feat: code agent with web search

app.py

@@ -1,23 +1,143 @@
 import os
+from traceback import print_tb
 import gradio as gr
 import requests
 import inspect
 import pandas as pd
+from smolagents import CodeAgent, DuckDuckGoSearchTool, InferenceClientModel, OpenAIServerModel, ToolCallingAgent, VisitWebpageTool
+from dotenv import load_dotenv
+from utils import detect_file_category
+from tools import transcribe_audio
+from PIL import Image
+
+load_dotenv()
 
 # (Keep Constants as is)
 # --- Constants ---
 DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
+DASHSCOPE_API_BASE = os.getenv("DASHSCOPE_API_BASE")
+DASHSCOPE_API_KEY = os.getenv("DASHSCOPE_API_KEY")  
 
 # --- Basic Agent Definition ---
 # ----- THIS IS WERE YOU CAN BUILD WHAT YOU WANT ------
 class BasicAgent:
     def __init__(self):
         print("BasicAgent initialized.")
-    def __call__(self, question: str) -> str:
-        print(f"Agent received question (first 50 chars): {question[:50]}...")
-        fixed_answer = "This is a default answer."
-        print(f"Agent returning fixed answer: {fixed_answer}")
-        return fixed_answer
+
+    def __call__(self, question: str, file_path: str) -> str:
+        if not file_path or "ERROR" in file_path:
+            file_path = None
+
+        if file_path:
+            category = detect_file_category(file_path)
+        else:
+            category = "none"
+
+        if category == "none":
+            agent = CodeAgent(tools=[DuckDuckGoSearchTool(), VisitWebpageTool()], model=OpenAIServerModel(
+                model_id="qwen3-coder-flash",
+                api_base=DASHSCOPE_API_BASE,
+                api_key=DASHSCOPE_API_KEY,
+            ))
+            return agent.run(question)
+        
+        if category == "audio":
+            agent = CodeAgent(
+                tools=[DuckDuckGoSearchTool(), VisitWebpageTool(), transcribe_audio], 
+                model=OpenAIServerModel(
+                    model_id="qwen3-coder-flash",
+                    api_base=DASHSCOPE_API_BASE,
+                    api_key=DASHSCOPE_API_KEY,
+            ))
+            return agent.run(question + f"\n\nfile_path：{file_path}")
+
+        if category == "image":
+            agent = CodeAgent(
+                model=OpenAIServerModel(
+                    model_id="qwen3-vl-flash",
+                    api_base=DASHSCOPE_API_BASE,
+                    api_key=DASHSCOPE_API_KEY,
+                ),
+                max_steps=20,
+                verbosity_level=2
+            )
+            # agent = CodeAgent(
+            #     tools=[DuckDuckGoSearchTool(), VisitWebpageTool()], 
+            #     model=OpenAIServerModel(
+            #         model_id="qwen3-coder-flash",
+            #         api_base=DASHSCOPE_API_BASE,
+            #         api_key=DASHSCOPE_API_KEY,
+            #     ),
+            #     managed_agents=[image_agent])
+            return agent.run(question, images=[Image.open(file_path).convert("RGB")])
+        
+        agent = CodeAgent(
+            additional_authorized_imports=["pandas"],
+            tools=[DuckDuckGoSearchTool(), VisitWebpageTool()], 
+            model=OpenAIServerModel(
+                model_id="qwen3-coder-flash",
+                api_base=DASHSCOPE_API_BASE,
+                api_key=DASHSCOPE_API_KEY,
+        ))
+        return agent.run(question + f"\n\nfile_path：{file_path}")
+
+# 新增：下载与 task_id 关联的文件的辅助函数
+import re
+
+def download_task_file(api_url: str, task_id: str, output_dir: str = "downloads") -> str:
+    files_url = f"{api_url}/files/{task_id}"
+    try:
+        os.makedirs(output_dir, exist_ok=True)
+
+        # 快速预检：如果 downloads 里已存在以 task_id 命名的文件则直接返回
+        try:
+            for name in os.listdir(output_dir):
+                base, _ext = os.path.splitext(name)
+                candidate = os.path.join(output_dir, name)
+                if base == task_id and os.path.isfile(candidate):
+                    print(f"File for task {task_id} already exists: {candidate}")
+                    return candidate
+        except FileNotFoundError:
+            pass
+
+        with requests.get(files_url, stream=True, timeout=30) as r:
+            r.raise_for_status()
+            filename = None
+            cd = r.headers.get("content-disposition")
+            if cd:
+                m = re.search('filename="?([^";]+)"?', cd)
+                if m:
+                    filename = m.group(1)
+            if not filename:
+                filename = r.headers.get("x-filename")
+            if not filename:
+                filename = f"{task_id}.download"
+            dest_path = os.path.join(output_dir, filename)
+
+            # 二次检查：若目标文件已存在则跳过重新下载
+            if os.path.exists(dest_path):
+                print(f"File for task {task_id} already exists: {dest_path}")
+                return dest_path
+
+            with open(dest_path, "wb") as f:
+                for chunk in r.iter_content(chunk_size=8192):
+                    if chunk:
+                        f.write(chunk)
+        print(f"Downloaded file for task {task_id} to: {dest_path}")
+        return dest_path
+    except requests.exceptions.HTTPError as e:
+        status = getattr(e.response, 'status_code', 'unknown')
+        print(f"File download HTTP error for task {task_id}: {e}")
+        return f"ERROR: HTTP {status} for task {task_id}"
+    except requests.exceptions.Timeout:
+        print(f"File download timed out for task {task_id}")
+        return f"ERROR: Timeout downloading task {task_id}"
+    except requests.exceptions.RequestException as e:
+        print(f"File download network error for task {task_id}: {e}")
+        return f"ERROR: Network error downloading task {task_id}: {e}"
+    except Exception as e:
+        print(f"Unexpected error downloading file for task {task_id}: {e}")
+        return f"ERROR: Unexpected error downloading task {task_id}: {e}"
 
