app.py

import csv
import itertools
import random
import json
import os
import uuid
from datetime import datetime, timedelta
from io import BytesIO
from typing import Dict, List, Tuple, Optional

import gradio as gr
try:
    from huggingface_hub import HfApi
except Exception:  # optional dependency at runtime
    HfApi = None  # type: ignore


BASE_DIR = os.path.dirname(__file__)
PERSIST_DIR = os.environ.get("PERSIST_DIR", "/data")
# Persistent local storage inside HF Spaces
PERSIST_DIR = os.environ.get("PERSIST_DIR", "/data")
# Evaluation knobs (can be overridden via env vars)
MIN_RATERS_PER_PAIR = int(os.environ.get("MIN_RATERS_PER_PAIR", 20))
BATCH_SIZE = int(os.environ.get("BATCH_SIZE", 20))
RELOAD_EVERY = int(os.environ.get("RELOAD_EVERY", 5))
REPEAT_RATE = float(os.environ.get("REPEAT_RATE", 0.05))  # fraction of repeats within batch
REPEAT_MIN_HOURS = float(os.environ.get("REPEAT_MIN_HOURS", 24))
FAST_MIN_SEC = float(os.environ.get("FAST_MIN_SEC", 2.0))
TASK_CONFIG = {
    "Scene Composition & Object Insertion": {
        "folder": "scene_composition_and_object_insertion",
        "score_fields": [
            ("physical_interaction_fidelity_score", "物理交互保真度 (Physical Interaction Fidelity)"),
            ("optical_effect_accuracy_score", "光学效应准确度 (Optical Effect Accuracy)"),
            ("semantic_functional_alignment_score", "语义/功能对齐度 (Semantic/Functional Alignment)"),
            ("overall_photorealism_score", "整体真实感 (Overall Photorealism)"),
        ],
    },
}


def _csv_path_for_task(task_name: str, filename: str) -> str:
    folder = TASK_CONFIG[task_name]["folder"]
    return os.path.join(BASE_DIR, folder, filename)


def _persist_csv_path_for_task(task_name: str) -> str:
    folder = TASK_CONFIG[task_name]["folder"]
    return os.path.join(PERSIST_DIR, folder, "evaluation_results.csv")


def _resolve_image_path(path: str) -> str:
    return path if os.path.isabs(path) else os.path.join(BASE_DIR, path)


def _file_exists_under_base(rel_or_abs_path: str) -> bool:
    """Check if file exists, resolving relative paths under BASE_DIR."""
    check_path = rel_or_abs_path if os.path.isabs(rel_or_abs_path) else os.path.join(BASE_DIR, rel_or_abs_path)
    return os.path.exists(check_path)


def _load_task_rows(task_name: str) -> List[Dict[str, str]]:
    csv_path = _csv_path_for_task(task_name, "results.csv")
    if not os.path.exists(csv_path):
        raise FileNotFoundError(f"未找到任务 {task_name} 的结果文件: {csv_path}")

    with open(csv_path, newline="", encoding="utf-8") as csv_file:
        reader = csv.DictReader(csv_file)
        rows: List[Dict[str, str]] = []
        for row in reader:
            # Trim whitespaces in all string fields to avoid path/key mismatches
            cleaned = {k.strip(): (v.strip() if isinstance(v, str) else v) for k, v in row.items()}
            rows.append(cleaned)
        return rows


def _build_image_pairs(rows: List[Dict[str, str]], task_name: str) -> List[Dict[str, str]]:
    grouped: Dict[Tuple[str, str], List[Dict[str, str]]] = {}
    for row in rows:
        key = (row["test_id"], row["org_img"])
        grouped.setdefault(key, []).append(row)

    pairs: List[Dict[str, str]] = []
    folder = TASK_CONFIG[task_name]["folder"]

    for (test_id, org_img), entries in grouped.items():
        for model_a, model_b in itertools.combinations(entries, 2):
            if model_a["model_name"] == model_b["model_name"]:
                continue

            org_path = os.path.join(folder, org_img)
            path_a = os.path.join(folder, model_a["path"])
            path_b = os.path.join(folder, model_b["path"])

            # Validate existence to avoid UI errors
            if not (_file_exists_under_base(org_path) and _file_exists_under_base(path_a) and _file_exists_under_base(path_b)):
                try:
                    print("[VisArena] Skipping invalid paths for test_id=", test_id, {
                        "org": org_path,
                        "a": path_a,
                        "b": path_b,
                    })
                except Exception:
                    pass
                continue

            pair = {
                "test_id": test_id,
                "org_img": org_path,
                "model1_name": model_a["model_name"],
                "model1_res": model_a["res"],
                "model1_path": path_a,
                "model2_name": model_b["model_name"],
                "model2_res": model_b["res"],
                "model2_path": path_b,
            }
            pairs.append(pair)

    def sort_key(item: Dict[str, str]):
        test_id = item["test_id"]
        try:
            test_id_key = int(test_id)
        except ValueError:
            test_id_key = test_id
        return (test_id_key, item["model1_name"], item["model2_name"])

    pairs.sort(key=sort_key)
    return pairs


def _read_eval_counts(task_name: str) -> Dict[Tuple[str, frozenset, str], int]:
    """Global counts per pair key across all annotators."""
    counts: Dict[Tuple[str, frozenset, str], int] = {}
    csv_path = _persist_csv_path_for_task(task_name)
    if not os.path.exists(csv_path):
        return counts
    try:
        with open(csv_path, newline="", encoding="utf-8") as f:
            reader = csv.DictReader(f)
            for r in reader:
                tid = str(r.get("test_id", "")).strip()
                m1 = str(r.get("model1_name", "")).strip()
                m2 = str(r.get("model2_name", "")).strip()
                org = str(r.get("org_img", "")).strip()
                if not (tid and m1 and m2 and org):
                    continue
                key = (tid, frozenset({m1, m2}), org)
                counts[key] = counts.get(key, 0) + 1
    except Exception:
        pass
    return counts


def _read_user_done_keys(task_name: str, annotator_id: str) -> set:
    """Keys already evaluated by the given annotator.
    If CSV has no annotator_id column (legacy rows), those rows are ignored for per-user filtering.
    """
    keys = set()
    if not annotator_id:
        return keys
    csv_path = _persist_csv_path_for_task(task_name)
    if not os.path.exists(csv_path):
        return keys
    try:
        with open(csv_path, newline="", encoding="utf-8") as f:
            reader = csv.DictReader(f)
            for r in reader:
                if str(r.get("annotator_id", "")).strip() != str(annotator_id).strip():
                    continue
                tid = str(r.get("test_id", "")).strip()
                m1 = str(r.get("model1_name", "")).strip()
                m2 = str(r.get("model2_name", "")).strip()
                org = str(r.get("org_img", "")).strip()
                if tid and m1 and m2 and org:
                    keys.add((tid, frozenset({m1, m2}), org))
    except Exception:
        pass
    return keys


def _read_user_last_times(task_name: str, annotator_id: str) -> Dict[Tuple[str, frozenset, str], datetime]:
    """Return the user's last evaluation datetime per pair key."""
    last: Dict[Tuple[str, frozenset, str], datetime] = {}
    if not annotator_id:
        return last
    csv_path = _persist_csv_path_for_task(task_name)
    if not os.path.exists(csv_path):
        return last
    try:
        with open(csv_path, newline="", encoding="utf-8") as f:
            reader = csv.DictReader(f)
            for r in reader:
                if str(r.get("annotator_id", "")).strip() != str(annotator_id).strip():
                    continue
                tid = str(r.get("test_id", "")).strip()
                m1 = str(r.get("model1_name", "")).strip()
                m2 = str(r.get("model2_name", "")).strip()
                org = str(r.get("org_img", "")).strip()
                dt = str(r.get("eval_date", "")).strip() or str(r.get("submit_ts", "")).strip()
                if not (tid and m1 and m2 and org and dt):
                    continue
                key = (tid, frozenset({m1, m2}), org)
                try:
                    t = datetime.fromisoformat(dt)
                except Exception:
                    continue
                if key not in last or t > last[key]:
                    last[key] = t
    except Exception:
        pass
    return last


def _schedule_round_robin_by_test_id(pairs: List[Dict[str, str]], seed: Optional[int] = None) -> List[Dict[str, str]]:
    """Interleave pairs across test_ids for balanced coverage; shuffle within each group.
    """
    groups: Dict[str, List[Dict[str, str]]] = {}
    for p in pairs:
        groups.setdefault(p["test_id"], []).append(p)
    rnd = random.Random(seed)
    for lst in groups.values():
        rnd.shuffle(lst)
    # round-robin drain
    ordered: List[Dict[str, str]] = []
    while True:
        progressed = False
        for tid in sorted(groups.keys(), key=lambda x: (int(x) if x.isdigit() else x)):
            if groups[tid]:
                ordered.append(groups[tid].pop())
                progressed = True
        if not progressed:
            break
    return ordered


def load_task(task_name: str, annotator_id: str = ""):
    if not task_name:
        raise gr.Error("Please select a task first.")

    rows = _load_task_rows(task_name)
    pairs_all = _build_image_pairs(rows, task_name)
    # Per-user filtering and global balancing
    def key_of(p: Dict[str, str]):
        return (p["test_id"], frozenset({p["model1_name"], p["model2_name"]}), p["org_img"])
    user_done_keys = _read_user_done_keys(task_name, annotator_id)
    user_last_times = _read_user_last_times(task_name, annotator_id)
    global_counts = _read_eval_counts(task_name)
    # Main eligible set: not done by this user and below min raters threshold
    pairs = [
        p for p in pairs_all
        if key_of(p) not in user_done_keys and global_counts.get(key_of(p), 0) < MIN_RATERS_PER_PAIR
    ]

    # Balanced schedule: prioritize low-count pairs, and within same count do round-robin by test_id
    seed_env = os.environ.get("SCHEDULE_SEED")
    seed = int(seed_env) if seed_env and seed_env.isdigit() else None
    def count_of(p: Dict[str, str]):
        return global_counts.get(key_of(p), 0)
    buckets: Dict[int, List[Dict[str, str]]] = {}
    for p in sorted(pairs, key=count_of):
        buckets.setdefault(count_of(p), []).append(p)
    ordered: List[Dict[str, str]] = []
    for c in sorted(buckets.keys()):
        ordered.extend(_schedule_round_robin_by_test_id(buckets[c], seed=seed))
    pairs = ordered

    # Deterministic rotation by user's progress to avoid always starting from the same pairs
    try:
        elig_keys = [key_of(p) for p in pairs]
        progress = len([k for k in user_done_keys if k in elig_keys])
        if pairs:
            rot = progress % len(pairs)
            pairs = pairs[rot:] + pairs[:rot]
    except Exception:
        pass

    # Limit batch size
    main_batch = pairs[: max(0, BATCH_SIZE)]

    # Small proportion of spaced repeats for test-retest
    repeats: List[Dict[str, str]] = []
    try:
        repeat_target = int(max(0, round(BATCH_SIZE * REPEAT_RATE)))
        if repeat_target > 0 and user_last_times:
            min_time = datetime.utcnow() - timedelta(hours=REPEAT_MIN_HOURS)
            candidates = [k for k, t in user_last_times.items() if t < min_time]
            def find_pair_from_key(k):
                tid, names, org = k
                for p in pairs_all:
                    if p["test_id"] == tid and p["org_img"] == org and frozenset({p["model1_name"], p["model2_name"]}) == names:
                        return p
                return None
            picked = 0
            used_keys = {key_of(p) for p in main_batch}
            for k in candidates:
                if picked >= repeat_target:
                    break
                p = find_pair_from_key(k)
                if not p:
                    continue
                if key_of(p) in used_keys:
                    continue
                repeats.append(p)
                used_keys.add(key_of(p))
                picked += 1
    except Exception:
        pass

    pairs = main_batch + repeats

    # Assign A/B order to counteract position bias: alternate after scheduling
    for idx, p in enumerate(pairs):
        p["swap"] = bool(idx % 2)  # True -> A=B's image; False -> A=A's image

    if not pairs:
        try:
            print("[VisArena] No pending pairs.")
            print("[VisArena] total_pairs=", len(pairs_all))
            print("[VisArena] already_done_by_user=", len(user_done_keys))
            print("[VisArena] persist_csv=", _persist_csv_path_for_task(task_name))
        except Exception:
            pass
        # Return empty list; UI will render an informative message instead of erroring out
        return []

    return pairs


def _format_pair_header(_pair: Dict[str, str]) -> str:
    # Mask model identity in UI; keep header neutral
    return ""


def _build_eval_row(pair: Dict[str, str], scores: Dict[str, int]) -> Dict[str, object]:
    row = {
        "eval_date": datetime.utcnow().isoformat(),
        "test_id": pair["test_id"],
        "model1_name": pair["model1_name"],
        "model2_name": pair["model2_name"],
        "org_img": pair["org_img"],
        "model1_res": pair["model1_res"],
        "model2_res": pair["model2_res"],
        "model1_path": pair["model1_path"],
        "model2_path": pair["model2_path"],
    }
    row.update(scores)
    return row


def _local_persist_csv_path(task_name: str) -> str:
    folder = TASK_CONFIG[task_name]["folder"]
    return os.path.join(PERSIST_DIR, folder, "evaluation_results.csv")


def _append_local_persist_csv(task_name: str, row: Dict[str, object]) -> bool:
    csv_path = _local_persist_csv_path(task_name)
    os.makedirs(os.path.dirname(csv_path), exist_ok=True)
    csv_exists = os.path.exists(csv_path)
    fieldnames = [
        "eval_date",
        "annotator_id",
        "session_id",
        "view_start_ts",
        "submit_ts",
        "duration_sec",
        "is_fast",
        "is_flat_a",
        "is_flat_b",
        "test_id",
        "model1_name",
        "model2_name",
        "org_img",
        "model1_res",
        "model2_res",
        "model1_path",
        "model2_path",
        "model1_physical_interaction_fidelity_score",
        "model1_optical_effect_accuracy_score",
        "model1_semantic_functional_alignment_score",
        "model1_overall_photorealism_score",
        "model2_physical_interaction_fidelity_score",
        "model2_optical_effect_accuracy_score",
        "model2_semantic_functional_alignment_score",
        "model2_overall_photorealism_score",
    ]
    try:
        with open(csv_path, "a", newline="", encoding="utf-8") as csv_file:
            writer = csv.DictWriter(csv_file, fieldnames=fieldnames)
            if not csv_exists:
                writer.writeheader()
            writer.writerow(row)
        return True
    except Exception:
        return False


def _upload_eval_record_to_dataset(task_name: str, row: Dict[str, object]) -> Tuple[bool, str]:
    """Upload a single-eval JSONL record to a dataset repo.
    Repo is taken from EVAL_REPO_ID env or defaults to 'peiranli0930/VisEval'.
    Returns (ok, message) for UI feedback and debugging.
    """
    if HfApi is None:
        return False, "huggingface_hub not installed"
    token = os.environ.get("HF_TOKEN") or os.environ.get("HUGGINGFACEHUB_API_TOKEN")
    repo_id = os.environ.get("EVAL_REPO_ID", "peiranli0930/VisEval")
    if not token:
        return False, "Missing write token (HF_TOKEN/HUGGINGFACEHUB_API_TOKEN)"
    if not repo_id:
        return False, "EVAL_REPO_ID is not set"
    try:
        from huggingface_hub import CommitOperationAdd

        api = HfApi(token=token)
        date_prefix = datetime.utcnow().strftime("%Y-%m-%d")
        folder = TASK_CONFIG[task_name]["folder"]
        uid = str(uuid.uuid4())
        path_in_repo = f"submissions/{folder}/{date_prefix}/{uid}.jsonl"
        payload = (json.dumps(row, ensure_ascii=False) + "\n").encode("utf-8")
        operations = [CommitOperationAdd(path_in_repo=path_in_repo, path_or_fileobj=BytesIO(payload))]
        api.create_commit(
            repo_id=repo_id,
            repo_type="dataset",
            operations=operations,
            commit_message=f"Add eval {folder} {row.get('test_id')} {uid}",
        )
        return True, f"Uploaded: {repo_id}/{path_in_repo}"
    except Exception as e:
        # Print to logs for debugging in Space
        try:
            print("[VisArena] Upload to dataset failed:", repr(e))
        except Exception:
            pass
        return False, f"Exception: {type(e).__name__}: {e}"


def _extract_annotator_id(request: Optional[gr.Request]) -> str:
    """Best-effort extraction of a stable user identifier on HF Spaces.
    Priority: request.username -> X-Forwarded-User header -> cookies/user-id -> env/session fallback.
    """
    try:
        if request is None:
            return ""
        # gradio>=4.0 may set username for Spaces-authenticated users
        username = getattr(request, "username", None)
        if username:
            return str(username)
        headers = getattr(request, "headers", {}) or {}
        for k in ("x-forwarded-user", "x-user", "x-hub-user"):
            v = headers.get(k) or headers.get(k.upper())
            if v:
                return str(v)
    except Exception:
        pass
    return ""


def on_task_change(task_name: str, _state_pairs: List[Dict[str, str]], request: gr.Request,
                   view_started_at: float, session_quota: int, reload_count: int, session_id: str):
    annotator_id = _extract_annotator_id(request)
    if not annotator_id:
        default_scores = [3, 3, 3, 3, 3, 3, 3, 3]
        return (
            [],
            gr.update(value=0, minimum=0, maximum=0, visible=False),
            gr.update(value=""),
            gr.update(value=None),
            gr.update(value=None),
            gr.update(value=None),
            *default_scores,
            gr.update(value="请先登录你的 Hugging Face 账户后再开始评测。"),
            float(datetime.utcnow().timestamp()),
            BATCH_SIZE,
            0,
            session_id or str(uuid.uuid4()),
        )

    pairs = load_task(task_name, annotator_id)
    # Defaults for A and B (8 sliders total)
    default_scores = [3, 3, 3, 3, 3, 3, 3, 3]
    if not pairs:
        return (
            [],
            gr.update(value=0, minimum=0, maximum=0, visible=False),
            gr.update(value=""),
            gr.update(value=None),
            gr.update(value=None),
            gr.update(value=None),
            *default_scores,
            gr.update(value="当前没有待评对（或已达到最小标注阈值）。"),
            float(datetime.utcnow().timestamp()),
            BATCH_SIZE,
            0,
            session_id or str(uuid.uuid4()),
        )

    pair = pairs[0]
    header = _format_pair_header(pair)
    # Pick display order according to swap flag
    a_path = pair["model2_path"] if pair.get("swap") else pair["model1_path"]
    b_path = pair["model1_path"] if pair.get("swap") else pair["model2_path"]
    max_index = max(0, len(pairs) - 1)
    return (
        pairs,
        gr.update(value=0, minimum=0, maximum=max_index, visible=(len(pairs) > 1)),
        gr.update(value=header),
        _resolve_image_path(pair["org_img"]),
        _resolve_image_path(a_path),
        _resolve_image_path(b_path),
        *default_scores,
        gr.update(value=f"本批次分配 {len(pairs)} 组；目标每对 {MIN_RATERS_PER_PAIR} 人。"),
        float(datetime.utcnow().timestamp()),
        BATCH_SIZE,
        0,
        session_id or str(uuid.uuid4()),
    )


def on_pair_navigate(index: int, pairs: List[Dict[str, str]], view_started_at: float):
    if not pairs:
        # Gracefully no-op when no pairs
        return (
            gr.update(value=0, minimum=0, maximum=0, visible=False),
            gr.update(value=""),
            gr.update(value=None),
            gr.update(value=None),
            gr.update(value=None),
            3, 3, 3, 3,  # A
            3, 3, 3, 3,  # B
            float(datetime.utcnow().timestamp()),
        )
    index = int(index)
    index = max(0, min(index, len(pairs) - 1))
    pair = pairs[index]
    header = _format_pair_header(pair)
    a_path = pair["model2_path"] if pair.get("swap") else pair["model1_path"]
    b_path = pair["model1_path"] if pair.get("swap") else pair["model2_path"]
    return (
        gr.update(value=index),
        gr.update(value=header),
        _resolve_image_path(pair["org_img"]),
        _resolve_image_path(a_path),
        _resolve_image_path(b_path),
        3, 3, 3, 3,  # A
        3, 3, 3, 3,  # B
        float(datetime.utcnow().timestamp()),
    )


