Complete corpus signal discovery fixes

.claude/skills/signal-discovery/SKILL.md

@@ -58,7 +58,10 @@ finding.
 3. Run the applicable fresh and historical discovery modes from the reference.
    Use `corpus_gap_scan.py --since YYYY-MM-DD` for the overlapping fresh window
    and omit `--since` for the complete history. Report counts per input shape
-   and per outcome; an example does not establish coverage.
+   and per outcome; an example does not establish coverage. Give every code and
+   corpus snapshot a distinct `--report` path. The adjacent
+   `.progress.jsonl` checkpoint is the live progress source and makes the same
+   command resumable; use `--restart` only when intentionally discarding it.
 4. Triage every candidate into one of: parser/serialization gap, missing vendor
    attribution, new visible mark, possible invisible signal, format/read error,
    expected non-AI provenance, or false candidate. Preserve unresolved items as

docs/supported-signals.md

@@ -92,8 +92,9 @@ The inspection and stripping code handles signals in these groups:
   MP4/MOV `moov.udta.meta.keys/ilst`, MKV/WebM
   `Segment.Tags.Tag.SimpleTag`, AVI `LIST/INFO/AIGC`, and FLV
   `script.onMetaData.AIGC` placements;
-- xAI and Grok EXIF signature fields in ordinary EXIF and ImageMagick PNG raw
-  EXIF profiles;
+- xAI and Grok signature-plus-UUID pairs in ordinary EXIF, ImageMagick PNG raw
+  EXIF profiles, XMP description/creator fields, PNG Description/Author text,
+  and IPTC Caption-Abstract/By-line fields;
 - Samsung AI editing markers;
 - Hugging Face job metadata;
 - open Stable Diffusion style DWT-DCT watermarks with the `detect` extra;

docs/verification-plan.md

@@ -498,6 +498,9 @@ the metadata-serialization probe inside that workflow. It now records raw
 markers, input shapes, identify errors, integrity clashes, unattributed results,
 and blind-marker candidates, but it does not cover visible marks, proprietary
 invisible marks, detection/removal parity, or historical crossfire.
+The scan runs files in parallel and checkpoints each completed row beside the
+requested CSV as `.progress.jsonl`; repeat the same command to resume an
+interrupted run, and use a distinct report path for every code/corpus snapshot.
 
 Run a fresh-delta discovery review weekly after the retained corpus is
 refreshed, using `corpus_gap_scan.py --since` with an overlap from the preceding

docs/watermarking-landscape.md

@@ -42,20 +42,20 @@ routes, not a claim that every Krea or Riverflow export is metadata-free.
 **ByteDance Volcano Engine (Volcengine)** — the cloud behind Doubao / Jimeng — signs its AI image output with a cert from `certificate_center@volcengine.com` + `trainedAlgorithmicMedia` (issuer `b"volcengine"` → "ByteDance (Volcano Engine)", platform "ByteDance Volcano Engine"); note this is the C2PA-signed surface, distinct from the XMP/PNG TC260 `AIGC` label Doubao also uses. ByteDance's **international brand (BytePlus / Seedream / Seededit)** signs the same content as **"Byteplus Pte. Ltd."**. The bare `volcengine` needle missed it, so BytePlus output was mis-attributed to "Adobe Firefly" through an incidental "Adobe XMP" toolkit string. Issuer `b"Byteplus"` maps directly to "BytePlus (ByteDance)". ByteDance's consumer app **Dreamina** (the international Jimeng brand) signs as **"Bytedance Pte. Ltd."** with a `Dreamina/x.y` claim generator but, unlike the Volcano Engine surface, ships **no `trainedAlgorithmicMedia`**. Issuer `b"Dreamina"` maps to "ByteDance Dreamina" with **`asserts_ai=True`**. Registering the broader **issuer** `b"Bytedance Pte"` was deliberately avoided because that same entity also signs non-AI CapCut edits; keying on the `Dreamina` generator token is precise.
 - **EXIF/XMP/PNG-text generator tag (caught by `exif_generator`):** **Ideogram** writes EXIF `Make="Ideogram AI"` (collected 2026-05-24 — no C2PA, no SynthID, no imwatermark; the Make tag is the only signal). Additional verified generator stamps include **NovelAI** (`Software`, `Source`, and `Title` PNG text chunks), **Reve** (`Software` or XMP `CreatorTool` = `reve.com`), and **Aphrodite AI** (`Make` or `Software` = `Aphrodite AI`). A JSON `generationParams.modelName` value is parsed structurally and reduced to a recognized model name, rather than exposing its surrounding prompt JSON as a platform label.
 - **App-export provenance and AIGC JSON:** supported ByteDance-family exports can place a JSON object in EXIF `ImageDescription` or `UserComment`, independently of C2PA or TC260. Exact `product` values for Doubao, Xinghui, and Dreamina are removable product provenance, but do not alone prove that the pixels were generated. Dreamina additionally requires `exportType=generation` for that verdict. A nested Aweme `aigc_type=1` or private ByteDance `aigc_label_type=1` / `2` is an AIGC disclosure; `0` is inconclusive and can occur on a Dreamina generation export. Plain Aweme, retouch, and `lv` exports are preserved. The lower-case private field is deliberately not interpreted as the normative TC260 `Label`, whose values `1` / `2` / `3` mean generated / possibly generated / suspected generated under [GB 45438-2025](https://www.tc260.org.cn/upload/2025-03-15/1742009439794081593.pdf).
-- **xAI / Grok legacy JPEG EXIF signature scheme (DETECTED by `metadata.xai_signature`, built 2026-05-26).**
+- **xAI / Grok signature scheme (DETECTED by `metadata.xai_signature`, built 2026-05-26, expanded 2026-09-13).**
 
-The sampled Grok JPEG downloads from the Aurora route carry **no C2PA, no XMP, no SynthID, no IPTC** — only EXIF `Artist` = a UUID and EXIF `ImageDescription` = `Signature: <base64>` (a crypto signature, unverifiable locally without xAI's public key). This is route-specific, not a product-wide no-C2PA claim: newer Grok Imagine output can carry a valid AI-source C2PA claim whose `claim_generator` is `Grok Imagine`. `exif_generator` misses the legacy pair (neither field holds an `AI_GENERATOR_TOKENS` token), so a dedicated detector `xai_signature(path)` matches it (`ImageDescription ~ ^Signature: [A-Za-z0-9+/=]{64,}` AND UUID `Artist`); wired into `has_ai_metadata`, `get_ai_metadata` (key `xai_signature`), and `identify` (signal `xai_signature`, platform "xAI (Grok / Aurora)").
+The sampled Grok JPEG downloads from the Aurora route carry **no C2PA, no XMP, no SynthID, no IPTC** — only EXIF `Artist` = a UUID and EXIF `ImageDescription` = `Signature: <base64>` (a crypto signature, unverifiable locally without xAI's public key). This is route-specific, not a product-wide no-C2PA claim: newer Grok Imagine output can carry a valid AI-source C2PA claim whose `claim_generator` is `Grok Imagine`. `exif_generator` misses the legacy pair (neither field holds an `AI_GENERATOR_TOKENS` token), so a dedicated detector `xai_signature(path)` requires both the long signature shape and a UUID; it is wired into `has_ai_metadata`, `get_ai_metadata` (key `xai_signature`), and `identify` (signal `xai_signature`, platform "xAI (Grok / Aurora)").
+
+Metadata-preserving transcodes can move the same two values out of EXIF. Confirmed equivalents are XMP `dc:description` + `dc:creator`, PNG `Description` + `Author`, and IPTC `Caption-Abstract` + `By-line`. Detection requires one of those exact field pairs and keeps XMP values within the same RDF description, so an unrelated document UUID cannot complete a signature-shaped caption. A signature-shaped caption without its paired creator UUID, or a UUID without the signature blob, remains ordinary metadata.
 
 **Format confirmed stable across n=3 genuine generations:** exactly three EXIF tags (`Artist`, `ExifOffset`, `ImageDescription`), `Signature:` prefix constant, base64 payload 300-1004 chars. Two capture facts: (a) the `Artist` UUID **equals the public image id** in the asset URL (`https://imagine-public.x.ai/imagine-public/images/<uuid>.jpg`), so it is NOT a private per-user secret — only the `Signature` blob is; (b) the Grok web-UI image is a re-encoded **WebP with no signature** — the EXIF survives only in the *original* JPEG (download button or that public tokenless URL), which is why screenshots / re-encodes are metadata-stripped. A real fixture `data/fixtures/provenance/grok-1.jpg` plus **synthetic** JPEG fixtures (fake UUID + fake `Signature:` blob) cover the detector; never add a real Grok image carrying private content (the repo is public).
 
