File size: 21,352 Bytes
6a2ca60
 
e58f9a0
 
f294685
e58f9a0
5390756
6a2ca60
af82024
5390756
e58f9a0
6a2ca60
f294685
5390756
 
 
 
af82024
5390756
6a2ca60
 
 
 
 
 
 
969d75c
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
e58f9a0
 
 
 
 
f294685
 
 
 
 
 
e58f9a0
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
969d75c
 
 
 
 
e58f9a0
 
 
 
 
969d75c
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
6a2ca60
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
5390756
 
 
af82024
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
5390756
 
 
 
 
af82024
5390756
6a2ca60
 
 
 
 
 
969d75c
 
 
 
 
6a2ca60
969d75c
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
6a2ca60
 
969d75c
af82024
 
 
 
 
 
 
 
 
 
 
 
 
6a2ca60
 
 
 
 
af82024
 
 
6a2ca60
af82024
6a2ca60
 
af82024
 
 
 
6a2ca60
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
5390756
 
 
 
 
 
 
 
cc21aaf
6a2ca60
 
 
 
 
 
 
 
 
 
 
af82024
cc21aaf
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
6a2ca60
cc21aaf
af82024
6a2ca60
969d75c
6a2ca60
 
 
 
 
cc21aaf
 
 
 
5390756
 
 
af82024
 
 
 
6a2ca60
 
 
 
 
 
 
 
5390756
 
cc21aaf
6a2ca60
 
 
 
5390756
af82024
 
cc21aaf
 
5390756
 
 
 
 
6a2ca60
969d75c
 
cc21aaf
 
 
 
 
 
 
 
 
 
 
 
 
 
 
6a2ca60
 
 
 
 
 
969d75c
6a2ca60
 
969d75c
cc21aaf
6a2ca60
 
5390756
 
6a2ca60
af82024
 
 
 
 
 
 
6a2ca60
af82024
6a2ca60
af82024
5390756
 
cc21aaf
5390756
 
 
 
af82024
5390756
 
af82024
 
5390756
 
 
 
 
6a2ca60
 
 
5390756
 
6a2ca60
5390756
6a2ca60
969d75c
 
 
6a2ca60
5390756
 
969d75c
 
 
5390756
 
 
 
6a2ca60
 
969d75c
5390756
6a2ca60
 
 
5390756
6a2ca60
 
 
af82024
5390756
 
 
 
 
 
 
 
969d75c
 
5390756
 
 
 
969d75c
 
5390756
 
 
 
 
 
969d75c
cc21aaf
af82024
 
 
cc21aaf
6a2ca60
 
 
 
 
 
 
 
 
 
 
cc21aaf
6a2ca60
af82024
5390756
af82024
 
5390756
 
 
 
 
 
 
 
 
 
af82024
 
5390756
 
6a2ca60
969d75c
 
 
 
 
 
 
 
 
 
 
6a2ca60
 
 
 
 
 
 
969d75c
6a2ca60
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
#!/usr/bin/env python3
# sync_library_and_hf.py
''''
RUN BELOW FOR NEW HTML FILES TO UPDATE OLD ONES ON DFATASET REPO
python sync_library_and_hf.py --db-path library.csv  --repo-id akazemian/audio-html --model-name wavcoch_audio-preds-sr=16000   --index-filename index.csv   --wipe-remote   --wipe-local
'''
import argparse, datetime, uuid, posixpath, sys, traceback, os, hashlib
from pathlib import Path
from typing import List, Tuple, Set
from urllib.parse import unquote
import os
import pandas as pd
import numpy as np
from huggingface_hub import (
    HfApi,
    hf_hub_download,
    CommitOperationAdd,
    CommitOperationDelete,
)
from huggingface_hub.utils import HfHubHTTPError

REQUIRED_DB_COLS = [
    "id","filename","path","tags","keywords","notes","uploaded_at","category","dataset","hf_path"
]
INDEX_COLS = ["id","filename","relpath","category","dataset","tags","keywords","notes","uploaded_at"]


