Update app.py

app.py

@@ -2,75 +2,125 @@ import os
 import gradio as gr
 import requests
 import pandas as pd
-from transformers import pipeline
-import torch
-import re
+from smolagents import CodeAgent, DuckDuckGoSearchTool, HfApiModel
+import logging
+
+# Настройка логирования
+logging.basicConfig(level=logging.INFO)
+logger = logging.getLogger(__name__)
 
 # --- Константы ---
 DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
 
-class DiagnosticAgent:
+# --- Internet-Enabled Agent Definition ---
+class InternetAgent:
     def __init__(self):
-        print("DiagnosticAgent initializing...")
-        
-    def test_api_connection(self, api_url: str) -> dict:
-        """Тестирует соединение с API и возвращает статус"""
-        results = {}
+        print("🌐 InternetAgent initializing with web search capabilities...")
         
-        # Тестируем базовый URL
         try:
-            response = requests.get(api_url, timeout=10)
-            results['base_url'] = {
-                'status_code': response.status_code,
-                'headers': dict(response.headers),
-                'content_preview': response.text[:500] if response.text else 'No content'
+            # Используем модель от Hugging Face и инструмент поиска
+            self.model = HfApiModel("Qwen/Qwen2.5-Coder-32B-Instruct")
+            self.search_tool = DuckDuckGoSearchTool()
+            
+            # Создаем агента с доступом к поиску
+            self.agent = CodeAgent(
+                tools=[self.search_tool],
+                model=self.model,
+                max_steps=6,  # Ограничиваем шаги для скорости
+                add_base_tools=False  # Используем только наши инструменты
+            )
+            print("✅ InternetAgent initialized successfully with web search")
+            
+        except Exception as e:
+            print(f"❌ Error initializing InternetAgent: {e}")
+            self.agent = None
+            # Резервная база знаний на случай ошибки
+            self.fallback_knowledge = {
+                "capital of france": "Paris",
+                "capital of germany": "Berlin", 
+                "capital of uk": "London",
+                "capital of usa": "Washington D.C.",
+                "2+2": "4",
+                "largest planet": "Jupiter",
             }
+
+    def __call__(self, question: str) -> str:
+        print(f"🤖 Processing: {question}")
+        
+        if not self.agent:
+            # Используем резервную базу знаний если агент не инициализирован
+            question_lower = question.lower()
+            for key, answer in self.fallback_knowledge.items():
+                if key in question_lower:
+                    return answer
+            return "I need internet access to answer this question properly."
+        
+        try:
+            # Создаем оптимизированный промпт для лучших результатов
+            optimized_prompt = f"""
+            Please provide a clear, concise, and accurate answer to the following question.
+            If you need to search for information, use the search tool.
+            Keep your answer brief and to the point.
+            
+            Question: {question}
+            
+            Answer:
+            """
+            
+            # Запускаем агента
+            response = self.agent.run(optimized_prompt)
+            
+            # Очищаем ответ
+            clean_response = self.clean_response(response)
+            print(f"✅ Answer: {clean_response[:100]}...")
+            
+            return clean_response
+            
         except Exception as e:
-            results['base_url'] = {'error': str(e)}
+            print(f"❌ Error in agent execution: {e}")
+            return f"I encountered an error while searching for the answer: {str(e)}"
+
+    def clean_response(self, response: str) -> str:
+        """Очищает ответ от лишней информации"""
+        # Удаляем мета-комментарии агента
+        lines = response.split('\n')
+        clean_lines = []
         
-        # Тестируем различные возможные endpoints
-        endpoints = [
-            '/questions',
-            '/api/questions',
-            '/v1/questions',
-            '/get-questions',
-            '/'
-        ]
+        for line in lines:
+            # Пропускаем строки с инструментами или процессами
+            if any(term in line.lower() for term in ['tool:', 'searching', 'step', 'using tool']):
+                continue
+            # Пропускаем пустые строки в начале
+            if not clean_lines and not line.strip():
+                continue
+            clean_lines.append(line)
         
-        for endpoint in endpoints:
-            try:
-                test_url = f"{api_url}{endpoint}"
-                response = requests.get(test_url, timeout=10)
-                results[endpoint] = {
-                    'status_code': response.status_code,
-                    'content_preview': response.text[:200] if response.text else 'No content'
-                }
-            except Exception as e:
-                results[endpoint] = {'error': str(e)}
+        clean_response = '\n'.join(clean_lines).strip()
         
-        return results
+        # Если ответ слишком длинный, берем первую часть
+        if len(clean_response) > 500:
+            clean_response = clean_response[:497] + "..."
+            
+        return clean_response if clean_response else "I couldn't find a clear answer to that question."
 
