"""
CROVIA OMISSION ORACLE v2.1.0
The Evidence Protocol for AI Training Compliance

Building the global standard for AI trust — in public, from day one.
"""

import json
import hashlib
import gradio as gr
import requests
from datetime import datetime, timezone
from typing import Dict, List, Any, Optional

VERSION = "2.1.0"
PROTOCOL_VERSION = "v1.0"
TPR_API_URL = "https://registry.croviatrust.com"
DATASET_ID = "Crovia/global-ai-training-omissions"

# Fallback cards for display
FALLBACK_CARDS = [
    {"model_id": "meta-llama/Llama-3.1-8B", "score": 72, "badge": "BRONZE", "violations": ["NEC#1", "NEC#7"], "author": "meta-llama", "downloads": "N/A"},
    {"model_id": "mistralai/Mistral-7B-v0.3", "score": 94, "badge": "GOLD", "violations": [], "author": "mistralai", "downloads": "N/A"},
]

try:
    from huggingface_hub import hf_hub_download
    HF_AVAILABLE = True
except ImportError:
    HF_AVAILABLE = False


def load_latest_cards_from_registry() -> List[Dict[str, Any]]:
    """Load latest observations from live registry API with Shadow Scores."""
    try:
        # Fetch live data from registry
        response = requests.get(f"{TPR_API_URL}/api/registry/recent?limit=12", timeout=5)
        if response.status_code == 200:
            data = response.json()
            observations = data.get('observations', [])
            
            # Get Shadow Scores from dataset (updated by cron)
            shadow_scores = {}
            try:
                if HF_AVAILABLE:
                    index_path = hf_hub_download(
                        repo_id=DATASET_ID,
                        filename="badges/index.json",
                        repo_type="dataset"
                    )
                    with open(index_path, 'r') as f:
                        badges_data = json.load(f)
                    for badge in badges_data.get("badges", []):
                        shadow_scores[badge.get("model_id")] = badge.get("score", 0)
            except:
                pass
            
            # Convert observations to card format
            cards = []
            for obs in observations:
                # Skip test/internal files
                target_id = obs.get('target_id', '')
                if any(pattern in target_id.lower() for pattern in ['test-', 'crovia-', 'debug-', 'demo-']):
                    continue
                
                # Get Shadow Score if available
                score = shadow_scores.get(target_id, 0)
                
                # Determine badge based on score
                if score >= 90:
                    badge = "GOLD"
                elif score >= 70:
                    badge = "SILVER"  
                elif score >= 50:
                    badge = "BRONZE"
                else:
                    badge = "UNVERIFIED"
                    
                cards.append({
                    "model_id": target_id,
                    "score": score,
                    "badge": badge,
                    "violations": [],
                    "author": target_id.split("/")[0] if "/" in target_id else "Unknown",
                    "downloads": "N/A",
                    "observation_type": obs.get('observation_type', 'unknown'),
                    "observed_at": obs.get('observed_at', '')
                })
            
            if cards:
                return cards
        
        # Fallback to dataset if API fails
        return load_latest_cards_from_dataset()
        
    except Exception as e:
        print(f"Error loading from registry: {e}")
        return load_latest_cards_from_dataset()


def load_latest_cards_from_dataset() -> List[Dict[str, Any]]:
    """Load pre-computed model cards from public dataset (fallback)."""
    try:
        if not HF_AVAILABLE:
            return FALLBACK_CARDS
        
        index_path = hf_hub_download(
            repo_id=DATASET_ID,
            filename="badges/index.json",
            repo_type="dataset"
        )
        
        with open(index_path, 'r') as f:
            data = json.load(f)
        
        cards = []
        for badge in data.get("badges", [])[:12]:
            model_id = badge.get("model_id", "Unknown")
            score = badge.get("score", 0)
            badge_type = badge.get("badge", "UNVERIFIED")
            
            violations = []
            try:
                card_filename = f"cards/{model_id.replace('/', '__')}.json"
                card_path = hf_hub_download(
                    repo_id=DATASET_ID,
                    filename=card_filename,
                    repo_type="dataset"
                )
                with open(card_path, 'r') as cf:
                    card_data = json.load(cf)
                    violations = card_data.get("violations", [])
            except:
                pass
            
            author = model_id.split("/")[0] if "/" in model_id else "Unknown"
            
            cards.append({
                "model_id": model_id,
                "score": score,
                "badge": badge_type,
                "violations": violations,
                "author": author,
                "downloads": "N/A"
            })
        
        return cards if cards else FALLBACK_CARDS
        
    except Exception as e:
        print(f"Error loading cards: {e}")
        return FALLBACK_CARDS


