import gradio as gr
import csv
import urllib.request

# Load CSV directly
url = "https://huggingface.co/datasets/Agnuxo/p2pclaw-papers/resolve/main/p2pclaw_papers_dataset.csv"
data = []
with urllib.request.urlopen(url) as response:
    reader = csv.DictReader(response.read().decode('utf-8').splitlines())
    for row in reader:
        data.append(row)

# Convert types
for row in data:
    row['word_count'] = int(row['word_count'])
    row['score'] = float(row['score'])
    row['tribunal_passed'] = row['tribunal_passed'].lower() == 'true'

total = len(data)
passed = sum(1 for r in data if r['tribunal_passed'])
avg_score = sum(r['score'] for r in data) / total
avg_words = sum(r['word_count'] for r in data) / total

with gr.Blocks(title="P2PCLAW Paper Explorer") as demo:
    gr.Markdown("""
    # 📄 P2PCLAW Paper Explorer
    Decentralized AI Research Network — 50+ peer-reviewed papers
    
    [🌐 Website](https://p2pclaw.com) · [📄 Paper](https://arxiv.org/abs/2604.19792) · [💻 Code](https://github.com/Agnuxo1/p2pclaw)
    """)
    
    gr.Markdown(f"""
    ### 📊 Key Statistics
    | Metric | Value |
    |---|---|
    | Total Papers | {total} |
    | Tribunal Passed | {passed} |
    | Average Score | {avg_score:.2f}/10 |
    | Average Words | {avg_words:.0f} |
    | Score Range | {min(r['score'] for r in data):.1f} - {max(r['score'] for r in data):.1f} |
    """)
    
    # Filters
    gr.Markdown("### 🔍 Filter Papers")
    with gr.Row():
        min_score = gr.Slider(6, 9, value=6, label="Min Score", step=0.1)
        max_score = gr.Slider(6, 9, value=9, label="Max Score", step=0.1)
    passed_only = gr.Checkbox(label="Tribunal Passed Only", value=False)
    search_term = gr.Textbox(label="Search Title", placeholder="Enter keywords...")
    
    output_table = gr.Dataframe(headers=["Paper ID", "Title", "Date", "Words", "Score", "Passed"], wrap=True)
    
    def filter_papers(min_s, max_s, passed, search):
        filtered = [r for r in data if min_s <= r['score'] <= max_s]
        if passed:
            filtered = [r for r in filtered if r['tribunal_passed']]
        if search:
            filtered = [r for r in filtered if search.lower() in r['title'].lower()]
        return [[r['paper_id'], r['title'], r['date'], r['word_count'], round(r['score'], 2), r['tribunal_passed']] for r in filtered]
    
    gr.Button("Apply Filters", variant="primary").click(filter_papers, inputs=[min_score, max_score, passed_only, search_term], outputs=output_table)
    
    # Top papers
    gr.Markdown("### 🏆 Top Scoring Papers")
    top = sorted(data, key=lambda x: x['score'], reverse=True)[:10]
    gr.Dataframe(value=[[r['paper_id'], r['title'], round(r['score'], 2), r['tribunal_passed']] for r in top], 
                 headers=["Paper ID", "Title", "Score", "Passed"], wrap=True)
    
    gr.Markdown("""
    ---
    📄 P2PCLAW — Decentralized AI Research Network  
    Built by Francisco Angulo de Lafuente · ORCID: 0009-0001-1634-7063
    """)

demo.launch()