agent with search, file read, youtube

.gitignore

@@ -0,0 +1,4 @@
+.env
+__pycache__/
+*.py[cod]
+gaia_files

agent.py

@@ -0,0 +1,69 @@
+from dotenv import load_dotenv
+from smolagents import CodeAgent
+from smolagents import OpenAIServerModel
+from tool import fetch_webpage, read_file_tool, get_youtube_transcript
+
+from smolagents import VisitWebpageTool, WikipediaSearchTool, PythonInterpreterTool, DuckDuckGoSearchTool, WebSearchTool, SpeechToTextTool
+
+from prompt import gaia_prompt
+
+load_dotenv()
+
+openai_nano_model = OpenAIServerModel(
+    model_id="gpt-4.1-nano-2025-04-14",
+    # model_id="o3-mini-2025-01-31",
+)
+
+gaia_agent = CodeAgent(
+    model=openai_nano_model,
+    tools=[fetch_webpage, DuckDuckGoSearchTool(), PythonInterpreterTool(), read_file_tool, get_youtube_transcript], # WikipediaSearchTool(), VisitWebpageTool(max_output_length=60000)
+    max_steps=5,
+    verbosity_level=2,
+    additional_authorized_imports=["requests", "bs4", "pandas", "numpy", "markdownify"]
+)
+
+class GAIA_Agent:
+    def __init__(self):
+        self.system_prompt = gaia_prompt
+        self.agent = gaia_agent
+
+    def __call__(self, question: str) -> str:
+
+        try:
+
+            full_context = self.system_prompt + "\nTHE QUESTION:\n" + question
+
+            final_answer = self.agent.run(full_context)
+            return final_answer
+        except Exception as e:
+            error = f"An error occurred while processing the question: {e}"
+            print(error)
+            return error
+
+# build context + append instructions and all
+
+# clean answer function
+
+if __name__ == "__main__":
+    pass
+    # gaia_agent.run("What is the weather in Mumbai?")
+#     answer = gaia_agent.run(
+# f"""
+# You are a general AI assistant. I will ask you a question. You can 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. Remember: GAIA requires exact answer matching. Just provide the factual answer.
+
+# How many studio albums were published by Mercedes Sosa between 2000 and 2009 (included)? You can use the latest 2022 version of english wikipedia.
+# """
+#     )
+
+    # print(gaia_prompt)
+#     answer = gaia_agent.run("""
+#         You are a general AI assistant. I will ask you a question. Report your thoughts, and finish your answer 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 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 can search for results and then visit a webpage to get more information. Break down the problem into smaller sub-problems and solve them one by one.
+#         Think like a human.
+
+#         What is the final numeric output from the attached Python code?
+                            
+#         ----
+
+# """)
+#     print(f"this is the final answer the gaia agent gave ---> {answer}")

app.py

@@ -3,6 +3,8 @@ import gradio as gr
 import requests
 import inspect
 import pandas as pd
+from utils.fetch_file import download_file_if_any
+from agent import GAIA_Agent
 
 # (Keep Constants as is)
 # --- Constants ---
@@ -37,10 +39,12 @@ def run_and_submit_all( profile: gr.OAuthProfile | None):
     api_url = DEFAULT_API_URL
     questions_url = f"{api_url}/questions"
     submit_url = f"{api_url}/submit"
+    file_path = f"{api_url}/files"
 
     # 1. Instantiate Agent ( modify this part to create your agent)
     try:
-        agent = BasicAgent()
+        # agent = BasicAgent()
+        agent = GAIA_Agent()
     except Exception as e:
         print(f"Error instantiating agent: {e}")
         return f"Error initializing agent: {e}", None
@@ -73,19 +77,39 @@ def run_and_submit_all( profile: gr.OAuthProfile | None):
     results_log = []
     answers_payload = []
     print(f"Running agent on {len(questions_data)} questions...")
-    for item in questions_data:
-        task_id = item.get("task_id")
-        question_text = item.get("question")
-        if not task_id or question_text is None:
-            print(f"Skipping item with missing task_id or question: {item}")
-            continue
-        try:
-            submitted_answer = agent(question_text)
-            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})
-        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}"})
+    to_answer = [1, 3, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20]
+    for index, item in enumerate(questions_data):
+
+        if index + 1 in to_answer:
+            task_id = item.get("task_id")
+            question_text = item.get("question")
+            file_name = item.get('file_name')
+            file_path = None
+
+            if file_name:
+                try:
+                    file_path = download_file_if_any(task_id)
+                except Exception as e:
+                    file_path = None
+
+            if not task_id or question_text is None:
+                print(f"Skipping item with missing task_id or question: {item}")
+                continue
+            try:
+                agent_question = question_text
+                if file_path:
+                    agent_question += f"\n\nA file was downloaded for this task and saved locally at:\n {file_path}\n"
+
+                submitted_answer = agent(agent_question)
+
+                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})
+            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}"})
+
+        else:
+            print("Do not answer")
 
     if not answers_payload:
         print("Agent did not produce any answers to submit.")
@@ -95,6 +119,7 @@ def run_and_submit_all( profile: gr.OAuthProfile | None):
     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}'..."
     print(status_update)
+    print(submission_data)
 
