # 確保列表長度一致，並為空列表提供默認值
+            num_queries = len(ids_list)
+            if not documents_list or len(documents_list) != num_queries:
+                documents_list = [[] for _ in range(num_queries)]
+            if not metadatas_list or len(metadatas_list) != num_queries:
+                metadatas_list = [[{}] * len(ids_list[i]) for i in range(num_queries)]
+            if not distances_list or len(distances_list) != num_queries:
+                distances_list = [[0.0] * len(ids_list[i]) for i in range(num_queries)]
+
+            # 對於多查詢文本的情況，需要分別處理每個查詢的結果
+            for query_idx, (ids, documents, metadatas, distances) in enumerate(zip(
+                ids_list, 
+                documents_list,
+                metadatas_list,
+                distances_list
+            )):
+                # 處理每個查詢結果
+                for i, (doc_id, document, metadata, distance) in enumerate(zip(
+                    ids, documents, 
+                    metadatas if metadatas else [{}] * len(ids), # 再次確保元數據存在
+                    distances if distances else [0.0] * len(ids) # 再次確保距離存在
+                )):
+                    # 計算相似度分數
+                    similarity = 1.0 - min(float(distance) if distance is not None else 1.0, 1.0)
+                    
+                    result_item = {
+                        "rank": i + 1,
+                        "query_index": query_idx,
+                        "id": doc_id,
+                        "document": document,
+                        "metadata": metadata if metadata else {}, # 確保 metadata 是字典
+                        "similarity": similarity,
+                        "distance": float(distance) if distance is not None else 0.0,
+                        "query_type": query_type
+                    }
+                    
+                    if query_type == "hybrid":
+                        result_item["hybrid_alpha"] = hybrid_alpha
+                    
+                    processed_results.append(result_item)
             
             self.query_results = processed_results
-            self.logger.info(f"查詢完成，找到 {len(processed_results)} 個結果")
+            self.logger.info(f"查詢完成，找到 {len(processed_results)} 個結果，查詢類型: {query_type}")
             return processed_results
             
         except Exception as e:
             self.logger.error(f"執行查詢時出錯: {str(e)}")
             self.query_results = []
             return []
+
+    def get_documents_by_ids(self, doc_ids: List[str]) -> List[Dict]:
+        """按文檔ID列表獲取文檔"""
+        if not self.current_collection:
+            self.logger.warning("沒有選擇集合，無法按 ID 獲取文檔。")
+            return []
+        if not doc_ids:
+            self.logger.warning("未提供文檔 ID。")
+            return []
+
+        try:
+            results = self.current_collection.get(
+                ids=doc_ids,
+                include=["documents", "metadatas"] 
+            )
+            
+            processed_results = []
+            retrieved_ids = results.get('ids', [])
+            retrieved_documents = results.get('documents', [])
+            retrieved_metadatas = results.get('metadatas', [])
+
+            # 創建一個字典以便快速查找已檢索到的文檔信息
+            found_docs_map = {}
+            for i, r_id in enumerate(retrieved_ids):
+                found_docs_map[r_id] = {
+                    "document": retrieved_documents[i] if i < len(retrieved_documents) else None,
+                    "metadata": retrieved_metadatas[i] if i < len(retrieved_metadatas) else {}
+                }
+
+            rank_counter = 1
+            for original_id in doc_ids: # 遍歷原始請求的ID，以保持某種順序感，並標記未找到的
+                if original_id in found_docs_map:
+                    doc_data = found_docs_map[original_id]
+                    if doc_data["document"] is not None:
+                        processed_results.append({
+                            "rank": rank_counter,
+                            "id": original_id,
+                            "document": doc_data["document"],
+                            "metadata": doc_data["metadata"],
+                            "similarity": None, # Not applicable
+                            "distance": None,   # Not applicable
+                            "query_type": "id_lookup" 
+                        })
+                        rank_counter += 1
+                    else: # ID 存在但文檔為空（理論上不應發生在 get 中，除非 include 設置問題）
+                        self.logger.warning(f"ID {original_id} 找到但文檔內容為空。")
+                # else: # ID 未在返回結果中找到，可以選擇不添加到 processed_results 或添加一個標記
+                #    self.logger.info(f"ID {original_id} 未在集合中找到。")
+            
+            self.query_results = processed_results
+            self.logger.info(f"按 ID 查詢完成，從請求的 {len(doc_ids)} 個ID中，實際找到 {len(processed_results)} 個文檔。")
+            return processed_results
+
+        except Exception as e:
+            self.logger.error(f"按 ID 獲取文檔時出錯: {str(e)}")
+            # traceback.print_exc() # For debugging
+            self.query_results = []
+            return []
     