-**Stripped on removal too:** `remove_ai_metadata` calls `_scrub_ai_exif` on
-JPEG EXIF, which deletes the xAI Signature and UUID Artist pair plus supported
-AI generator values while retaining unrelated camera and editor EXIF. The
-shared `xai_signature_pair` helper is the single source of truth for the
-pair. Converted PNGs can retain the same pair in an ImageMagick raw EXIF
-profile; Pillow exposes that form through `getexif()` even when `info["exif"]`
-is absent, so detection, removal, and portable records share that fallback. On
-the ISOBMFF path, `blank_ai_exif_tokens` provides the corresponding
+**Stripped on removal too:** `remove_ai_metadata` deletes both members of the
+pair from EXIF, XMP, PNG text, or IPTC while retaining unrelated camera and
+editor metadata. Converted PNGs can retain the original pair in an ImageMagick
+raw EXIF profile; Pillow exposes that form through `getexif()` even when
+`info["exif"]` is absent, so detection, removal, and portable records share that
+fallback. On the ISOBMFF path, `blank_ai_exif_tokens` provides the corresponding
 in-place scrub for supported EXIF values, TC260 AIGC blocks, and the xAI pair.
 - **China TC260 AIGC label (caught by `AIGC_MARKERS` / `metadata.aigc_label`, surfaced by `identify` as the `aigc` signal):** China-served generators embed an XMP `<TC260:AIGC>{"Label":"1","ContentProducer":...}` block — China's mandatory AI-content labeling (TC260 namespace `tc260.org.cn/ns/AIGC`). The label says only "this is AI", but its `ContentProducer` names the signing entity — `001` + `1` + an 18-char Unified Social Credit Code + a 5-digit product suffix, normalized by `metadata.uscc_of`, or for a few generators a bare product name. `KnownMark.tc260_producer_codes` maps the codes settled per vendor by `scripts/vendor_cohort_harvest.py` to registry mark keys, so an AIGC image relaxes only the detector of the product it actually carries; an unmapped or absent producer leaves every product strict. The same mapping supplies the manufacturer for provenance-conflict checks. A code identifies a legal entity, not necessarily one brand, so a hosting or aggregating platform that signs for several apps is a recall bet rather than a proof.
 

pyproject.toml

@@ -4,7 +4,7 @@
 # on it, including the ComfyUI node package. The console script below carries
 # the same weight, since users have it on PATH.
 name = "remove-ai-watermarks"
-version = "0.40.1"
+version = "0.40.2"
 description = "AI watermark remover for visible, invisible, and provenance marks in images and video"
 readme = "README.md"
 requires-python = ">=3.11,<3.15"

scripts/corpus_gap_scan.py

@@ -2,9 +2,10 @@
 
 Two jobs in one pass:
 
-1. **Report** -- run ``identify`` over every image and write one CSV row per file
-   (verdict, platform, confidence, watermarks, signals, raw metadata markers,
-   candidate classes, integrity clashes, and errors).
+1. **Report** -- run metadata-only ``identify`` over every file and write one CSV
+   row per file (verdict, platform, confidence, watermarks, signals, raw metadata
+   markers, candidate classes, integrity clashes, and errors). Visible and
+   invisible pixel detectors belong to their own corpus audits.
 2. **Gap audit** -- for every ``unknown``-verdict file, scan only its *metadata
    region* (PNG text/eXIf chunks, JPEG APPn segments before SOS, or the file
    head for other containers) for known provenance markers. A marker found there
@@ -20,17 +21,25 @@ This is how new detector gaps get found (it is what surfaced the JPEG-EXIF
 Usage:
     uv run python scripts/corpus_gap_scan.py --corpus .local-eval/originals
     uv run python scripts/corpus_gap_scan.py --corpus .local-eval/originals \\
-        --report .local-eval/detector-report.csv
+        --workers 8 --report .local-eval/detector-report.csv
     uv run python scripts/corpus_gap_scan.py --corpus .local-eval/originals \\
         --since 2026-09-01 --report .local-eval/detector-report-weekly.csv
+
+Rows stream into ``<report>.progress.jsonl`` and resume by relative path after an
+interruption. Use a new report name for a new code/corpus snapshot, or pass
+``--restart`` when intentionally replacing the checkpoint. The final CSV is
+written atomically after every selected row is present.
 """
 
 from __future__ import annotations
 
 import csv
+import itertools
+import json
 import logging
-import re
+import os
 from collections import Counter
+from concurrent.futures import FIRST_COMPLETED, Future, ProcessPoolExecutor, wait
 from datetime import date
 from importlib.metadata import version
 from pathlib import Path
@@ -48,6 +57,7 @@ from remove_ai_watermarks.metadata import (
     c2pa_marker_in,
     samsung_genai_in,
     scan_head,
+    xai_signature_in_metadata,
 )
 
 log = logging.getLogger(__name__)
@@ -79,9 +89,6 @@ MARKERS: tuple[bytes, ...] = (
     b"Nano Banana",
     b"Stability AI",
 )
-_XAI_SIGNATURE_RE = re.compile(rb"Signature:\s*[A-Za-z0-9+/=]{64,}")
-_UUID_RE = re.compile(rb"[0-9a-f]{8}-[0-9a-f]{4}-[0-9a-f]{4}-[0-9a-f]{4}-[0-9a-f]{12}", re.IGNORECASE)
-
 REPORT_FIELDS: tuple[str, ...] = (
     "path",
     "suffix",
@@ -96,6 +103,8 @@ REPORT_FIELDS: tuple[str, ...] = (
     "candidate_classes",
     "error",
 )
+_DISPLAY_CANDIDATE_LIMIT = 50
+_WORKER_BATCH_SIZE = 16
 
 
 def _base_row(path: str, suffix: str, lib_version: str) -> dict[str, str]:
@@ -128,7 +137,7 @@ def _marker_hits(region: bytes) -> list[str]:
         hits.add("IPTC AI disclosure")
     if samsung_genai_in(region) is not None:
         hits.add("Samsung genAIType")
-    if _XAI_SIGNATURE_RE.search(region) and _UUID_RE.search(region):
+    if xai_signature_in_metadata(region):
         hits.add("xAI signature pair")
     return sorted(hits)
 
