Upload 2 files

README.md

@@ -16,7 +16,7 @@ A Gradio-based AI-powered web application for querying SEC financial data throug
 
 ## ✨ Features
 
-- 🤖 **Intelligent AI Assistant**: Chat naturally with Qwen/Qwen2.5-72B-Instruct model
+- 🤖 **Intelligent AI Assistant**: Chat naturally with **Qwen/Qwen2.5-VL-72B-Instruct:nebius** model
 - 🛠️ **Automatic Tool Calling**: AI automatically selects and calls MCP tools based on your questions
 - 🔍 Search companies by name or ticker symbol
 - 📈 View latest financial data
@@ -26,7 +26,24 @@ A Gradio-based AI-powered web application for querying SEC financial data throug
 
 ## 🚀 Quick Start
 
-No configuration needed! Just start asking questions:
+### Prerequisites
+
+- Python 3.9+
+- Hugging Face API token (get one at https://huggingface.co/settings/tokens)
+
+### Installation
+
+1. Clone the repository
+2. Install dependencies: `pip install -r requirements.txt`
+3. Create a `.env` file with your HF token:
+   ```
+   HF_TOKEN=your_token_here
+   ```
+4. Run: `python app.py`
+
+### Usage
+
+No additional configuration needed! Just start asking questions:
 
 ### AI Assistant Tab
 
@@ -79,7 +96,7 @@ SEC EDGAR data via MCP Server: https://huggingface.co/spaces/JC321/EasyReportDat
 
 - **Frontend**: Gradio 6.0.1
 - **Backend**: Python with requests
-- **AI Model**: Qwen/Qwen2.5-72B-Instruct (via Hugging Face Inference API)
+- **AI Model**: Qwen/Qwen2.5-VL-72B-Instruct:nebius (via Hugging Face Inference API)
 - **MCP Protocol**: FastMCP with HTTP transport (stateless)
 - **Data Source**: SEC EDGAR via MCP Server
 

app.py

@@ -35,10 +35,18 @@ def create_session_with_retry():
 session = create_session_with_retry()
 
 # 初始化 Hugging Face Inference Client
-# 直接使用免费的公开 API，无需 token
+# 使用 Qwen/Qwen2.5-VL-72B-Instruct:nebius 模型（更强大的视觉语言模型）
 try:
-    client = InferenceClient()
-    print("✅ Hugging Face client initialized (free tier)")
+    # 从 Hugging Face Space 环境变量读取 token
+    # 在 HF Space 设置中配置 HF_TOKEN secret
+    hf_token = os.environ.get("HF_TOKEN")
+    if hf_token:
+        client = InferenceClient(api_key=hf_token)
+        print(f"✅ Hugging Face client initialized with Qwen/Qwen2.5-VL-72B-Instruct:nebius model")
+    else:
+        print("⚠️ HF_TOKEN not found in environment variables")
+        print("ℹ️ Please add HF_TOKEN in your Hugging Face Space Settings > Repository secrets")
+        client = None
 except Exception as e:
     print(f"❌ Warning: Failed to initialize Hugging Face client: {e}")
     client = None
@@ -617,36 +625,21 @@ def chatbot_response(message, history):
         # 构建对话历史
         messages = []
         
-        # 系统提示词 - 财报分析专家
-        system_prompt = """You are a financial reporting analysis expert, specializing in analyzing company financial data and providing professional insights.
+        # 系统提示词 - 自由智能的财报分析助手
+        system_prompt = """You are an intelligent financial analysis assistant with access to real-time SEC EDGAR data.
 