     def get_collection_info(self, collection_name: str) -> Dict:
         """獲取集合的詳細信息"""
@@ -235,7 +444,19 @@ class ChromaDBReaderUI:
         # 設置窗口
         self.root.title("ChromaDB 備份讀取器")
         self.root.geometry("1280x800")
-        self.setup_ui()
+        
+        # 初始化嵌入模型相關變量
+        self.embedding_model_var = tk.StringVar(value="預設 (ChromaDB)") # 顯示名稱
+        self.embedding_models = {
+            "預設 (ChromaDB)": "default",
+            "all-MiniLM-L6-v2 (ST)": "all-MiniLM-L6-v2",
+            "paraphrase-multilingual-MiniLM-L12-v2 (ST)": "paraphrase-multilingual-MiniLM-L12-v2"
+        }
+        # 初始化 reader 中的嵌入模型 (確保 reader 實例已創建)
+        # self.reader.set_query_embedding_model(self.embedding_models[self.embedding_model_var.get()])
+        # ^^^ 這行需要在 setup_ui 之後，或者在 on_embedding_model_changed 中處理首次設置
+
+        self.setup_ui() # setup_ui 會創建 reader 實例
         
         # 默認主題
         self.current_theme = "darkly"  # ttkbootstrap的深色主題
@@ -262,9 +483,13 @@ class ChromaDBReaderUI:
         # 右側面板 (查詢和結果)
         self.right_panel = ttk.Frame(self.main_frame)
         self.right_panel.pack(side=LEFT, fill=BOTH, expand=YES)
+
+        # 設置狀態欄 (提前，以確保 self.status_var 在其他地方使用前已定義)
+        self.setup_status_bar()
         
         # 設置左側面板
         self.setup_directory_frame()
+        self.setup_embedding_model_frame() # 新增嵌入模型選擇框架
         self.setup_backups_frame()
         self.setup_collections_frame()
         
@@ -272,9 +497,6 @@ class ChromaDBReaderUI:
         self.setup_query_frame()
         self.setup_results_frame()
         
-        # 設置狀態欄
-        self.setup_status_bar()
-        
         # 設置菜單
         self.setup_menu()
     
@@ -314,6 +536,24 @@ class ChromaDBReaderUI:
         ttk.Entry(dir_frame, textvariable=self.backups_dir_var).pack(side=LEFT, fill=X, expand=YES)
         ttk.Button(dir_frame, text="瀏覽", command=self.browse_directory).pack(side=LEFT, padx=(5, 0))
         ttk.Button(dir_frame, text="載入", command=self.load_backups_directory).pack(side=LEFT, padx=(5, 0))
+
+    def setup_embedding_model_frame(self):
+        """設置查詢嵌入模型選擇框架"""
+        embedding_frame = ttk.LabelFrame(self.left_panel, text="查詢嵌入模型", padding=10)
+        embedding_frame.pack(fill=X, pady=(0, 10))
+
+        self.embedding_model_combo = ttk.Combobox(
+            embedding_frame,
+            textvariable=self.embedding_model_var,
+            values=list(self.embedding_models.keys()),
+            state="readonly"
+        )
+        self.embedding_model_combo.pack(fill=X, expand=YES)
+        self.embedding_model_combo.set(list(self.embedding_models.keys())[0]) # 設置預設顯示值
+        self.embedding_model_combo.bind("<<ComboboxSelected>>", self.on_embedding_model_changed)
+
+        # 初始化Reader中的嵌入模型選擇
+        self.on_embedding_model_changed() 
     
     def setup_backups_frame(self):
         """設置備份列表框架"""
@@ -388,12 +628,46 @@ class ChromaDBReaderUI:
         query_frame = ttk.LabelFrame(self.right_panel, text="查詢", padding=10)
         query_frame.pack(fill=X, pady=(0, 10))
         
-        # 查詢文本輸入
+        # 創建一個 Notebook 以包含不同的查詢類型標籤頁
-        ttk.Label(query_frame, text="查詢文本:").pack(anchor=W)
+        self.query_notebook = ttk.Notebook(query_frame)
-        self.query_text = tk.Text(query_frame, height=4, width=50)
+        self.query_notebook.pack(fill=X, pady=5)
-        self.query_text.pack(fill=X, pady=5)
         
