Upload app.py

app.py

@@ -37,7 +37,11 @@ session = create_session_with_retry()
 # 初始化 Hugging Face Inference Client
 # 使用环境变量或者免费的公开模型
 HF_TOKEN = os.getenv("HF_TOKEN", None)  # 可选：如果需要访问私有模型
-client = InferenceClient(token=HF_TOKEN)
+try:
+    client = InferenceClient(token=HF_TOKEN)
+except Exception as e:
+    print(f"Warning: Failed to initialize Hugging Face client: {e}")
+    client = None
 
 # 定义可用的 MCP 工具
 MCP_TOOLS = [
@@ -279,14 +283,9 @@ def query_financial_data(company_name, query_type):
         # 检查是否有错误
         if "error" in search_result:
             return f"❌ Server Error: {search_result.get('error')}\n\nResponse: {search_result.get('detail', 'N/A')}\n\nURL: {search_result.get('url', MCP_URL)}"
-            return f"❌ Server Error: HTTP {search_resp.status_code}\n\nResponse: {search_resp.text[:500]}"
         
-        try:
-            result = search_resp.json()
-            # 使用统一的 MCP 响应解析函数
-            company = parse_mcp_response(result)
-        except (ValueError, KeyError, json.JSONDecodeError) as e:
-            return f"❌ JSON Parse Error: {str(e)}\n\nResponse: {search_resp.text[:500]}"
+        # 解析搜索结果
+        company = parse_mcp_response(search_result)
         
         if isinstance(company, dict) and company.get("error"):
             return f"❌ Error: {company['error']}"
@@ -665,6 +664,10 @@ Always be helpful, accurate, and cite the data sources when providing financial
             
             # 使用支持工具调用的模型（如 Qwen, Llama 等）
             try:
+                # 检查 client 是否可用
+                if client is None:
+                    return fallback_chatbot_response(message)
+                
                 response = client.chat_completion(
                     messages=messages,
                     model="Qwen/Qwen2.5-72B-Instruct",  # 支持工具调用的模型