     # 5. Submit
     print(f"Submitting {len(answers_payload)} answers to: {submit_url}")

prompt.py

@@ -0,0 +1,72 @@
+
+
+final_answer = """When answering, provide ONLY the precise answer requested.
+Do not include explanations, steps, reasoning, or additional text. 
+Be direct and specific. GAIA benchmark requires exact matching answers.
+"""
+
+final_answer1 = prompt = """
+    Here is a user-given task and the agent steps: {agent_memory.get_succinct_steps()}. Now here is the FINAL ANSWER that was given: 
+    {final_answer}
+    Ensure the FINAL ANSWER is in the right format as asked for by the task.  Here are the instructions that you need to evaluate:
+    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 commas to write your number.  Don't use units such as $ or percent sign unless specified otherwise. Write your number in Arabic numbers (such as 9 or 3 or 1093) unless specified otherwise.
+    If you are asked for a currency in your answer, use the symbol for that currency.  For example, if you are asked for the answers in USD, an example answer would be $40.00
+    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.
+    If you are asked for a comma separated list, ensure you only return the content of that list, and NOT the brackets '[]'
+    First list reasons why it is/is not in the correct format and then write your final decision: PASS in caps lock if it is satisfactory, FAIL if it is not.
+    """
+
+
+sys_prompt = """You are the top agent of a multi-agent system that can answer questions by coordinating the work of other agents.
+You will receive a question and you will decide which agent to use to answer it.
+You can use the web_agent to search the web for information and for fetching the content of a web page, or the audiovideo_agent to extract information from video or audio files.
+You can also use your own knowledge to answer the question.
+You need to respect the output format that is given to you.
+Finding the correct answer to the question need reasoning and plannig, read the question carrefully, think step by step and do not skip any steps.
+"""
+
+sys_prompt1 = """
+You are a helpful assistant tasked with answering questions using a set of tools.
+
+Your final answer must strictly follow this format:
+FINAL ANSWER: [ANSWER]
+
+Only write the answer in that exact format. Do not explain anything. Do not include any other text.
+
+If you are provided with a similar question and its final answer, and the current question is **exactly the same**, then simply return the same final answer without using any tools.
+
+Only use tools if the current question is different from the similar one.
+
+Examples:
+- FINAL ANSWER: FunkMonk
+- FINAL ANSWER: Paris
+- FINAL ANSWER: 128
+
+If you do not follow this format exactly, your response will be considered incorrect."""
+
+
+gaia_prompt = """
+You are a helpful general AI assistant. You are tasked with answering questions from GAIA benchmark.
+You can answer questions using a set of tools or you can also use your own knowledge to answer them.
+
+Finding the correct answer to the question need reasoning and plannig, read the question carrefully, think step by step and do not skip any steps.
+You have access to various tools, including, but not limited to, web search, visiting webpage, executing python code, reading files, transcribing audio, analysing images.
+Break down the problem into smaller sub-problems and solve them one by one.
+
+If the question has associated file, you can be use it to answer the question.
+
+Your final answer must strictly follow this format:
+[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.
+Reminder! When giving final answer, provide ONLY the precise answer requested. Do not include explanations, steps, reasoning, or additional text. Be direct and specific. 
+GAIA benchmark requires exact matching answers. If you do not follow this format exactly, your response will be considered incorrect.
+
+BEFORE GIVING THE FINAL ANSWER DOUBLE CHECK THE EXACT FORMAT IN WHICH THE ANSWER IS NEEDED.
+
+"""