-        # 查詢參數
+        # 基本查詢標籤頁
+        self.basic_query_frame = ttk.Frame(self.query_notebook)
+        self.query_notebook.add(self.basic_query_frame, text="基本查詢")
+        
+        # 元數據查詢標籤頁
+        self.metadata_query_frame = ttk.Frame(self.query_notebook)
+        self.query_notebook.add(self.metadata_query_frame, text="元數據查詢")
+        
+        # 混合查詢標籤頁
+        self.hybrid_query_frame = ttk.Frame(self.query_notebook)
+        self.query_notebook.add(self.hybrid_query_frame, text="混合查詢")
+        
+        # 多向量查詢標籤頁
+        self.multi_vector_frame = ttk.Frame(self.query_notebook)
+        self.query_notebook.add(self.multi_vector_frame, text="多向量查詢")
+
+        # ID 查詢標籤頁 (新增)
+        self.id_query_frame = ttk.Frame(self.query_notebook)
+        self.query_notebook.add(self.id_query_frame, text="ID 查詢")
+        
+        # 設置基本查詢頁面
+        self.setup_basic_query_tab()
+        
+        # 設置元數據查詢頁面
+        self.setup_metadata_query_tab()
+        
+        # 設置混合查詢頁面
+        self.setup_hybrid_query_tab()
+        
+        # 設置多向量查詢頁面
+        self.setup_multi_vector_tab()
+
+        # 設置 ID 查詢頁面 (新增)
+        self.setup_id_query_tab()
+        
+        # 查詢參數（共用部分）
         params_frame = ttk.Frame(query_frame)
         params_frame.pack(fill=X)
         
@@ -405,9 +679,102 @@ class ChromaDBReaderUI:
         ttk.Button(
             query_frame, 
             text="執行查詢", 
-            command=self.execute_query,
+            command=self.execute_query, # 注意：這個 execute_query 方法將被新的替換
             style="Accent.TButton"
         ).pack(pady=10)
+
+    def setup_basic_query_tab(self):
+        """設置基本查詢標籤頁"""
+        ttk.Label(self.basic_query_frame, text="查詢文本:").pack(anchor=W)
+        self.basic_query_text = tk.Text(self.basic_query_frame, height=4, width=50)
+        self.basic_query_text.pack(fill=X, pady=5)
+
+    def setup_metadata_query_tab(self):
+        """設置元數據查詢標籤頁"""
+        ttk.Label(self.metadata_query_frame, text="查詢文本:").pack(anchor=W)
+        self.metadata_query_text = tk.Text(self.metadata_query_frame, height=4, width=50)
+        self.metadata_query_text.pack(fill=X, pady=5)
+        
+        ttk.Label(self.metadata_query_frame, text="元數據過濾條件 (JSON 格式):").pack(anchor=W)
+        self.metadata_filter_text = tk.Text(self.metadata_query_frame, height=4, width=50)
+        self.metadata_filter_text.pack(fill=X, pady=5)
+        self.metadata_filter_text.insert("1.0", '{"key": "value"}')
+        
+        # 添加一個幫助按鈕，顯示元數據過濾語法的說明
+        ttk.Button(
+            self.metadata_query_frame,
+            text="?",
+            width=2,
+            command=self.show_metadata_help
+        ).pack(anchor=E)
+
+    def setup_hybrid_query_tab(self):
+        """設置混合查詢標籤頁"""
+        ttk.Label(self.hybrid_query_frame, text="查詢文本:").pack(anchor=W)
+        self.hybrid_query_text = tk.Text(self.hybrid_query_frame, height=4, width=50)
+        self.hybrid_query_text.pack(fill=X, pady=5)
+        
+        alpha_frame = ttk.Frame(self.hybrid_query_frame)
+        alpha_frame.pack(fill=X)
+        
+        ttk.Label(alpha_frame, text="Alpha 值 (0-1):").pack(side=LEFT)
+        self.hybrid_alpha_var = tk.DoubleVar(value=0.5)
+        ttk.Scale(
+            alpha_frame, 
+            from_=0.0, to=1.0, 
+            variable=self.hybrid_alpha_var, 
+            orient=tk.HORIZONTAL,
+            length=200
+        ).pack(side=LEFT, padx=5, fill=X, expand=YES)
+        
+        # 創建一個Label來顯示Scale的當前值
+        self.hybrid_alpha_label = ttk.Label(alpha_frame, text=f"{self.hybrid_alpha_var.get():.2f}")
+        self.hybrid_alpha_label.pack(side=LEFT)
+        # 綁定Scale的變動到更新Label的函數
+        self.hybrid_alpha_var.trace_add("write", lambda *args: self.hybrid_alpha_label.config(text=f"{self.hybrid_alpha_var.get():.2f}"))
+
+        ttk.Label(self.hybrid_query_frame, text="注意: Alpha=0 完全使用向量搜索，Alpha=1 完全使用關鍵詞搜索").pack(pady=2)
+        ttk.Label(self.hybrid_query_frame, text="混合查詢將使用集合原始嵌入模型，忽略上方選擇的查詢嵌入模型。", font=("TkDefaultFont", 8)).pack(pady=2)
+
+
+    def setup_multi_vector_tab(self):
+        """設置多向量查詢標籤頁"""
+        ttk.Label(self.multi_vector_frame, text="多個查詢文本 (每行一個，或使用 ||| 分隔):").pack(anchor=W)
+        self.multi_vector_text = tk.Text(self.multi_vector_frame, height=6, width=50)
+        self.multi_vector_text.pack(fill=X, pady=5)
+        self.multi_vector_text.insert("1.0", "查詢文本 1\n|||查詢文本 2\n|||查詢文本 3")
+        
+        ttk.Label(self.multi_vector_frame, text="用於比較多個查詢之間的相似性").pack(pady=5)
+
+    def setup_id_query_tab(self):
+        """設置ID查詢標籤頁"""
+        ttk.Label(self.id_query_frame, text="文檔 ID (每行一個，或用逗號/空格分隔):").pack(anchor=tk.W)
+        self.id_query_text = tk.Text(self.id_query_frame, height=6, width=50)
+        self.id_query_text.pack(fill=tk.X, pady=5)
+        self.id_query_text.insert("1.0", "id1\nid2,id3 id4") # 示例
+        ttk.Label(self.id_query_frame, text="此查詢將獲取指定ID的文檔，忽略上方“結果數量”設置。").pack(pady=5)
+
+
+    def show_metadata_help(self):
+        """顯示元數據過濾語法說明"""
+        help_text = """元數據過濾語法示例:
+
+基本過濾:
+{"category": "文章"}  # 精確匹配
+
+範圍過濾:
+{"date": {"$gt": "2023-01-01"}}  # 大於
+{"date": {"$lt": "2023-12-31"}}  # 小於
+{"count": {"$gte": 10}}  # 大於等於
+{"count": {"$lte": 100}}  # 小於等於
+
+多條件過濾:
+{"$and": [{"category": "文章"}, {"author": "張三"}]}  # AND 條件
+{"$or": [{"category": "文章"}, {"category": "新聞"}]}  # OR 條件
+
+注意: 此處語法遵循 ChromaDB 的過濾語法，非標準 JSON 查詢語法。
+"""
+        messagebox.showinfo("元數據過濾語法說明", help_text)
     