def on_submit(
    task_name: str,
    index: int,
    pairs: List[Dict[str, str]],
    a_physical_score: int,
    a_optical_score: int,
    a_semantic_score: int,
    a_overall_score: int,
    b_physical_score: int,
    b_optical_score: int,
    b_semantic_score: int,
    b_overall_score: int,
    request: gr.Request,
    view_started_at: float,
    session_quota: int,
    reload_count: int,
    session_id: str,
):
    if not task_name:
        return (
            pairs,
            gr.update(value=0),
            gr.update(value=""),
            gr.update(value=None),
            gr.update(value=None),
            gr.update(value=None),
            3, 3, 3, 3,
            3, 3, 3, 3,
            gr.update(value="Please select a task first."),
            float(datetime.utcnow().timestamp()),
            session_quota,
            reload_count,
            session_id,
        )

    if not pairs:
        return (
            pairs,
            gr.update(value=0, minimum=0, maximum=0, visible=False),
            gr.update(value=""),
            gr.update(value=None),
            gr.update(value=None),
            gr.update(value=None),
            3, 3, 3, 3,
            3, 3, 3, 3,
            gr.update(value="No pending pairs to submit."),
            float(datetime.utcnow().timestamp()),
            session_quota,
            reload_count,
            session_id,
        )

    # Resolve annotator id from request
    annotator_id = _extract_annotator_id(request)

    pair = pairs[index]
    score_map = {
        # Model A
        "model1_physical_interaction_fidelity_score": int(a_physical_score),
        "model1_optical_effect_accuracy_score": int(a_optical_score),
        "model1_semantic_functional_alignment_score": int(a_semantic_score),
        "model1_overall_photorealism_score": int(a_overall_score),
        # Model B
        "model2_physical_interaction_fidelity_score": int(b_physical_score),
        "model2_optical_effect_accuracy_score": int(b_optical_score),
        "model2_semantic_functional_alignment_score": int(b_semantic_score),
        "model2_overall_photorealism_score": int(b_overall_score),
    }
    # Map A/B scores to the correct model columns depending on swap
    if pair.get("swap"):
        # UI A == model2, UI B == model1
        score_map = {
            "model1_physical_interaction_fidelity_score": int(b_physical_score),
            "model1_optical_effect_accuracy_score": int(b_optical_score),
            "model1_semantic_functional_alignment_score": int(b_semantic_score),
            "model1_overall_photorealism_score": int(b_overall_score),
            "model2_physical_interaction_fidelity_score": int(a_physical_score),
            "model2_optical_effect_accuracy_score": int(a_optical_score),
            "model2_semantic_functional_alignment_score": int(a_semantic_score),
            "model2_overall_photorealism_score": int(a_overall_score),
        }
    # Build record
    row = _build_eval_row(pair, score_map)
    row["annotator_id"] = annotator_id
    # timing + heuristics
    submit_ts = datetime.utcnow()
    try:
        started = datetime.utcfromtimestamp(float(view_started_at)) if view_started_at else submit_ts
    except Exception:
        started = submit_ts
    duration = max(0.0, (submit_ts - started).total_seconds())
    row["view_start_ts"] = started.isoformat()
    row["submit_ts"] = submit_ts.isoformat()
    row["duration_sec"] = round(duration, 3)
    row["is_fast"] = bool(duration < FAST_MIN_SEC)
    row["is_flat_a"] = bool(len({int(a_physical_score), int(a_optical_score), int(a_semantic_score), int(a_overall_score)}) == 1)
    row["is_flat_b"] = bool(len({int(b_physical_score), int(b_optical_score), int(b_semantic_score), int(b_overall_score)}) == 1)
    row["session_id"] = session_id or str(uuid.uuid4())

    # Idempotency: check if this pair already evaluated; if so, skip writing
    done_keys = _read_user_done_keys(task_name, annotator_id)
    eval_key = (pair["test_id"], frozenset({pair["model1_name"], pair["model2_name"]}), pair["org_img"])
    if eval_key in done_keys:
        ok_local = False
        ok_hub, hub_msg = (False, "Skipped duplicate; already evaluated.")
        info_prefix = "Skipped duplicate submission."
    else:
        ok_local = _append_local_persist_csv(task_name, row)
        # add key locally for subsequent filtering in this call
        if ok_local:
            done_keys.add(eval_key)
        ok_hub, hub_msg = _upload_eval_record_to_dataset(task_name, row)
        info_prefix = "Saved evaluation."

