Update app.py

app.py

@@ -5,13 +5,14 @@ import gradio as gr
 import time
 import os
 import json
+import PyPDF2
+import io
+import asyncio
+import aiohttp
+import aiofiles
 import re
 from datetime import datetime
-import torch
-from transformers import AutoTokenizer, AutoModel
-import networkx as nx
-from pyvis.network import Network
-import matplotlib.pyplot as plt
+import base64
 
 # 🧙‍♂️ Magical Utility Functions 🧙‍♂️
 
@@ -19,6 +20,31 @@ def safe_filename(title):
     """Convert a string to a safe filename. No more 'file not found' nightmares! 🙅‍♂️📁"""
     return re.sub(r'[^\w\-_\. ]', '_', title)
 
+def create_date_directory():
+    """Create a directory named with the current date. It's like a time capsule for your downloads! 🗓️📦"""
+    date_str = datetime.now().strftime("%Y-%m-%d")
+    os.makedirs(date_str, exist_ok=True)
+    return date_str
+
+def get_base64_download_link(content, filename):
+    """Create a base64 download link for text content. It's like teleportation for your files! 🌟📲"""
+    b64 = base64.b64encode(content.encode()).decode()
+    return f'<a href="data:text/plain;base64,{b64}" download="{filename}">Download {filename}</a>'
+
+# 🎬 Animated Banner Messages 🎬
+def animated_banner(message, emoji):
+    """Create an animated banner message. It's like a tiny parade for your console! 🎉🚩"""
+    frames = [
+        f"╔════ {emoji} ════╗\n║ {message:^16} ║\n╚════════════╝",
+        f"╔════ {emoji} ════╗\n║ {message:^16} ║\n╚════════════╝",
+        f"╔════{emoji}════╗\n║ {message:^14} ║\n╚══════════╝",
+        f"╔═══{emoji}═══╗\n║ {message:^12} ║\n╚════════╝",
+        f"╔══{emoji}══╗\n║ {message:^10} ║\n╚══════╝",
+        f"╔═{emoji}═╗\n║ {message:^8} ║\n╚════╝",
+        f"╔{emoji}╗\n║ {message:^6} ║\n╚══╝",
+    ]
+    return frames
+
 # 🕵️‍♂️ Data Fetching and Caching Shenanigans 🕵️‍♂️
 
 def get_rank_papers(url, progress=gr.Progress(track_tqdm=True)):
@@ -54,8 +80,19 @@ def get_rank_papers(url, progress=gr.Progress(track_tqdm=True)):
             
             Github_Star = ppr.find('span', class_='badge badge-secondary').text.strip().replace(',', '') if ppr.find('span', class_='badge badge-secondary') else "0"
             
+            pdf_link = ''
+            try:
+                response_link = session.get(link, headers=headers)
+                soup_link = BeautifulSoup(response_link.text, 'html.parser')
+                paper_info_link = soup_link.find_all('div', class_='paper-abstract')
+                pdf_link = paper_info_link[0].find('div', class_='col-md-12').find('a')['href']
+            except Exception as e:
+                print(f"Failed to retrieve PDF link for {title}: {e}")
+            
+            print(f"Title: {title}, Link: {link}, Github Star: {Github_Star}, PDF Link: {pdf_link}")
+            
             if title not in data_list:
-                data_list[title] = {'link': link, 'Github Star': int(Github_Star), 'title': title}
+                data_list[title] = {'link': link, 'Github Star': int(Github_Star), 'pdf_link': pdf_link.strip()}
             else:
                 break_duplicate -= 1
                 if break_duplicate == 0:
@@ -90,53 +127,119 @@ def load_and_cache_data(url, cache_file):
     save_cached_data(new_data, cache_file)
     return new_data
 
-# 🚀 Transformer-based Word and Context Analysis 🚀
+# 📊 Data Processing and Display Magic 📊
 
-def generate_embeddings(titles):
-    """Generate word embeddings using a transformer model."""
-    model_name = "sentence-transformers/all-MiniLM-L6-v2"
-    tokenizer = AutoTokenizer.from_pretrained(model_name)
-    model = AutoModel.from_pretrained(model_name)
+def format_dataframe(data):
+    """Format data into a pretty DataFrame. It's like giving your data a makeover! 💅📈"""
+    if not data:
+        print("No data found to format.")
+        return pd.DataFrame()
     
-    embeddings = []
-    with torch.no_grad():
-        for title in titles:
-            tokens = tokenizer(title, return_tensors="pt", padding=True, truncation=True)
-            output = model(**tokens)
-            embeddings.append(output.last_hidden_state.mean(dim=1).squeeze())
+    df = pd.DataFrame(data).T
+    df['title'] = df.index
     
