ADD DOWNLOAD_UNIVERSAL

app.py

@@ -4,6 +4,8 @@ import requests
 import inspect
 import pandas as pd
 import time
+import mimetypes
+from pathlib import Path
 
 # (Keep Constants as is)
 # --- Constants ---
@@ -15,33 +17,123 @@ DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
 from smolagents import CodeAgent, LiteLLMModel 
 from my_tools import my_tool_list  
 
-def download_file(task_id, filename, save_dir="attachments"):
+import mimetypes
+from pathlib import Path
+
+def download_file_universal(task_id, save_dir="attachments"):
+    """
+    通用文件下载，自动检测文件类型和扩展名
+    """
     os.makedirs(save_dir, exist_ok=True)
-    url = f"https://agents-course-unit4-scoring.hf.space/files/{task_id}/{filename}"
-    save_path = os.path.join(save_dir, filename)
-    print(f"[DEBUG] Try download: url={url} save_path={save_path}")
+    
+    url = f"https://agents-course-unit4-scoring.hf.space/files/{task_id}"
+    
+    print(f"[DEBUG] Downloading from: {url}")
+    
     try:
-        resp = requests.get(url, timeout=30)  
-        print(f"[DEBUG] HTTP {resp.status_code} for {url}")
+        headers = {
+            'Accept': '*/*',
+            'User-Agent': 'Mozilla/5.0 (compatible; Agent/1.0)'
+        }
+        
+        resp = requests.get(url, headers=headers, timeout=30, stream=True)
+        print(f"[DEBUG] HTTP {resp.status_code}")
+        print(f"[DEBUG] Content-Type: {resp.headers.get('content-type', 'Unknown')}")
+        print(f"[DEBUG] Content-Disposition: {resp.headers.get('content-disposition', 'Unknown')}")
+        
         resp.raise_for_status()
         
-        if resp.headers.get('content-type', '').startswith('text/html'):
-            print(f"Warning: Received HTML response, file might not exist: {filename}")
-            return None
-            
+        # 从Content-Disposition获取原始文件名
+        filename = None
+        content_disp = resp.headers.get('content-disposition', '')
+        if 'filename=' in content_disp:
+            filename = content_disp.split('filename=')[1].strip('"\'')
+        
+        # 如果没有文件名，根据Content-Type推断
+        if not filename:
+            content_type = resp.headers.get('content-type', '').lower()
+            ext = mimetypes.guess_extension(content_type.split(';')[0])
+            if not ext:
+                # 手动映射常见类型
+                type_map = {
+                    'image/png': '.png',
+                    'image/jpeg': '.jpg',
+                    'image/gif': '.gif',
+                    'video/mp4': '.mp4',
+                    'video/avi': '.avi',
+                    'video/mov': '.mov',
+                    'audio/mp3': '.mp3',
+                    'audio/wav': '.wav',
+                    'audio/mpeg': '.mp3',
+                    'application/pdf': '.pdf',
+                    'text/plain': '.txt',
+                    'application/json': '.json',
+                    'text/csv': '.csv'
+                }
+                ext = type_map.get(content_type.split(';')[0], '.bin')
+            filename = f"{task_id}{ext}"
+        
+        save_path = os.path.join(save_dir, filename)
+        print(f"[DEBUG] Saving as: {save_path}")
+        
+        # 流式下载，适合大文件
         with open(save_path, "wb") as f:
-            f.write(resp.content)
-        print(f"Downloaded attachment for task {task_id} -> {save_path}")
-        return save_path
-    except requests.exceptions.HTTPError as e:
-        if e.response.status_code == 404:
-            print(f"File not found: {filename} for task {task_id}")
-        else:
-            print(f"HTTP error downloading {filename}: {e}")
-        return None
+            for chunk in resp.iter_content(chunk_size=8192):
+                f.write(chunk)
+        
+        file_size = os.path.getsize(save_path)
+        print(f"[DEBUG] Successfully saved: {filename} ({file_size} bytes)")
+        
+        return save_path, filename
+        
     except Exception as e:
-        print(f"Attachment download failed for {task_id}/{filename}: {e}")
-        return None
+        print(f"[DEBUG] Download error: {e}")
+        return None, None
+
+def download_task_files_on_demand(task_id, file_list, save_dir="attachments"):
+    """
+    按需下载：处理每个问题时才下载对应文件
+    """
+    os.makedirs(save_dir, exist_ok=True)
+    downloaded_files = []
+    
+    if not file_list:
+        print(f"[INFO] No files listed for task {task_id}, attempting direct download...")
+        file_path, filename = download_file_universal(task_id, save_dir)
+        if file_path:
+            downloaded_files.append(file_path)
+    else:
+        print(f"[INFO] Task {task_id} has {len(file_list)} files to download")
+        for expected_filename in file_list:
+            # 先检查是否已经下载过
+            potential_path = os.path.join(save_dir, expected_filename)
+            if os.path.exists(potential_path):
+                print(f"[CACHE] File already exists: {expected_filename}")
+                downloaded_files.append(potential_path)
+                continue
+            
+            # 下载文件
+            file_path, actual_filename = download_file_universal(task_id, save_dir)
+            if file_path:
+                # 如果实际文件名与期望不符，重命名
+                if actual_filename != expected_filename:
+                    new_path = os.path.join(save_dir, expected_filename)
+                    try:
+                        os.rename(file_path, new_path)
+                        file_path = new_path
+                        print(f"[INFO] Renamed {actual_filename} to {expected_filename}")
+                    except:
+                        print(f"[WARN] Could not rename file, keeping as {actual_filename}")
+                
+                downloaded_files.append(file_path)
+                print(f"[SUCCESS] Downloaded: {os.path.basename(file_path)}")
+            else:
+                print(f"[FAIL] Could not download: {expected_filename}")
+            
+            # 添加小延迟避免速率限制
+            time.sleep(0.5)
+    
+    return downloaded_files
 