     def setup_results_frame(self):
         """設置結果顯示框架"""
@@ -442,6 +809,26 @@ class ChromaDBReaderUI:
         self.status_var = tk.StringVar(value="就緒")
         status_label = ttk.Label(status_frame, textvariable=self.status_var, relief=tk.SUNKEN, anchor=W)
         status_label.pack(fill=X)
+
+    def on_embedding_model_changed(self, event=None):
+        """處理查詢嵌入模型選擇變更事件"""
+        selected_display_name = self.embedding_model_var.get()
+        model_name_key = self.embedding_models.get(selected_display_name, "default")
+        
+        if hasattr(self, 'reader') and self.reader:
+            self.reader.set_query_embedding_model(model_name_key) # 更新Reader中的模型
+            
+            # 更新狀態欄提示
+            if model_name_key == "default":
+                self.status_var.set("查詢將使用集合內部嵌入模型。")
+            elif self.reader.query_embedding_function: # 檢查模型是否成功加載
+                self.status_var.set(f"查詢將使用外部模型: {selected_display_name}")
+            else: # 加載失敗
+                self.status_var.set(f"模型 {selected_display_name} 加載失敗/無效，將使用集合內部模型。")
+        else:
+            # Reader尚未初始化，這通常在UI初始化早期發生
+            # self.reader.set_query_embedding_model 會在 setup_embedding_model_frame 中首次調用時處理
+            pass
     
     def browse_directory(self):
         """瀏覽選擇備份目錄"""
@@ -527,27 +914,38 @@ class ChromaDBReaderUI:
         
         # 獲取選定項的索引
         item_id = selection[0]
-        item_index = self.backups_tree.index(item_id)
+        # item_index = self.backups_tree.index(item_id) # 這個索引是相對於當前顯示的項目的
-        
+
-        # 獲取所有顯示的備份項目
+        # 直接從 Treeview item 中獲取備份名稱，然後在 self.reader.backups 中查找
-        visible_items = self.backups_tree.get_children()
+        try:
-        if item_index >= len(visible_items):
+            backup_name_from_tree = self.backups_tree.item(item_id)["values"][0]
+        except IndexError:
+            self.logger.error("無法從 Treeview 獲取備份名稱")
             return
+
+        actual_backup_index = -1
+        for i, backup_info in enumerate(self.reader.backups):
+            if backup_info["name"] == backup_name_from_tree:
+                actual_backup_index = i
+                break
         
-        # 查找此顯示項對應的實際備份索引
+        if actual_backup_index == -1:
-        backup_name = self.backups_tree.item(visible_items[item_index])["values"][0]
+            self.logger.error(f"在備份列表中未找到名為 {backup_name_from_tree} 的備份")
-        backup_index = next((i for i, b in enumerate(self.reader.backups) if b["name"] == backup_name), -1)
-        
-        if backup_index == -1:
             return
         
