Update miniapp_leaderboard.py

miniapp_leaderboard.py

@@ -6,6 +6,411 @@ import re
 import uuid
 from urllib.parse import urlparse
 
+import gradio as gr
+import pandas as pd
+import requests
+from huggingface_hub import HfApi, hf_hub_download
+from huggingface_hub.errors import RepositoryNotFoundError
+
+
+APP_NAME = "miniapp"
+
+# Space Secrets:
+# - HF_TOKEN: Hugging Face access token with WRITE access to the dataset repo
+# - LEADERBOARD_DATASET: dataset repo id like "ha251/miniapp-leaderboard"
+HF_TOKEN = os.environ.get("HF_TOKEN") or os.environ.get("TOKEN") or os.environ.get("HUGGINGFACE_TOKEN")
+LEADERBOARD_DATASET = (os.environ.get("LEADERBOARD_DATASET") or "").strip()
+MAX_ENTRIES = int(os.environ.get("MAX_ENTRIES", "500"))
+
+ENTRIES_PREFIX = "entries/"
+
+LEADERBOARD_COLUMNS = [
+    "Model name",
+    "Avg",
+    "Easy",
+    "Mid",
+    "Hard",
+    "Games",
+    "Science",
+    "Tools",
+    "Humanities",
+    "Viz",
+    "Lifestyle",
+    "Submitted at",
+    "Submitter",
+]
+
+NUMERIC_COLS = [
+    "Avg",
+    "Easy",
+    "Mid",
+    "Hard",
+    "Games",
+    "Science",
+    "Tools",
+    "Humanities",
+    "Viz",
+    "Lifestyle",
+]
+
+IN_SPACES = bool(
+    os.environ.get("SPACE_ID")
+    or os.environ.get("SPACE_REPO_NAME")
+    or os.environ.get("SPACE_AUTHOR_NAME")
+    or os.environ.get("system", "") == "spaces"
+)
+
+
+def _api() -> HfApi:
+    return HfApi(token=HF_TOKEN)
+
+
+def _is_valid_http_url(url: str) -> bool:
+    try:
+        parsed = urlparse(url)
+        return parsed.scheme in ("http", "https") and bool(parsed.netloc)
+    except Exception:
+        return False
+
+
+def _slug(s: str, max_len: int = 60) -> str:
+    s = (s or "").strip().lower()
+    s = re.sub(r"[^a-z0-9]+", "-", s)
+    s = re.sub(r"-{2,}", "-", s).strip("-")
+    return (s[:max_len] or "x")
+
+
+def _empty_df() -> pd.DataFrame:
+    return pd.DataFrame(columns=LEADERBOARD_COLUMNS)
+
+
+def _ensure_dataset_readable() -> tuple[bool, str]:
+    if not HF_TOKEN:
+        return False, "Space is missing HF_TOKEN (Secrets)."
+    if not LEADERBOARD_DATASET:
+        return False, "Space is missing LEADERBOARD_DATASET (Secrets)."
+    api = _api()
+    try:
+        api.repo_info(repo_id=LEADERBOARD_DATASET, repo_type="dataset")
+        return True, ""
+    except RepositoryNotFoundError:
+        return False, (
+            f"Dataset repo not found: {LEADERBOARD_DATASET}. "
+            "Create it first (as a dataset) or fix LEADERBOARD_DATASET."
+        )
+    except Exception:
+        return False, "Cannot access the dataset repo. Check token permissions."