# --- manifest helpers ---
AUDIO_EXTS = {".wav", ".mp3"}  # extend if needed: ".flac", ".ogg", etc.

def _strip_ext(name: str, exts: set[str]) -> str:
    n = name
    for ext in exts:
        if n.lower().endswith(ext):
            return n[: -len(ext)]
    return n

def key_from_html_filename(fname: str) -> str:
    # e.g. "foo_bar.html" -> "foo_bar"
    base = Path(fname).name
    if base.lower().endswith(".html"):
        base = base[:-5]
    return base

def key_from_manifest_filename(fname: str) -> str:
    # e.g. "foo_bar.wav" or "foo_bar.mp3" -> "foo_bar"
    base = Path(fname).name
    return _strip_ext(base, AUDIO_EXTS)

def create_file_specific_manifest(csv_path: Path) -> pd.DataFrame:

    audio_dir = "/data/atlask/BAU-Quant/val"
    manifest = pd.read_csv(csv_path)

    mask = manifest['dataset'].eq('TUT_urban_acoustic_scenes')
    manifest['audio_category'] = np.where(mask, manifest['dataset'], manifest['audio_category'])
    manifest = manifest.assign(
        audio_category = manifest['audio_category'].where(~mask, manifest['dataset'])
    )

    # 1) Build a files dataframe
    files = pd.DataFrame({"file_name": os.listdir(audio_dir)})
    # keep only audio files if needed
    files = files[files["file_name"].str.lower().str.endswith((".wav", ".mp3", ".flac", ".ogg", ".m4a", ".opus"))].copy()
    files["file_path"] = files["file_name"].apply(lambda f: os.path.join(audio_dir, f))

    # Normalize to a join key: drop extension, then strip `_chunk...`
    files["key"] = (
        files["file_name"]
            .str.replace(r"\.[^.]+$", "", regex=True)      # remove extension
            .str.replace(r"_chunk.*$", "", regex=True)     # remove _chunk suffix if present
    )

    # 2) Prepare manifest with the same key
    man = manifest.copy()

    # If manifest['file_name'] includes extensions / chunk suffixes, normalize the same way:
    man["key"] = (
        man["file_name"]
            .str.replace(r"\.[^.]+$", "", regex=True)
            .str.replace(r"_chunk.*$", "", regex=True)
    )

    # If duplicates exist in manifest for the same key, decide how to resolve:
    # e.g., keep first occurrence
    man = man.drop_duplicates(subset="key", keep="first")

    # 3) Merge once (vectorized)
    cols_to_take = ["sr", "dataset", "audio_category", "split", "duration_s"]
    out = files.merge(man[["key"] + cols_to_take], on="key", how="left")

    # 4) Final column order
    return out[["sr", "file_name", "file_path", "dataset", "audio_category", "split"]]

def load_manifest_map(csv_path: Path) -> dict[str, tuple[str, str]]:
    """
    Returns {basename_key: (dataset, category)} from the manifest.
    Manifest must have columns: file_name, dataset, category
    """
    # if not csv_path.exists():
    #     print(f"[manifest] WARNING: not found: {csv_path}")
    #     return {}
    # dfm = pd.read_csv(csv_path)
    dfm = create_file_specific_manifest(csv_path)
    dfm = dfm.rename(columns={'audio_category':'category'})
    required = {"file_name", "dataset", "category"}
    missing = required - set(dfm.columns)
    if missing:
        raise ValueError(f"manifest missing columns: {sorted(missing)}")

    m = {}
    for _, r in dfm.iterrows():
        k = key_from_manifest_filename(str(r["file_name"]))
        ds = str(r["dataset"]) if pd.notna(r["dataset"]) else ""
        cat = str(r["category"]) if pd.notna(r["category"]) else ""
        if k and (ds or cat):
            m[k] = (ds, cat)
    print(f"[manifest] loaded {len(m)} keys from {csv_path}")
    return m

def now_iso() -> str:
    return datetime.datetime.now().isoformat(timespec="seconds")

def ensure_cols(df: pd.DataFrame, cols: list) -> pd.DataFrame:
    for c in cols:
        if c not in df.columns:
            df[c] = ""
    for c in cols:
        df[c] = df[c].fillna("").astype(str)
    return df[cols]

def load_db(db_path: Path) -> pd.DataFrame:
    if db_path.exists():
        df = pd.read_csv(db_path)
    else:
        df = pd.DataFrame(columns=REQUIRED_DB_COLS)
    return ensure_cols(df, REQUIRED_DB_COLS)

def save_db(df: pd.DataFrame, db_path: Path):
    db_path.parent.mkdir(parents=True, exist_ok=True)
    df.to_csv(db_path, index=False)

def load_hf_index(repo_id: str, index_filename: str) -> Tuple[pd.DataFrame, bool]:
    try:
        p = hf_hub_download(repo_id=repo_id, repo_type="dataset", filename=index_filename)
        df = pd.read_csv(p)
        return ensure_cols(df, INDEX_COLS), True
    except HfHubHTTPError as e:
        if e.response is not None and e.response.status_code == 404:
            return ensure_cols(pd.DataFrame(columns=INDEX_COLS), INDEX_COLS), False
        raise

def relpath_posix(local_path: Path, root: Path) -> str:
    rel = local_path.resolve().relative_to(root.resolve())
    parts = [unquote(p) for p in rel.as_posix().split("/")]
    return posixpath.join(*parts)

# --- model prefix + sharding helpers ---
def ensure_model_prefix(relpath: str, model_name: str | None) -> str:
    """
    If model_name is provided and relpath doesn't start with "<model_name>/",
    prepend it. Otherwise return relpath unchanged.
    """
    if not model_name:
        return relpath
    model = model_name.strip()
    if not model:
        return relpath
    if relpath.startswith(model + "/"):
        return relpath
    return f"{model}/{relpath}"

def shard_relpath_under_model(relpath: str, hexdigits: int = 2) -> str:
    """
    Insert shard bucket immediately after the *model* segment (first path part).
    If there is only 1 segment, just return relpath.
    """
    parts = relpath.split("/")
    if len(parts) < 2:
        return relpath
    filename = parts[-1]
    bucket = hashlib.sha1(filename.encode("utf-8")).hexdigest()[:hexdigits]
    # parts[0] = model, parts[1:] = rest of path
    return "/".join([parts[0], bucket] + parts[1:])

def discover_new_local_htmls(reports_root: Path, df_db: pd.DataFrame) -> List[Path]:
    all_htmls = list(reports_root.rglob("*.html"))
    existing_paths = set(df_db["path"].astype(str))
    return sorted([p for p in all_htmls if str(p) not in existing_paths])

def rows_from_files(
    files: List[Path],
    reports_root: Path,
    manifest_map: dict[str, tuple[str,str]],
) -> pd.DataFrame:
    ts = now_iso()
    rows = []
    for p in files:
        k = key_from_html_filename(p.name)
        ds, cat = manifest_map.get(k, ("", ""))
        rows.append({
            "id": uuid.uuid4().hex[:8],
            "filename": p.name,
            "path": str(p),
            "tags": "",
            "keywords": "",
            "notes": "",
            "uploaded_at": ts,
            "category": cat,
            "dataset": ds,
            "hf_path": "",
        })
    return pd.DataFrame(rows, columns=REQUIRED_DB_COLS) if rows else pd.DataFrame(columns=REQUIRED_DB_COLS)


def backfill_hf_paths_by_relpath(
    df_db: pd.DataFrame,
    reports_root: Path,
    hf_repo: str,
    idx: pd.DataFrame,
    model_name: str | None,
    do_shard: bool,
    shard_digits: int,
) -> int:
    """
    For each local file path, compute the *target* repo relpath exactly as we upload it
    (model prefix + optional shard). If that relpath appears in index.csv, backfill hf_path.
    """
    rel_set = set(idx["relpath"].astype(str))
    updated = 0
    for i, p in enumerate(df_db["path"].astype(str).tolist()):
        if not p:
            continue
        lp = Path(p)
        if not lp.exists():
            continue
        try:
            base_rp = relpath_posix(lp, reports_root)                 # e.g. "file.html" or "model/.../file.html"
        except Exception:
            continue
        base_rp = ensure_model_prefix(base_rp, model_name)            # ensure "<model>/..."
        rp_target = shard_relpath_under_model(base_rp, shard_digits) if do_shard else base_rp
        if rp_target in rel_set and not df_db.at[i, "hf_path"]:
            df_db.at[i, "hf_path"] = f"hf://{hf_repo}/{rp_target}"
            updated += 1
    return updated

def backfill_hf_paths_by_filename(df_db: pd.DataFrame, hf_repo: str, idx: pd.DataFrame) -> int:
    updated = 0
    rel_by_fname = dict(zip(idx["filename"].astype(str), idx["relpath"].astype(str)))
    mask = df_db["hf_path"].astype(str) == ""
    for i in df_db.index[mask]:
        fn = str(df_db.at[i, "filename"])
        rp = rel_by_fname.get(fn)
        if rp:
            df_db.at[i, "hf_path"] = f"hf://{hf_repo}/{rp}"
            updated += 1
    return updated

def append_to_remote_index(remote_index: pd.DataFrame, new_rows: List[dict]) -> pd.DataFrame:
    if not new_rows:
        return remote_index
    add_df = pd.DataFrame(new_rows, columns=INDEX_COLS)
    merged = pd.concat([remote_index, add_df], ignore_index=True)
    merged = merged.drop_duplicates(subset=["relpath"], keep="first")
    return merged[INDEX_COLS]

def list_remote_relpaths(api: HfApi, repo_id: str) -> Set[str]:
    files = api.list_repo_files(repo_id=repo_id, repo_type="dataset")
    out = set()
    for f in files:
        parts = [unquote(s) for s in f.split("/")]
        out.add("/".join(parts))
    return out

def commit_ops_in_batches(api: HfApi, repo_id: str, ops: List, batch_size: int, msg_prefix: str):
    if not ops:
        return
    for start in range(0, len(ops), batch_size):
        batch = ops[start:start+batch_size]
        api.create_commit(
            repo_id=repo_id,
            repo_type="dataset",
            operations=batch,
            commit_message=f"{msg_prefix} (n={len(batch)})"
        )

# ---------- Wipe helpers ----------
def wipe_remote_dataset(api: HfApi, repo_id: str, keep: Set[str], batch_size: int, dry: bool):
    files = api.list_repo_files(repo_id=repo_id, repo_type="dataset")
    to_delete = []
    for f in files:
        f_norm = "/".join([unquote(s) for s in f.split("/")])
        if f_norm in keep:
            continue
        to_delete.append(CommitOperationDelete(path_in_repo=f_norm))
    if not to_delete:
        print("[wipe] nothing to delete")
        return
    if dry:
        print(f"[dry-run] would delete {len(to_delete)} files from {repo_id}")
        return
    print(f"[wipe] deleting {len(to_delete)} files from {repo_id} ...")
    commit_ops_in_batches(api, repo_id, to_delete, batch_size, "Wipe dataset")

def main():
    ap = argparse.ArgumentParser(description="Reset and sync HF dataset from local HTMLs (optionally wipe repo), shard to avoid 10k/dir limit, update index.csv, backfill hf_path.")
    ap.add_argument("--reports-root", default='/data/atlask/Model-Preds-Html/AudioSet-Audio', type=Path, help="Root containing {model}/.../*.html (or just the model dir)")
    ap.add_argument("--db-path", required=True, type=Path, help="Path to local library.csv")
    ap.add_argument("--manifest-csv", default="/data/atlask/BAU-Quant/manifest_val.csv", type=Path, help="CSV with columns file_name,dataset,category; matched by basename without extension")
    ap.add_argument("--repo-id", required=True, help="HF dataset repo id, e.g. USER/audio-html")
    ap.add_argument("--index-filename", default="index.csv", help="Index filename in the HF dataset (default: index.csv)")
    ap.add_argument("--batch-size", type=int, default=1000, help="Files per commit when uploading to HF")
    ap.add_argument("--dry-run", action="store_true", help="Print actions; do not write or push")
    ap.add_argument("--commit-message", default="Sync: add new HTMLs + update index.csv", help="Commit message prefix")
    # Reset/Wipe options
    ap.add_argument("--wipe-remote", action="store_true", help="Delete ALL files in the HF dataset before uploading")
    ap.add_argument("--keep", action="append", default=[], help="Paths to keep during wipe (can be passed multiple times)")
    ap.add_argument("--wipe-local", action="store_true", help="Delete local library.csv before scanning")
    # SHARD controls
    ap.add_argument("--no-shard", action="store_true", help="Disable sharding (NOT recommended; risk 10k/dir limit)")
    ap.add_argument("--shard-hexdigits", type=int, default=2, help="Digits of SHA1 prefix for bucket (default: 2 -> 256 buckets)")
    # Model prefix
    ap.add_argument("--model-name", type=str, default=None,
        help="Force all uploaded relpaths to be prefixed with this model folder (use if reports-root is already inside the model).")