-Your expertise:
-- Deep understanding of financial statements (10-K, 10-Q filings)
-- Expert at identifying trends, patterns, and anomalies in financial data
-- Skilled at comparing metrics across different periods and competitors
-- Proficient in calculating and interpreting key financial ratios
+You have 3 powerful tools to fetch financial data:
+1. advanced_search_company(company_input) - Search for any US company
+2. get_latest_financial_data(cik) - Get latest financial report
+3. extract_financial_metrics(cik, years) - Get multi-year trends (3 or 5 years)
 
-Your available tools:
-1. **advanced_search_company(company_input)** - Find US companies by name or ticker symbol
-2. **get_latest_financial_data(cik)** - Retrieve the most recent financial report
-3. **extract_financial_metrics(cik, years)** - Get multi-year financial trends (3 or 5 years)
+When users ask about a company:
+- Automatically use tools to fetch the data they need
+- Analyze the numbers and provide insights
+- Explain trends, growth rates, and what they mean
+- Be conversational and helpful
 
-How to analyze:
-- When users ask about a company, automatically fetch relevant data
-- Provide professional analysis with specific numbers and percentages
-- Calculate YoY growth rates, profit margins, and other key metrics
-- Identify trends and explain their business implications
-- Compare current performance with historical data
-- Highlight both strengths and concerns in the financial data
-
-Analysis framework:
-- Revenue analysis: Growth rate, trend direction, stability
-- Profitability: Net income, profit margins, efficiency
-- Cash flow: Operating cash flow adequacy, cash generation ability
-- Per-share metrics: EPS trends, shareholder value creation
-- Overall assessment: Financial health and future outlook
-
-Remember: Provide data-driven insights, not just raw numbers. Help users understand what the numbers really mean for the business."""
+You can handle any financial question - from simple data queries to complex multi-company comparisons. Be creative and thorough in your analysis."""
         
         messages.append({"role": "system", "content": system_prompt})
         
@@ -685,10 +678,11 @@ Remember: Provide data-driven insights, not just raw numbers. Help users underst
                 
                 response = client.chat_completion(
                     messages=messages,
-                    model="Qwen/Qwen2.5-72B-Instruct",  # 支持工具调用的模型
+                    model="Qwen/Qwen2.5-VL-72B-Instruct:nebius",  # 使用更强大的视觉语言模型
                     tools=MCP_TOOLS,
-                    max_tokens=2500,  # 增加token数，支持更详细的分析
-                    temperature=0.8  # 较高的temperature使回答更自然、更有创意
+                    max_tokens=3000,
+                    temperature=0.7,
+                    tool_choice="auto"  # 让模型自动决定是否使用工具
                 )
                 
                 print(f"✅ LLM response received (iteration {iteration})")
@@ -735,23 +729,38 @@ Remember: Provide data-driven insights, not just raw numbers. Help users underst
                     break
                     
             except ValueError as ve:
-                # API Key 相关错误（虽然我们不使用 token，但仍然可能出现）
                 error_msg = str(ve)
+                print(f"⚠️ LLM API error: {ve}")
                 if "api_key" in error_msg.lower() or "token" in error_msg.lower():
-                    print(f"⚠️ LLM API rate limit or authentication issue: {ve}")
                     print("ℹ️ Falling back to simple response logic")
                     return fallback_chatbot_response(message)
                 else:
                     raise
             except Exception as e:
-                # 其他 LLM API 错误（如速率限制）
-                print(f"⚠️ LLM API error (possibly rate limited): {e}")
+                print(f"⚠️ LLM API error: {type(e).__name__} - {e}")
+                # 如果是工具调用相关错误，尝试不使用工具重新请求
+                if "tool" in str(e).lower() or "function" in str(e).lower():
+                    print("🔄 Tool calling failed, retrying without tools...")
+                    try:
+                        response = client.chat_completion(
+                            messages=messages,
+                            model="Qwen/Qwen2.5-VL-72B-Instruct:nebius",
+                            max_tokens=3000,
+                            temperature=0.7
+                        )
+                        response_text = response.choices[0].message.content
+                        return response_text
+                    except:
+                        pass
                 print("ℹ️ Falling back to simple response logic")
                 return fallback_chatbot_response(message)
         