+
+
+def _list_entry_files() -> list[str]:
+    ok, _ = _ensure_dataset_readable()
+    if not ok:
+        return []
+    api = _api()
+    try:
+        files = api.list_repo_files(repo_id=LEADERBOARD_DATASET, repo_type="dataset")
+    except Exception:
+        return []
+
+    entry_files = [f for f in files if f.startswith(ENTRIES_PREFIX) and f.endswith(".json")]
+    entry_files.sort(reverse=True)
+    return entry_files[:MAX_ENTRIES]
+
+
+def _load_entries_df() -> pd.DataFrame:
+    ok, _ = _ensure_dataset_readable()
+    if not ok:
+        return _empty_df()
+
+    rows: list[dict] = []
+    for filename in _list_entry_files():
+        try:
+            path = hf_hub_download(
+                repo_id=LEADERBOARD_DATASET,
+                repo_type="dataset",
+                filename=filename,
+                token=HF_TOKEN,
+            )
+            with open(path, "r", encoding="utf-8") as fp:
+                row = json.load(fp)
+                rows.append(row)
+        except Exception:
+            continue
+
+    if not rows:
+        return _empty_df()
+
+    df = pd.DataFrame(rows)
+    for c in LEADERBOARD_COLUMNS:
+        if c not in df.columns:
+            df[c] = ""
+    df = df[LEADERBOARD_COLUMNS]
+    for c in NUMERIC_COLS:
+        df[c] = pd.to_numeric(df[c], errors="coerce")
+    df = df.sort_values(by=["Submitted at"], ascending=False, kind="stable")
+    return df
+
+
+def _apply_search_and_sort(
+    df: pd.DataFrame,
+    search_text: str,
+    search_in: str,
+    sort_by: str,
+    sort_order: str,
+) -> pd.DataFrame:
+    search_text = (search_text or "").strip().lower()
+    if search_text:
+        if search_in == "Model name":
+            df = df[df["Model name"].astype(str).str.lower().str.contains(search_text, na=False)]
+        elif search_in == "Submitter":
+            df = df[df["Submitter"].astype(str).str.lower().str.contains(search_text, na=False)]
+        else:
+            mask = (
+                df["Model name"].astype(str).str.lower().str.contains(search_text, na=False)
+                | df["Submitter"].astype(str).str.lower().str.contains(search_text, na=False)
+            )
+            df = df[mask]
+
+    ascending = sort_order == "Ascending"
+    if sort_by in df.columns:
+        df = df.sort_values(by=[sort_by], ascending=ascending, kind="stable", na_position="last")
+    return df
+
+
+def refresh(search_text: str, search_in: str, sort_by: str, sort_order: str):
+    df = _load_entries_df()
+    return _apply_search_and_sort(df, search_text, search_in, sort_by, sort_order)
+
+
+def _parse_hf_created_at(created_at: str) -> datetime.datetime | None:
+    try:
+        if created_at.endswith("Z"):
+            created_at = created_at[:-1] + "+00:00"
+        return datetime.datetime.fromisoformat(created_at)
+    except Exception:
+        return None
+
+
+def _check_user_eligibility(username: str) -> tuple[bool, str]:
+    """
+    - Must be older than ~4 months (>= 120 days)
+    """
+    try:
+        r = requests.get(f"https://huggingface.co/api/users/{username}/overview", timeout=10)
+        r.raise_for_status()
+        created_at = r.json().get("createdAt")
+        if not created_at:
+            return False, "Cannot verify account creation date."
+        dt = _parse_hf_created_at(created_at)
+        if not dt:
+            return False, "Cannot parse account creation date."
+        now = datetime.datetime.now(datetime.timezone.utc)
+        if dt.tzinfo is None:
+            dt = dt.replace(tzinfo=datetime.timezone.utc)
+        age_days = (now - dt).days
+        if age_days < 120:
+            return False, "Account must be older than 4 months to submit."
+        return True, ""
+    except Exception:
+        return False, "Cannot verify Hugging Face account. Please try again later."
+
+
+def _submitted_today(username: str) -> bool:
+    df = _load_entries_df()
+    if df.empty:
+        return False
+    today = datetime.datetime.utcnow().date().isoformat()
+    user_rows = df[df["Submitter"].astype(str) == username]
+    if user_rows.empty:
+        return False
+    return any(str(v).startswith(today) for v in user_rows["Submitted at"].tolist())
+
+
+def submit(
+    model_name: str,
+    model_api: str,
+    api_key: str,
+    avg: float,
+    easy: float,
+    mid: float,
+    hard: float,
+    games: float,
+    science: float,
+    tools: float,
+    humanities: float,
+    viz: float,
+    lifestyle: float,
+    search_text: str,
+    search_in: str,
+    sort_by: str,
+    sort_order: str,
+    profile: gr.OAuthProfile | None,
+):
+    # Login required to submit (in Spaces)
+    if IN_SPACES and (profile is None or not getattr(profile, "username", None)):
+        return "You must log in to submit.", refresh(search_text, search_in, sort_by, sort_order)
+
+    submitter = (getattr(profile, "username", None) if profile is not None else "local") or "anonymous"
+
+    model_name = (model_name or "").strip()
+    model_api = (model_api or "").strip()
+    api_key = (api_key or "").strip()
+
+    if not model_name:
+        return "Model name is required.", refresh(search_text, search_in, sort_by, sort_order)
+    if not model_api:
+        return "Model API URL is required.", refresh(search_text, search_in, sort_by, sort_order)
+    if not _is_valid_http_url(model_api):
+        return "Model API must be a valid http(s) URL.", refresh(search_text, search_in, sort_by, sort_order)
+    if not api_key:
+        return "API key is required.", refresh(search_text, search_in, sort_by, sort_order)
+
+    ok, msg = _ensure_dataset_readable()
+    if not ok:
+        return msg, refresh(search_text, search_in, sort_by, sort_order)
+
+    if IN_SPACES:
+        ok, msg = _check_user_eligibility(submitter)
+        if not ok:
+            return msg, refresh(search_text, search_in, sort_by, sort_order)
+
+    if _submitted_today(submitter):
+        return "You have already submitted today. Please try again tomorrow.", refresh(search_text, search_in, sort_by, sort_order)
+
+    now = datetime.datetime.utcnow().replace(microsecond=0).isoformat() + "Z"
+    nonce = uuid.uuid4().hex[:8]
+    safe_model = _slug(model_name)
+    safe_user = _slug(submitter)
+    path_in_repo = f"{ENTRIES_PREFIX}{now[:10]}/{now}-{safe_user}-{safe_model}-{nonce}.json"
+
+    # NOTE: api_key is collected but NOT stored.
