app.py

from smolagents import CodeAgent,DuckDuckGoSearchTool, HfApiModel,load_tool,tool
import datetime
import requests
import yaml
from tools.final_answer import FinalAnswerTool
from bs4 import BeautifulSoup
import pandas as pd
import time
import datetime
import random
from requests.adapters import HTTPAdapter
from urllib3.util.retry import Retry

from Gradio_UI import GradioUI


# ✅ Tool wrapper function for SmolAgent
@tool
def scrape_drug_reviews_tool(drug_name: str, max_pages: int = 3) -> dict :
    """
    Scrapes reviews from the website Drugs.com using Playwright for a given drug name.
    Args:
        drug_name: the name of the target drug for which I want to retrieve reviews,
        max_pages: the number of pages of reviews from Drugs.com that I want to collect

    Output: a dictionary url:review mapping the url of a review to the text of the review
    """
    try:
        df = scrape_drugs_com_reviews_requests(drug_name, max_pages)
        return df.to_dict(orient="records")
    except Exception as e:
        return {"error": str(e)}


        
# List of User-Agents for rotation
USER_AGENTS = [
    "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/121 Safari/537.36",
    "Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/120 Safari/537.36",
    "Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/119 Safari/537.36",
    "Mozilla/5.0 (Windows NT 10.0; Win64; x64; rv:115.0) Gecko/20100101 Firefox/115.0"
]

# Retry logic wrapper
def requests_retry_session(retries=3, backoff_factor=0.5, status_forcelist=(500, 502, 503, 504), session=None):
    session = session or requests.Session()
    retry = Retry(
        total=retries,
        read=retries,
        connect=retries,
        backoff_factor=backoff_factor,
        status_forcelist=status_forcelist,
    )
    adapter = HTTPAdapter(max_retries=retry)
    session.mount("http://", adapter)
    session.mount("https://", adapter)
    return session

# Scraper function using requests
def scrape_drugs_com_reviews_requests(drug_name, max_pages=3, delay=2):
    base_url = f"https://www.drugs.com/comments/{drug_name}/"
    all_reviews = []

    session = requests_retry_session()

    for page_num in range(1, max_pages + 1):
        url = base_url if page_num == 1 else f"{base_url}?page={page_num}"
        headers = {"User-Agent": random.choice(USER_AGENTS)}

        try:
            response = session.get(url, headers=headers, timeout=10)
            response.raise_for_status()

            soup = BeautifulSoup(response.text, "html.parser")
            review_blocks = soup.find_all("div", class_="ddc-comment ddc-box ddc-mgb-2")

            if not review_blocks:
                print(f"No reviews found on page {page_num}.")
                break

            for block in review_blocks:
                review_paragraph = block.find("p")
                review_text = None
                if review_paragraph:
                    if review_paragraph.b:
                        review_paragraph.b.extract()  # remove category (e.g., "For Back Pain")
                    review_text = review_paragraph.get_text(strip=True)

                all_reviews.append({
                    "review": review_text,
                    "source": url
                })

            time.sleep(delay)  # Polite delay

        except Exception as e:
            print(f"Error scraping {url}: {e}")
            continue

    return pd.DataFrame(all_reviews)


final_answer = FinalAnswerTool()

# If the agent does not answer, the model is overloaded, please use another model or the following Hugging Face Endpoint that also contains qwen2.5 coder:
# model_id='https://pflgm2locj2t89co.us-east-1.aws.endpoints.huggingface.cloud' 

model = HfApiModel(
max_tokens=2096,
temperature=0.5,
model_id='Qwen/Qwen2.5-Coder-32B-Instruct',# it is possible that this model may be overloaded
custom_role_conversions=None,
)



with open("prompts.yaml", 'r') as stream:
    prompt_templates = yaml.safe_load(stream)
    
agent = CodeAgent(
    model=model,
    tools=[scrape_drug_reviews_tool,final_answer], ## add your tools here (don't remove final answer)
    max_steps=6,
    verbosity_level=1,
    grammar=None,
    planning_interval=None,
    name="DrugReviewScraperAgent",
    description="Agent that can scrape drug reviews and analyze causal relations",
    prompt_templates=prompt_templates
)


GradioUI(agent).launch()