         # 載入備份
-        self.status_var.set(f"正在載入備份: {backup_name}...")
+        self.status_var.set(f"正在載入備份: {backup_name_from_tree}...")
         self.root.update_idletasks()
         
+        # 確保 Reader 中的嵌入模型是最新的 (雖然 on_embedding_model_changed 應該已經處理了)
+        # selected_display_name = self.embedding_model_var.get()
+        # model_key = self.embedding_models.get(selected_display_name, "default")
+        # self.reader.set_query_embedding_model(model_key) # 這行不需要，因為模型選擇是獨立的
+
         def load_backup_thread():
-            success = self.reader.load_backup(backup_index)
+            # load_backup 不再需要 embedding_model_name 參數，因為嵌入模型選擇是針對查詢的
-            self.root.after(0, lambda: self.finalize_backup_loading(success, backup_name))
+            success = self.reader.load_backup(actual_backup_index)
+            self.root.after(0, lambda: self.finalize_backup_loading(success, backup_name_from_tree))
         
         threading.Thread(target=load_backup_thread).start()
     
@@ -618,7 +1016,7 @@ class ChromaDBReaderUI:
             # 獲取集合詳細信息並顯示
             info = self.reader.get_collection_info(collection_name)
             info_text = f"集合: {info['name']}\n文檔數: {info['document_count']}\n向量維度: {info['dimension']}"
-            messagebox.showinfo("集合信息", info_text)
+            # messagebox.showinfo("集合信息", info_text) # 暫時註解掉，避免每次選集合都彈窗
         else:
             self.status_var.set(f"載入集合失敗: {collection_name}")
             messagebox.showerror("錯誤", f"無法載入集合: {collection_name}")
@@ -629,25 +1027,170 @@ class ChromaDBReaderUI:
             messagebox.showinfo("提示", "請先選擇一個集合")
             return
         
-        query_text = self.query_text.get("1.0", tk.END).strip()
+        # 根據當前選擇的標籤頁確定查詢類型
-        if not query_text:
+        try:
-            messagebox.showinfo("提示", "請輸入查詢文本")
+            current_tab_widget = self.query_notebook.nametowidget(self.query_notebook.select())
-            return
+            if current_tab_widget == self.basic_query_frame:
-        
+                current_tab = 0
+            elif current_tab_widget == self.metadata_query_frame:
+                current_tab = 1
+            elif current_tab_widget == self.hybrid_query_frame:
+                current_tab = 2
+            elif current_tab_widget == self.multi_vector_frame:
+                current_tab = 3
+            elif current_tab_widget == self.id_query_frame: # 新增 ID 查詢頁判斷
+                current_tab = 4
+            else:
+                messagebox.showerror("錯誤", "未知的查詢標籤頁")
+                return
+        except tk.TclError: # Notebook可能還沒有任何分頁被選中
+             messagebox.showerror("錯誤", "請選擇一個查詢類型標籤頁")
+             return
+
+        # 獲取查詢參數
         try:
             n_results = int(self.n_results_var.get())
         except ValueError:
             messagebox.showerror("錯誤", "結果數量必須是整數")
             return
         