+# --- Упрощенная версия для тестирования ---
+class LiteInternetAgent:
+    def __init__(self):
+        print("🌐 LiteInternetAgent initializing...")
+        self.search_tool = DuckDuckGoSearchTool()
+        
     def __call__(self, question: str) -> str:
-        # Простой агент для тестирования
-        return f"Test answer for: {question}"
-
-def diagnose_api():
-    """Функция для диагностики API"""
-    agent = DiagnosticAgent()
-    results = agent.test_api_connection(DEFAULT_API_URL)
-    
-    print("=== API DIAGNOSTICS ===")
-    for endpoint, result in results.items():
-        print(f"\n{endpoint}:")
-        for key, value in result.items():
-            print(f"  {key}: {value}")
-    
-    return results
+        try:
+            # Прямой поиск через инструмент
+            result = self.search_tool(question)
+            return f"According to web search: {result[:300]}..." if len(result) > 300 else result
+        except Exception as e:
+            return f"Search failed: {str(e)}"
 
-# --- Основная функция с улучшенной обработкой ошибок ---
+# --- Основная функция ---
 def run_and_submit_all(profile: gr.OAuthProfile | None):
     """
-    Fetches all questions, runs the Agent on them, submits all answers,
+    Fetches all questions, runs the InternetAgent on them, submits all answers,
     and displays the results.
     """
     space_id = os.getenv("SPACE_ID")
@@ -83,60 +133,16 @@ def run_and_submit_all(profile: gr.OAuthProfile | None):
         return "Please Login to Hugging Face with the button.", None
 
     api_url = DEFAULT_API_URL
-    
-    # Сначала выполним диагностику
-    print("🔍 Running API diagnostics...")
-    diagnostic_results = diagnose_api()
-    
-    # Пробуем разные возможные endpoints для вопросов
-    possible_question_endpoints = [
-        '/questions',
-        '/api/questions', 
-        '/v1/questions',
-        '/get-questions',
-        '/'
-    ]
-    
-    questions_data = None
-    used_endpoint = None
-    
-    for endpoint in possible_question_endpoints:
-        questions_url = f"{api_url}{endpoint}"
-        print(f"Trying endpoint: {questions_url}")
-        
-        try:
-            response = requests.get(questions_url, timeout=30)
-            if response.status_code == 200:
-                questions_data = response.json()
-                used_endpoint = endpoint
-                print(f"✅ Success with endpoint: {endpoint}")
-                break
-            else:
-                print(f"❌ Endpoint {endpoint} returned status: {response.status_code}")
-        except Exception as e:
-            print(f"❌ Error with endpoint {endpoint}: {e}")
-    
-    if questions_data is None:
-        error_msg = f"""
-🚨 Could not fetch questions from any endpoint!
-
-Diagnostic results:
-{diagnostic_results}
+    questions_url = f"{api_url}/questions"
+    submit_url = f"{api_url}/submit"
 
-Possible solutions:
-1. Check if the API URL is correct: {DEFAULT_API_URL}
-2. The API might be temporarily down
-3. The endpoint might have changed
-4. There might be network restrictions
-
-Please check the assignment instructions for the correct API URL.
-"""
-        print(error_msg)
-        return error_msg, None
-
-    # Instantiate Agent
+    # 1. Instantiate our Internet Agent
     try:
-        agent = DiagnosticAgent()
+        # Пробуем полную версию, если не работает - упрощенную
+        agent = InternetAgent()
+        if agent.agent is None:
+            agent = LiteInternetAgent()
+        print("✅ InternetAgent created successfully")
     except Exception as e:
         print(f"Error instantiating agent: {e}")
         return f"Error initializing agent: {e}", None
@@ -144,9 +150,32 @@ Please check the assignment instructions for the correct API URL.
     agent_code = f"https://huggingface.co/spaces/{space_id}/tree/main"
     print(f"Agent code URL: {agent_code}")
 