    # Recompute remaining pairs by filtering current state against done_keys
    def key_of(p: Dict[str, str]):
        return (p["test_id"], frozenset({p["model1_name"], p["model2_name"]}), p["org_img"])
    remaining_pairs = [p for p in pairs if key_of(p) not in done_keys]

    info = f"{info_prefix} Local persistence " + ("succeeded" if ok_local else "skipped/failed") + "."
    info += " Dataset upload " + ("succeeded" if ok_hub else "failed") + (f" ({hub_msg})" if hub_msg else "") + "."

    # Quota + reload
    session_quota = max(0, int(session_quota) - 1)
    reload_count = int(reload_count) + 1
    # Periodic reload to absorb new results.csv / re-balance
    if reload_count >= RELOAD_EVERY:
        fresh_pairs = load_task(task_name, annotator_id)
        remaining_pairs = fresh_pairs
        reload_count = 0

    if session_quota <= 0:
        return (
            [],
            gr.update(value=0, minimum=0, maximum=0, visible=False),
            gr.update(value=""),
            gr.update(value=None),
            gr.update(value=None),
            gr.update(value=None),
            3, 3, 3, 3,
            3, 3, 3, 3,
            gr.update(value=info + " 本批次已完成 20 组，请刷新页面获取下一批次。"),
            float(datetime.utcnow().timestamp()),
            session_quota,
            reload_count,
            row["session_id"],
        )

    if remaining_pairs:
        next_index = min(index, len(remaining_pairs) - 1)
        pair = remaining_pairs[next_index]
        header = _format_pair_header(pair)
        a_path = pair["model2_path"] if pair.get("swap") else pair["model1_path"]
        b_path = pair["model1_path"] if pair.get("swap") else pair["model2_path"]
        return (
            remaining_pairs,
            gr.update(value=next_index),
            gr.update(value=header),
            _resolve_image_path(pair["org_img"]),
            _resolve_image_path(a_path),
            _resolve_image_path(b_path),
            3, 3, 3, 3,
            3, 3, 3, 3,
            gr.update(value=info + f" Next pair ({next_index + 1}/{len(remaining_pairs)})."),
            float(datetime.utcnow().timestamp()),
            session_quota,
            reload_count,
            row["session_id"],
        )

    # No remaining pairs: clear UI, hide slider, and return updated empty state
    return (
        [],
        gr.update(value=0, minimum=0, maximum=0, visible=False),
        gr.update(value=""),
        gr.update(value=None),
        gr.update(value=None),
        gr.update(value=None),
        3, 3, 3, 3,
        3, 3, 3, 3,
        gr.update(value=info + " All pairs completed."),
        float(datetime.utcnow().timestamp()),
        session_quota,
        reload_count,
        row["session_id"],
    )


with gr.Blocks(title="VisArena Human Evaluation") as demo:
    gr.Markdown(
        """
        # VisArena Human Evaluation
        Please select a task and rate the generated images. Each score ranges from 1 (poor) to 5 (excellent).
        """
    )

    with gr.Row():
        task_selector = gr.Dropdown(
            label="Task",
            choices=list(TASK_CONFIG.keys()),
            interactive=True,
            value="Scene Composition & Object Insertion",
        )
        index_slider = gr.Slider(
            label="Pair Index",
            value=0,
            minimum=0,
            maximum=0,
            step=1,
            interactive=True,
            visible=False,
        )

    pair_state = gr.State([])
    # Hidden states for session control and metrics
    view_started_at_state = gr.State(0.0)
    session_quota_state = gr.State(BATCH_SIZE)
    reload_count_state = gr.State(0)
    session_id_state = gr.State("")

    pair_header = gr.Markdown("")