 class BasicAgent:
     def __init__(self):
@@ -66,19 +158,35 @@ class BasicAgent:
         else:
             print(f"{self.agent_name} received question: {question[:80]}...")
         try:
-            # system prompt + question
             system_prompt = """
             You are Celum, an advanced agent skilled at using external tools and step-by-step reasoning to solve real-world problems.
             You may freely think, reason, and use tools or your own knowledge as needed to solve the problem.
-            When you are ready to submit your answer,ONLY output your final answer in the exact format required by the question.DO NOT add any extra context.
-            If the answer is a number, output only the number (no units, no commas).
-            If the answer is a word or string, do not use articles or abbreviations, and write digits as plain numbers.
-            If the answer is a comma-separated list, apply the same rules to each item.
-            If you cannot answer, return the word 'unknown'.           
+            
+            IMPORTANT OUTPUT INSTRUCTIONS:
+            - When you need to return your final answer, use print() to output it
+            - For example: print("Qxf7") or print("42") or print("unknown")
+            - DO NOT just write the answer as a standalone variable
+            - If you cannot solve the problem, use: print("unknown")
+            
+            Answer format requirements:
+            - If the answer is a number, output only the number (no units, no commas)
+            - If the answer is a word or string, do not use articles or abbreviations, and write digits as plain numbers
+            - If the answer is a comma-separated list, apply the same rules to each item
+            - If you cannot answer, return the word 'unknown'
             """
+            
             files_prompt = ""
             if files:
-                files_prompt = f"\n[You have the following attached files: {', '.join(files)}]\n"
+                files_prompt = f"\n[You have the following attached files available: {', '.join(files)}]\n"
+                # 添加文件处理指导
+                files_prompt += """
+    To work with image files, you can:
+    1. Use PIL (from PIL import Image) to load and analyze images
+    2. Use cv2 (import cv2) for image processing
+    3. For chess positions, you might need to analyze the board position pixel by pixel
+    4. Consider using OCR libraries if there's text in the image
+    """
+            
             full_question = system_prompt + files_prompt + "\n\n" + question
             return self.agent.run(full_question)
         except Exception as e:
@@ -153,39 +261,63 @@ def run_and_submit_all( profile: gr.OAuthProfile | None):
     results_log = []
     answers_payload = []
     print(f"Running agent on {len(questions_data)} questions...")
+    
     for idx, item in enumerate(questions_data[3:4]):
         task_id = item.get("task_id")
         question_text = item.get("question")
         file_list = item.get("files", [])
-        print(f"[DEBUG] Q{idx+1}/{len(questions_data)} | task_id={task_id} | files={file_list}")
         
+        print(f"\n{'='*60}")
+        print(f"Processing Question {idx+1}/{len(questions_data)}")
+        print(f"Task ID: {task_id}")
+        print(f"Question: {question_text[:100]}...")
+        print(f"Expected files: {file_list}")
+        print(f"{'='*60}")
+        
+        # 按需下载文件
         local_files = []
-        for fname in file_list:
-            fpath = download_file(task_id, fname)
-            if fpath:
-                local_files.append(fpath)
-        if local_files:
-            print(f"Downloaded attachments: {local_files}")
+        if file_list or True:  # 总是尝试下载，因为有些任务可能没有在file_list中列出
+            print(f"[DOWNLOAD] Starting download for task {task_id}...")
+            local_files = download_task_files_on_demand(task_id, file_list)
+            
+            if local_files:
+                print(f"[DOWNLOAD] Successfully got {len(local_files)} files:")
+                for f in local_files:
+                    size = os.path.getsize(f)
+                    print(f"  - {os.path.basename(f)} ({size} bytes)")
+            else:
+                print(f"[DOWNLOAD] No files downloaded for task {task_id}")
         
-        print(f"===== [Celum is answering No. {idx+1}/{len(questions_data)} ] =====")
+        # 运行Agent
+        print(f"[AGENT] Running Celum on question {idx+1}...")
         try:
             submitted_answer = safe_run_agent(agent, question_text, local_files, idx, len(questions_data))
+            
             answers_payload.append({"task_id": task_id, "submitted_answer": submitted_answer})
             results_log.append({
                 "Task ID": task_id,
-                "Question": question_text,
+                "Question": question_text[:100] + "...",
                 "Submitted Answer": submitted_answer,
-                "Files": local_files
+                "Files": [os.path.basename(f) for f in local_files] if local_files else []
             })
+            
+            print(f"[AGENT] Answer: {submitted_answer}")
+            
         except Exception as e:
-            print(f"[Celum Error at Q{idx+1}]: {e}")
+            error_msg = f"AGENT ERROR: {e}"
+            print(f"[ERROR] {error_msg}")
+            answers_payload.append({"task_id": task_id, "submitted_answer": "unknown"})
             results_log.append({
                 "Task ID": task_id,
-                "Question": question_text,
-                "Submitted Answer": f"AGENT ERROR: {e}",
-                "Files": local_files
+                "Question": question_text[:100] + "...",
+                "Submitted Answer": error_msg,
+                "Files": []
             })
-        time.sleep(10)
+        
+        # 每题之间的延迟
+        if idx < len(questions_data) - 1:
+            print(f"[WAIT] Waiting before next question...")
+            time.sleep(2)
     
     if not answers_payload:
         print("Agent did not produce any answers to submit.")