Upload app.py

app.py

@@ -1384,27 +1384,25 @@ def _chat_fragment():
             )
 
             with st.chat_message("assistant"):
-                container = st.empty()
-                meta_placeholder = st.empty()  # 用empty占位，避免提前渲染
-
                 # 联网模式：大模型 + 网络搜索，不查知识库
                 # 知识库模式：RAG 检索本地知识库
                 if _web_on:
-                    container.markdown("*🌐 正在联网搜索...*")
-                    relevant_docs = []
+                    with st.status("🌐 正在联网搜索...", expanded=False):
+                        relevant_docs = []
                 else:
-                    container.markdown("*🔍 正在搜索知识库...*")
-                    relevant_docs = search_local(q, _top_k, _threshold)
+                    with st.status("🔍 正在搜索知识库...", expanded=False):
+                        relevant_docs = search_local(q, _top_k, _threshold)
 
-                container.markdown("*🤔 正在组织语言...*")
+                response_container = st.empty()
+                response_container.markdown("*🤔 正在组织语言...*")
 
                 try:
-                    full_response = container.write_stream(
+                    full_response = response_container.write_stream(
                         llm_answer(q, relevant_docs, _model_name, _web_on)
                     )
                     meta_info = st.session_state.get("last_meta", "")
                     if meta_info:
-                        meta_placeholder.caption(meta_info)  # 流式完成后才显示meta
+                        st.caption(meta_info)
                     st.session_state.messages.append(
                         {"role": "assistant", "content": full_response, "meta": meta_info}
                     )