app.py

import gradio as gr
import requests
import os
import arxiv
import io
import re
from pypdf import PdfReader

# -----------------------------
# 1. CONFIGURATION & LLM ENGINE
# -----------------------------
API_KEY = os.getenv("XAI_API_KEY")
API_URL = "https://api.x.ai/v1/chat/completions"
MODEL_NAME = "grok-4-1-fast-non-reasoning"

def call_grok(prompt):
    if not API_KEY:
        return "❌ ERROR: XAI_API_KEY not found in Hugging Face Secrets."
    
    headers = {
        "Authorization": f"Bearer {API_KEY}",
        "Content-Type": "application/json"
    }
    payload = {
        "model": MODEL_NAME,
        "messages": [
            {"role": "system", "content": "You are a world-class research scientist. Use LaTeX for math and format the equations properly, the equations and math symbols should all be clear. Use exact headers: [SUMMARY], [PROBLEM], [IDEAS], [THEORY], [ALGO], [FINDINGS], [AUTHORS], [VERDICT]. EXPLAIN EACH SECTION in a detailed manner for the audiences to understand the paper deeply. Explain the math and problem statement in a simple manner"},
            {"role": "user", "content": prompt}
        ],
        "temperature": 0.1
    }
    
    try:
        response = requests.post(API_URL, headers=headers, json=payload, timeout=60)
        data = response.json()
        
        if "error" in data:
            return f"❌ API ERROR: {data['error']['message']}"
        if "choices" not in data:
            return f"❌ Unexpected Response: {str(data)}"
            
        return data["choices"][0]["message"]["content"]
    except Exception as e:
        return f"❌ Connection Failed: {str(e)}"

# -----------------------------
# 2. CORE LOGIC
# -----------------------------
def analyze_paper(url):
    try:
        # Extract ID
        match = re.search(r"(\d{4}\.\d{4,5})", url)
        if not match:
            return ["❌ Invalid URL. Please use a standard arXiv link."] + [""] * 8
        paper_id = match.group(1)

        search = arxiv.Search(id_list=[paper_id])
        paper = next(search.results())
        
        resp = requests.get(paper.pdf_url)
        reader = PdfReader(io.BytesIO(resp.content))
        num_pages = len(reader.pages)

        text_pages = [p.extract_text() for p in reader.pages[:7]]
        if num_pages > 7:
            text_pages.append(reader.pages[-1].extract_text())
        content = " ".join(text_pages)

        prompt = f"Title: {paper.title}\nAuthors: {', '.join([a.name for a in paper.authors])}\n\nFull Text Snippet: {content[:18000]}"
        
        raw_analysis = call_grok(prompt)
        
        if "❌" in raw_analysis:
            return [paper.title, raw_analysis] + [""] * 7

        markers = ["SUMMARY", "PROBLEM", "IDEAS", "THEORY", "ALGO", "FINDINGS", "AUTHORS", "VERDICT"]
        results = {}
        for i, m in enumerate(markers):
            start = rf"\[{m}\]"
            end = rf"\[{markers[i+1]}\]" if i+1 < len(markers) else "$"
            match = re.search(f"{start}(.*?){end}", raw_analysis, re.DOTALL | re.IGNORECASE)
            results[m] = match.group(1).strip() if match else "Section could not be parsed."

        return [paper.title] + [results[m] for m in markers]

    except Exception as e:
        return [f"❌ Error: {str(e)}"] + [""] * 8

# -----------------------------
# 3. GRADIO UI
# -----------------------------
with gr.Blocks(theme=gr.themes.Soft(primary_hue="cyan"), title="arXivForMe Ultra") as demo:
    gr.Markdown("# 🔬 arXivForMe")
    gr.Markdown("### A Research Deconstruction Application")
    
    with gr.Row():
        url_input = gr.Textbox(label="arXiv URL", placeholder="https://arxiv.org/abs/2401.12345", scale=4)
        run_btn = gr.Button("ANALYZE", variant="primary", scale=1)

    paper_display = gr.HTML("<center><h3>Enter a link to begin analysis</h3></center>")

    with gr.Tabs():
        with gr.Tab("📋 Summary & Problem"):
            with gr.Row():
                out_sum = gr.Markdown(label="Summary")
                out_prob = gr.Markdown(label="Research Problem")
        
        with gr.Tab("💡 Innovation"):
            out_idea = gr.Markdown(label="Main Ideas")
            
        with gr.Tab("📐 Math & Theory"):
            out_theo = gr.Markdown(label="Formal Framework")
            
        with gr.Tab("💻 Algorithm"):
            out_algo = gr.Markdown(label="Implementation Logic")
            
        with gr.Tab("📊 Findings"):
            out_find = gr.Markdown(label="Results")
            
        with gr.Tab("🕵️ Author Reputation"):
            out_auth = gr.Markdown(label="Author Context")
            
        with gr.Tab("⚖️ The Verdict"):
            out_verd = gr.Markdown(label="AI Critical Opinion")

    outputs = [paper_display, out_sum, out_prob, out_idea, out_theo, out_algo, out_find, out_auth, out_verd]
    run_btn.click(fn=analyze_paper, inputs=url_input, outputs=outputs)

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