Create app.py

app.py

@@ -0,0 +1,212 @@
+import os  # 导入操作系统模块
+import gradio as gr  # 导入Gradio库用于创建Web界面
+import requests  # 导入请求库用于API调用
+import inspect  # 导入检查模块用于函数检查
+import pandas as pd  # 导入pandas用于数据处理
+
+from smolagents import CodeAgent, DuckDuckGoSearchTool, OpenAIServerModel, WikipediaSearchTool  # 从smolagents导入所需组件
+
+# (Keep Constants as is)  # 保持常量不变
+# --- Constants ---  # 常量部分
+DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space" 
+deepseek_api_key = os.environ.get("DEEPSEEK_API_KEY")  # 从环境变量中获取 DeepSeek API Key
+deepseek_api_base_url = "https://api.deepseek.com"
+# --- Basic Agent Definition ---  
+# ----- THIS IS WERE YOU CAN BUILD WHAT YOU WANT ------  
+class BasicAgent: 
+    def __init__(self):  
+        # model = OpenAIServerModel(model_id="gpt-4o") 
+        if not deepseek_api_key:
+            raise ValueError("DeepSeek API Key not found in environment variables.") 
+        model = OpenAIServerModel(
+            model_id="deepseek-chat", 
+            api_base=deepseek_api_base_url,
+            api_headers={"Authorization": f"Bearer {deepseek_api_key}", "Content-Type": "application/json"}
+        )
+        search_tool = DuckDuckGoSearchTool()  
+        wiki_search = WikipediaSearchTool()  
+
+        self.agent = CodeAgent(  # 初始化代码代理
+            model = model,  # 设置模型
+            tools=[  # 设置可用工具列表
+                search_tool,  # 搜索工具
+                wiki_search  # 维基百科工具
+            ]
+        )
+
+    def __call__(self, question: str) -> str:  # 定义调用方法，使对象可直接调用
+        return self.agent.run(question) # type: ignore  # 运行代理并返回结果
+
+def run_and_submit_all( profile: gr.OAuthProfile | None):  # 运行并提交所有回答的函数
+    """
+    Fetches all questions, runs the BasicAgent on them, submits all answers,
+    and displays the results.
+    """ 
+    # --- Determine HF Space Runtime URL and Repo URL ---  
+    space_id = os.getenv("SPACE_ID") # Get the SPACE_ID for sending link to the code  # 获取SPACE_ID用于发送代码链接
+
+    if profile:  # 如果用户已登录
+        username= f"{profile.username}"  # 获取用户名
+        print(f"User logged in: {username}")  # 打印已登录的用户名
+    else:  # 如果用户未登录
+        print("User not logged in.")  # 打印用户未登录
+        return "Please Login to Hugging Face with the button.", None  # 返回请求登录的消息
+
+    api_url = DEFAULT_API_URL  # 设置API URL
+    questions_url = f"{api_url}/questions"  # 构建获取问题的URL
+    submit_url = f"{api_url}/submit"  # 构建提交答案的URL
+
+    # 1. Instantiate Agent ( modify this part to create your agent)  # 1. 实例化代理（修改此部分来创建你的代理）
+    try:  # 尝试
+        agent = BasicAgent()  # 创建基本代理实例
+    except Exception as e:  # 捕获异常
+        print(f"Error instantiating agent: {e}")  # 打印代理实例化错误
+        return f"Error initializing agent: {e}", None  # 返回初始化错误信息
+    # In the case of an app running as a hugging Face space, this link points toward your codebase ( usefull for others so please keep it public)  # 当应用程序作为Hugging Face空间运行时，此链接指向你的代码库（对其他人有用，请保持公开）
+    agent_code = f"https://huggingface.co/spaces/{space_id}/tree/main"  # 构建代码库URL
+    print(agent_code)  # 打印代码库URL
+
+    # 2. Fetch Questions  # 2. 获取问题
+    print(f"Fetching questions from: {questions_url}")  # 打印正在从哪个URL获取问题
+    try:  # 尝试
+        response = requests.get(questions_url, timeout=15)  # 发送GET请求获取问题，超时设置为15秒
+        response.raise_for_status()  # 检查HTTP错误
+        questions_data = response.json()  # 解析JSON响应
+        if not questions_data:  # 如果问题列表为空
+             print("Fetched questions list is empty.")  # 打印问题列表为空
+             return "Fetched questions list is empty or invalid format.", None  # 返回问题列表为空或格式无效的消息
+        print(f"Fetched {len(questions_data)} questions.")  # 打印获取到的问题数量
+    except requests.exceptions.RequestException as e:  # 捕获请求异常
+        print(f"Error fetching questions: {e}")  # 打印获取问题错误
+        return f"Error fetching questions: {e}", None  # 返回获取问题错误
+    except requests.exceptions.JSONDecodeError as e:  # 捕获JSON解码错误
+         print(f"Error decoding JSON response from questions endpoint: {e}")  # 打印JSON解码错误