-        self.status_var.set("正在執行查詢...")
+        # 執行不同類型的查詢
+        if current_tab == 0:  # 基本查詢
+            query_text = self.basic_query_text.get("1.0", tk.END).strip()
+            if not query_text:
+                messagebox.showinfo("提示", "請輸入查詢文本")
+                return
+            
+            self.status_var.set("正在執行基本查詢...")
+            self.execute_basic_query(query_text, n_results)
+            
+        elif current_tab == 1:  # 元數據查詢
+            query_text = self.metadata_query_text.get("1.0", tk.END).strip()
+            metadata_filter_text = self.metadata_filter_text.get("1.0", tk.END).strip()
+            
+            if not query_text: # 元數據查詢的文本也可以是空的，如果只想用metadata_filter
+                # messagebox.showinfo("提示", "請輸入查詢文本")
+                # return
+                pass # 允許空查詢文本
+            
+            try:
+                metadata_filter = json.loads(metadata_filter_text) if metadata_filter_text else None
+            except json.JSONDecodeError:
+                messagebox.showerror("錯誤", "元數據過濾條件必須是有效的 JSON 格式")
+                return
+            
+            if not query_text and not metadata_filter:
+                messagebox.showinfo("提示", "請輸入查詢文本或元數據過濾條件")
+                return
+
+            self.status_var.set("正在執行元數據查詢...")
+            self.execute_metadata_query(query_text, n_results, metadata_filter)
+            
+        elif current_tab == 2:  # 混合查詢
+            query_text = self.hybrid_query_text.get("1.0", tk.END).strip()
+            hybrid_alpha = self.hybrid_alpha_var.get()
+            
+            if not query_text:
+                messagebox.showinfo("提示", "請輸入查詢文本")
+                return
+            
+            self.status_var.set("正在執行混合查詢...")
+            self.execute_hybrid_query(query_text, n_results, hybrid_alpha)
+            
+        elif current_tab == 3:  # 多向量查詢
+            query_text = self.multi_vector_text.get("1.0", tk.END).strip()
+            
+            if not query_text:
+                messagebox.showinfo("提示", "請輸入查詢文本")
+                return
+            
+            self.status_var.set("正在執行多向量查詢...")
+            self.execute_multi_vector_query(query_text, n_results)
+
+        elif current_tab == 4: # ID 查詢
+            id_input_str = self.id_query_text.get("1.0", tk.END).strip()
+            if not id_input_str:
+                messagebox.showinfo("提示", "請輸入文檔 ID。")
+                return
+
+            # 解析 ID: 支持逗號、空格、換行符分隔
+            doc_ids = [id_val.strip() for id_val in re.split(r'[,\s\n]+', id_input_str) if id_val.strip()]
+            
+            if not doc_ids:
+                messagebox.showinfo("提示", "未解析到有效的文檔 ID。")
+                return
+            
+            self.status_var.set("正在按 ID 獲取文檔...")
+            self.execute_id_lookup_query(doc_ids)
+
+
+    def execute_basic_query(self, query_text, n_results):
+        """執行基本查詢"""
+        self.status_var.set(f"正在執行基本查詢: {query_text[:30]}...")
         self.root.update_idletasks()
-        
         def query_thread():
-            results = self.reader.execute_query(query_text, n_results)
+            results = self.reader.execute_query(
+                query_text=query_text, 
+                n_results=n_results,
+                query_type="basic"
+            )
             self.root.after(0, lambda: self.display_results(results))
         
-        threading.Thread(target=query_thread).start()
+        threading.Thread(target=query_thread, daemon=True).start()
+
+    def execute_metadata_query(self, query_text, n_results, metadata_filter):
+        """執行元數據查詢"""
+        self.status_var.set(f"正在執行元數據查詢: {query_text[:30]}...")
+        self.root.update_idletasks()
+        def query_thread():
+            results = self.reader.execute_query(
+                query_text=query_text, 
+                n_results=n_results,
+                query_type="metadata", # 這裡應該是 "metadata" 但後端邏輯會轉為 where
+                metadata_filter=metadata_filter
+            )
+            self.root.after(0, lambda: self.display_results(results))
+        
+        threading.Thread(target=query_thread, daemon=True).start()
+
+    def execute_hybrid_query(self, query_text, n_results, hybrid_alpha):
+        """執行混合查詢"""
+        self.status_var.set(f"正在執行混合查詢 (α={hybrid_alpha:.2f}): {query_text[:30]}...")
+        self.root.update_idletasks()
+        def query_thread():
+            results = self.reader.execute_query(
+                query_text=query_text, 
+                n_results=n_results,
+                query_type="hybrid",
+                hybrid_alpha=hybrid_alpha
+            )
+            self.root.after(0, lambda: self.display_results(results))
+        
+        threading.Thread(target=query_thread, daemon=True).start()
+
+    def execute_multi_vector_query(self, query_text, n_results):
+        """執行多向量查詢"""
+        self.status_var.set(f"正在執行多向量查詢: {query_text.splitlines()[0][:30] if query_text.splitlines() else ''}...")
+        self.root.update_idletasks()
+        def query_thread():
+            results = self.reader.execute_query(
+                query_text=query_text, 
+                n_results=n_results,
+                query_type="multi_vector"
+            )
+            self.root.after(0, lambda: self.display_results(results))
+        
+        threading.Thread(target=query_thread, daemon=True).start()
+
+    def execute_id_lookup_query(self, doc_ids: List[str]):
+        """執行ID查找查詢"""
+        self.status_var.set(f"正在按 ID 獲取 {len(doc_ids)} 個文檔...")
+        self.root.update_idletasks()
+        def query_thread():
+            results = self.reader.get_documents_by_ids(doc_ids)
+            self.root.after(0, lambda: self.display_results(results))
+        
+        threading.Thread(target=query_thread, daemon=True).start()
     
     def display_results(self, results):
         """顯示查詢結果"""
@@ -679,27 +1222,49 @@ class ChromaDBReaderUI:
             widget.destroy()
             
