本地RAG尝试

frontend/components/chat_area.py

@@ -119,6 +119,7 @@ def _handle_ai_response():
         api_thought = ""
         display_text = ""
         display_thought = ""
+        rag_sources = None  # 存储 RAG 检索来源信息
         
         # 调用流式 API
         stream = api_client.chat_stream(
@@ -213,6 +214,25 @@ def _handle_ai_response():
                     last_msg = messages_update[-1] if messages_update else {}
                     if isinstance(last_msg, dict) and last_msg.get("role") == "tool":
                         tool_name = last_msg.get("name", "unknown")
+                        tool_content = last_msg.get("content", "")
+                        # 存储 RAG 检索结果
+                        if tool_name == "search_knowledge_base":
+                            # 尝试解析 tool_content，它可能是 JSON 字符串
+                            sources = []
+                            try:
+                                if isinstance(tool_content, str):
+                                    import json
+                                    data = json.loads(tool_content)
+                                else:
+                                    data = tool_content
+                                # 提取来源列表
+                                if isinstance(data, dict) and "sources" in data:
+                                    sources = data["sources"]
+                                else:
+                                    sources = [str(data)]
+                            except Exception:
+                                sources = [str(tool_content)]
+                            rag_sources = sources
                         tool_status_placeholder.success(f"✅ 工具 {tool_name} 执行完成")
                         # 短暂显示后清除，保持界面清爽
                         import time
@@ -270,6 +290,31 @@ def _handle_ai_response():
         # 移除光标
         message_placeholder.markdown(display_text)
         
+        # 显示 RAG 检索来源（如果有）
+        if rag_sources:
+            with st.expander("🔍 检索来源", expanded=False):
+                # 格式化来源列表
+                if isinstance(rag_sources, list):
+                    for i, source in enumerate(rag_sources, 1):
+                        if isinstance(source, dict):
+                            content = source.get("page_content", source.get("content", str(source)))
+                            metadata = source.get("metadata", {})
+                            filename = metadata.get("filename", metadata.get("source", "未知文件"))
+                            page = metadata.get("page", metadata.get("page_number", ""))
+                            if page:
+                                source_info = f"**来源 {i}:** {filename} (第{page}页)"
+                            else:
+                                source_info = f"**来源 {i}:** {filename}"
+                            st.markdown(source_info)
+                            # 显示内容预览（前200字符）
+                            preview = content[:200] + "..." if len(content) > 200 else content
+                            st.markdown(f"> {preview}")
+                            st.markdown("---")
+                        else:
+                            st.markdown(f"**来源 {i}:** {str(source)}")
+                else:
+                    st.markdown(str(rag_sources))
+        
         # 拼装包含思考过程的完整内容，以便后续在历史中正确渲染
         final_content = display_text
         if display_thought: