app.py

# app.py

import gradio as gr
from bs4 import BeautifulSoup
from sentence_transformers import SentenceTransformer
import faiss
import numpy as np
import requests
import time
import re
import base64
import logging
import os
import sys

# Import OpenAI library
import openai

# Set up logging to output to the console
logger = logging.getLogger(__name__)
logger.setLevel(logging.INFO)

# Create a console handler
console_handler = logging.StreamHandler(sys.stdout)
console_handler.setLevel(logging.INFO)

# Create a formatter and set it for the handler
formatter = logging.Formatter('%(asctime)s %(levelname)s %(name)s %(message)s')
console_handler.setFormatter(formatter)

# Add the handler to the logger
logger.addHandler(console_handler)

# Initialize models and variables
logger.info("Initializing models and variables")
embedding_model = SentenceTransformer('all-MiniLM-L6-v2')
faiss_index = None
bookmarks = []
fetch_cache = {}

# Define the categories
CATEGORIES = [
    "Social Media",
    "News and Media",
    "Education and Learning",
    "Entertainment",
    "Shopping and E-commerce",
    "Finance and Banking",
    "Technology",
    "Health and Fitness",
    "Travel and Tourism",
    "Food and Recipes",
    "Sports",
    "Arts and Culture",
    "Government and Politics",
    "Business and Economy",
    "Science and Research",
    "Personal Blogs and Journals",
    "Job Search and Careers",
    "Music and Audio",
    "Videos and Movies",
    "Reference and Knowledge Bases",
    "Dead Link",
    "Uncategorized",
]

# Set up Groq Cloud API key and base URL
GROQ_API_KEY = os.getenv('GROQ_API_KEY')

if not GROQ_API_KEY:
    logger.error("GROQ_API_KEY environment variable not set.")

openai.api_key = GROQ_API_KEY
openai.api_base = "https://api.groq.com/openai/v1"

def extract_main_content(soup):
    """
    Extract the main content from a webpage while filtering out boilerplate content.
    """
    if not soup:
        return ""

    # Remove unwanted elements
    for element in soup(['script', 'style', 'header', 'footer', 'nav', 'aside', 'form', 'noscript']):
        element.decompose()

    # Extract text from <p> tags
    p_tags = soup.find_all('p')
    if p_tags:
        content = ' '.join([p.get_text(strip=True, separator=' ') for p in p_tags])
    else:
        # Fallback to body content
        content = soup.get_text(separator=' ', strip=True)

    # Clean up the text
    content = re.sub(r'\s+', ' ', content)  # Remove multiple spaces

    # Truncate content to a reasonable length (e.g., 1500 words)
    words = content.split()
    if len(words) > 1500:
        content = ' '.join(words[:1500])

    return content

def get_page_metadata(soup):
    """
    Extract metadata from the webpage including title, description, and keywords.
    """
    metadata = {
        'title': '',
        'description': '',
        'keywords': ''
    }

    if not soup:
        return metadata

    # Get title
    title_tag = soup.find('title')
    if title_tag and title_tag.string:
        metadata['title'] = title_tag.string.strip()

    # Get meta description
    meta_desc = (
        soup.find('meta', attrs={'name': 'description'}) or
        soup.find('meta', attrs={'property': 'og:description'}) or
        soup.find('meta', attrs={'name': 'twitter:description'})
    )
    if meta_desc:
        metadata['description'] = meta_desc.get('content', '').strip()

    # Get meta keywords
    meta_keywords = soup.find('meta', attrs={'name': 'keywords'})
    if meta_keywords:
        metadata['keywords'] = meta_keywords.get('content', '').strip()

    # Get OG title if main title is empty
    if not metadata['title']:
        og_title = soup.find('meta', attrs={'property': 'og:title'})
        if og_title:
            metadata['title'] = og_title.get('content', '').strip()

    return metadata

def generate_summary(bookmark):
    """
    Generate a concise summary for a bookmark using available content and LLM via the Groq Cloud API.
    """
    logger.info(f"Generating summary for bookmark: {bookmark.get('url')}")

    try:
        html_content = bookmark.get('html_content', '')

        # Get the HTML soup object from the bookmark
        soup = BeautifulSoup(html_content, 'html.parser')

        # Extract metadata and main content
        metadata = get_page_metadata(soup)
        main_content = extract_main_content(soup)