-    print(f"Fetched {len(questions_data)} questions from endpoint: {used_endpoint}")
+    # 2. Fetch Questions
+    print(f"Fetching questions from: {questions_url}")
+    try:
+        response = requests.get(questions_url, timeout=30)
+        response.raise_for_status()
+        questions_data = response.json()
+        if not questions_data:
+            print("Fetched questions list is empty.")
+            return "Fetched questions list is empty or invalid format.", None
+        print(f"✅ Fetched {len(questions_data)} questions.")
+    except Exception as e:
+        error_msg = f"❌ Error fetching questions: {e}"
+        print(error_msg)
+        
+        # Демо-режим с разными типами вопросов
+        demo_questions = [
+            {"task_id": "demo1", "question": "What is the capital of France?"},
+            {"task_id": "demo2", "question": "What is the current weather in Tokyo?"},
+            {"task_id": "demo3", "question": "Who won the Nobel Prize in Physics in 2023?"},
+            {"task_id": "demo4", "question": "What is the population of Brazil?"},
+            {"task_id": "demo5", "question": "Explain quantum computing in simple terms"},
+        ]
+        questions_data = demo_questions
+        print("🚨 Using demo questions since API is unavailable")
 
-    # Run your Agent
+    # 3. Run your Agent
     results_log = []
     answers_payload = []
     print(f"Running agent on {len(questions_data)} questions...")
@@ -159,7 +188,7 @@ Please check the assignment instructions for the correct API URL.
             print(f"Skipping item with missing task_id or question: {item}")
             continue
             
-        print(f"🔍 Processing question {i+1}/{len(questions_data)}")
+        print(f"🔍 Processing question {i+1}/{len(questions_data)}: {question_text[:50]}...")
         
         try:
             submitted_answer = agent(question_text)
@@ -173,94 +202,103 @@ Please check the assignment instructions for the correct API URL.
         print("Agent did not produce any answers to submit.")
         return "Agent did not produce any answers to submit.", pd.DataFrame(results_log)
 
-    # Prepare Submission
+    # 4. Prepare Submission 
     submission_data = {
         "username": username.strip(), 
         "agent_code": agent_code, 
         "answers": answers_payload
     }
     
-    status_update = f"✅ Agent finished. Submitting {len(answers_payload)} answers for user '{username}'..."
+    status_update = f"✅ Agent finished. Processed {len(answers_payload)} answers for user '{username}'"
     print(status_update)
 
-    # Try different submit endpoints
-    possible_submit_endpoints = [
-        '/submit',
-        '/api/submit',
-        '/v1/submit',
-        '/post-answers'
-    ]
-    
-    result_data = None
-    
-    for endpoint in possible_submit_endpoints:
-        submit_url = f"{api_url}{endpoint}"
-        print(f"Trying submit endpoint: {submit_url}")
-        
+    # 5. Submit answers (только если вопросы реальные, не демо)
+    if "demo" not in str(questions_data[0].get("task_id", "")):
+        print(f"Submitting {len(answers_payload)} answers to: {submit_url}")
         try:
             response = requests.post(submit_url, json=submission_data, timeout=120)
-            if response.status_code == 200:
-                result_data = response.json()
-                print(f"✅ Submit successful with endpoint: {endpoint}")
-                break
-            else:
-                print(f"❌ Submit endpoint {endpoint} returned status: {response.status_code}")
+            response.raise_for_status()
+            result_data = response.json()
+            
+            final_status = (
+                f"🎉 Submission Successful!\n"
+                f"👤 User: {result_data.get('username')}\n"
+                f"📊 Overall Score: {result_data.get('score', 'N/A')}% "
+                f"({result_data.get('correct_count', '?')}/{result_data.get('total_attempted', '?')} correct)\n"
+                f"💬 Message: {result_data.get('message', 'No message received.')}"
+            )
+            print("✅ Submission successful.")
+            results_df = pd.DataFrame(results_log)
+            return final_status, results_df
+            
         except Exception as e:
-            print(f"❌ Error with submit endpoint {endpoint}: {e}")
-
-    if result_data is None:
-        error_message = "❌ Submission Failed: Could not submit to any endpoint"
-        print(error_message)
+            error_message = f"❌ Submission Failed: {str(e)}"
+            print(error_message)
+            results_df = pd.DataFrame(results_log)
+            return error_message, results_df
+    else:
+        # Демо-режим: показываем ответы но не отправляем
+        demo_status = (
+            f"🧪 DEMO MODE (API Unavailable)\n"
+            f"👤 User: {username}\n"
+            f"📊 Processed: {len(answers_payload)} demo questions\n"
+            f"🌐 Agent used web search for answers\n"
+            f"💬 Real submission disabled - API not accessible\n\n"
+            f"Check the web-powered answers below!"
+        )
+        print("✅ Demo completed - showing results without submission")
         results_df = pd.DataFrame(results_log)
-        return error_message, results_df
-        
-    final_status = (
-        f"🎉 Submission Successful!\n"
-        f"👤 User: {result_data.get('username')}\n"
-        f"📊 Overall Score: {result_data.get('score', 'N/A')}% "
-        f"({result_data.get('correct_count', '?')}/{result_data.get('total_attempted', '?')} correct)\n"
-        f"💬 Message: {result_data.get('message', 'No message received.')}"
-    )
-    print("✅ Submission successful.")
-    results_df = pd.DataFrame(results_log)
-    return final_status, results_df
+        return demo_status, results_df
 
