app.py

import gradio as gr
from datetime import datetime
from typing import Any, Dict, Iterable, List, Optional, Tuple
from collections import Counter
import json
import os
import html as html_lib

from huggingface_hub import HfApi, InferenceClient


def _created_year(obj):
    if hasattr(obj, "created_at"):
        dt = getattr(obj, "created_at")
        return dt.year

def _repo_id(obj: Any) -> str:
    if isinstance(obj, dict):
        return obj.get("id") or obj.get("modelId") or obj.get("repoId") or "N/A"
    return (
        getattr(obj, "id", None)
        or getattr(obj, "modelId", None)
        or getattr(obj, "repoId", None)
        or getattr(obj, "repo_id", None)
        or "N/A"
    )

def _repo_likes(obj: Any) -> int:
    return int(getattr(obj, "likes", 0) or 0)

def _repo_tags(obj: Any) -> List[str]:
    tags = getattr(obj, "tags", None) or []
    return [t for t in tags if isinstance(t, str)]

def _repo_pipeline_tag(obj: Any) -> Optional[str]:
    val = getattr(obj, "pipeline_tag", None)
    return val 

def _collect_2025_sorted_desc(items: Iterable[Any]) -> List[Any]:
    """
    We rely on API-side sorting (createdAt desc) + early-stop once we hit < 2025.
    This avoids pulling a user's entire history.
    """
    out: List[Any] = []
    for item in items:
        yr = _created_year(item)
        if yr is None:
            continue
        if yr < 2025:
            break
        if yr == 2025:
            out.append(item)
    return out

def fetch_user_data_2025(username: str, token: Optional[str] = None) -> Dict[str, List[Any]]:
    """Fetch user's models/datasets/spaces created in 2025 (API-side sort + paginated early-stop)."""
    api = HfApi(token=token)
    data: Dict[str, List[Any]] = {"models": [], "datasets": [], "spaces": []}

    data["models"] = _collect_2025_sorted_desc(
        api.list_models(author=username, full=True, sort="createdAt", direction=-1)
    )

    data["datasets"] = _collect_2025_sorted_desc(
        api.list_datasets(author=username, full=True, sort="createdAt", direction=-1)
    )


    data["spaces"] = _collect_2025_sorted_desc(
        api.list_spaces(author=username, full=True, sort="createdAt", direction=-1)
    )

    return data

def get_most_liked_item(items: List[Dict]) -> Optional[Dict]:
    """Get the item with most likes"""
    return max(items, key=lambda x: x.get("likes", 0))

def _normalize_task_tag(tag: str) -> Optional[str]:
    t = (tag or "").strip()
    if not t:
        return None
    for prefix in ("task_categories:", "task_ids:", "pipeline_tag:"):
        if t.startswith(prefix):
            t = t[len(prefix):].strip()
    t = t.strip().lower()
    return t or None

def _suggested_nickname_for_task(task: Optional[str]) -> Optional[str]:
    t = task.strip().lower()
    mapping = {
        "text-generation": "LLM Whisperer 🗣️",
        "image-text-to-text": "VLM Nerd 🤓",
        "text-to-speech": "Full‑time Yapper 🗣️",
        "automatic-speech-recognition": "Subtitle Goblin 🎧",
        "text-to-image": "Diffusion Gremlin 🎨",
        "image-classification": "Pixel Judge 👁️",
        "token-classification": "NERd Lord 🤓",
        "text-classification": "Opinion Machine 🧠",
        "translation": "Language Juggler 🗺️",
        "summarization": "TL;DR Dealer ✍️",
        "image-to-text": "Caption Connoisseur 🖼️",
        "zero-shot-classification": "Label Wizard 🪄",
    }
    return mapping.get(t)