         # 創建表格
-        columns = ("rank", "similarity", "id", "document")
+        columns = ("rank", "similarity", "query_type", "id", "document")
         tree = ttk.Treeview(self.list_view, columns=columns, show="headings")
         tree.heading("rank", text="#")
         tree.heading("similarity", text="相似度")
+        tree.heading("query_type", text="查詢類型")
         tree.heading("id", text="文檔ID")
         tree.heading("document", text="文檔內容")
         
         tree.column("rank", width=50, anchor=CENTER)
         tree.column("similarity", width=100, anchor=CENTER)
-        tree.column("id", width=200)
+        tree.column("query_type", width=120, anchor=CENTER) # 調整寬度以適應更長的類型名稱
-        tree.column("document", width=600)
+        tree.column("id", width=150)
+        tree.column("document", width=530) # 調整寬度
+        
+        # 確定查詢類型名稱映射
+        query_type_names = {
+            "basic": "基本查詢",
+            "metadata": "元數據查詢", 
+            "hybrid": "混合查詢",
+            "multi_vector": "多向量查詢",
+            "id_lookup": "ID 查詢" # 新增
+        }
         
         # 添加結果到表格
         for result in results:
+            raw_query_type = result.get("query_type", "basic")
+            display_query_type = query_type_names.get(raw_query_type, raw_query_type.capitalize())
+
+            if raw_query_type == "hybrid" and "hybrid_alpha" in result:
+                display_query_type += f" (α={result['hybrid_alpha']:.2f})"
+            if raw_query_type == "multi_vector" and "query_index" in result:
+                display_query_type += f" (Q{result['query_index']+1})"
+            
+            similarity_display = f"{result.get('similarity', 0.0):.4f}" if result.get('similarity') is not None else "N/A"
+            
             tree.insert(
                 "", "end",
                 values=(
-                    result["rank"],
+                    result.get("rank", "-"),
-                    f"{result['similarity']:.4f}",
+                    similarity_display,
-                    result["id"],
+                    display_query_type,
-                    result["document"][:100] + ("..." if len(result["document"]) > 100 else "")
+                    result.get("id", "N/A"),
+                    result.get("document", "")[:100] + ("..." if len(result.get("document", "")) > 100 else "")
                 )
             )
         
@@ -710,7 +1275,6 @@ class ChromaDBReaderUI:
         # 雙擊項目顯示完整內容
         tree.bind("<Double-1>", lambda event: self.show_full_document(tree))
         
-        # 使用 Frame 容器來實現滾動功能
         # 佈局
         tree.pack(side=LEFT, fill=BOTH, expand=YES)
         scrollbar.pack(side=RIGHT, fill=Y)
@@ -739,7 +1303,10 @@ class ChromaDBReaderUI:
                 
                 # 添加文檔信息
                 info_text = f"文檔ID: {result['id']}\n"
-                info_text += f"相似度: {result['similarity']:.4f}\n"
+                if result.get('similarity') is not None:
+                    info_text += f"相似度: {result['similarity']:.4f}\n"
+                else:
+                    info_text += "相似度: N/A\n"
                 
                 if result['metadata']:
                     info_text += "\n元數據:\n"
@@ -806,9 +1373,10 @@ class ChromaDBReaderUI:
             title_frame = ttk.Frame(card)
             title_frame.pack(fill=X)
             
+            similarity_text_detail = f"{result['similarity']:.4f}" if result.get('similarity') is not None else "N/A"
             ttk.Label(
                 title_frame, 
-                text=f"#{result['rank']} - 相似度: {result['similarity']:.4f}", 
+                text=f"#{result['rank']} - 相似度: {similarity_text_detail}", 
                 font=("TkDefaultFont", 10, "bold")
             ).pack(side=LEFT)
             
@@ -881,7 +1449,10 @@ class ChromaDBReaderUI:
         
         # 添加文檔信息
         info_text = f"文檔ID: {result['id']}\n"
-        info_text += f"相似度: {result['similarity']:.4f}\n"
+        if result.get('similarity') is not None:
+            info_text += f"相似度: {result['similarity']:.4f}\n"
+        else:
+            info_text += "相似度: N/A\n"
         
         if result['metadata']:
             info_text += "\n元數據:\n"
@@ -1250,4 +1821,4 @@ def main():
     root.mainloop()
 
 if __name__ == "__main__":
-    main()
+    main()

ui_interaction.py

@@ -4,6 +4,8 @@
 import pyautogui
 import cv2 # opencv-python
 import numpy as np
+import sys # Added for special character handling
+import io  # Added for special character handling
 import pyperclip
 import time
 import os
@@ -22,6 +24,26 @@ import math # Added for distance calculation in dual method
 # Or could use threading.Event()
 monitoring_paused_flag = [False] # List containing a boolean
 