tool.py

@@ -0,0 +1,81 @@
+from smolagents import Tool, tool
+from youtube_transcript_api import YouTubeTranscriptApi
+
+@tool
+def fetch_webpage(url: str, convert_to_markdown: bool = True) -> str:
+    """
+    Visit a website / url and fetch the content of the webpage. 
+    if markdown conversion is enabled, it will remove script and style and return the text content as markdown else return raw unfiltered HTML
+    Args:
+        url (str): The URL to fetch.
+        convert_to_markdown (bool): If True, convert the HTML content to Markdown format. else return the raw HTML.
+    Returns:
+        str: The HTML content of the URL.
+    """
+    import requests
+    from bs4 import BeautifulSoup
+    from markdownify import markdownify as md
+
+    content = None
+    headers = {
+        'User-Agent': 'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/91.0.4472.124 Safari/537.36'
+    }
+    response = requests.get(url, timeout=30, headers=headers)
+    
+    if (convert_to_markdown):
+        soup = BeautifulSoup(response.text, "html.parser")
+
+        # remove script and style tags
+        for script in soup(["script", "style"]):
+            script.extract()
+
+        # for wikipedia only keep the main content
+        if "wikipedia.org" in url:
+            main_content = soup.find("main",{"id":"content"})
+            if main_content:
+                content = md(str(main_content),strip=['script', 'style'], heading_style="ATX").strip()
+            else:
+                content = md(response.text,strip=['script', 'style'], heading_style="ATX").strip()
+        else:
+            # Fallback for all other sites - from chatgpt - not tested
+            content = md(str(soup), strip=['script', 'style'], heading_style="ATX").strip()
+    else:
+        content = response.text
+           
+    return content
+
+
+@tool
+def read_file_tool(file_path: str) -> str:
+    """
+    Tool to read a file and return its content.
+
+    Args:
+        file_path (str): Path to the file to read.
+
+    Returns:
+        str: Content of the file or error message.
+    """
+    try:
+        with open(file_path, "r") as file:
+            return file.read()
+    except Exception as e:
+        return f"Error reading file: {str(e)}"
+    
+
+@tool
+def get_youtube_transcript(video_id: str) -> str:
+    """
+    Fetches the transcript of a YouTube video given its video ID.    
+    Args:
+        video_id (str): The ID of the YouTube video. Pass in the video ID, NOT the video URL. For a video with the URL https://www.youtube.com/watch?v=12345 the ID is 12345.
+    Returns:
+        str: The transcript of the YouTube video. as a single string with each line separated by a newline character.
+    """
+    # Initialize the YouTubeTranscriptApi
+    ytt_api = YouTubeTranscriptApi()
+    fetched_transcript = ytt_api.fetch(video_id)
+    raw_data = fetched_transcript.to_raw_data()
+    # raw data is in the form of [{        'text': 'Hey there',        'start': 0.0,        'duration': 1.54    },    {        'text': 'how are you',,        'start': 1.54,        'duration': 4.16    },   ... ] we will return ony the text element as lines
+    transcript = "\n".join([item['text'] for item in raw_data])
+    return transcript

tools/web_search.py

@@ -0,0 +1,47 @@
+from smolagents import tool
+
+@tool
+def fetch_webpage(url: str, convert_to_markdown: bool = True) -> str:
+    """
+    Visits a website and fetches the content of a given URL / webpage. 
+    if markdown conversion is enabled, it will remove script and style and return the text content as markdown else return raw unfiltered HTML
+    Args:
+        url (str): The URL to fetch.
+        convert_to_markdown (bool): If True, convert the HTML content to Markdown format. else return the raw HTML.
+    Returns:
+        str: The HTML content of the URL.
+    """
+    import requests
+    from bs4 import BeautifulSoup
+    from markdownify import markdownify as md
+
+    content = None
+    headers = {
+        'User-Agent': 'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/91.0.4472.124 Safari/537.36'
+    }
+    response = requests.get(url, timeout=30, headers=headers)
+    # print(response.text)
+    if response.text is not None:
+        print("not none")
+    if (convert_to_markdown):
+        soup = BeautifulSoup(response.text, "html.parser")
+        # remove script and style tags
+        for script in soup(["script", "style"]):
+            script.extract()
+
+        # for wikipedia only keep the main content
+        if "wikipedia.org" in url:
+            main_content = soup.find("main",{"id":"content"})
+            if main_content:
+                content = md(str(main_content),strip=['script', 'style'], heading_style="ATX").strip()
+            else:
+                content = md(response.text,strip=['script', 'style'], heading_style="ATX").strip()
+        else:
+            # Fallback for all other sites - from chatgpt - not tested
+            content = md(str(soup), strip=['script', 'style'], heading_style="ATX").strip()
+    else:
+        content = response.text
+    
+    # save_file_with_timestamp(content, "webpage", ".md" if convert_to_markdown else ".html")
+           
+    return content

utils/fetch_file.py

@@ -0,0 +1,38 @@
+import requests
+import os
+
+DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
+files_folder = "gaia_files"
+os.makedirs(files_folder, exist_ok=True)
+
+def download_file_if_any(task_id: str) -> str:
+    url = f"{DEFAULT_API_URL}/files/{task_id}"
+    try:
+        response = requests.get(url)
+        response.raise_for_status()
+
+        content_disposition = response.headers.get("content-disposition")
+        filename = content_disposition.split("filename=")[1].strip('"') if content_disposition and "filename=" in content_disposition else None
+
+        if filename:
+            # path relative from app.py
+            file_path = os.path.join(files_folder, filename)
+
+            with open(file_path, "wb") as file:
+                file.write(response.content)
+
+            return str(file_path)
+        else:
+            return None
+            
+    except requests.exceptions.RequestException as e:
+        print(f"Error making request: {e}")
+        return None
+    except ValueError as e:
+        print(f"Error decoding JSON response: {e}")
+        return None
+    
+
+
+if __name__ == "__main__":
+    print(download_file_if_any("f918266a-b3e0-4914-865d-4faa564f1aef"))