Update app.py

app.py

@@ -1,187 +1,91 @@
 import gradio as gr
-import fitz  # PyMuPDF
+import fitz
 import torch
 from transformers import pipeline
-import time, logging, re
-import matplotlib
-matplotlib.use('Agg')
-import matplotlib.pyplot as plt
-import io
+import re, time, io, uuid, os
 from PIL import Image
+import matplotlib.pyplot as plt
+matplotlib.use('Agg')
 from concurrent.futures import ThreadPoolExecutor
-import numpy as np
 
-logging.basicConfig(level=logging.INFO)
-logger = logging.getLogger(__name__)
-
-# Set device and optimize for speed
 device = 0 if torch.cuda.is_available() else -1
-logger.info(f"🔧 Using {'GPU' if device == 0 else 'CPU'}")
-
-# Load a lighter model for faster inference
-try:
-    summarizer = pipeline(
-        "summarization",
-        model="t5-small",  # Lightweight model
-        device=device,
-        framework="pt",
-        torch_dtype=torch.float16 if device == 0 else torch.float32  # Half-precision on GPU
-    )
-except Exception as e:
-    logger.error(f"❌ Model loading failed: {str(e)}")
-    exit(1)
-
-def visualize_chunk_status(chunk_data):
-    status_colors = {'summarized': 'green', 'skipped': 'orange', 'error': 'red'}
-    labels = [f"C{i['chunk']}" for i in chunk_data]
-    colors = [status_colors.get(i['status'], 'gray') for i in chunk_data]
-    times = [i.get('time', 0.1) for i in chunk_data]
-
-    fig, ax = plt.subplots(figsize=(8, 2))  # Smaller figure size
-    ax.barh(labels, times, color=colors, height=0.4)  # Reduced bar height
-    ax.set_xlabel("Time (s)")
-    ax.set_title("Chunk Status")
-    plt.tight_layout(pad=0.5)  # Minimal padding
-    buf = io.BytesIO()
-    plt.savefig(buf, format='png', dpi=100)  # Lower DPI for speed
-    plt.close(fig)
-    buf.seek(0)
-    return Image.open(buf)
+summarizer = pipeline("summarization", model="sshleifer/distilbart-cnn-12-6", device=device)
+
+def process_chunk(i, chunk):
+    if sum(1 for c in chunk if not c.isalnum()) / len(chunk) > 0.5:
+        return None
+    try:
+        summary = summarizer(chunk, max_length=80, min_length=15, do_sample=False)[0]['summary_text']
+        return summary
+    except:
+        return None
 
 def create_summary_flowchart(summaries):
-    filtered = [s for s in summaries if s.startswith("**Chunk") and "Skipped" not in s and "Error" not in s]
+    filtered = [s for s in summaries if s]
     if not filtered:
-        return None
+        return None, None
 
-    fig_height = max(1.5, len(filtered) * 0.5)  # Reduced height scaling
-    fig, ax = plt.subplots(figsize=(5, fig_height))
+    fig_height = max(2, len(filtered) * 0.8 + 1)
+    fig, ax = plt.subplots(figsize=(6, fig_height))
     ax.axis('off')
 
-    ypos = np.arange(len(filtered) * 1.5, 0, -1.5)  # Tighter spacing
-    for i, (y, summary) in enumerate(zip(ypos, filtered)):
-        summary_text = summary.split("**Chunk")[1].split("\n", 1)[-1].strip()[:100] + ("..." if len(summary.split("**Chunk")[1].split("\n", 1)[-1].strip()) > 100 else "")
-        ax.text(0.5, y, summary_text, ha='center', va='center', fontsize=8, bbox=dict(facecolor="lightblue", edgecolor="black", pad=0.2))
+    ypos = list(range(len(filtered) * 2, 0, -2))
+    boxprops = dict(boxstyle="round,pad=0.5", facecolor="lightblue", edgecolor="black")
+
+    for i, (y, summary_text) in enumerate(zip(ypos, filtered)):
+        if len(summary_text) > 120:
+            summary_text = summary_text[:120] + "..."
+        ax.text(0.5, y, summary_text, ha='center', va='center', bbox=boxprops, fontsize=9)
         if i < len(filtered) - 1:
-            ax.arrow(0.5, y - 0.2, 0, -1.1, head_width=0.02, head_length=0.1, fc='black', ec='black')
+            ax.annotate('', xy=(0.5, y - 1.2), xytext=(0.5, y - 0.3),
+                        arrowprops=dict(arrowstyle="->", lw=1.5))
 
-    plt.tight_layout(pad=0.1)
+    # Save to in-memory buffer (for display)
     buf = io.BytesIO()
-    fig.savefig(buf, format='png', dpi=100, bbox_inches='tight')
+    fig.savefig(buf, format='png', bbox_inches='tight')
     plt.close(fig)
     buf.seek(0)
-    return Image.open(buf)
+    image = Image.open(buf)
 
