Upload app.py

app.py

@@ -213,36 +213,45 @@ def chatbot_response(message, history):
                 
                 for tool_call in choice.message.tool_calls:
                     tool_name = tool_call.function.name
-                    tool_args = json.loads(tool_call.function.arguments)
+                    try:
+                        tool_args = json.loads(tool_call.function.arguments)
+                    except json.JSONDecodeError:
+                        tool_args = {}
                     
                     # 调用 MCP 工具
                     tool_result = call_mcp_tool(tool_name, tool_args)
                     
-                    # 限制返回结果大小
-                    result_str = json.dumps(tool_result, ensure_ascii=False)
-                    
-                    if len(result_str) > MAX_TOOL_RESULT_CHARS:
-                        if isinstance(tool_result, dict) and "text" in tool_result:
-                            truncated_text = truncate_text(tool_result["text"], MAX_TOOL_RESULT_CHARS - 50)
-                            tool_result_truncated = {"text": truncated_text, "_truncated": True}
-                        elif isinstance(tool_result, dict):
-                            truncated = {}
-                            char_count = 0
-                            for k, v in list(tool_result.items())[:8]:  # 保留前8个字段
-                                v_str = str(v)[:300]  # 每个值最多300字符
-                                truncated[k] = v_str
-                                char_count += len(k) + len(v_str)
-                                if char_count > MAX_TOOL_RESULT_CHARS:
-                                    break
-                            tool_result_truncated = {**truncated, "_truncated": True}
-                        else:
-                            tool_result_truncated = {"preview": truncate_text(result_str, MAX_TOOL_RESULT_CHARS), "_truncated": True}
-                        result_for_llm = json.dumps(tool_result_truncated, ensure_ascii=False)
+                    # 检查错误
+                    if isinstance(tool_result, dict) and "error" in tool_result:
+                        # 工具调用失败，记录错误
+                        tool_calls_log.append({"name": tool_name, "arguments": tool_args, "result": tool_result, "error": True})
+                        result_for_llm = json.dumps({"error": tool_result.get("error", "Unknown error")}, ensure_ascii=False)
                     else:
-                        result_for_llm = result_str
-                    
-                    # 记录工具调用(包含完整结果用于UI显示)
-                    tool_calls_log.append({"name": tool_name, "arguments": tool_args, "result": tool_result})
+                        # 限制返回结果大小
+                        result_str = json.dumps(tool_result, ensure_ascii=False)
+                        
+                        if len(result_str) > MAX_TOOL_RESULT_CHARS:
+                            if isinstance(tool_result, dict) and "text" in tool_result:
+                                truncated_text = truncate_text(tool_result["text"], MAX_TOOL_RESULT_CHARS - 50)
+                                tool_result_truncated = {"text": truncated_text, "_truncated": True}
+                            elif isinstance(tool_result, dict):
+                                truncated = {}
+                                char_count = 0
+                                for k, v in list(tool_result.items())[:8]:  # 保留前8个字段
+                                    v_str = str(v)[:300]  # 每个值最多300字符
+                                    truncated[k] = v_str
+                                    char_count += len(k) + len(v_str)
+                                    if char_count > MAX_TOOL_RESULT_CHARS:
+                                        break
+                                tool_result_truncated = {**truncated, "_truncated": True}
+                            else:
+                                tool_result_truncated = {"preview": truncate_text(result_str, MAX_TOOL_RESULT_CHARS), "_truncated": True}
+                            result_for_llm = json.dumps(tool_result_truncated, ensure_ascii=False)
+                        else:
+                            result_for_llm = result_str
+                        
+                        # 记录成功的工具调用
+                        tool_calls_log.append({"name": tool_name, "arguments": tool_args, "result": tool_result})
                     
                     messages.append({
                         "role": "tool",
@@ -269,13 +278,21 @@ def chatbot_response(message, history):
 """.format(len(tool_calls_log))
             