+         print(f"Response text: {response.text[:500]}")  # 打印响应文本（前500个字符）
+         return f"Error decoding server response for questions: {e}", None  # 返回服务器响应解码错误
+    except Exception as e:  # 捕获其他异常
+        print(f"An unexpected error occurred fetching questions: {e}")  # 打印获取问题时发生意外错误
+        return f"An unexpected error occurred fetching questions: {e}", None  # 返回获取问题时发生意外错误
+
+    # 3. Run your Agent  # 3. 运行你的代理
+    results_log = []  # 初始化结果日志列表
+    answers_payload = []  # 初始化答案载荷列表
+    print(f"Running agent on {len(questions_data)} questions...")  # 打印正在多少个问题上运行代理
+    for item in questions_data:  # 遍历每个问题数据项
+        task_id = item.get("task_id")  # 获取任务ID
+        question_text = item.get("question")  # 获取问题文本
+        if not task_id or question_text is None:  # 如果任务ID或问题文本缺失
+            print(f"Skipping item with missing task_id or question: {item}")  # 打印跳过缺少任务ID或问题的项
+            continue  # 继续下一个循环
+        try:  # 尝试
+            submitted_answer = agent(question_text)  # 使用代理回答问题
+            answers_payload.append({"task_id": task_id, "submitted_answer": submitted_answer})  # 将答案添加到载荷中
+            results_log.append({"Task ID": task_id, "Question": question_text, "Submitted Answer": submitted_answer})  # 将结果添加到日志中
+        except Exception as e:  # 捕获异常
+             print(f"Error running agent on task {task_id}: {e}")  # 打印在任务上运行代理时出错
+             results_log.append({"Task ID": task_id, "Question": question_text, "Submitted Answer": f"AGENT ERROR: {e}"})  # 将错误添加到结果日志中
+
+    if not answers_payload:  # 如果答案载荷为空
+        print("Agent did not produce any answers to submit.")  # 打印代理没有产生任何答案可提交
+        return "Agent did not produce any answers to submit.", pd.DataFrame(results_log)  # 返回代理没有产生任何答案可提交
+
+    # 4. Prepare Submission  # 4. 准备提交
+    submission_data = {"username": username.strip(), "agent_code": agent_code, "answers": answers_payload}  # 准备提交数据
+    status_update = f"Agent finished. Submitting {len(answers_payload)} answers for user '{username}'..."  # 构建状态更新消息
+    print(status_update)  # 打印状态更新
+
+    # 5. Submit  # 5. 提交
+    print(f"Submitting {len(answers_payload)} answers to: {submit_url}")  # 打印正在将多少个答案提交到哪个URL
+    try:  # 尝试
+        response = requests.post(submit_url, json=submission_data, timeout=60)  # 发送POST请求提交数据，超时设置为60秒
+        response.raise_for_status()  # 检查HTTP错误
+        result_data = response.json()  # 解析JSON响应
+        final_status = (  # 构建最终状态消息
+            f"Submission Successful!\n"  # 提交成功
+            f"User: {result_data.get('username')}\n"  # 用户名
+            f"Overall Score: {result_data.get('score', 'N/A')}% "  # 总分
+            f"({result_data.get('correct_count', '?')}/{result_data.get('total_attempted', '?')} correct)\n"  # 正确数/总尝试数
+            f"Message: {result_data.get('message', 'No message received.')}"  # 消息
+        )
+        print("Submission successful.")  # 打印提交成功
+        results_df = pd.DataFrame(results_log)  # 创建结果数据框
+        return final_status, results_df  # 返回最终状态和结果数据框
+    except requests.exceptions.HTTPError as e:  # 捕获HTTP错误
+        error_detail = f"Server responded with status {e.response.status_code}."  # 服务器响应状态
+        try:  # 尝试
+            error_json = e.response.json()  # 解析错误响应的JSON
+            error_detail += f" Detail: {error_json.get('detail', e.response.text)}"  # 添加错误详情
+        except requests.exceptions.JSONDecodeError:  # 捕获JSON解码错误
+            error_detail += f" Response: {e.response.text[:500]}"  # 添加原始响应文本
+        status_message = f"Submission Failed: {error_detail}"  # 构建提交失败消息
+        print(status_message)  # 打印状态消息
+        results_df = pd.DataFrame(results_log)  # 创建结果数据框
+        return status_message, results_df  # 返回状态消息和结果数据框
+    except requests.exceptions.Timeout:  # 捕获超时异常
+        status_message = "Submission Failed: The request timed out."  # 构建请求超时消息
+        print(status_message)  # 打印状态消息
+        results_df = pd.DataFrame(results_log)  # 创建结果数据框
+        return status_message, results_df  # 返回状态消息和结果数据框