+    payload = {
+        "Model name": model_name,
+        "Avg": avg,
+        "Easy": easy,
+        "Mid": mid,
+        "Hard": hard,
+        "Games": games,
+        "Science": science,
+        "Tools": tools,
+        "Humanities": humanities,
+        "Viz": viz,
+        "Lifestyle": lifestyle,
+        "Submitted at": now,
+        "Submitter": submitter,
+        "Model API": model_api,
+    }
+
+    api = _api()
+    data = (json.dumps(payload, ensure_ascii=False, indent=2) + "\n").encode("utf-8")
+    api.upload_file(
+        repo_id=LEADERBOARD_DATASET,
+        repo_type="dataset",
+        path_or_fileobj=io.BytesIO(data),
+        path_in_repo=path_in_repo,
+        commit_message=f"miniapp: submit {submitter}/{model_name}",
+        token=HF_TOKEN,
+    )
+
+    return "Submitted successfully.", refresh(search_text, search_in, sort_by, sort_order)
+
+
+with gr.Blocks(title=f"{APP_NAME} leaderboard") as demo:
+    # Layout: Overview -> Leaderboard -> Submission instructions
+    gr.Markdown(
+        f"## {APP_NAME} leaderboard\n\n"
+        "### Overview\n"
+        "_Placeholder overview text. Replace this with your benchmark description, rules, and links._\n\n"
+        "### Leaderboard\n"
+        "_The leaderboard below supports sorting and searching._\n\n"
+        "### Submission\n"
+        "_Placeholder submission instructions. Requires login. One submission per user per day. "
+        "Account must be older than 4 months._\n"
+    )
+
+    with gr.Row():
+        with gr.Column(scale=3):
+            with gr.Row():
+                search_text = gr.Textbox(label="Search", placeholder="Model name or submitter")
+                search_in = gr.Dropdown(
+                    label="Search in",
+                    choices=["Both", "Model name", "Submitter"],
+                    value="Both",
+                )
+            with gr.Row():
+                sort_by = gr.Dropdown(label="Sort by", choices=LEADERBOARD_COLUMNS, value="Avg")
+                sort_order = gr.Radio(label="Order", choices=["Descending", "Ascending"], value="Descending")
+                refresh_btn = gr.Button("Refresh")
+
+            leaderboard = gr.Dataframe(
+                label="Leaderboard",
+                value=_load_entries_df(),
+                interactive=False,
+                wrap=True,
+            )
+
+        with gr.Column(scale=2):
+            with gr.Accordion("Submit (login required)", open=True):
+                model_name = gr.Textbox(label="Model name (required)")
+                model_api = gr.Textbox(label="Model API (required)", placeholder="https://...")
+                api_key = gr.Textbox(label="API key (required)", type="password", placeholder="Will not be stored")
+
+                with gr.Row():
+                    avg = gr.Number(label="Avg", value=0)
+                    easy = gr.Number(label="Easy", value=0)
+                    mid = gr.Number(label="Mid", value=0)
+                    hard = gr.Number(label="Hard", value=0)
+
+                with gr.Row():
+                    games = gr.Number(label="Games", value=0)
+                    science = gr.Number(label="Science", value=0)
+                    tools = gr.Number(label="Tools", value=0)
+
+                with gr.Row():
+                    humanities = gr.Number(label="Humanities", value=0)
+                    viz = gr.Number(label="Viz", value=0)
+                    lifestyle = gr.Number(label="Lifestyle", value=0)
+
+                with gr.Row():
+                    gr.LoginButton()
+                    submit_btn = gr.Button("Submit", variant="primary")
+                status = gr.Markdown()
+
+    refresh_btn.click(refresh, inputs=[search_text, search_in, sort_by, sort_order], outputs=[leaderboard])
+    submit_btn.click(
+        submit,
+        inputs=[
+            model_name,
+            model_api,
+            api_key,
+            avg,
+            easy,
+            mid,
+            hard,
+            games,
+            science,
+            tools,
+            humanities,
+            viz,
+            lifestyle,
+            search_text,
+            search_in,
+            sort_by,
+            sort_order,
+        ],
+        outputs=[status, leaderboard],
+    )
+
+demo.launch()
+
+import datetime
+import io
+import json
+import os
+import re
+import uuid
+from urllib.parse import urlparse
+
 import gradio as gr
 import pandas as pd
 from huggingface_hub import HfApi, hf_hub_download