         # 构建最终响应
         final_response = ""
         
+        # 显示调试信息：使用的模型
+        final_response += f"<div style='padding: 8px; background: #e3f2fd; border-left: 3px solid #2196f3; margin-bottom: 10px; font-size: 0.9em;'>🤖 <strong>Model:</strong> Qwen/Qwen2.5-VL-72B-Instruct:nebius | <strong>Iterations:</strong> {iteration}</div>\n\n"
+        
         # 如果有工具调用，显示调用日志
         if tool_calls_log:
             final_response += "**🛠️ MCP Tools Used:**\n\n"
@@ -769,6 +778,9 @@ Remember: Provide data-driven insights, not just raw numbers. Help users underst
 
 def fallback_chatbot_response(message):
     """退回策略：当 LLM API 不可用时使用的简单逻辑"""
+    # 显示警告信息
+    fallback_warning = "<div style='padding: 10px; background: #fff3cd; border-left: 4px solid #ffc107; margin-bottom: 15px;'>⚠️ <strong>Notice:</strong> LLM API unavailable, using fallback logic.</div>\n\n"
+    
     # 检查是否是财务查询相关问题
     financial_keywords = ['financial', 'revenue', 'income', 'earnings', 'cash flow', 'expenses', '财务', '收入', '利润', 'data', 'trend', 'performance']
     
@@ -800,12 +812,12 @@ def fallback_chatbot_response(message):
             
             # 调用财务查询函数
             result = query_financial_data(detected_company, query_type)
-            return f"Here's the financial information for {detected_company}:\n\n{result}"
+            return fallback_warning + f"Here's the financial information for {detected_company}:\n\n{result}"
         else:
-            return "I can help you query financial data! Please specify a company name. For example: 'Show me Apple's latest financial data' or 'What's NVIDIA's 3-year trend?' \n\nSupported companies include: Apple, Microsoft, NVIDIA, Tesla, Alibaba, Google, Amazon, and more."
+            return fallback_warning + "I can help you query financial data! Please specify a company name. For example: 'Show me Apple's latest financial data' or 'What's NVIDIA's 3-year trend?' \n\nSupported companies include: Apple, Microsoft, NVIDIA, Tesla, Alibaba, Google, Amazon, and more."
     
     # 如果不是财务查询，返回通用回复
-    return "Hello! I'm a financial data assistant powered by SEC EDGAR data. I can help you query financial information for US listed companies.\n\n**What I can do:**\n- Get latest financial data (revenue, income, EPS, etc.)\n- Show 3-year or 5-year financial trends\n- Provide detailed financial metrics\n\n**Try asking:**\n- 'Show me Apple's latest financial data'\n- 'What's NVIDIA's 3-year financial trend?'\n- 'How is Microsoft performing financially?'"
+    return fallback_warning + "Hello! I'm a financial data assistant powered by SEC EDGAR data. I can help you query financial information for US listed companies.\n\n**What I can do:**\n- Get latest financial data (revenue, income, EPS, etc.)\n- Show 3-year or 5-year financial trends\n- Provide detailed financial metrics\n\n**Try asking:**\n- 'Show me Apple's latest financial data'\n- 'What's NVIDIA's 3-year financial trend?'\n- 'How is Microsoft performing financially?'"
 
 # 包装函数，显示加载状态
 def query_with_status(company, query_type):
@@ -831,7 +843,7 @@ with gr.Blocks(title="SEC Financial Data Query Assistant") as demo:
     # 显示 AI 功能说明
     gr.Markdown("""
     <div style='padding: 15px; background: #d4edda; border-left: 4px solid #28a745; margin: 10px 0; border-radius: 4px;'>
-        <strong>✅ AI Assistant Enabled:</strong> Powered by Qwen/Qwen2.5-72B-Instruct model with automatic MCP tool calling.
+        <strong>✅ AI Assistant Enabled:</strong> Powered by Qwen/Qwen2.5-VL-72B-Instruct:nebius model with automatic MCP tool calling.
         <br>
         <strong>💬 Ask me anything:</strong> I can understand natural language and automatically fetch financial data when needed!
     </div>