+    except requests.exceptions.RequestException as e:  # 捕获请求异常
+        status_message = f"Submission Failed: Network error - {e}"  # 构建网络错误消息
+        print(status_message)  # 打印状态消息
+        results_df = pd.DataFrame(results_log)  # 创建结果数据框
+        return status_message, results_df  # 返回状态消息和结果数据框
+    except Exception as e:  # 捕获其他异常
+        status_message = f"An unexpected error occurred during submission: {e}"  # 构建意外错误消息
+        print(status_message)  # 打印状态消息
+        results_df = pd.DataFrame(results_log)  # 创建结果数据框
+        return status_message, results_df  # 返回状态消息和结果数据框
+
+# --- Build Gradio Interface using Blocks --- 
+with gr.Blocks() as demo:  # 创建Gradio Blocks界面
+    gr.Markdown("# Basic Agent Evaluation Runner")  # 添加标题
+    gr.Markdown(  # 添加说明文本
+        """
+        Modified by niku....  # 由niku修改
+        **Instructions:**  # 指示
+        1.  Please clone this space, then modify the code to define your agent's logic, the tools, the necessary packages, etc ...  # 请克隆此空间，然后修改代码以定义代理逻辑、工具和必要的包等
+        2.  Log in to your Hugging Face account using the button below. This uses your HF username for submission.  # 使用下面的按钮登录您的Hugging Face账户。这将使用您的HF用户名进行提交
+        3.  Click 'Run Evaluation & Submit All Answers' to fetch questions, run your agent, submit answers, and see the score.  # 点击"运行评估并提交所有答案"以获取问题，运行您的代理，提交答案，并查看分数
+        ---
+        **Disclaimers:**  # 免责声明
+        Once clicking on the "submit button, it can take quite some time ( this is the time for the agent to go through all the questions).  # 点击"提交"按钮后，可能需要相当长的时间（这是代理回答所有问题所需的时间）
+        This space provides a basic setup and is intentionally sub-optimal to encourage you to develop your own, more robust solution. For instance for the delay process of the submit button, a solution could be to cache the answers and submit in a seperate action or even to answer the questions in async.  # 此空间提供了一个基本设置，故意设计成不是最优的，以鼓励您开发自己的更强大的解决方案。例如，对于提交按钮的延迟过程，一个解决方案可能是缓存答案并在单独的操作中提交，或者甚至异步回答问题
+        """
+    )
+
+    gr.LoginButton()  # 添加登录按钮
+
+    run_button = gr.Button("Run Evaluation & Submit All Answers")  # 添加运行评估按钮
+
+    status_output = gr.Textbox(label="Run Status / Submission Result", lines=5, interactive=False)  # 添加状态输出文本框
+    # Removed max_rows=10 from DataFrame constructor  # 从DataFrame构造函数中移除了max_rows=10
+    results_table = gr.DataFrame(label="Questions and Agent Answers", wrap=True)  # 添加结果表格
+
+    run_button.click(  # 设置按钮点击事件
+        fn=run_and_submit_all,  # 点击时运行的函数
+        outputs=[status_output, results_table]  # 输出组件
+    )
+
+if __name__ == "__main__":  # 如果这个脚本作为主程序运行
+    print("\n" + "-"*30 + " App Starting " + "-"*30)  # 打印应用启动分隔线
+    # Check for SPACE_HOST and SPACE_ID at startup for information  # 启动时检查SPACE_HOST和SPACE_ID信息
+    space_host_startup = os.getenv("SPACE_HOST")  # 获取SPACE_HOST环境变量
+    space_id_startup = os.getenv("SPACE_ID") # Get SPACE_ID at startup  # 获取SPACE_ID环境变量
+
+    if space_host_startup:  # 如果找到SPACE_HOST
+        print(f"✅ SPACE_HOST found: {space_host_startup}")  # 打印找到SPACE_HOST
+        print(f"   Runtime URL should be: https://{space_host_startup}.hf.space")  # 打印运行时URL
+    else:  # 如果没找到SPACE_HOST
+        print("ℹ️  SPACE_HOST environment variable not found (running locally?).")  # 打印SPACE_HOST环境变量未找到
+
+    if space_id_startup: # Print repo URLs if SPACE_ID is found  # 如果找到SPACE_ID，打印代码库URL
+        print(f"✅ SPACE_ID found: {space_id_startup}")  # 打印找到SPACE_ID
+        print(f"   Repo URL: https://huggingface.co/spaces/{space_id_startup}")  # 打印代码库URL
+        print(f"   Repo Tree URL: https://huggingface.co/spaces/{space_id_startup}/tree/main")  # 打印代码库树URL
+    else:  # 如果没找到SPACE_ID
+        print("ℹ️  SPACE_ID environment variable not found (running locally?). Repo URL cannot be determined.")  # 打印SPACE_ID环境变量未找到
+
+    print("-"*(60 + len(" App Starting ")) + "\n")  # 打印结束分隔线
+
+    print("Launching Gradio Interface for Basic Agent Evaluation...")  # 打印正在启动Gradio界面
+    demo.launch(debug=True, share=False)  # 启动Gradio演示，启用调试模式，不共享