# image source: https://www.globesign.com/blog/a-beginners-guide-to-google-website-analyzer/
#ref: i) https://blog.dailydoseofds.com/p/building-a-multi-agent-internet-research , ii) https://blog.dailydoseofds.com/p/build-a-multi-agent-research-assistant
import streamlit as st
from swarm import Swarm, Agent
from bs4 import BeautifulSoup
import requests
import os
import io
from reportlab.pdfgen import canvas
# Function to fetch OpenAI API key
def fetch_openai_api_key():
"""Fetch the OpenAI API key from Hugging Face secrets."""
try:
secret_key = st.secrets.get("OPENAI_API_KEY", "")
if secret_key:
os.environ['OPENAI_API_KEY'] = secret_key
else:
st.warning("⚠️ OpenAI API Key is missing! Please check your Hugging Face secrets configuration.")
except Exception as e:
st.error(f"Error retrieving OpenAI API Key: {str(e)}")
# Initialize the Swarm client
def initialize_swarm_client():
return Swarm()
# Define the scraping function
def scrape_website(url):
"""Scrapes the content of the website."""
try:
response = requests.get(url)
response.raise_for_status()
soup = BeautifulSoup(response.text, 'html.parser')
return soup.get_text() # Return the text content from the HTML
except requests.exceptions.RequestException as e:
return f"Error during scraping: {str(e)}"
# Scraper Agent
scraper_agent = Agent(
name="Scraper Agent",
instructions="You are an agent that scrapes content from websites.",
functions=[scrape_website]
)
# Define the analysis function
def analyze_content(content):
"""Dynamically analyzes the content and extracts key insights."""
# Generic analysis prompt for flexibility
summary = (
"📜 Final Report:\n\n"
"Based on the website content, here are the key takeaways:\n\n"
"Offerings and Highlights:\n"
"- Summarize key offerings, products, or services.\n\n"
"Prominent Features:\n"
"- Identify any standout features or unique aspects.\n\n"
"Additional Notes:\n"
"- Provide other insights that might be useful for the user."
)
return summary
# Research Agent
research_agent = Agent(
name="Research Agent",
instructions="You are an agent that highlights key insights by dynamically analyzing content and adapting to the available information.",
functions=[analyze_content]
)
# Define the writing function
def write_summary(context_variables):
"""Writes a summary based on the analysis."""
analysis = context_variables.get('analysis', '')
summary = f"Here's a detailed report based on the research: {analysis}"
return summary
# Writer Agent
writer_agent = Agent(
name="Writer Agent",
instructions="You are an agent that writes summaries of research.",
functions=[write_summary]
)
# Orchestrate the workflow
def orchestrate_workflow(client, url):
# Step 1: Scrape the website
scrape_result = client.run(
agent=scraper_agent,
messages=[{"role": "user", "content": f"Scrape the following website: {url}"}]
)
scraped_content = scrape_result.messages[-1]["content"]
# Check for any error during scraping
if "Error during scraping" in scraped_content:
return scraped_content
# Step 2: Analyze the scraped content
research_result = client.run(
agent=research_agent,
messages=[{"role": "user", "content": f"Analyze the following content: {scraped_content}"}]
)
analysis_summary = research_result.messages[-1]["content"]
# Step 3: Write the summary based on the analysis
writer_result = client.run(
agent=writer_agent,
messages=[{"role": "user", "content": f"Write a summary based on this analysis: {analysis_summary}"}],
context_variables={"analysis": analysis_summary}
)
final_summary = writer_result.messages[-1]["content"]
return final_summary
# Helper functions to create text and PDF files
def create_text_file(content):
"""Create a downloadable text file."""
return content # Return plain text
def create_pdf_file(content):
"""Create a downloadable PDF file."""
buffer = io.BytesIO()
c = canvas.Canvas(buffer)
c.drawString(100, 750, "Generated Report")
c.drawString(100, 730, "--------------------")
lines = content.split("\n")
y = 700
for line in lines:
if y < 50: # Create a new page if the content overflows
c.showPage()
y = 750
c.drawString(100, y, line)
y -= 20
c.save()
buffer.seek(0)
return buffer.getvalue() # Return binary content
# Streamlit App UI
st.markdown(
"""
""",
unsafe_allow_html=True,
)
# 1. Add the title at the top
st.markdown('Swarm-based Web Content Analyzer 🧐
', unsafe_allow_html=True)
# 2. Add the description below the title
st.markdown('Effortlessly extract, analyze, and summarize web content using multi-agent.
', unsafe_allow_html=True)
# 3. Add the image below the description
st.image("./image-4.png", use_container_width=True)
# 4. Add Acknowledgement
st.markdown(
"""
""",
unsafe_allow_html=True
)
# 5. Add one line-spacing after the acknowledgment
st.markdown('', unsafe_allow_html=True)
fetch_openai_api_key()
if 'OPENAI_API_KEY' in os.environ and os.environ['OPENAI_API_KEY']:
client = initialize_swarm_client()
# Add interface for URL input
st.subheader("Enter the Website URL 🔗")
url = st.text_input("Enter the URL of the website you want to scrape", placeholder="https://huggingface.co/models")
# Add some spacing
st.markdown('', unsafe_allow_html=True)
# Add the "Run Workflow" button
if st.button("🚀 Run Workflow", key="run"):
if url:
with st.spinner("Running the multi-agent workflow... This may take a moment."):
final_report = orchestrate_workflow(client, url)
st.success("✅ Workflow complete!")
st.write("### 📜 Final Report:")
st.write(final_report)
# Add download buttons for the report
text_file = create_text_file(final_report)
pdf_file = create_pdf_file(final_report)
st.download_button(
label="Download Report as Text",
data=text_file,
file_name="report.txt",
mime="text/plain",
)
st.download_button(
label="Download Report as PDF",
data=pdf_file,
file_name="report.pdf",
mime="application/pdf",
)
else:
st.error("❌ Please enter a valid URL.")
else:
st.sidebar.warning("⚠️ OpenAI API Key not set. Please check your Hugging Face secrets configuration.")