-    return embeddings
-
-def build_graph(titles, embeddings, threshold=0.7):
-    """Build a graph of words based on similarity between titles."""
-    G = nx.Graph()
+    # Check if required columns are present
+    if 'Github Star' in df.columns and 'link' in df.columns and 'pdf_link' in df.columns:
+        df = df[['title', 'Github Star', 'link', 'pdf_link']]
+        df = df.sort_values(by='Github Star', ascending=False)
+        df['link'] = df['link'].apply(lambda x: f'<a href="{x}" target="_blank">Link</a>')
+        df['pdf_link'] = df['pdf_link'].apply(lambda x: f'<a href="{x}" target="_blank">{x}</a>')
+    else:
+        print("Required columns are missing in the dataframe.")
+        print(f"Columns available: {df.columns}")
+    
+    return df
 
-    for i, title in enumerate(titles):
-        G.add_node(i, label=title)
+def update_display(category):
+    """Update the display for a category. Freshen up your data like it's spring cleaning! 🧹🌸"""
+    cache_file = f"{category}_papers_cache.json"
+    url = f"https://paperswithcode.com/{category}" if category != "top" else "https://paperswithcode.com/"
+    
+    data = load_and_cache_data(url, cache_file)
+    df = format_dataframe(data)
+    
+    return len(df), df.to_html(escape=False, index=False)
 
-    for i in range(len(embeddings)):
-        for j in range(i+1, len(embeddings)):
-            sim = torch.cosine_similarity(embeddings[i], embeddings[j], dim=0).item()
-            if sim > threshold:
-                G.add_edge(i, j, weight=sim)
+def load_all_data():
+    """Load data for all categories. It's like a buffet for your brain! 🧠🍽️"""
+    top_count, top_html = update_display("top")
+    new_count, new_html = update_display("latest")
+    greatest_count, greatest_html = update_display("greatest")
+    return top_count, top_html, new_count, new_html, greatest_count, greatest_html
 
-    return G
+# 🚀 Asynchronous Paper Processing Wizardry 🚀
 
-def visualize_graph(G, titles):
-    """Visualize the graph using pyvis and show it as a mind map."""
-    net = Network(height="750px", width="100%", notebook=True)
+async def download_and_process_pdf(session, title, paper_info, directory):
+    """Download and process a PDF. It's like turning lead into gold, but with papers! 📜✨"""
+    pdf_url = paper_info['pdf_link']
+    if not pdf_url:
+        return f"🚫 No PDF link for: {title}. It's playing hide and seek! 🙈", None, None
     
-    for node in G.nodes(data=True):
-        net.add_node(node[0], label=titles[node[0]])
+    try:
+        timeout = aiohttp.ClientTimeout(total=60)  # 60 seconds timeout
+        async with session.get(pdf_url, timeout=timeout) as response:
+            if response.status != 200:
+                return f"🚨 Failed to grab PDF for {title}: HTTP {response.status}. The internet gremlins strike again! 👹", None, None
+            pdf_content = await response.read()
+        
+        file_length = len(pdf_content)
+        if file_length < 5000:  # Check if the PDF is less than 5KB
+            return f"🐜 PDF for {title} is tiny ({file_length} bytes). It's like a paper for ants! 🐜📄", None, None
+        
+        # Convert PDF to text
+        pdf_file = io.BytesIO(pdf_content)
+        pdf_reader = PyPDF2.PdfReader(pdf_file)
+        text = ""
+        for page in pdf_reader.pages:
+            text += page.extract_text()
+        
+        if len(text) < 5000:  # Check if the extracted text is less than 5KB
+            return f"📉 Extracted text for {title} is too small ({len(text)} characters). It's not you, it's the PDF! 💔", None, None
+        
+        safe_title = safe_filename(title)
+        txt_filename = f"{safe_title}.txt"
+        txt_filepath = os.path.join(directory, txt_filename)
+        
+        async with aiofiles.open(txt_filepath, 'w', encoding='utf-8') as f:
+            await f.write(text)
+        
+        return f"🎉 Successfully processed: {txt_filename} (File length: {file_length} bytes). It's alive! 🧬", txt_filepath, text
+    except asyncio.TimeoutError:
+        return f"⏳ Timeout for {title}. The PDF is playing hard to get! 💃", None, None
+    except Exception as e:
+        return f"💥 Oops! Error processing {title}: {str(e)}. Gremlins in the system! 🛠️", None, None
 
-    for edge in G.edges(data=True):
-        net.add_edge(edge[0], edge[1], value=edge[2]['weight'])
-
-    net.show("paper_network.html")
-    return "paper_network.html"
+async def process_papers(data, directory, progress=gr.Progress()):
+    """Process multiple papers asynchronously. It's like juggling papers, but faster! 🤹‍♂️📚"""
+    async with aiohttp.ClientSession() as session:
+        tasks = []
+        for title, paper_info in data.items():
+            task = asyncio.ensure_future(download_and_process_pdf(session, title, paper_info, directory))
+            tasks.append(task)
+        
+        results = []
+        successful_downloads = []
+        errors = []
+        for i, task in enumerate(asyncio.as_completed(tasks), start=1):
+            result, filepath, text = await task
+            results.append(result)
+            if filepath and text:
+                successful_downloads.append((filepath, text))
+            else:
+                errors.append(result)
+            progress(i / len(tasks), f"🚀 Processed {i}/{len(tasks)} papers. Science waits for no one!")
+            
+            # Display animated banner
+            banner_frames = animated_banner("Processing", "📄")
+            for frame in banner_frames:
+                print(frame, end='\r')
+                await asyncio.sleep(0.1)
+            print(" " * len(banner_frames[-1]), end='\r')  # Clear the last frame
+    
+    return results, successful_downloads, errors
 