             for idx, tool_call in enumerate(tool_calls_log):
+                # 预先计算 JSON 字符串，避免重复调用
+                args_json = json.dumps(tool_call['arguments'], ensure_ascii=False)
+                result_json = json.dumps(tool_call.get('result', {}), ensure_ascii=False, indent=2)
+                result_preview = result_json[:1500] + ('...' if len(result_json) > 1500 else '')
+                
+                # 显示错误状态
+                error_indicator = " ❌ Error" if tool_call.get('error') else ""
+                
                 # 使用原生 HTML5 details/summary 标签（不需要 JavaScript）
                 response_prefix += f"""<details style='margin: 8px 0; border: 1px solid #ddd; border-radius: 6px; overflow: hidden;'>
   <summary style='background: #fff; padding: 10px; cursor: pointer; user-select: none; list-style: none;'>
     <div style='display: flex; justify-content: space-between; align-items: center;'>
       <div style='flex: 1;'>
-        <strong style='color: #2c5aa0;'>📌 {idx+1}. {tool_call['name']}</strong>
-        <div style='font-size: 0.85em; color: #666; margin-top: 4px;'>📥 Input: <code style='background: #f5f5f5; padding: 2px 6px; border-radius: 3px;'>{json.dumps(tool_call['arguments'], ensure_ascii=False)}</code></div>
+        <strong style='color: #2c5aa0;'>📌 {idx+1}. {tool_call['name']}{error_indicator}</strong>
+        <div style='font-size: 0.85em; color: #666; margin-top: 4px;'>📥 Input: <code style='background: #f5f5f5; padding: 2px 6px; border-radius: 3px;'>{args_json}</code></div>
       </div>
       <span style='font-size: 1.2em; color: #999; margin-left: 10px;'>▶</span>
     </div>
@@ -283,7 +300,7 @@ def chatbot_response(message, history):
   <div style='background: #f9f9f9; padding: 12px; border-top: 1px solid #eee;'>
     <div style='font-size: 0.9em; color: #333;'>
       <strong>📤 Output:</strong>
-      <pre style='background: #fff; padding: 10px; border-radius: 4px; overflow-x: auto; margin-top: 6px; font-size: 0.85em; border: 1px solid #e0e0e0; max-height: 400px; white-space: pre-wrap;'>{json.dumps(tool_call.get('result', {}), ensure_ascii=False, indent=2)[:1500]}{'...' if len(json.dumps(tool_call.get('result', {}), ensure_ascii=False)) > 1500 else ''}</pre>
+      <pre style='background: #fff; padding: 10px; border-radius: 4px; overflow-x: auto; margin-top: 6px; font-size: 0.85em; border: 1px solid #e0e0e0; max-height: 400px; white-space: pre-wrap;'>{result_preview}</pre>
     </div>
   </div>
 </details>
@@ -299,26 +316,38 @@ def chatbot_response(message, history):
         # 流式输出最终答案
         yield response_prefix
         
-        # 如果已经有最终答案，直接流式输出
+        # 如果已经有最终答案，直接输出
         if final_response_content:
             # 已经从循环中获得了最终答案，直接输出
             yield response_prefix + final_response_content
         else:
             # 如果循环结束但没有最终答案（达到最大迭代次数），需要再调用一次让模型总结
-            stream = client.chat.completions.create(
-                model="Qwen/Qwen3-32B:groq",
-                messages=messages,
-                tools=None,  # 不再允许调用工具
-                max_tokens=MAX_OUTPUT_TOKENS,
-                temperature=0.5,
-                stream=True
-            )
-            
-            accumulated_text = ""
-            for chunk in stream:
-                if chunk.choices and len(chunk.choices) > 0 and chunk.choices[0].delta.content:
-                    accumulated_text += chunk.choices[0].delta.content
-                    yield response_prefix + accumulated_text
+            try:
+                stream = client.chat.completions.create(
+                    model="Qwen/Qwen3-32B:groq",
+                    messages=messages,
+                    tools=None,  # 不再允许调用工具
+                    max_tokens=MAX_OUTPUT_TOKENS,
+                    temperature=0.5,
+                    stream=True
+                )
+                
+                accumulated_text = ""
+                for chunk in stream:
+                    if chunk.choices and len(chunk.choices) > 0 and chunk.choices[0].delta.content:
+                        accumulated_text += chunk.choices[0].delta.content
+                        yield response_prefix + accumulated_text
+            except Exception as stream_error:
+                # 流式输出失败，尝试非流式
+                final_resp = client.chat.completions.create(
+                    model="Qwen/Qwen3-32B:groq",
+                    messages=messages,
+                    tools=None,
+                    max_tokens=MAX_OUTPUT_TOKENS,
+                    temperature=0.5,
+                    stream=False
+                )
+                yield response_prefix + final_resp.choices[0].message.content
         
     except Exception as e:
         import traceback
@@ -347,7 +376,6 @@ with gr.Blocks(title="Financial AI Assistant") as demo:
 # 启动应用
 if __name__ == "__main__":
     import sys
-    import asyncio
     
     # 修复 asyncio 事件循环问题
     if sys.platform == 'linux':