    # Layout: Original on top, two outputs below with their own sliders
    with gr.Row():
        with gr.Column(scale=12):
            orig_image = gr.Image(type="filepath", label="Original", interactive=False)

    with gr.Row():
        with gr.Column(scale=6):
            model1_image = gr.Image(type="filepath", label="Output A", interactive=False)
            a_physical_input = gr.Slider(1, 5, value=3, step=1, label="A: Physical Interaction Fidelity")
            a_optical_input = gr.Slider(1, 5, value=3, step=1, label="A: Optical Effect Accuracy")
            a_semantic_input = gr.Slider(1, 5, value=3, step=1, label="A: Semantic/Functional Alignment")
            a_overall_input = gr.Slider(1, 5, value=3, step=1, label="A: Overall Photorealism")
        with gr.Column(scale=6):
            model2_image = gr.Image(type="filepath", label="Output B", interactive=False)
            b_physical_input = gr.Slider(1, 5, value=3, step=1, label="B: Physical Interaction Fidelity")
            b_optical_input = gr.Slider(1, 5, value=3, step=1, label="B: Optical Effect Accuracy")
            b_semantic_input = gr.Slider(1, 5, value=3, step=1, label="B: Semantic/Functional Alignment")
            b_overall_input = gr.Slider(1, 5, value=3, step=1, label="B: Overall Photorealism")

        submit_button = gr.Button("Submit Evaluation", variant="primary")
        feedback_box = gr.Markdown("")

        # Event bindings
        task_selector.change(
            fn=on_task_change,
            inputs=[task_selector, pair_state, view_started_at_state, session_quota_state, reload_count_state, session_id_state],
            outputs=[
                pair_state,
                index_slider,
                pair_header,
                orig_image,
                model1_image,
                model2_image,
                a_physical_input,
                a_optical_input,
                a_semantic_input,
                a_overall_input,
                b_physical_input,
                b_optical_input,
                b_semantic_input,
                b_overall_input,
                feedback_box,
                view_started_at_state,
                session_quota_state,
                reload_count_state,
                session_id_state,
            ],
        )

        index_slider.release(
            fn=on_pair_navigate,
            inputs=[index_slider, pair_state, view_started_at_state],
            outputs=[
                index_slider,
                pair_header,
                orig_image,
                model1_image,
                model2_image,
                a_physical_input,
                a_optical_input,
                a_semantic_input,
                a_overall_input,
                b_physical_input,
                b_optical_input,
                b_semantic_input,
                b_overall_input,
                view_started_at_state,
            ],
        )

        submit_button.click(
            fn=on_submit,
            inputs=[
                task_selector,
                index_slider,
                pair_state,
                a_physical_input,
                a_optical_input,
                a_semantic_input,
                a_overall_input,
                b_physical_input,
                b_optical_input,
                b_semantic_input,
                b_overall_input,
                view_started_at_state,
                session_quota_state,
                reload_count_state,
                session_id_state,
            ],
            outputs=[
                pair_state,
                index_slider,
                pair_header,
                orig_image,
                model1_image,
                model2_image,
                a_physical_input,
                a_optical_input,
                a_semantic_input,
                a_overall_input,
                b_physical_input,
                b_optical_input,
                b_semantic_input,
                b_overall_input,
                feedback_box,
                view_started_at_state,
                session_quota_state,
                reload_count_state,
                session_id_state,
            ],
        )

        # Auto-load default task on startup
        demo.load(
            fn=on_task_change,
            inputs=[task_selector, pair_state, view_started_at_state, session_quota_state, reload_count_state, session_id_state],
            outputs=[
                pair_state,
                index_slider,
                pair_header,
                orig_image,
                model1_image,
                model2_image,
                a_physical_input,
                a_optical_input,
                a_semantic_input,
                a_overall_input,
                b_physical_input,
                b_optical_input,
                b_semantic_input,
                b_overall_input,
                feedback_box,
                view_started_at_state,
                session_quota_state,
                reload_count_state,
                session_id_state,
            ],
        )


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