-def analyze_and_generate_graph(progress=gr.Progress()):
-    """Analyze papers, generate embeddings, and visualize the relationship graph."""
+def download_all_papers(progress=gr.Progress()):
+    """Download and process all papers. It's like hosting a paper party, and everyone's invited! 🎉📚"""
     all_data = {}
     for category in ["top", "latest", "greatest"]:
         cache_file = f"{category}_papers_cache.json"
@@ -144,35 +247,27 @@ def analyze_and_generate_graph(progress=gr.Progress()):
         if data:
             all_data.update(data)
     
-    titles = [paper['title'] for paper in all_data.values()]
+    date_directory = create_date_directory()
+    results, successful_downloads, errors = asyncio.run(process_papers(all_data, date_directory, progress))
     
-    # Generate embeddings
-    embeddings = generate_embeddings(titles)
+    summary = f"📊 Papers processed: {len(all_data)} (We're basically librarians now!)\n"
+    summary += f"✅ Successfully downloaded and converted: {len(successful_downloads)} (Take that, PDF gremlins!)\n"
+    summary += f"❌ Errors: {len(errors)} (Even superheroes have off days)\n\n"
+    summary += "🚨 Error List (AKA 'The Wall of Shame'):\n" + "\n".join(errors) if errors else "No errors occurred. It's a miracle! 🙌"
     
-    # Build a similarity graph based on the embeddings
-    G = build_graph(titles, embeddings)
+    download_links = []
+    text_contents = []
+    for filepath, text in successful_downloads:
+        filename = os.path.basename(filepath)
+        download_links.append(get_base64_download_link(text, filename))
+        text_contents.append(f"--- {filename} ---\n\n{text[:1000]}... (There's more, but we don't want to spoil the ending! 📚🔮)\n\n")
     
-    # Visualize the graph as a mind map
-    graph_file = visualize_graph(G, titles)
-    
-    summary = f"📊 Papers analyzed: {len(titles)}\n"
-    summary += f"✅ Graph generated and visualized.\n"
-    
-    return summary, graph_file
-
-# 🚀 Define load_all_data Properly 🚀
-
-def load_all_data():
-    """Load data for all categories and prepare for display."""
-    top_count, top_html = update_display("top")
-    new_count, new_html = update_display("latest")
-    greatest_count, greatest_html = update_display("greatest")
-    return top_count, top_html, new_count, new_html, greatest_count, greatest_html
+    return summary, "<br>".join(download_links), "\n".join(text_contents)
 
 # 🎭 Gradio Interface: Where the Magic Happens 🎭
 
 with gr.Blocks() as demo:
-    gr.Markdown("<h1><center>Papers Leaderboard with Context Analysis</center></h1>")
+    gr.Markdown("<h1><center>Papers Leaderboard</center></h1>")
     
     with gr.Tab("Top Trending Papers"):
         top_count = gr.Textbox(label="Number of Papers Fetched")
@@ -192,15 +287,16 @@ with gr.Blocks() as demo:
         greatest_button = gr.Button("Refresh Leaderboard")
         greatest_button.click(fn=lambda: update_display("greatest"), inputs=None, outputs=[greatest_count, greatest_html])
 
-    analyze_button = gr.Button("🔍 Analyze and Generate Graph", variant="primary")
-    analyze_output = gr.Textbox(label="Analysis Status")
-    graph_output = gr.HTML(label="Graph Visualization")
-    
-    analyze_button.click(fn=analyze_and_generate_graph, inputs=None, outputs=[analyze_output, graph_output])
+    download_button = gr.Button("📚 Download All Papers", variant="primary")
+    download_output = gr.Textbox(label="Download Status")
+    download_links = gr.HTML(label="Download Links")
+    # Updated this to gr.Code with Python as language for displaying paper contents
+    text_output = gr.Code(label="Paper Contents", language="python")
+    download_button.click(fn=download_all_papers, inputs=None, outputs=[download_output, download_links, text_output])
 
     # Load initial data for all tabs
     demo.load(fn=load_all_data, outputs=[top_count, top_html, new_count, new_html, greatest_count, greatest_html])
 
 # 🚀 Launch the Gradio interface with a public link
-print("🎭 Launching the Papers Leaderboard with Context Analysis! Get ready to explore the relationships between papers! 🎢📚")
+print("🎭 Launching the Papers Leaderboard! Get ready for a wild ride through the land of academia! 🎢📚")
 demo.launch(share=True)