def infer_task_and_modality(models: List[Any], datasets: List[Any], spaces: List[Any]) -> Tuple[Optional[str], Optional[str], Counter]:
    """
    Returns: (most_common_task, task_counter)
    - Task is primarily inferred from model `pipeline_tag`, then from task-ish tags on all artifacts.
    """
    model_tasks: List[str] = []
    for m in models:
        pt = _repo_pipeline_tag(m)
        if pt:
            model_tasks.append(pt.strip().lower())

    tag_tasks: List[str] = []
    for obj in (models + datasets + spaces):
        for tag in _repo_tags(obj):
            nt = _normalize_task_tag(tag)
            if nt:
                tag_tasks.append(nt)

    counts = Counter(model_tasks if model_tasks else tag_tasks)
    top_task = counts.most_common(1)[0][0] if counts else None

    return top_task, counts

def _k2_model_candidates() -> List[str]:
    """
    Kimi K2 repo IDs can vary; allow override via env and try a small list.
    """
    env_model = "moonshotai/Kimi-K2-Instruct"
    # de-dupe while preserving order
    seen = set()
    out = []
    for c in candidates:
        if c and c not in seen:
            out.append(c)
            seen.add(c)
    return out

def _esc(value: Any) -> str:
    if value is None:
        return ""
    return html_lib.escape(str(value), quote=True)

def _profile_username(profile: Any) -> Optional[str]:
    if profile is None:
        return None
    for key in ("username", "preferred_username", "name", "user", "handle"):
        val = getattr(profile, key, None)
        if isinstance(val, str) and val.strip():
            return val.strip().lstrip("@")
    data = getattr(profile, "data", None)
    if isinstance(data, dict):
        for key in ("username", "preferred_username", "name"):
            val = data.get(key)
            if isinstance(val, str) and val.strip():
                return val.strip().lstrip("@")
        for container in ("profile", "user"):
            blob = data.get(container)
            if isinstance(blob, dict):
                val = blob.get("username") or blob.get("preferred_username") or blob.get("name")
                if isinstance(val, str) and val.strip():
                    return val.strip().lstrip("@")
    if isinstance(profile, dict):
        val = profile.get("username") or profile.get("preferred_username") or profile.get("name")
        if isinstance(val, str) and val.strip():
            return val.strip().lstrip("@")
    return None

def _profile_token(profile: Any) -> Optional[str]:
    """
    Gradio's OAuth payload varies by version.
    We try common attribute names and `.data` shapes.
    """
    if profile is None:
        return None
    for key in ("token", "access_token", "hf_token", "oauth_token", "oauth_access_token"):
        val = getattr(profile, key, None)
        if isinstance(val, str) and val.strip():
            return val.strip()
    data = getattr(profile, "data", None)
    if isinstance(data, dict):
        for key in ("token", "access_token", "hf_token", "oauth_token", "oauth_access_token"):
            val = data.get(key)
            if isinstance(val, str) and val.strip():
                return val.strip()
        # Common nested objects
        oauth_info = data.get("oauth_info") or data.get("oauth") or data.get("oauthInfo") or {}
        if isinstance(oauth_info, dict):
            val = oauth_info.get("access_token") or oauth_info.get("token")
            if isinstance(val, str) and val.strip():
                return val.strip()
    if isinstance(profile, dict):
        val = profile.get("token") or profile.get("access_token")
        if isinstance(val, str) and val.strip():
            return val.strip()
    return None

def generate_roast_and_nickname_with_k2(
    *,
    username: str,
    total_artifacts_2025: int,
    models_2025: int,
    datasets_2025: int,
    spaces_2025: int,
    top_task: Optional[str],
) -> Tuple[Optional[str], Optional[str]]:
    """
    Calls Kimi K2 via Hugging Face Inference Providers (via huggingface_hub InferenceClient).
    Returns (nickname, roast). If call fails, returns (None, None).
    """
    token = (os.getenv("HF_TOKEN") or "").strip()
    if not token:
        return None, None

    vibe = top_task or "mysterious vibes"
    above_below = "above" if total_artifacts_2025 > 20 else "below"
    suggested = _suggested_nickname_for_task(top_task)

    system = (
        "You are a witty, playful roast-comedian. Keep it fun, not cruel. "
        "No slurs, no hate, no harassment. Avoid profanity. Keep it short."
    )
    user = f"""
Create TWO things about this Hugging Face user, based on their 2025 activity stats.

User: @{username}
Artifacts created in 2025: {total_artifacts_2025} (models={models_2025}, datasets={datasets_2025}, spaces={spaces_2025}) which is {above_below} 20.
Top task (pipeline_tag): {top_task or "unknown"}

Nickname guidance (examples you SHOULD follow when applicable):
- text-generation -> LLM Whisperer 🗣️
- image-text-to-text -> VLM Nerd 🤓
- text-to-speech -> Full‑time Yapper 🗣️

If top task is known and you have a strong matching idea, pick a nickname like the examples. {f'If unsure, you may use this suggested nickname: {suggested}' if suggested else ''}
Roast should reference the task and whether they are above/below 20 artifacts.
Most common vibe: {vibe}

Return ONLY valid JSON with exactly these keys:
{{
  "nickname": "...",   // short, funny, can include 1 emoji
  "roast": "..."       // 1-2 sentences max, playful, no bullying
}}
""".strip()

    client = InferenceClient(model="moonshotai/Kimi-K2-Instruct", token=token)
    resp = client.chat.completions.create(
        model="moonshotai/Kimi-K2-Instruct",
        messages=[
            {"role": "system", "content": system},
            {"role": "user", "content": user},
        ],
        max_tokens=180,
        temperature=0.8,
    )
    content = (resp.choices[0].message.content or "").strip()

    payload = json.loads(content)
    nickname = payload.get("nickname")
    roast = payload.get("roast")
    nickname_out = nickname.strip() if isinstance(nickname, str) else None
    roast_out = roast.strip() if isinstance(roast, str) else None
    return nickname_out, roast_out

def generate_wrapped_report(profile: gr.OAuthProfile) -> str:
    """Generate the HF Wrapped 2025 report"""
    username = _profile_username(profile) or "unknown"
    token = _profile_token(profile)

    # Fetch 2025 data (API-side sort + early stop)
    user_data_2025 = fetch_user_data_2025(username, token)
    models_2025 = user_data_2025["models"]
    datasets_2025 = user_data_2025["datasets"]
    spaces_2025 = user_data_2025["spaces"]

    most_liked_model = max(models_2025, key=_repo_likes) if models_2025 else None
    most_liked_dataset = max(datasets_2025, key=_repo_likes) if datasets_2025 else None
    most_liked_space = max(spaces_2025, key=_repo_likes) if spaces_2025 else None

    total_likes = sum(_repo_likes(x) for x in (models_2025 + datasets_2025 + spaces_2025))

    top_task, _task_counts = infer_task_and_modality(models_2025, datasets_2025, spaces_2025)

    total_artifacts_2025 = len(models_2025) + len(datasets_2025) + len(spaces_2025)
    nickname, roast = generate_roast_and_nickname_with_k2(
        username=username,
        total_artifacts_2025=total_artifacts_2025,
        models_2025=len(models_2025),
        datasets_2025=len(datasets_2025),
        spaces_2025=len(spaces_2025),
        top_task=top_task,
    )

    # Create HTML report
    html = f"""
    <!DOCTYPE html>
    <html>
    <head>
        <style>
            @import url('https://fonts.googleapis.com/css2?family=Poppins:wght@300;400;600;700;800&display=swap');

            body {{
                font-family: 'Poppins', sans-serif;
                background: linear-gradient(135deg, #667eea 0%, #764ba2 50%, #f093fb 100%);
                margin: 0;
                padding: 20px;
                min-height: 100vh;
            }}

            .container {{
                max-width: 800px;
                margin: 0 auto;
                background: rgba(255, 255, 255, 0.95);
                border-radius: 30px;
                padding: 40px;
                box-shadow: 0 20px 60px rgba(0, 0, 0, 0.3);
                animation: fadeIn 0.8s ease-in;
            }}

            @keyframes fadeIn {{
                from {{ opacity: 0; transform: translateY(20px); }}
                to {{ opacity: 1; transform: translateY(0); }}
            }}

            .header {{
                text-align: center;
                margin-bottom: 40px;
            }}

            .header h1 {{
                font-size: 3em;
                background: linear-gradient(135deg, #667eea 0%, #764ba2 100%);
                -webkit-background-clip: text;
                -webkit-text-fill-color: transparent;
                margin: 0;
                font-weight: 800;
                animation: slideDown 0.6s ease-out;
            }}

            @keyframes slideDown {{
                from {{ transform: translateY(-30px); opacity: 0; }}
                to {{ transform: translateY(0); opacity: 1; }}
            }}

            .username {{
                font-size: 1.5em;
                color: #764ba2;
                margin-top: 10px;
                font-weight: 600;
            }}

            .nickname {{
                font-size: 1.1em;
                color: #111 !important;
                margin-top: 8px;
                font-weight: 700;
                background: #ffffff !important;
                display: inline-block;
                padding: 6px 12px;
                border-radius: 999px;
                border: 1px solid rgba(245, 87, 108, 0.25);
                box-shadow: 0 8px 18px rgba(0, 0, 0, 0.08);
            }}

            .year {{
                font-size: 4em;
                font-weight: 800;
                text-align: center;
                background: linear-gradient(135deg, #f093fb 0%, #f5576c 100%);
                -webkit-background-clip: text;
                -webkit-text-fill-color: transparent;
                margin: 30px 0;
                animation: pulse 2s ease-in-out infinite;
            }}

            @keyframes pulse {{
                0%, 100% {{ transform: scale(1); }}
                50% {{ transform: scale(1.05); }}
            }}

            .stats-grid {{
                display: grid;
                grid-template-columns: repeat(auto-fit, minmax(200px, 1fr));
                gap: 20px;
                margin: 30px 0;
            }}

            .stat-card {{
                background: linear-gradient(135deg, #667eea 0%, #764ba2 100%);
                color: white;
                padding: 25px;
                border-radius: 20px;
                text-align: center;
                box-shadow: 0 10px 25px rgba(102, 126, 234, 0.3);
                transition: transform 0.3s ease, box-shadow 0.3s ease;
                animation: popIn 0.5s ease-out backwards;
            }}

            .stat-card:nth-child(1) {{ animation-delay: 0.1s; }}
            .stat-card:nth-child(2) {{ animation-delay: 0.2s; }}
            .stat-card:nth-child(3) {{ animation-delay: 0.3s; }}

            @keyframes popIn {{
                from {{ transform: scale(0.8); opacity: 0; }}
                to {{ transform: scale(1); opacity: 1; }}
            }}

            .stat-card:hover {{
                transform: translateY(-5px) scale(1.05);
                box-shadow: 0 15px 35px rgba(102, 126, 234, 0.4);
            }}

            .stat-number {{
                font-size: 3em;
                font-weight: 800;
                margin: 10px 0;
            }}

            .stat-label {{
                font-size: 1em;
                font-weight: 600;
                text-transform: uppercase;
                letter-spacing: 1px;
            }}

            .section {{
                margin: 40px 0;
                padding: 25px;
                background: #ffffff !important;
                border-radius: 20px;
                animation: slideIn 0.6s ease-out;
                color: #111 !important;
                border: 1px solid rgba(17, 17, 17, 0.08);
                box-shadow: 0 12px 30px rgba(0, 0, 0, 0.10);
                border-top: 6px solid rgba(102, 126, 234, 0.85);
            }}

            @keyframes slideIn {{
                from {{ transform: translateX(-30px); opacity: 0; }}
                to {{ transform: translateX(0); opacity: 1; }}
            }}

            .section h2 {{
                color: #1f1b5a !important;
                font-size: 1.8em;
                margin-top: 0;
                font-weight: 700;
                display: flex;
                align-items: center;
                gap: 10px;
            }}

            .trophy {{
                font-size: 1.5em;
            }}

            .item {{
                background: #ffffff !important;
                padding: 20px;
                margin: 15px 0;
                border-radius: 15px;
                box-shadow: 0 5px 15px rgba(0, 0, 0, 0.1);
                transition: transform 0.2s ease;
                border: 1px solid rgba(17, 17, 17, 0.08);
            }}

            .item:hover {{
                transform: translateX(10px);
            }}

            .item-name {{
                font-weight: 600;
                font-size: 1.2em;
                color: #111 !important;
                margin-bottom: 5px;
            }}

            .item-likes {{
                color: #d92d20 !important;
                font-weight: 600;
                font-size: 1.1em;
            }}

            .item-sub {{
                color: #1f2937 !important;
                font-weight: 600;
                font-size: 1.05em;
            }}

            .emoji {{
                font-size: 1.5em;
                margin-right: 10px;
            }}

            .total-likes {{
                text-align: center;
                margin: 40px 0;
                padding: 30px;
                background: linear-gradient(135deg, #f093fb 0%, #f5576c 100%);
                border-radius: 20px;
                color: white;
            }}

            .total-likes-number {{
                font-size: 4em;
                font-weight: 800;
                margin: 10px 0;
            }}

            .total-likes-label {{
                font-size: 1.3em;
                font-weight: 600;
            }}

            .footer {{
                text-align: center;
                margin-top: 40px;
                color: #111 !important;
                font-weight: 600;
                background: #ffffff !important;
                border: 1px solid rgba(17, 17, 17, 0.08);
                border-radius: 16px;
                padding: 16px 18px;
                box-shadow: 0 10px 24px rgba(0, 0, 0, 0.08);
            }}

            .footer p {{
                margin: 8px 0;
                color: #111 !important;
                opacity: 1 !important;
                font-size: 1.05em;
                line-height: 1.35;
            }}

            .no-data {{
                text-align: center;
                color: #111 !important;
                font-style: italic;
                padding: 20px;
            }}

            .roast {{
                font-size: 1.15em;
                line-height: 1.5;
                color: #111 !important;
                background: #fff0f3 !important;
                border-left: 6px solid #f5576c;
                padding: 18px 18px;
                border-radius: 14px;
                margin-top: 10px;
                border: 1px solid rgba(245, 87, 108, 0.25);
            }}
        </style>
    </head>
    <body>
        <div class="container">
            <div class="header">
                <h1>🎉 HF WRAPPED 🎉</h1>
                <div class="username">@{username}</div>
                {f'<div class="nickname">You are a {_esc(nickname)}</div>' if nickname else ''}
            </div>

            <div class="year">2025</div>

            <div class="stats-grid">
                <div class="stat-card">
                    <div class="stat-number">{len(models_2025)}</div>
                    <div class="stat-label">🤖 Models</div>
                </div>
                <div class="stat-card">
                    <div class="stat-number">{len(datasets_2025)}</div>
                    <div class="stat-label">📊 Datasets</div>
                </div>
                <div class="stat-card">
                    <div class="stat-number">{len(spaces_2025)}</div>
                    <div class="stat-label">🚀 Spaces</div>
                </div>
            </div>

            <div class="section">
                <h2><span class="trophy">🧠</span> Your Signature Vibe</h2>
                {f'''
                <div class="item">
                    <div class="item-name"><span class="emoji">🎯</span>Most common task: {_esc(top_task)}</div>
                    <div class="item-sub">Total 2025 artifacts: {total_artifacts_2025}</div>
                </div>
                ''' if top_task else '<div class="no-data">Not enough metadata to infer your vibe (yet).</div>'}
            </div>

            <div class="total-likes">
                <div class="total-likes-number">❤️ {total_likes}</div>
                <div class="total-likes-label">Total Likes Received</div>
            </div>

            <div class="section">
                <h2><span class="trophy">🔥</span> Roast (Kimi K2)</h2>
                {f'<div class="roast">{_esc(roast)}</div>' if roast else '<div class="no-data">Couldn’t generate a roast (missing token or Kimi K2 not reachable).</div>'}
            </div>

            <div class="section">
                <h2><span class="trophy">🏆</span> Most Liked Model</h2>
                {f'''
                <div class="item">
                    <div class="item-name"><span class="emoji">🤖</span>{_repo_id(most_liked_model)}</div>
                    <div class="item-likes">❤️ {_repo_likes(most_liked_model)} likes</div>
                </div>
                ''' if most_liked_model else '<div class="no-data">No models yet</div>'}
            </div>

            <div class="section">
                <h2><span class="trophy">🏆</span> Most Liked Dataset</h2>
                {f'''
                <div class="item">
                    <div class="item-name"><span class="emoji">📊</span>{_repo_id(most_liked_dataset)}</div>
                    <div class="item-likes">❤️ {_repo_likes(most_liked_dataset)} likes</div>
                </div>
                ''' if most_liked_dataset else '<div class="no-data">No datasets yet</div>'}
            </div>

            <div class="section">
                <h2><span class="trophy">🏆</span> Most Liked Space</h2>
                {f'''
                <div class="item">
                    <div class="item-name"><span class="emoji">🚀</span>{_repo_id(most_liked_space)}</div>
                    <div class="item-likes">❤️ {_repo_likes(most_liked_space)} likes</div>
                </div>
                ''' if most_liked_space else '<div class="no-data">No spaces yet</div>'}
            </div>

            <div class="footer">
                <p>🎊 Thank you for being part of the Hugging Face community! 🎊</p>
                <p>Keep building amazing things in 2025! 🚀</p>
                <p>Made with Inference Providers with love 💖</p>
            </div>
        </div>
    </body>
    </html>
    """

    return html

def show_login_message():
    """Show message for non-logged-in users"""
    return """
    <div style="text-align: center; padding: 50px; font-family: 'Poppins', sans-serif;">
        <h1 style="color: #667eea; font-size: 3em;">🎉 Welcome to HF Wrapped 2025! 🎉</h1>
        <p style="font-size: 1.5em; color: #764ba2;">
            Please log in with your Hugging Face account to see your personalized report!
        </p>
        <p style="font-size: 1.2em; color: #666;">
            Click the "Sign in with Hugging Face" button above 👆
        </p>
    </div>
    """

# Create Gradio interface
with gr.Blocks(theme=gr.themes.Soft(), css="""
    .gradio-container {
        background: linear-gradient(135deg, #667eea 0%, #764ba2 50%, #f093fb 100%);
    }
""") as demo:
    gr.HTML("""
        <div style="text-align: center; padding: 20px; color: white;">
            <h1 style="font-size: 3em; margin: 0;">🎉 HF Wrapped 2025 🎉</h1>
            <p style="font-size: 1.2em;">Discover your Hugging Face journey this year!</p>
        </div>
    """)

    with gr.Row():
        with gr.Column():
            login_button = gr.LoginButton()
            output = gr.HTML(value=show_login_message())

    def _render(profile_obj: Optional[gr.OAuthProfile] = None):
        # In Gradio versions that support OAuth, `profile_obj` is injected after login.
        return generate_wrapped_report(profile_obj) if profile_obj is not None else show_login_message()

    # On load show the login message (and in some Gradio versions, this also receives the injected profile)
    demo.load(fn=_render, inputs=None, outputs=output)

    # After login completes, clicking the login button will trigger a rerender.
    # Older Gradio treats LoginButton as a button (click event), not a value component (change event).
    if hasattr(login_button, "click"):
        login_button.click(fn=_render, inputs=None, outputs=output)

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