-def process_chunk(i, chunk):
-    chunk_result = {'chunk': i + 1, 'status': '', 'time': 0}
-    start_time = time.time()
-
-    if not chunk.strip() or sum(1 for c in chunk if not c.isalnum()) / len(chunk) > 0.5:
-        result = f"**Chunk {i+1}**: Skipped (empty or equation-heavy)"
-        chunk_result['status'] = 'skipped'
-    else:
-        try:
-            summary = summarizer(chunk, max_length=60, min_length=10, do_sample=False)[0]['summary_text']
-            result = f"**Chunk {i+1}**:\n{summary}"
-            chunk_result['status'] = 'summarized'
-        except Exception as e:
-            result = f"**Chunk {i+1}**: Error: {str(e)}"
-            chunk_result['status'] = 'error'
-
-    chunk_result['time'] = time.time() - start_time
-    return result, chunk_result
-
-def summarize_file(file_bytes):
-    start = time.time()
-    summaries = []
-    chunk_info = []
-
-    # Stream text extraction
+    # Save to disk (for download)
+    filename = f"/tmp/flowchart_{uuid.uuid4().hex}.png"
+    image.save(filename)
+
+    return image, filename
+
+def generate_flowchart(file_bytes):
     try:
         doc = fitz.open(stream=file_bytes, filetype="pdf")
-        text = ""
-        for page in doc:
-            page_text = page.get_text("text")
-            if not page_text.strip():
-                continue
-            text += page_text
-            if len(text) > 30000:  # Early cutoff
-                text = text[:30000]
-                break
-        doc.close()
-        
-        # Fast text cleaning
-        text = re.sub(r"\$\s*[^$]+\s*\$|\\cap|\s+", lambda m: "intersection" if m.group(0) == "\\cap" else " ", text)
-        text = "".join(c for c in text if ord(c) < 128)[:30000]
+        text = "".join(page.get_text("text") for page in doc)
+        text = re.sub(r"\$\s*([^$]+)\s*\$", r"\1", text)
+        text = re.sub(r"\\cap", "intersection", text)
+        text = re.sub(r"\s+", " ", text).strip()
+        text = "".join(c for c in text if ord(c) < 128)
     except Exception as e:
-        return f"Text extraction failed: {str(e)}", None, None
-
-    if not text.strip():
-        return "No text found", None, None
-
-    # Smaller, sentence-aware chunks
-    chunks = []
-    current_chunk = ""
-    for sentence in re.split(r'(?<=[.!?])\s+', text):
-        if len(current_chunk) + len(sentence) <= 1000:
-            current_chunk += sentence
-        else:
-            if current_chunk:
-                chunks.append(current_chunk)
-            current_chunk = sentence
-        if len(chunks) >= 30:  # Limit chunks
-            break
-    if current_chunk:
-        chunks.append(current_chunk)
-
-    # Dynamic worker count based on CPU/GPU
-    max_workers = min(8, max(2, torch.cuda.device_count() * 4 if device == 0 else 4))
-    with ThreadPoolExecutor(max_workers=max_workers) as executor:
-        results = list(executor.map(lambda ic: process_chunk(*ic), enumerate(chunks)))
-
-    for summary, info in results:
-        summaries.append(summary)
-        chunk_info.append(info)
-
-    final_summary = f"**Chunks**: {len(chunks)}\n**Time**: {time.time() - start:.2f}s\n\n" + "\n\n".join(summaries)
-    process_img = visualize_chunk_status(chunk_info)
-    flow_img = create_summary_flowchart(summaries)
-    return final_summary, process_img, flow_img
+        return f"❌ Error reading PDF: {str(e)}", None
+
+    chunks = [text[i:i+1500] for i in range(0, min(len(text), 30000), 1500)]
+
+    with ThreadPoolExecutor(max_workers=4) as executor:
+        summaries = list(executor.map(lambda ic: process_chunk(*ic), enumerate(chunks)))
+
+    img, filepath = create_summary_flowchart(summaries)
+    if img is None:
+        return "❌ No valid summaries to display.", None
+    return img, filepath
 
 demo = gr.Interface(
-    fn=summarize_file,
-    inputs=gr.File(label="Upload PDF", type="binary"),
+    fn=generate_flowchart,
+    inputs=gr.File(label="📄 Upload PDF", type="binary"),
     outputs=[
-        gr.Textbox(label="Summary", lines=15),
-        gr.Image(label="Chunk Status", type="pil"),
-        gr.Image(label="Flow Summary", type="pil")
+        gr.Image(label="🔁 Flowchart", type="pil"),
+        gr.File(label="📥 Download Flowchart (PNG)")
     ],
-    title="PDF Summarizer",
-    description="Summarizes PDFs up to 30,000 characters with visualizations."
+    title="📘 Summary Flowchart Generator",
+    description="Uploads a PDF, generates summary chunks, and provides a downloadable visual flowchart."
 )
 
 if __name__ == "__main__":
-    try:
-        logger.info("Starting Gradio on http://127.0.0.1:7860")
-        demo.launch(
-            share=False,
-            server_name="127.0.0.1",
-            server_port=7860,
-            debug=False  # Disable debug for speed
-        )
-    except Exception as e:
-        logger.error(f"Failed on port 7860: {str(e)}")
-        logger.info("Trying port 7861...")
-        try:
-            demo.launch(
-                share=False,
-                server_name="127.0.0.1",
-                server_port=7861,
-                debug=False
-            )
-        except Exception as e2:
-            logger.error(f"Failed on port 7861: {str(e2)}")
-            raise
-
-
+    demo.launch(share=False, server_port=7860)