Update src/streamlit_app.py

src/streamlit_app.py

@@ -115,8 +115,8 @@ if "messages" not in st.session_state:
     st.session_state.messages = []
 
 # === 修改：檔案處理邏輯 (轉為 FAISS) ===
-def process_file_to_faiss(uploaded_file, chunk_size, chunk_overlap=50):
-    """讀取檔案 -> 切塊 -> 建立 FAISS 索引"""
+def process_file_to_faiss(uploaded_file, chunk_overlap=0):
+    """讀取檔案 -> 以 </Event> 分割 -> 建立 FAISS 索引"""
     text_content = ""
     try:
         # 1. 讀取文字
@@ -134,17 +134,13 @@ def process_file_to_faiss(uploaded_file, chunk_size, chunk_overlap=50):
         if not text_content.strip():
             return None, "File is empty"
 
-        # 2. 切分文本 (Chunking)
-        text_splitter = RecursiveCharacterTextSplitter(
-            chunk_size=chunk_size,
-            chunk_overlap=chunk_overlap,
-            separators=["\n\n", "\n", " ", ""]
-        )
-        # 將純文字包裝成 LangChain Document 物件
-        docs = [Document(page_content=x) for x in text_splitter.split_text(text_content)]
+        # 2. 以 </Event> 做分割
+        events = [e + "</Event>" for e in text_content.split("</Event>") if e.strip()]
+        
+        # 3. 封裝成 Document
+        docs = [Document(page_content=e) for e in events]
         
-        # 3. 建立/更新 FAISS
-        # 如果 session 中已有 vector_store，可以選擇合併，這裡示範每次上傳新檔就重建一個新的
+        # 4. 建立 FAISS
         vector_store = FAISS.from_documents(docs, embedding_model)
         
         return vector_store, f"Success: {len(docs)} chunks created."