+# --- Global Error Handling Setup for Text Encoding ---
+def handle_text_encoding(text, default_text="[無法處理的文字]"):
+    """安全處理任何文字，確保不會因編碼問題而崩潰程序"""
+    if text is None:
+        return default_text
+    
+    try:
+        # 嘗試使用 utf-8 編碼
+        return text
+    except UnicodeEncodeError:
+        try:
+            # 嘗試將特殊字符替換為可顯示字符
+            return text.encode('utf-8', errors='replace').decode('utf-8')
+        except:
+            # 最後手段：忽略任何無法處理的字符
+            try:
+                return text.encode('utf-8', errors='ignore').decode('utf-8')
+            except:
+                return default_text
+
 # --- Color Config Loading ---
 def load_bubble_colors(config_path='bubble_colors.json'):
     """Loads bubble color configuration from a JSON file."""
@@ -1068,7 +1090,13 @@ class InteractionModule:
 
         if copied and copied_text and copied_text != "___MCP_CLEAR___":
             print(f"Successfully copied text, length: {len(copied_text)}")
-            return copied_text.strip()
+            # 添加編碼安全處理
+            try:
+                safe_text = handle_text_encoding(copied_text.strip())
+                return safe_text
+            except Exception as e:
+                print(f"Error handling copied text encoding: {str(e)}")
+                return copied_text.strip()  # 即使有問題也嘗試返回原始文字
         else:
             print("Error: Copy operation unsuccessful or clipboard content invalid.")
             return None
@@ -2115,17 +2143,31 @@ def run_ui_monitoring_loop(trigger_queue: queue.Queue, command_queue: queue.Queu
 
                     # 7. Send Trigger Info to Main Thread
                     print("\n>>> Putting trigger info in Queue <<<")
-                    print(f"   Sender: {sender_name}")
+                    try:
-                    print(f"   Content: {bubble_text[:100]}...")
+                        # 安全地處理和顯示發送者名稱
+                        safe_sender_display = handle_text_encoding(sender_name, "[未知發送者]")
+                        print(f"   Sender: {safe_sender_display}")
+                        
+                        # 安全地處理和顯示消息內容
+                        if bubble_text:
+                            display_text = bubble_text[:100] + "..." if len(bubble_text) > 100 else bubble_text
+                            safe_content_display = handle_text_encoding(display_text, "[無法處理的文字內容]")
+                            print(f"   Content: {safe_content_display}")
+                        else:
+                            print("   Content: [空]")
+                    except Exception as e_display:
+                        print(f"Error displaying message info: {str(e_display)}")
+                    
                     print(f"   Bubble Region: {bubble_region}") # Original region for context
                     print(f"   Reply Context Activated: {reply_context_activated}")
                     try:
+                        # 確保所有文字數據都經過安全處理
                         data_to_send = {
-                            'sender': sender_name,
+                            'sender': handle_text_encoding(sender_name, "[未知發送者]"),
-                            'text': bubble_text,
+                            'text': handle_text_encoding(bubble_text, "[無法處理的文字內容]"),
-                            'bubble_region': bubble_region, # Send original region for context if needed
+                            'bubble_region': bubble_region,
                             'reply_context_activated': reply_context_activated,
-                            'bubble_snapshot': bubble_snapshot, # Send the snapshot used
+                            'bubble_snapshot': bubble_snapshot,
                             'search_area': search_area
                         }
                         trigger_queue.put(data_to_send)
@@ -2136,13 +2178,26 @@ def run_ui_monitoring_loop(trigger_queue: queue.Queue, command_queue: queue.Queu
                         break # Exit the 'for target_bubble_info in sorted_bubbles' loop
 
                     except Exception as q_err:
-                        print(f"Error putting data in Queue: {q_err}")
+                        print(f"Error preparing or enqueueing data: {q_err}")
-                        # Don't break if queue put fails, maybe try next bubble? Or log and break?
+                        # 嘗試使用最小數據集合保證功能性
+                        try:
+                            minimal_data = {
+                                'sender': "[數據處理錯誤]",
+                                'text': handle_text_encoding(bubble_text[:100] if bubble_text else "[內容獲取失敗]"), # Apply encoding here too
+                                'bubble_region': bubble_region,
+                                'reply_context_activated': False, # Sensible default
+                                'bubble_snapshot': bubble_snapshot, # Keep snapshot if available
+                                'search_area': search_area
+                            }
+                            trigger_queue.put(minimal_data)
+                            print("Minimal fallback data placed in Queue after error.")
+                        except Exception as min_q_err:
+                            print(f"Critical failure: Could not place any data in queue: {min_q_err}")
                         # Let's break here too, as something is wrong.
                         print("Breaking scan cycle due to queue error.")
                         break
 
-                # End of keyword found block (if keyword_coords:)
+                # End of keyword found block (if result:)
             # End of loop through sorted bubbles (for target_bubble_info...)
 
             # If the loop finished without breaking (i.e., no trigger processed), wait the full interval.