 # --- Gradio Interface ---
-with gr.Blocks() as demo:
-    gr.Markdown("# 🔍 API Diagnostic Agent")
+with gr.Blocks(title="Internet-Enabled AI Agent") as demo:
     gr.Markdown("""
-    This agent will diagnose API connectivity issues and try multiple endpoints.
+    # 🌐 Internet-Enabled AI Agent
+    **Powered by web search and large language models**
     
-    **If you're getting 404 errors:**
-    1. Check if the API URL is correct
-    2. Try using VPN if there are network restrictions
-    3. Check the assignment instructions for updated URLs
-    """)
+    ### 🔧 Capabilities:
+    - **Web Search**: Real-time information from DuckDuckGo
+    - **LLM Power**: Qwen2.5-32B model for understanding
+    - **Multi-step Reasoning**: Complex question answering
     
-    gr.Markdown(f"**Current API URL:** `{DEFAULT_API_URL}`")
+    ### 📚 Example questions:
+    - *"Current weather in any city"*
+    - *"Latest news headlines"* 
+    - *"Historical facts and data"*
+    - *"Scientific explanations"*
+    - *"Complex calculations"*
+    """)
     
-    # Добавляем кнопку для диагностики
-    diagnose_btn = gr.Button("🔍 Run API Diagnostics", variant="secondary")
-    diagnostic_output = gr.Textbox(label="Diagnostic Results", lines=10, interactive=False)
+    gr.Markdown("""
+    ### ⚠️ Important Notes:
+    - This agent requires internet access
+    - Responses may take longer due to web searches
+    - Some questions might not have clear online answers
+    """)
     
-    gr.LoginButton()
-    run_button = gr.Button("🚀 Run Evaluation & Submit All Answers", variant="primary")
+    with gr.Row():
+        with gr.Column():
+            gr.LoginButton()
+            run_button = gr.Button("🚀 Run Evaluation", variant="primary")
     
-    status_output = gr.Textbox(label="📊 Run Status / Submission Result", lines=5, interactive=False)
-    results_table = gr.DataFrame(label="📝 Questions and Agent Answers", wrap=True)
+    with gr.Row():
+        with gr.Column():
+            status_output = gr.Textbox(
+                label="Status", 
+                lines=4, 
+                interactive=False
+            )
+        with gr.Column():
+            results_table = gr.DataFrame(
+                label="Questions & Answers", 
+                wrap=True
+            )
 
-    # Обработчик для диагностики
-    diagnose_btn.click(
-        fn=lambda: str(diagnose_api()),
-        outputs=[diagnostic_output]
-    )
-    
-    # Обработчик для основной функции
     run_button.click(
         fn=run_and_submit_all,
         outputs=[status_output, results_table]
     )
 
 if __name__ == "__main__":
-    print("🚀 Launching Diagnostic Agent Interface...")
+    print("🚀 Starting Internet-Enabled AI Agent...")
     demo.launch(debug=True, share=False)