    args = ap.parse_args()

    reports_root: Path = args.reports_root
    db_path: Path = args.db_path
    hf_repo: str = args.repo_id
    index_filename: str = args.index_filename
    bs: int = args.batch_size
    dry: bool = args.dry_run
    do_shard: bool = not args.no_shard
    shard_digits: int = max(1, args.shard_hexdigits)
    keep_set: Set[str] = set(args.keep)

    print(f"[config] reports_root={reports_root}")
    print(f"[config] db_path={db_path}")
    print(f"[config] repo_id={hf_repo}, index={index_filename}")
    print(f"[config] batch_size={bs}, dry_run={dry}, shard={'on' if do_shard else 'off'}:{shard_digits}")
    if args.model_name:
        print(f"[config] model_name={args.model_name}")
    if keep_set:
        print(f"[config] wipe keep-list: {sorted(keep_set)}")

    if os.environ.get("HF_HUB_ENABLE_HF_TRANSFER") != "1":
        print("[tip] For faster uploads, install `hf-transfer` and set HF_HUB_ENABLE_HF_TRANSFER=1")

    api = HfApi()

    manifest_map = load_manifest_map(args.manifest_csv)

    # 0) Optional wipes
    if args.wipe_remote:
        wipe_remote_dataset(api, hf_repo, keep_set, bs, dry)

    if args.wipe_local and db_path.exists():
        if dry:
            print(f"[dry-run] would remove local DB: {db_path}")
        else:
            print(f"[wipe] removing local DB: {db_path}")
            try:
                db_path.unlink()
            except FileNotFoundError:
                pass

    # 1) Load DB (fresh if wiped)
    df_db = load_db(db_path)

    # 2) Append new local *.html files to DB
    new_local_files = discover_new_local_htmls(reports_root, df_db)
    print(f"[scan] new local HTML files: {len(new_local_files)}")
    if new_local_files:
        df_new = rows_from_files(new_local_files, reports_root, manifest_map)
        df_db = pd.concat([df_db, df_new], ignore_index=True)


    # 3) Load remote index + list files (will be empty after wipe)
    remote_index, existed = load_hf_index(hf_repo, index_filename)
    print(f"[index] remote exists={existed}, rows={len(remote_index)}")
    remote_files_set = list_remote_relpaths(api, hf_repo)
    print(f"[remote] files in repo: {len(remote_files_set)}")

    # 4) Backfill hf_path (now uses model prefix + shard)
    n1 = backfill_hf_paths_by_relpath(
        df_db, reports_root, hf_repo, remote_index,
        model_name=args.model_name,
        do_shard=do_shard,
        shard_digits=shard_digits,
    )
    n2 = backfill_hf_paths_by_filename(df_db, hf_repo, remote_index)
    print(f"[hf] backfilled hf_path: by_relpath={n1}, by_filename={n2}")

    # 5) Decide which rows to upload (and target relpaths, sharded under model)
    need_upload = []
    for i, r in df_db.iterrows():
        # If you've wiped, hf_path will be empty; we only upload files that exist locally
        local = Path(str(r["path"]))
        if (not local) or (not local.exists()):
            continue
        try:
            base_rp = relpath_posix(local, reports_root)                  # "file.html" or "model/.../file.html"
        except Exception:
            continue
        base_rp = ensure_model_prefix(base_rp, args.model_name)           # ensure "<model>/..."
        rp = shard_relpath_under_model(base_rp, shard_digits) if do_shard else base_rp
        if rp not in remote_files_set:
            need_upload.append((i, r.to_dict(), rp))

    print(f"[hf] rows needing upload (not present in repo): {len(need_upload)}")

    ops: List[CommitOperationAdd] = []
    new_index_rows: List[dict] = []

    for i, rdict, rp in need_upload:
        local = Path(rdict["path"])
        if not local.exists():
            continue
        ops.append(CommitOperationAdd(path_in_repo=rp, path_or_fileobj=str(local)))
        # derive from HTML filename using manifest map
        k = key_from_html_filename(rdict["filename"])
        ds, cat = manifest_map.get(k, (str(rdict["dataset"]), str(rdict["category"])))
        new_index_rows.append({
            "id": rdict["id"] or uuid.uuid4().hex[:8],
            "filename": rdict["filename"],
            "relpath": rp,
            "category": cat,
            "dataset": ds,
            "tags": rdict["tags"],
            "keywords": rdict["keywords"],
            "notes": rdict["notes"],
            "uploaded_at": rdict["uploaded_at"] or now_iso(),
        })