# Load cards from live registry (real-time data)
PRELOADED_CARDS = load_latest_cards_from_registry()

NECESSITY_CANON = {
    "NEC#1": {"name": "Missing data provenance", "severity": 75},
    "NEC#2": {"name": "Missing license attribution", "severity": 80},
    "NEC#7": {"name": "Missing usage scope", "severity": 45},
    "NEC#10": {"name": "Missing temporal validity", "severity": 40},
    "NEC#13": {"name": "Missing accountable entity", "severity": 70},
}


def analyze_model_live(model_id: str) -> Optional[Dict[str, Any]]:
    """Live analysis via TPR API (if available)."""
    if not model_id or not model_id.strip():
        return None
    
    try:
        response = requests.post(
            f"{TPR_API_URL}/api/analyze",
            json={"model_id": model_id.strip()},
            timeout=30
        )
        
        if response.status_code == 200:
            result = response.json()
            result["analyzed_via"] = "tpr-api"
            result["timestamp"] = datetime.now(timezone.utc).strftime("%Y-%m-%d %H:%M UTC")
            return result
        else:
            return None
            
    except:
        return None


def analyze_model_fallback(model_id: str) -> Optional[Dict[str, Any]]:
    """Fallback analysis using HuggingFace API directly."""
    if not model_id or not model_id.strip():
        return None
    
    model_id = model_id.strip()
    
    if not HF_AVAILABLE:
        return {"error": "Analysis unavailable - HuggingFace Hub not accessible"}
    
    try:
        from huggingface_hub import HfApi
        
        api = HfApi()
        info = api.model_info(model_id)
        
        card_data = info.card_data or {}
        license_val = card_data.get("license") or getattr(info, "license", None)
        datasets = card_data.get("datasets", []) or []
        author = info.author or ""
        tags = info.tags or []
        
        violations = []
        
        if not datasets:
            violations.append("NEC#1")
        if not license_val:
            violations.append("NEC#2")
        has_use = any("task" in t.lower() or "use" in t.lower() for t in tags)
        if not has_use:
            violations.append("NEC#7")
        if not author:
            violations.append("NEC#13")
        
        total_severity = sum(NECESSITY_CANON.get(v, {}).get("severity", 30) for v in violations)
        score = max(0, min(100, 100 - int(total_severity * 0.15)))
        
        if score >= 90:
            badge = "GOLD"
        elif score >= 75:
            badge = "SILVER"
        elif score >= 60:
            badge = "BRONZE"
        else:
            badge = "UNVERIFIED"
        
        return {
            "model_id": model_id,
            "score": score,
            "badge": badge,
            "violations": violations,
            "author": author or "Unknown",
            "license": license_val or "Not declared",
            "evidence": hashlib.sha256(f"{model_id}:{score}".encode()).hexdigest()[:12],
            "timestamp": datetime.now(timezone.utc).strftime("%Y-%m-%d %H:%M UTC"),
            "analyzed_via": "hf-fallback"
        }
        
    except Exception as e:
        return {"error": f"Model not found: {model_id}"}


def analyze_model(model_id: str) -> Optional[Dict[str, Any]]:
    """Main analysis function with TPR API and fallback."""
    result = analyze_model_live(model_id)
    
    if not result:
        result = analyze_model_fallback(model_id)
    
    return result


def get_badge_color(badge: str) -> str:
    colors = {
        "GOLD": "#F59E0B",
        "SILVER": "#94A3B8",
        "BRONZE": "#D97706",
        "UNVERIFIED": "#EF4444"
    }
    return colors.get(badge, "#6B7280")


def run_analysis(model_id: str):
    result = analyze_model(model_id)
    