        # Prepare content for the prompt
        content_parts = []
        if metadata['title']:
            content_parts.append(f"Title: {metadata['title']}")
        if metadata['description']:
            content_parts.append(f"Description: {metadata['description']}")
        if metadata['keywords']:
            content_parts.append(f"Keywords: {metadata['keywords']}")
        if main_content:
            content_parts.append(f"Main Content: {main_content}")

        content_text = '\n'.join(content_parts)

        # Detect insufficient or erroneous content
        error_keywords = ['Access Denied', 'Security Check', 'Cloudflare', 'captcha', 'unusual traffic']
        if not content_text or len(content_text.split()) < 50:
            use_prior_knowledge = True
            logger.info(f"Content for {bookmark.get('url')} is insufficient. Instructing LLM to use prior knowledge.")
        elif any(keyword.lower() in content_text.lower() for keyword in error_keywords):
            use_prior_knowledge = True
            logger.info(f"Content for {bookmark.get('url')} contains error messages. Instructing LLM to use prior knowledge.")
        else:
            use_prior_knowledge = False

        if use_prior_knowledge:
            # Construct prompt to use prior knowledge
            prompt = f"""
You are a knowledgeable assistant with up-to-date information as of 2023.

The user provided a URL: {bookmark.get('url')}

Please provide a concise summary in **no more than two sentences** about this website.

Focus on:
- The main purpose or topic of the website.
- Key information or features.

Be concise and objective.
"""
        else:
            # Construct the prompt with the extracted content
            prompt = f"""
You are a helpful assistant that creates concise webpage summaries.

Analyze the following webpage content:

{content_text}

Provide a concise summary in **no more than two sentences** focusing on:
- The main purpose or topic of the page.
- Key information or features.

Be concise and objective.
"""

        # Call the LLM via Groq Cloud API
        response = openai.ChatCompletion.create(
            model='llama-3.1-70b-versatile',
            messages=[
                {"role": "user", "content": prompt}
            ],
            max_tokens=100,
            temperature=0.5,
        )
        summary = response['choices'][0]['message']['content'].strip()
        if not summary:
            raise ValueError("Empty summary received from the model.")
        logger.info("Successfully generated LLM summary")
        bookmark['summary'] = summary
        time.sleep(3)  # Wait to respect rate limits

    except Exception as e:
        logger.error(f"Error generating summary: {e}", exc_info=True)
        bookmark['summary'] = 'No summary available.'

def assign_category(bookmark):
    """
    Assign a category to a bookmark using the LLM based on its summary via the Groq Cloud API.
    """
    if bookmark.get('dead_link'):
        bookmark['category'] = 'Dead Link'
        logger.info(f"Assigned category 'Dead Link' to bookmark: {bookmark.get('url')}")
        return

    summary = bookmark.get('summary', '')
    if not summary:
        bookmark['category'] = 'Uncategorized'
        return

    # Prepare the prompt
    categories_str = ', '.join([f'"{cat}"' for cat in CATEGORIES if cat != 'Dead Link'])
    prompt = f"""
You are a helpful assistant that categorizes webpages.

Based on the following summary, assign the most appropriate category from the list below.

Summary:
{summary}

Categories:
{categories_str}

Respond with only the category name.
"""

    try:
        response = openai.ChatCompletion.create(
            model='llama-3.1-70b-versatile',
            messages=[
                {"role": "user", "content": prompt}
            ],
            max_tokens=10,
            temperature=0,
        )
        category = response['choices'][0]['message']['content'].strip().strip('"')
        # Validate the category
        if category in CATEGORIES:
            bookmark['category'] = category
            logger.info(f"Assigned category '{category}' to bookmark: {bookmark.get('url')}")
        else:
            bookmark['category'] = 'Uncategorized'
            logger.warning(f"Invalid category '{category}' returned by LLM for bookmark: {bookmark.get('url')}")
        time.sleep(3)  # Wait to respect rate limits

    except Exception as e:
        logger.error(f"Error assigning category: {e}", exc_info=True)
        bookmark['category'] = 'Uncategorized'

def parse_bookmarks(file_content):
    """
    Parse bookmarks from HTML file.
    """
    logger.info("Parsing bookmarks")
    try:
        soup = BeautifulSoup(file_content, 'html.parser')
        extracted_bookmarks = []
        for link in soup.find_all('a'):
            url = link.get('href')
            title = link.text.strip()
            if url and title:
                extracted_bookmarks.append({'url': url, 'title': title})
        logger.info(f"Extracted {len(extracted_bookmarks)} bookmarks")
        return extracted_bookmarks
    except Exception as e:
        logger.error("Error parsing bookmarks: %s", e, exc_info=True)
        raise

def fetch_url_info(bookmark):
    """
    Fetch information about a URL.
    """
    url = bookmark['url']
    if url in fetch_cache:
        bookmark.update(fetch_cache[url])
        return

    try:
        logger.info(f"Fetching URL info for: {url}")
        headers = {
            'User-Agent': 'Mozilla/5.0',
            'Accept-Language': 'en-US,en;q=0.9',
        }
        response = requests.get(url, headers=headers, timeout=5, verify=False, allow_redirects=True)
        bookmark['etag'] = response.headers.get('ETag', 'N/A')
        bookmark['status_code'] = response.status_code

        content = response.text
        logger.info(f"Fetched content length for {url}: {len(content)} characters")

        # Handle status codes
        if response.status_code >= 500:
            # Server error, consider as dead link
            bookmark['dead_link'] = True
            bookmark['description'] = ''
            bookmark['html_content'] = ''
            logger.warning(f"Dead link detected: {url} with status {response.status_code}")
        else:
            bookmark['dead_link'] = False
            bookmark['html_content'] = content
            bookmark['description'] = ''
            logger.info(f"Fetched information for {url}")

    except requests.exceptions.Timeout:
        bookmark['dead_link'] = False  # Mark as 'Unknown' instead of 'Dead'
        bookmark['etag'] = 'N/A'
        bookmark['status_code'] = 'Timeout'
        bookmark['description'] = ''
        bookmark['html_content'] = ''
        bookmark['slow_link'] = True  # Custom flag to indicate slow response
        logger.warning(f"Timeout while fetching {url}. Marking as 'Slow'.")
    except Exception as e:
        bookmark['dead_link'] = True
        bookmark['etag'] = 'N/A'
        bookmark['status_code'] = 'Error'
        bookmark['description'] = ''
        bookmark['html_content'] = ''
        logger.error(f"Error fetching URL info for {url}: {e}", exc_info=True)
    finally:
        fetch_cache[url] = {
            'etag': bookmark.get('etag'),
            'status_code': bookmark.get('status_code'),
            'dead_link': bookmark.get('dead_link'),
            'description': bookmark.get('description'),
            'html_content': bookmark.get('html_content', ''),
            'slow_link': bookmark.get('slow_link', False),
        }

def vectorize_and_index(bookmarks_list):
    """
    Create vector embeddings for bookmarks and build FAISS index with ID mapping.
    """
    logger.info("Vectorizing summaries and building FAISS index")
    try:
        summaries = [bookmark['summary'] for bookmark in bookmarks_list]
        embeddings = embedding_model.encode(summaries)
        dimension = embeddings.shape[1]
        index = faiss.IndexIDMap(faiss.IndexFlatL2(dimension))
        # Assign unique IDs to each bookmark
        ids = np.array([bookmark['id'] for bookmark in bookmarks_list], dtype=np.int64)
        index.add_with_ids(np.array(embeddings).astype('float32'), ids)
        logger.info("FAISS index built successfully with IDs")
        return index
    except Exception as e:
        logger.error(f"Error in vectorizing and indexing: {e}", exc_info=True)
        raise

def display_bookmarks():
    """
    Generate HTML display for bookmarks.
    """
    logger.info("Generating HTML display for bookmarks")
    cards = ''
    for i, bookmark in enumerate(bookmarks):
        index = i + 1
        if bookmark.get('dead_link'):
            status = "❌ Dead Link"
            card_style = "border: 2px solid red;"
            text_style = "color: red;"
        elif bookmark.get('slow_link'):
            status = "⏳ Slow Response"
            card_style = "border: 2px solid orange;"
            text_style = "color: orange;"
        else:
            status = "✅ Active"
            card_style = "border: 2px solid green;"
            text_style = "color: black;"

        title = bookmark['title']
        url = bookmark['url']
        etag = bookmark.get('etag', 'N/A')
        summary = bookmark.get('summary', '')
        category = bookmark.get('category', 'Uncategorized')

        # Escape HTML content to prevent XSS attacks
        from html import escape
        title = escape(title)
        url = escape(url)
        summary = escape(summary)
        category = escape(category)

        card_html = f'''
        <div class="card" style="{card_style}; padding: 10px; margin: 10px; border-radius: 5px;">
            <div class="card-content">
                <h3 style="{text_style}">{index}. {title} {status}</h3>
                <p style="{text_style}"><strong>Category:</strong> {category}</p>
                <p style="{text_style}"><strong>URL:</strong> <a href="{url}" target="_blank" style="{text_style}">{url}</a></p>
                <p style="{text_style}"><strong>ETag:</strong> {etag}</p>
                <p style="{text_style}"><strong>Summary:</strong> {summary}</p>
            </div>
        </div>
        '''
        cards += card_html
    logger.info("HTML display generated")
    return cards

def process_uploaded_file(file):
    """
    Process the uploaded bookmarks file.
    """
    global bookmarks, faiss_index
    logger.info("Processing uploaded file")

    if file is None:
        logger.warning("No file uploaded")
        return "Please upload a bookmarks HTML file.", '', gr.update(choices=[]), display_bookmarks()

    try:
        file_content = file.decode('utf-8')
    except UnicodeDecodeError as e:
        logger.error(f"Error decoding the file: {e}", exc_info=True)
        return "Error decoding the file. Please ensure it's a valid HTML file.", '', gr.update(choices=[]), display_bookmarks()

    try:
        bookmarks = parse_bookmarks(file_content)
    except Exception as e:
        logger.error(f"Error parsing bookmarks: {e}", exc_info=True)
        return "Error parsing the bookmarks HTML file.", '', gr.update(choices=[]), display_bookmarks()

    if not bookmarks:
        logger.warning("No bookmarks found in the uploaded file")
        return "No bookmarks found in the uploaded file.", '', gr.update(choices=[]), display_bookmarks()

    # Assign unique IDs to bookmarks
    for idx, bookmark in enumerate(bookmarks):
        bookmark['id'] = idx

    # Fetch bookmark info sequentially
    for bookmark in bookmarks:
        fetch_url_info(bookmark)

    # Process bookmarks sequentially with LLM
    for bookmark in bookmarks:
        generate_summary(bookmark)
        assign_category(bookmark)

    try:
        faiss_index = vectorize_and_index(bookmarks)
    except Exception as e:
        logger.error(f"Error building FAISS index: {e}", exc_info=True)
        return "Error building search index.", '', gr.update(choices=[]), display_bookmarks()

    message = f"✅ Successfully processed {len(bookmarks)} bookmarks."
    logger.info(message)

    # Generate displays and updates
    bookmark_html = display_bookmarks()
    choices = [f"{i+1}. {bookmark['title']} (Category: {bookmark['category']})"
               for i, bookmark in enumerate(bookmarks)]

    return message, bookmark_html, gr.update(choices=choices), bookmark_html

def delete_selected_bookmarks(selected_indices):
    """
    Delete selected bookmarks and remove their vectors from the FAISS index.
    """
    global bookmarks, faiss_index
    if not selected_indices:
        return "⚠️ No bookmarks selected.", gr.update(choices=[]), display_bookmarks()

    ids_to_delete = []
    indices_to_delete = []
    for s in selected_indices:
        idx = int(s.split('.')[0]) - 1
        if 0 <= idx < len(bookmarks):
            bookmark_id = bookmarks[idx]['id']
            ids_to_delete.append(bookmark_id)
            indices_to_delete.append(idx)
            logger.info(f"Deleting bookmark at index {idx + 1}")

    # Remove vectors from FAISS index
    if faiss_index is not None and ids_to_delete:
        faiss_index.remove_ids(np.array(ids_to_delete, dtype=np.int64))

    # Remove bookmarks from the list (reverse order to avoid index shifting)
    for idx in sorted(indices_to_delete, reverse=True):
        bookmarks.pop(idx)

    message = "🗑️ Selected bookmarks deleted successfully."
    logger.info(message)
    choices = [f"{i+1}. {bookmark['title']} (Category: {bookmark['category']})"
               for i, bookmark in enumerate(bookmarks)]

    return message, gr.update(choices=choices), display_bookmarks()

def edit_selected_bookmarks_category(selected_indices, new_category):
    """
    Edit category of selected bookmarks.
    """
    if not selected_indices:
        return "⚠️ No bookmarks selected.", gr.update(choices=[]), display_bookmarks()
    if not new_category:
        return "⚠️ No new category selected.", gr.update(choices=[]), display_bookmarks()

    indices = [int(s.split('.')[0])-1 for s in selected_indices]
    for idx in indices:
        if 0 <= idx < len(bookmarks):
            bookmarks[idx]['category'] = new_category
            logger.info(f"Updated category for bookmark {idx + 1} to {new_category}")

    message = "✏️ Category updated for selected bookmarks."
    logger.info(message)

    # Update choices and display
    choices = [f"{i+1}. {bookmark['title']} (Category: {bookmark['category']})"
               for i, bookmark in enumerate(bookmarks)]

    return message, gr.update(choices=choices), display_bookmarks()

def export_bookmarks():
    """
    Export bookmarks to HTML file.
    """
    if not bookmarks:
        logger.warning("No bookmarks to export")
        return "⚠️ No bookmarks to export."

    try:
        logger.info("Exporting bookmarks to HTML")
        soup = BeautifulSoup("<!DOCTYPE NETSCAPE-Bookmark-file-1><Title>Bookmarks</Title><H1>Bookmarks</H1>", 'html.parser')
        dl = soup.new_tag('DL')
        for bookmark in bookmarks:
            dt = soup.new_tag('DT')
            a = soup.new_tag('A', href=bookmark['url'])
            a.string = bookmark['title']
            dt.append(a)
            dl.append(dt)
        soup.append(dl)
        html_content = str(soup)
        b64 = base64.b64encode(html_content.encode()).decode()
        href = f'data:text/html;base64,{b64}'
        logger.info("Bookmarks exported successfully")
        return f'<a href="{href}" download="bookmarks.html">💾 Download Exported Bookmarks</a>'
    except Exception as e:
        logger.error(f"Error exporting bookmarks: {e}", exc_info=True)
        return "⚠️ Error exporting bookmarks."

def chatbot_response(user_query):
    """
    Generate chatbot response using the FAISS index and embeddings.
    """
    if not bookmarks or faiss_index is None:
        logger.warning("No bookmarks available for chatbot")
        return "⚠️ No bookmarks available. Please upload and process your bookmarks first."

    logger.info(f"Chatbot received query: {user_query}")

    try:
        # Encode the user query
        query_vector = embedding_model.encode([user_query]).astype('float32')

        # Search the FAISS index
        k = 5  # Number of results to return
        distances, ids = faiss_index.search(query_vector, k)
        ids = ids.flatten()

        # Retrieve the bookmarks
        id_to_bookmark = {bookmark['id']: bookmark for bookmark in bookmarks}
        matching_bookmarks = [id_to_bookmark.get(id) for id in ids if id in id_to_bookmark]

        if not matching_bookmarks:
            return "No relevant bookmarks found for your query."

        # Format the response
        bookmarks_info = "\n".join([
            f"Title: {bookmark['title']}\nURL: {bookmark['url']}\nSummary: {bookmark['summary']}"
            for bookmark in matching_bookmarks
        ])

        # Use the LLM via Groq Cloud API to generate a response
        prompt = f"""
A user asked: "{user_query}"

Based on the bookmarks below, provide a helpful answer to the user's query, referencing the relevant bookmarks.

Bookmarks:
{bookmarks_info}

Provide a concise and helpful response.
"""

        response = openai.ChatCompletion.create(
            model='llama-3.1-70b-versatile',
            messages=[
                {"role": "user", "content": prompt}
            ],
            max_tokens=500,
            temperature=0.7,
        )
        answer = response['choices'][0]['message']['content'].strip()
        logger.info("Chatbot response generated")
        time.sleep(3)  # Wait to respect rate limits
        return answer

    except Exception as e:
        error_message = f"⚠️ Error processing your query: {str(e)}"
        logger.error(error_message, exc_info=True)
        return error_message

def build_app():
    """
    Build and launch the Gradio app.
    """
    try:
        logger.info("Building Gradio app")
        with gr.Blocks(css="app.css") as demo:
            # General Overview
            gr.Markdown("""
            # 📚 SmartMarks - AI Browser Bookmarks Manager

            Welcome to **SmartMarks**, your intelligent assistant for managing browser bookmarks. SmartMarks leverages AI to help you organize, search, and interact with your bookmarks seamlessly.

            ---

            ## 🚀 **How to Use SmartMarks**

            SmartMarks is divided into three main sections:

            1. **📂 Upload and Process Bookmarks:** Import your existing bookmarks and let SmartMarks analyze and categorize them for you.
            2. **💬 Chat with Bookmarks:** Interact with your bookmarks using natural language queries to find relevant links effortlessly.
            3. **🛠️ Manage Bookmarks:** View, edit, delete, and export your bookmarks with ease.
            """)

            # Upload and Process Bookmarks Tab
            with gr.Tab("Upload and Process Bookmarks"):
                gr.Markdown("""
                ## 📂 **Upload and Process Bookmarks**

                ### 📝 **Steps:**
                1. Click on the "Upload Bookmarks HTML File" button
                2. Select your bookmarks file
                3. Click "Process Bookmarks" to analyze and organize your bookmarks
                """)

                upload = gr.File(label="📁 Upload Bookmarks HTML File", type='binary')
                process_button = gr.Button("⚙️ Process Bookmarks")
                output_text = gr.Textbox(label="✅ Output", interactive=False)
                bookmark_display = gr.HTML(label="📄 Processed Bookmarks")

            # Chat with Bookmarks Tab
            with gr.Tab("Chat with Bookmarks"):
                gr.Markdown("""
                ## 💬 **Chat with Bookmarks**

                Ask questions about your bookmarks and get relevant results.
                """)

                user_input = gr.Textbox(
                    label="✍️ Ask about your bookmarks",
                    placeholder="e.g., Do I have any bookmarks about AI?"
                )
                chat_button = gr.Button("📨 Send")
                chat_output = gr.Textbox(label="💬 Response", interactive=False)

            # Manage Bookmarks Tab
            with gr.Tab("Manage Bookmarks"):
                gr.Markdown("""
                ## 🛠️ **Manage Bookmarks**
                Select bookmarks to delete or edit their categories.
                """)

                manage_output = gr.Textbox(label="🔄 Status", interactive=False)
                bookmark_selector = gr.CheckboxGroup(
                    label="✅ Select Bookmarks",
                    choices=[]
                )
                new_category = gr.Dropdown(
                    label="🆕 New Category",
                    choices=CATEGORIES,
                    value="Uncategorized"
                )
                bookmark_display_manage = gr.HTML(label="📄 Bookmarks")

                with gr.Row():
                    delete_button = gr.Button("🗑️ Delete Selected")
                    edit_category_button = gr.Button("✏️ Edit Category")
                    export_button = gr.Button("💾 Export")

                download_link = gr.HTML(label="📥 Download")

            # Set up event handlers
            process_button.click(
                process_uploaded_file,
                inputs=upload,
                outputs=[output_text, bookmark_display, bookmark_selector, bookmark_display_manage]
            )

            chat_button.click(
                chatbot_response,
                inputs=user_input,
                outputs=chat_output
            )

            delete_button.click(
                delete_selected_bookmarks,
                inputs=bookmark_selector,
                outputs=[manage_output, bookmark_selector, bookmark_display_manage]
            )

            edit_category_button.click(
                edit_selected_bookmarks_category,
                inputs=[bookmark_selector, new_category],
                outputs=[manage_output, bookmark_selector, bookmark_display_manage]
            )

            export_button.click(
                export_bookmarks,
                outputs=download_link
            )

        logger.info("Launching Gradio app")
        demo.launch(debug=True)
    except Exception as e:
        logger.error(f"Error building the app: {e}", exc_info=True)
        print(f"Error building the app: {e}")

if __name__ == "__main__":
    build_app()