    # 6) Upload in batches
    if ops and not dry:
        print(f"[hf] uploading {len(ops)} files in batches of {bs}...")
        commit_ops_in_batches(api, hf_repo, ops, bs, args.commit_message)
        remote_files_set = list_remote_relpaths(api, hf_repo)  # refresh
    elif ops and dry:
        print(f"[dry-run] would upload {len(ops)} files")

    # 7) Compose index.csv (fresh if wiped)
    current_index_rel = set(remote_index["relpath"].astype(str))
    current_index_rel.update([row["relpath"] for row in new_index_rows])
    missing_in_index = [rp for rp in remote_files_set if rp.endswith(".html") and rp not in current_index_rel]

    if missing_in_index:
        print(f"[index] adding {len(missing_in_index)} repo files that were missing from index.csv")
        for rp in missing_in_index:
            fname = Path(rp).name
            k = key_from_html_filename(fname)
            ds, cat = manifest_map.get(k, ("", ""))
            new_index_rows.append({
                "id": uuid.uuid4().hex[:8],
                "filename": fname,
                "relpath": rp,
                "category": cat,
                "dataset": ds,
                "tags": "",
                "keywords": "",
                "notes": "",
                "uploaded_at": now_iso(),
            })


    if new_index_rows or args.wipe_remote:
        # If wiped, overwrite index.csv with just merged content
        base_index = remote_index if not args.wipe_remote else pd.DataFrame(columns=INDEX_COLS)
        merged_index = append_to_remote_index(base_index, new_index_rows)
        merged_index = ensure_cols(merged_index, INDEX_COLS)
        if not dry:
            tmp = Path("index.updated.csv")
            merged_index.to_csv(tmp, index=False)
            api.create_commit(
                repo_id=hf_repo,
                repo_type="dataset",
                operations=[CommitOperationAdd(path_in_repo=index_filename, path_or_fileobj=str(tmp))],
                commit_message=f"{args.commit_message} (update {index_filename}, rows={len(merged_index)})"
            )
            tmp.unlink(missing_ok=True)
        else:
            print(f"[dry-run] would write fresh {index_filename} with {len(merged_index)} rows")

    # 8) Update local hf_path for rows now on HF (sharded + model-prefixed)
    for i, r in df_db.iterrows():
        if str(r.get("hf_path", "")):
            continue
        local = str(r["path"])
        if not local:
            continue
        p = Path(local)
        if not p.exists():
            continue
        try:
            base_rp = relpath_posix(p, reports_root)
        except Exception:
            continue
        base_rp = ensure_model_prefix(base_rp, args.model_name)
        rp = shard_relpath_under_model(base_rp, shard_digits) if do_shard else base_rp
        if rp in remote_files_set:
            df_db.at[i, "hf_path"] = f"hf://{hf_repo}/{rp}"

    # 8.5) Backfill dataset/category in DB from manifest if missing
    mask_missing = (df_db["dataset"].astype(str) == "") | (df_db["category"].astype(str) == "")
    for i, r in df_db[mask_missing].iterrows():
        k = key_from_html_filename(str(r["filename"]))
        if k in manifest_map:  # manifest_map was loaded earlier: load_manifest_map(args.manifest_csv)
            ds, cat = manifest_map[k]
            if not str(r["dataset"]):
                df_db.at[i, "dataset"] = ds
            if not str(r["category"]):
                df_db.at[i, "category"] = cat

    # 9) Save DB
    if dry:
        print("[dry-run] not writing library.csv")
    else:
        save_db(df_db, db_path)
        print(f"[done] wrote {len(df_db)} rows to {db_path}")


if __name__ == "__main__":
    try:
        main()
    except Exception as e:
        traceback.print_exc()
        sys.exit(1)