 def run_and_submit_all( profile: gr.OAuthProfile | None):
     """
@@ -79,13 +199,15 @@ def run_and_submit_all( profile: gr.OAuthProfile | None):
         if not task_id or question_text is None:
             print(f"Skipping item with missing task_id or question: {item}")
             continue
+        # 新增：下载与该 task_id 关联的文件
+        downloaded_path = download_task_file(api_url, task_id)
         try:
-            submitted_answer = agent(question_text)
+            submitted_answer = agent(question_text, downloaded_path)
             answers_payload.append({"task_id": task_id, "submitted_answer": submitted_answer})
-            results_log.append({"Task ID": task_id, "Question": question_text, "Submitted Answer": submitted_answer})
+            results_log.append({"Task ID": task_id, "Question": question_text, "Submitted Answer": submitted_answer, "Downloaded File": downloaded_path})
         except Exception as e:
              print(f"Error running agent on task {task_id}: {e}")
-             results_log.append({"Task ID": task_id, "Question": question_text, "Submitted Answer": f"AGENT ERROR: {e}"})
+             results_log.append({"Task ID": task_id, "Question": question_text, "Submitted Answer": f"AGENT ERROR: {e}", "Downloaded File": downloaded_path})
 
     if not answers_payload:
         print("Agent did not produce any answers to submit.")

downloads/f918266a-b3e0-4914-865d-4faa564f1aef.py

@@ -0,0 +1,35 @@
+from random import randint
+import time
+
+class UhOh(Exception):
+    pass
+
+class Hmm:
+    def __init__(self):
+        self.value = randint(-100, 100)
+
+    def Yeah(self):
+        if self.value == 0:
+            return True
+        else:
+            raise UhOh()
+
+def Okay():
+    while True:
+        yield Hmm()
+
+def keep_trying(go, first_try=True):
+    maybe = next(go)
+    try:
+        if maybe.Yeah():
+            return maybe.value
+    except UhOh:
+        if first_try:
+            print("Working...")
+            print("Please wait patiently...")
+        time.sleep(0.1)
+        return keep_trying(go, first_try=False)
+
+if __name__ == "__main__":
+    go = Okay()
+    print(f"{keep_trying(go)}")

requirements.txt

@@ -1,2 +1,7 @@
 gradio
-requests
+requests
+smolagents[all]
+faster-whisper
+filetype
+torch
+Pillow

tools.py

@@ -0,0 +1,36 @@
+from smolagents import tool
+import os
+
+@tool
+def transcribe_audio(file_path: str) -> str:
+    """
+    Transcribes an audio file using faster-whisper.
+
+    Args:
+        file_path: The path to the audio file.
+
+    Returns:
+        The transcribed text, or an error message if transcription fails.
+    """
+    # Use faster-whisper if available
+    try:
+        from faster_whisper import WhisperModel
+        import torch
+
+        if torch.cuda.is_available():
+            device = "cuda"
+        else:
+            device = "cpu"
+
+        model = WhisperModel("base", device=device)
+        segments, info = model.transcribe(file_path)
+        text_parts = []
+        for seg in segments:
+            try:
+                text_parts.append(seg.text)
+            except Exception:
+                pass
+        text = " ".join(text_parts).strip()
+        return text or "[ASR result is empty]"
+    except Exception as e:
+        return f"[ASR is not available] Please install `faster-whisper` to enable audio recognition. Error: {e}"

utils.py

@@ -0,0 +1,44 @@
+import os
+import mimetypes
+
+def detect_file_category(file_path: str) -> str:
+    if not file_path or not os.path.exists(file_path):
+        return "none"
+    mime = None
+    try:
+        import filetype  # optional; if not installed, fallback to mimetypes
+        kind = filetype.guess(file_path)
+        mime = kind.mime if kind else None
+    except Exception:
+        mime = None
+    if not mime:
+        mime, _ = mimetypes.guess_type(file_path)
+    ext = os.path.splitext(file_path)[1].lower()
+    if mime:
+        if mime.startswith("image/"):
+            return "image"
+        if mime.startswith("audio/"):
+            return "audio"
+        if mime.startswith("video/"):
+            return "video"
+        if mime in (
+            "application/vnd.ms-excel",
+            "application/vnd.openxmlformats-officedocument.spreadsheetml.sheet",
+            "text/csv",
+        ):
+            return "spreadsheet"
+        if mime == "application/pdf":
+            return "document"
+        if mime.startswith("text/"):
+            if ext in (".py", ".md", ".txt", ".csv"):
+                return "text"
+    # extension fallback
+    if ext in (".png", ".jpg", ".jpeg", ".gif", ".bmp", ".webp", ".tiff"):
+        return "image"
+    if ext in (".mp3", ".wav", ".m4a", ".flac", ".ogg"):
+        return "audio"
+    if ext in (".mp4", ".mov", ".mkv", ".webm"):
+        return "video"
+    if ext in (".xls", ".xlsx", ".csv"):
+        return "spreadsheet"
+    return "unknown"