app.py

import gradio as gr
import pdfplumber
import re
from transformers import pipeline

# Load models once
summarizer = pipeline("summarization", model="facebook/bart-large-cnn")
critic = pipeline("text2text-generation", model="google/flan-t5-large")

# Extract and clean PDF text
def extract_text(file):
    with pdfplumber.open(file) as pdf:
        return "\n".join([page.extract_text() or "" for page in pdf.pages])

# Split text into sections based on common headings
def extract_sections(text):
    pattern = r'\n(?P<header>[A-Z][A-Za-z ]{3,40})\n'
    matches = list(re.finditer(pattern, text))
    sections = {}
    for i, match in enumerate(matches):
        start = match.end()
        end = matches[i + 1].start() if i + 1 < len(matches) else len(text)
        sections[match.group("header").strip().lower()] = text[start:end].strip()
    return sections or {"full text": text}

# Limit content for faster inference
def truncate(text, max_tokens=1000):
    return text[:max_tokens * 5]  # approx. chars per token

# Run analysis
def analyze(file, pasted_text, custom_prompt):
    raw_text = extract_text(file.name) if file else pasted_text.strip()
    if not raw_text:
        return "No text provided.", "", "", ""

    summary = summarizer(truncate(raw_text), max_length=250, min_length=100, do_sample=False)[0]['summary_text']
    sections = extract_sections(raw_text)

    critique_output = ""
    gap_output = ""

    for sec, content in sections.items():
        snippet = truncate(content)
        critique_out = critic(f"Critique this section:\n{snippet}", max_length=256)[0]['generated_text']
        gaps_out = critic(f"Identify research gaps in this section:\n{snippet}", max_length=256)[0]['generated_text']
        critique_output += f"### {sec.title()}\n{critique_out}\n\n"
        gap_output += f"### {sec.title()}\n{gaps_out}\n\n"

    custom_out = ""
    if custom_prompt.strip():
        custom_out = critic(custom_prompt.strip(), max_length=256)[0]['generated_text']

    return summary, critique_output.strip(), gap_output.strip(), custom_out

# Gradio interface
demo = gr.Interface(
    fn=analyze,
    inputs=[
        gr.File(label="📄 Upload PDF", file_types=[".pdf"]),
        gr.Textbox(lines=10, label="✍️ Or paste paper content"),
        gr.Textbox(lines=2, label="💬 Optional: Custom prompt")
    ],
    outputs=[
        gr.Markdown(label="📋 Summary"),
        gr.Markdown(label="🧠 Critique"),
        gr.Markdown(label="🔎 Research Gaps"),
        gr.Markdown(label="💡 Custom Prompt Response")
    ],
    title="Research Paper Critique Generator",
    description="Upload a PDF or paste text to receive an AI-generated summary, critique, and identification of research gaps.",
    theme="default"
)

if __name__ == "__main__":
    demo.launch()