Update app.py

app.py

@@ -2,20 +2,17 @@ import os
 import gradio as gr
 import requests
 import inspect
+import tempfile
 import pandas as pd
 from smolagents import CodeAgent, DuckDuckGoSearchTool, VisitWebpageTool, InferenceClientModel, tool,OpenAIServerModel
 import re
 from huggingface_hub import InferenceClient
 import base64
 
-# (Keep Constants as is)
 # --- Constants ---
 DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
 
-# --- Basic Agent Definition ---
-# ----- THIS IS WERE YOU CAN BUILD WHAT YOU WANT ------
-
-
+#--- tools ---
 
 @tool
 def analyze_image_tool(image_path: str, question: str) -> str:
@@ -73,44 +70,116 @@ def download_task_file(task_id: str) -> str:
                 if matches:
                     filename = matches[0]
                     
-                temp_dir = tempfile.gettempdir()
-                filepath = os.path.join(temp_dir, filename)  
-                
-                with open(filepath, 'wb') as f:
-                    f.write(response.content)
-                
+            temp_dir = tempfile.gettempdir()
+            filepath = os.path.join(temp_dir, filename)  
+            
+            with open(filepath, 'wb') as f:
+                f.write(response.content)
+            
             return f"File downloaded successfully! The file is located at: {filepath}. You can now write python code to open and read this file."
         else:
             return f"Failed to download file. HTTP Status: {response.status_code}. The file might not exist for this task."
     except Exception as e:
         return f"An error occurred while downloading: {e}"
 
+@tool
+def transcribe_audio_tool(audio_path: str) -> str:
+    """
+    Transcribes an audio or video file (mp3, mp4, wav) to text.
+    Use this tool AFTER downloading a media file using download_task_file if the question requires knowing what is spoken in the file.
+    
+    Args:
+        audio_path: The local path to the audio/video file.
+    """
+    try:
+        api_key = os.getenv("GROQ_API_KEY")
+        if not api_key:
+            return "Error: GROQ_API_KEY not found in environment."
+            
+        headers = {"Authorization": f"Bearer {api_key}"}
+        url = "https://api.groq.com/openai/v1/audio/transcriptions"
+        
+        print(f"Transcribing audio file {audio_path}...")
+        with open(audio_path, "rb") as f:
+            files = {"file": (os.path.basename(audio_path), f)}
+            data = {"model": "whisper-large-v3"}
+            response = requests.post(url, headers=headers, files=files, data=data)
+            
+        response.raise_for_status()
+        return response.json().get("text", "No text could be extracted.")
+    except Exception as e:
+        return f"Error transcribing audio: {str(e)}"
+
+@tool
+def read_webpage_clean_tool(url: str) -> str:
+    """
+    Reads a webpage and extracts clean Markdown text. 
+    Use this tool when VisitWebpageTool fails or when you need to read articles, documentation, or bypass JavaScript rendering.
+    
+    Args:
+        url: The full URL of the webpage to read.
+    """
+    try:
+        print(f"Fetching clean text from: {url}")
+        jina_url = f"https://r.jina.ai/{url}"
+        response = requests.get(jina_url, timeout=20)
+        response.raise_for_status()
+        # Restrict length to prevent context window overflow
+        return response.text[:8000] 
+    except Exception as e:
+        return f"Error reading webpage with clean tool: {str(e)}" 
 search_tool = DuckDuckGoSearchTool()
 visit_webpage_tool = VisitWebpageTool()
 
+#--- agent ---
+
 model = OpenAIServerModel(
-    model_id="llama-3.1-8b-instant",
+    model_id="llama-3.3-70b-versatile",
     api_base="https://api.groq.com/openai/v1",
     api_key=os.getenv("GROQ_API_KEY")
 )
 manager_system_prompt="""
-You are a general AI assistant. I will ask you a question. Report your thoughts,
- and finish your answer with the following template: [YOUR FINAL ANSWER].
- YOUR FINAL ANSWER should be a number OR as few words as possible OR a comma separated list of numbers and/or strings.
- If you are asked for a number, don't use comma to write your number
-neither use units such as $ or percent sign unless specified otherwise.
-If you are asked for a string, don't use articles, neither abbreviations (e.g. for cities),
-and write the digits in plain text unless specified otherwise.
- If you are asked for a comma separated list,
-apply the above rules depending of whether the element to be put in the list is a number or a string."""
+You are an advanced AI assistant designed to solve complex, multi-step tasks from the GAIA benchmark.
+You MUST strictly adhere to the ReAct (Reason, Act, Observe) framework for EVERY step you take.
+
+Your process must look like this:
+1. **Thought:** Carefully analyze the current state. What information do you have? What is missing? Plan your next exact move step-by-step.
+2. **Action:** Execute your plan by calling the appropriate tool or writing Python code. 
+3. **Observation:** Review the output of your tool or code. (If code fails, your next Thought must be how to fix it).
+
+Repeat this loop until you are 100% confident you have the final answer.
+
+**Specific Guidelines for Tools and Code:**
+1. **Information Gathering:** If the prompt mentions a file, use `download_task_file` FIRST to retrieve the local file path. 
+2. **Analysis via Code:** Write Python code to open, read, and analyze downloaded files (using pandas, json, PyPDF2, etc.). ALWAYS use `print()` in your code to output intermediate data, statistics, or rows so you can see the results in your next step.
+3. **Vision/Images:** If the task involves an image, pass the local image path and your specific question to `analyze_image_tool`.
+4. **Audio/Video:** If the task involves a spoken media file, use `transcribe_audio_tool` to get the transcript.
+5. **Web Scraping:** Use `read_webpage_clean_tool` to fetch clean text from URLs if standard web browsing fails.
+6. **Self-Correction:** Do not guess the final answer. If your Python code throws an error, read the traceback carefully, fix the code, and try again.
+
+Finish your answer strictly with the following template: FINAL ANSWER: [YOUR FINAL ANSWER].
+YOUR FINAL ANSWER should be a number OR as few words as possible OR a comma separated list of numbers and/or strings.
+If you are asked for a number, don't use a comma to write your number neither use units such as $ or percent sign unless specified otherwise.
+If you are asked for a string, don't use articles, neither abbreviations (e.g. for cities), and write the digits in plain text unless specified otherwise.
+If you are asked for a comma separated list, apply the above rules depending on whether the element to be put in the list is a number or a string."""
 def create_new_agent():
     return CodeAgent(
-        tools=[search_tool, download_task_file, visit_webpage_tool, analyze_image_tool],
+        tools=[
+            search_tool, 
+            download_task_file, 
+            visit_webpage_tool, 
+            analyze_image_tool, 
+            transcribe_audio_tool, 
+            read_webpage_clean_tool
+        ],
         model=model,
         instructions=manager_system_prompt,  
-        max_steps=5,
-        additional_authorized_imports=["pandas", "openpyxl", "PyPDF2", "json", "csv", "re"] 
+        max_steps=10,
+        additional_authorized_imports=["pandas", "openpyxl", "PyPDF2", "json", "csv", "re", "math", "datetime"] 
     )
+
+#--- run ---
+
 def run_and_submit_all( profile: gr.OAuthProfile | None):
     """
     Fetches all questions, runs the manager_agent on them, submits all answers,