    if not result:
        return "", "", ""
    
    if "error" in result:
        error_msg = result.get("message", result["error"])
        return f'<div class="error-box">{error_msg}</div>', "", ""
    
    color = get_badge_color(result["badge"])
    
    score_html = f'''
    <div class="result-card">
        <div class="score-circle" style="border-color: {color}; box-shadow: 0 0 30px {color}40;">
            <span class="score-value" style="color: {color};">{result["score"]}</span>
            <span class="score-label">SHADOW SCORE</span>
        </div>
        <div class="badge-pill" style="background: {color}20; color: {color};">{result["badge"]}</div>
    </div>
    '''
    
    info_html = f'''
    <div class="info-card">
        <div class="info-header">
            <span class="info-label">MODEL</span>
            <span class="info-value">{result["model_id"]}</span>
        </div>
        <div class="info-row">
            <span>Author: <strong>{result.get("author", "Unknown")}</strong></span>
        </div>
        <div class="info-row">
            <span>License: <strong>{result.get("license", "Not declared")}</strong></span>
        </div>
        <div class="evidence">
            Evidence: {result.get("evidence", "N/A")} · {result.get("timestamp", "")}
        </div>
    </div>
    '''
    
    if result.get("violations"):
        viol_items = "".join([f'<span class="viol-tag">{v}</span>' for v in result["violations"]])
        viol_html = f'<div class="violations-box"><span class="viol-label">VIOLATIONS:</span>{viol_items}</div>'
    else:
        viol_html = '<div class="clean-box">✓ ALL CHECKS PASSED</div>'
    
    if result.get("analyzed_via") == "hf-fallback":
        viol_html += '<div class="note-box">⚠️ Basic analysis mode (TPR API unavailable)</div>'
    
    return score_html, info_html, viol_html


def generate_cards_html(cards_data=None):
    if cards_data is None:
        cards_data = load_latest_cards_from_registry()
    
    cards = ""
    for c in cards_data:
        color = get_badge_color(c["badge"])
        viols = " ".join([f'<span class="card-viol">{v}</span>' for v in c["violations"]]) if c["violations"] else '<span class="card-clean">CLEAN</span>'
        cards += f'''
        <div class="model-card">
            <div class="card-header">
                <span class="card-score" style="color:{color};">{c["score"]}</span>
                <span class="card-badge" style="background:{color}20;color:{color};">{c["badge"]}</span>
            </div>
            <div class="card-model">{c["model_id"]}</div>
            <div class="card-meta">{c["author"]}</div>
            <div class="card-violations">{viols}</div>
        </div>
        '''
    return cards


CSS = """
.gradio-container {
    max-width: 100% !important;
    padding: 0 !important;
    margin: 0 !important;
    background: #030305 !important;
}
.main, .contain, .wrap {
    max-width: 100% !important;
    padding: 0 !important;
}
footer { display: none !important; }
.gr-form { background: transparent !important; border: none !important; }

.hero {
    background: linear-gradient(180deg, #0a0a12 0%, #030305 100%);
    padding: 60px 20px 40px;
    text-align: center;
    border-bottom: 1px solid #1a1a2e;
}
.hero-badge {
    display: inline-block;
    background: #8B5CF620;
    border: 1px solid #8B5CF640;
    color: #8B5CF6;
    padding: 6px 16px;
    border-radius: 20px;
    font-size: 11px;
    font-weight: 600;
    letter-spacing: 0.1em;
    text-transform: uppercase;
    margin-bottom: 20px;
}
.hero-logo {
    width: 70px;
    height: 70px;
    border-radius: 16px;
    margin: 0 auto 20px;
    background: linear-gradient(135deg, #8B5CF6 0%, #6D28D9 100%);
    display: flex;
    align-items: center;
    justify-content: center;
    box-shadow: 0 0 40px #8B5CF650;
}
.hero-logo img { width: 50px; height: 50px; }
.hero-title {
    font-size: 42px;
    font-weight: 900;
    color: #fff;
    letter-spacing: 0.15em;
    margin: 0 0 12px 0;
    text-shadow: 0 0 30px #8B5CF640;
}
.hero-subtitle {
    color: #9CA3AF;
    font-size: 18px;
    font-weight: 500;
    margin: 0 0 20px 0;
    letter-spacing: 0.02em;
}
.hero-desc {
    color: #6B7280;
    font-size: 15px;
    max-width: 700px;
    margin: 0 auto 30px;
    line-height: 1.7;
}
.hero-emphasis {
    color: #8B5CF6;
    font-weight: 600;
}
.hero-stats {
    display: flex;
    justify-content: center;
    gap: 40px;
    margin: 30px auto 0;
    max-width: 900px;
    flex-wrap: wrap;
}
.stat-item {
    text-align: center;
}
.stat-label {
    color: #6B7280;
    font-size: 11px;
    text-transform: uppercase;
    letter-spacing: 0.1em;
    font-weight: 600;
}
.stat-value {
    color: #fff;
    font-size: 16px;
    font-weight: 700;
    margin-top: 4px;
}
.stat-link {
    color: #8B5CF6;
    text-decoration: none;
    font-size: 16px;
    font-weight: 700;
}
.stat-link:hover {
    color: #A78BFA;
}

.nav {
    display: flex;
    justify-content: center;
    gap: 24px;
    padding: 16px;
    background: #050507;
    border-bottom: 1px solid #1a1a2e;
    flex-wrap: wrap;
}
.nav a {
    color: #9CA3AF;
    text-decoration: none;
    font-size: 13px;
    font-weight: 500;
    transition: color 0.2s;
}
.nav a:hover { color: #8B5CF6; }

.search-section {
    max-width: 700px;
    margin: 40px auto;
    padding: 0 20px;
}
.search-box {
    display: flex;
    gap: 12px;
}
input[type="text"], textarea {
    flex: 1;
    background: #0a0a12 !important;
    border: 2px solid #1a1a2e !important;
    color: #fff !important;
    border-radius: 12px !important;
    padding: 16px 20px !important;
    font-size: 16px !important;
}
input:focus, textarea:focus {
    border-color: #8B5CF6 !important;
    box-shadow: 0 0 20px #8B5CF620 !important;
    outline: none !important;
}
button.primary {
    background: linear-gradient(135deg, #8B5CF6 0%, #6D28D9 100%) !important;
    border: none !important;
    padding: 16px 32px !important;
    border-radius: 12px !important;
    font-weight: 700 !important;
    font-size: 14px !important;
    letter-spacing: 0.05em !important;
    color: #fff !important;
    cursor: pointer !important;
    transition: all 0.2s;
}
button.primary:hover {
    transform: translateY(-2px);
    box-shadow: 0 10px 30px #8B5CF650 !important;
}

.results-section {
    max-width: 800px;
    margin: 0 auto 40px;
    padding: 0 20px;
}
.result-card {
    background: #0a0a12;
    border: 1px solid #1a1a2e;
    border-radius: 16px;
    padding: 30px;
    text-align: center;
}
.score-circle {
    width: 140px;
    height: 140px;
    border-radius: 50%;
    border: 4px solid;
    display: flex;
    flex-direction: column;
    align-items: center;
    justify-content: center;
    margin: 0 auto 20px;
    background: #050507;
}
.score-value { font-size: 48px; font-weight: 900; }
.score-label { font-size: 10px; color: #6B7280; letter-spacing: 0.1em; }
.badge-pill {
    display: inline-block;
    padding: 8px 24px;
    border-radius: 20px;
    font-weight: 700;
    font-size: 14px;
    letter-spacing: 0.1em;
}
.info-card {
    background: #0a0a12;
    border: 1px solid #1a1a2e;
    border-radius: 16px;
    padding: 20px;
    margin-top: 16px;
}
.info-header { margin-bottom: 12px; }
.info-label { color: #8B5CF6; font-size: 10px; letter-spacing: 0.1em; display: block; }
.info-value { color: #fff; font-size: 18px; font-weight: 700; }
.info-row { color: #9CA3AF; font-size: 13px; margin: 8px 0; }
.info-row strong { color: #fff; }
.evidence { color: #4B5563; font-size: 11px; font-family: monospace; margin-top: 12px; padding-top: 12px; border-top: 1px solid #1a1a2e; }
.violations-box { background: #1a0a0a; border: 1px solid #EF444440; border-radius: 12px; padding: 16px; margin-top: 16px; }
.viol-label { color: #EF4444; font-size: 11px; letter-spacing: 0.1em; margin-right: 12px; }
.viol-tag { background: #EF444420; color: #EF4444; padding: 4px 12px; border-radius: 6px; font-size: 12px; font-weight: 600; margin: 0 4px; }
.clean-box { background: #0a1a0a; border: 1px solid #22C55E40; border-radius: 12px; padding: 16px; margin-top: 16px; color: #22C55E; text-align: center; font-weight: 600; }
.note-box { background: #1a1a0a; border: 1px solid #F59E0B40; border-radius: 12px; padding: 12px; margin-top: 12px; color: #F59E0B; text-align: center; font-size: 12px; }
.error-box { background: #1a0a0a; border: 1px solid #EF4444; border-radius: 12px; padding: 20px; color: #EF4444; text-align: center; }

.cards-section {
    background: #050507;
    padding: 40px 20px;
    border-top: 1px solid #1a1a2e;
}
.cards-title {
    text-align: center;
    color: #fff;
    font-size: 24px;
    font-weight: 700;
    margin-bottom: 8px;
}
.cards-sub {
    text-align: center;
    color: #6B7280;
    font-size: 13px;
    margin-bottom: 30px;
}
.cards-grid {
    display: grid;
    grid-template-columns: repeat(auto-fill, minmax(280px, 1fr));
    gap: 16px;
    max-width: 1200px;
    margin: 0 auto;
}
.model-card {
    background: #0a0a12;
    border: 1px solid #1a1a2e;
    border-radius: 12px;
    padding: 20px;
    transition: transform 0.2s, box-shadow 0.2s;
}
.model-card:hover {
    transform: translateY(-4px);
    box-shadow: 0 10px 30px rgba(139, 92, 246, 0.15);
    border-color: #8B5CF640;
}
.card-header { display: flex; justify-content: space-between; align-items: center; margin-bottom: 12px; }
.card-score { font-size: 32px; font-weight: 900; }
.card-badge { padding: 4px 12px; border-radius: 6px; font-size: 11px; font-weight: 700; }
.card-model { color: #fff; font-size: 14px; font-weight: 600; margin-bottom: 4px; word-break: break-all; }
.card-meta { color: #6B7280; font-size: 12px; margin-bottom: 12px; }
.card-violations { display: flex; flex-wrap: wrap; gap: 6px; }
.card-viol { background: #EF444420; color: #EF4444; padding: 2px 8px; border-radius: 4px; font-size: 10px; font-weight: 600; }
.card-clean { background: #22C55E20; color: #22C55E; padding: 2px 8px; border-radius: 4px; font-size: 10px; font-weight: 600; }

.site-footer {
    background: #030305;
    border-top: 1px solid #1a1a2e;
    padding: 40px 20px 30px;
    text-align: center;
}
.footer-tagline {
    color: #8B5CF6;
    font-size: 14px;
    font-weight: 600;
    margin-bottom: 20px;
    letter-spacing: 0.05em;
}
.footer-links { display: flex; justify-content: center; gap: 24px; margin-bottom: 20px; flex-wrap: wrap; }
.footer-links a { color: #6B7280; text-decoration: none; font-size: 13px; }
.footer-links a:hover { color: #8B5CF6; }
.footer-copy { color: #4B5563; font-size: 11px; margin-bottom: 8px; }
.footer-protocol { color: #374151; font-size: 10px; font-family: monospace; }

@media (max-width: 640px) {
    .hero-title { font-size: 32px; }
    .hero-subtitle { font-size: 16px; }
    .hero-stats { gap: 20px; }
    .nav { gap: 16px; }
    .search-box { flex-direction: column; }
    button.primary { width: 100%; }
    .cards-grid { grid-template-columns: 1fr; }
}
"""

HEADER_HTML = f'''
<div class="hero">
    <div class="hero-badge">🔴 FOUNDING STAGE · PROTOCOL {PROTOCOL_VERSION}</div>
    <div class="hero-logo">
        <img src="https://huggingface.co/spaces/Crovia/omission-oracle/resolve/main/crovia-logo.png" alt="Crovia" onerror="this.style.display='none'">
    </div>
    <h1 class="hero-title">CROVIA</h1>
    <p class="hero-subtitle">The Evidence Protocol for AI Training</p>
    <p class="hero-desc">
        We're building the <span class="hero-emphasis">global standard</span> for AI compliance verification 
        — in public, from day one. Cryptographic evidence. Multi-regulation. Open protocol.
    </p>
    <div class="hero-stats">
        <div class="stat-item">
            <div class="stat-label">Protocol</div>
            <div class="stat-value">{PROTOCOL_VERSION} · Jan 2026</div>
        </div>
        <div class="stat-item">
            <div class="stat-label">Public Dataset</div>
            <a href="https://huggingface.co/datasets/{DATASET_ID}" target="_blank" class="stat-link">
                Updated Every 6h →
            </a>
        </div>
        <div class="stat-item">
            <div class="stat-label">Open Source</div>
            <a href="https://github.com/croviatrust" target="_blank" class="stat-link">
                GitHub →
            </a>
        </div>
    </div>
</div>
<div class="nav">
    <a href="https://croviatrust.com" target="_blank">🏠 Home</a>
    <a href="https://huggingface.co/datasets/{DATASET_ID}" target="_blank">📊 Rankings</a>
    <a href="https://huggingface.co/datasets/Crovia/cep-capsules" target="_blank">📦 Evidence Capsules</a>
    <a href="https://huggingface.co/spaces/Crovia/cep-terminal" target="_blank">🔍 CEP Terminal</a>
    <a href="https://github.com/croviatrust" target="_blank">⚙️ GitHub</a>
    <a href="https://github.com/croviatrust/CROVIA_DEV/blob/master/CROVIA_PROTOCOL_SPEC.md" target="_blank">📄 Protocol Spec</a>
</div>
'''

CARDS_HTML = f'''
<div class="cards-section">
    <h2 class="cards-title">Latest Oracle Cards</h2>
    <p class="cards-sub">Auto-generated evidence · Public dataset · Updated every 6 hours</p>
    <div class="cards-grid">
        {generate_cards_html()}
    </div>
</div>
'''

FOOTER_HTML = '''
<div class="site-footer">
    <div class="footer-tagline">Evidence, not promises. Observation, not judgment.</div>
    <div class="footer-links">
        <a href="https://croviatrust.com">Official Site</a>
        <a href="https://github.com/croviatrust">GitHub</a>
        <a href="https://huggingface.co/datasets/Crovia/global-ai-training-omissions">Dataset</a>
        <a href="https://github.com/croviatrust/CROVIA_DEV/blob/master/CROVIA_PROTOCOL_SPEC.md">Protocol Spec</a>
        <a href="https://github.com/croviatrust/CROVIA_DEV/blob/master/TPR_RESOLUTION_CONTRACT.md">TPR Contract</a>
    </div>
    <p class="footer-copy">© 2026 Crovia Protocol · Apache 2.0 License</p>
    <p class="footer-protocol">Built in public by contributors worldwide.</p>
</div>
'''

with gr.Blocks(css=CSS, title="CROVIA · Omission Oracle") as app:
    
    gr.HTML(HEADER_HTML)
    
    with gr.Column(elem_classes="search-section"):
        with gr.Row(elem_classes="search-box"):
            model_input = gr.Textbox(
                placeholder="Enter model ID (e.g., meta-llama/Llama-3.1-8B)",
                show_label=False,
                container=False,
            )
            analyze_btn = gr.Button("🔍 ANALYZE", variant="primary")
    
    with gr.Column(elem_classes="results-section"):
        score_output = gr.HTML()
        info_output = gr.HTML()
        viol_output = gr.HTML()
    
    gr.HTML(CARDS_HTML)
    gr.HTML(FOOTER_HTML)
    
    analyze_btn.click(run_analysis, inputs=[model_input], outputs=[score_output, info_output, viol_output])
    model_input.submit(run_analysis, inputs=[model_input], outputs=[score_output, info_output, viol_output])

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