app.py

from smolagents import CodeAgent,DuckDuckGoSearchTool, HfApiModel,load_tool,tool
import datetime
import requests
import pytz
import yaml
from tools.final_answer import FinalAnswerTool
from Gradio_UI import GradioUI
from bs4 import BeautifulSoup
import arxiv
from PyPDF2 import PdfReader
from xml.etree import ElementTree
import io
# Below is an example of a tool that does nothing. Amaze us with your creativity !


@tool
def get_top_paper()-> str: 
    """A tool that fetches the most upvoted paper on Hugging Face daily papers.
    """
    url = "https://huggingface.co/papers"
    try:
        res = requests.get(url)
        res.raise_for_status() 
        # Parse the HTML response
        soup = BeautifulSoup(response.text, "html.parser")

        #inspect h3 the selector 
        top_paper = soup.find("h3")

        if top_paper_element:
            return top_paper.text.strip()
        else:
            return "Paper not found"

    except Exception as e:
        return f"Error fetching top paper: {str(e)}"
        
@tool
def get_paper_link(title:str)->str:
    """
    A Tool that finds the Hugging Face paper link given its title.
    Args:
        title: A string representing the title of the paper (eg., 'Competitive Programming with Large Reasoning Models').
    """
    url = "https://huggingface.co/papers"
    
    try:
        res = requests.get(url)
        res.raise_for_status()
        soup = BeautifulSoup(response.text, "html.parser")
        paper_links = soup.find("h3")
        
        for paper in paper_links:
            if paper.text.strip() == title:
                return "https://huggingface.co" + paper["href"] 
        
        return "Paper link not found."

    except Exception as e:
        return f"Error fetching paper link: {str(e)}"
        
@tool    
def get_paper_content(link:str)->str:
    """
    A tool that reads the first four pages of a paper and returns its content as a string given its link.
    Args:
        link: A string representing the URL of the paper (eg., 'https://huggingface.co/papers/2502.06807').
    
    """
    try:
        #Get the id from the Hugging face URL 
        paper_id = link.split("/papers/")[-1]
        paper = next(arxiv.Client().results(arxiv.Search(id_list=[paper_id])))

        #Get the PDF URL of the paper from arXiv
        pdf_url = paper.entry_id.replace("abs", "pdf") + ".pdf"
        response = requests.get(pdf_url)
        response.raise_for_status()
        pdf_buffer = io.BytesIO(response.content)

        # Extract text from the first four pages
        content = ""
        reader = PdfReader(pdf_buffer)
        pages = reader.pages[:4]
        for page in pages:
            content += page.extract_text() or ""  

        return content.strip()
    
    except Exception as e:
        return f"Error reading paper: {str(e)}"

        
@tool 
def get_related_papers(title:str, max_results:int)->list:
    """
    A tool that searches for related papers on arXiv based on the title of the query paper.
    Args:
        title: A string representing the title of the query paper to find related papers for.
        max_results: A integer representing the number of related papers to return.
    Returns:
        list: A list of dictionaries, each containing a related paper's title and URL.
    """
    try:
        search_url = f"http://export.arxiv.org/api/query?search_query=title:{title}&start=0&max_results={max_results}"
        resp = requests.get(search_url)
        if resp.status_code != 200:
            return f"Error: Failed to retrieve papers from arXiv. Status code: {response.status_code}"
        root = ElementTree.fromstring(resp.text)
        papers = []
        
        for entry in root.findall("{http://www.w3.org/2005/Atom}entry"):
            paper_title = entry.find("{http://www.w3.org/2005/Atom}title").text
            paper_url = entry.find("{http://www.w3.org/2005/Atom}id").text
            papers.append({"title": paper_title, "url": paper_url})
        
        return papers
    except Exception as e:
        return f"Error: {str(e)}"
        






final_answer = FinalAnswerTool()
model = HfApiModel(
max_tokens=2096,
temperature=0.5,
model_id='https://wxknx1kg971u7k1n.us-east-1.aws.endpoints.huggingface.cloud',
custom_role_conversions=None,
)




with open("prompts.yaml", 'r') as stream:
    prompt_templates = yaml.safe_load(stream)
    
agent = CodeAgent(
    model=model,
    tools=[final_answer,get_top_paper,get_paper_link,get_paper_content,get_related_papers], ## add your tools here (don't remove final answer)
    max_steps=6,
    verbosity_level=1,
    grammar=None,
    planning_interval=None,
    name=None,
    description=None,
    prompt_templates=prompt_templates
)


GradioUI(agent).launch()