@@ -176,6 +185,106 @@ def _files(corpus: Path, since: date | None) -> list[Path]:
     return sorted(path for root in roots for path in root.rglob("*") if path.is_file())
 
 
+def _scan_one(args: tuple[str, str, str]) -> dict[str, str]:
+    """Scan one path in a worker process and always return a report row."""
+    path_str, relative_path, lib_version = args
+    path = Path(path_str)
+    suffix = path.suffix.lower()
+    try:
+        rep = identify(path, check_visible=False, check_invisible=False)
+    except Exception as exc:
+        log.warning("identify failed on %s: %s", relative_path, exc)
+        row = _base_row(relative_path, suffix, lib_version)
+        row["candidate_classes"] = "identify_error"
+        row["error"] = f"{type(exc).__name__}: {exc}"[:300].replace("\n", " ")
+    else:
+        row = _row(rep, path=relative_path, suffix=suffix, lib_version=lib_version)
+        hits = _marker_hits_for_path(path)
+        row["markers"] = "|".join(hits)
+        row["candidate_classes"] = "|".join(_candidate_classes(rep, hits))
+        return row
+    row["markers"] = "|".join(_marker_hits_for_path(path))
+    return row
+
+
+def _scan_batch(args: tuple[tuple[str, str, str], ...]) -> list[dict[str, str]]:
+    """Scan a small path batch to amortize process-pool dispatch overhead."""
+    return [_scan_one(item) for item in args]
+
+
+def _read_checkpoint(path: Path) -> dict[str, dict[str, str]]:
+    """Read complete JSONL rows, tolerating an interrupted final write."""
+    rows: dict[str, dict[str, str]] = {}
+    if not path.exists():
+        return rows
+    with path.open(encoding="utf-8") as stream:
+        for line in stream:
+            try:
+                row = json.loads(line)
+            except (json.JSONDecodeError, UnicodeDecodeError):
+                continue
+            if isinstance(row, dict) and isinstance(row.get("path"), str):
+                rows[row["path"]] = {field: str(row.get(field, "")) for field in REPORT_FIELDS}
+    return rows
+
+
+def _repair_checkpoint(path: Path) -> None:
+    """Remove an interrupted final JSONL fragment before another append."""
+    if not path.exists():
+        return
+    with path.open("rb+") as stream:
+        data = stream.read()
+        if data and not data.endswith(b"\n"):
+            last_newline = data.rfind(b"\n")
+            stream.truncate(last_newline + 1)
+
+
+def _summarize(rows: list[dict[str, str]]) -> None:
+    """Print bounded aggregate output for a completed or resumed report."""
+    verdicts: Counter[str] = Counter()
+    platforms: Counter[str] = Counter()
+    shapes: Counter[str] = Counter()
+    errors = 0
+    for row in rows:
+        shapes[row["suffix"] or "(none)"] += 1
+        if row["error"]:
+            errors += 1
+        elif row["is_ai"] == "True":
+            verdicts["ai"] += 1
+            platforms[row["platform"] or "?"] += 1
+        else:
+            verdicts["unknown"] += 1
+
+    console.print(f"\n[bold]Verdicts:[/bold] AI {verdicts['ai']} | unknown {verdicts['unknown']} | errors {errors}")
+    console.print("[bold]Input shapes:[/bold] " + " | ".join(f"{name} {n}" for name, n in shapes.most_common()))
+    plat = Table(title="AI platforms", show_header=False)
+    for name, count in platforms.most_common():
+        plat.add_row(str(count), name)
+    console.print(plat)
+
+    candidates = [row for row in rows if row["candidate_classes"]]
+    if not candidates:
+        console.print("\n[green]No review candidates in this run.[/green]")
+        return
+    candidate_counts = Counter(candidate for row in candidates for candidate in row["candidate_classes"].split("|"))
+    gap_tokens = Counter(marker for row in rows for marker in row["markers"].split("|") if marker)
+    console.print(
+        f"\n[bold red]Review candidates[/bold red]: {len(candidates)} file(s) "
+        f"({', '.join(f'{name}={n}' for name, n in candidate_counts.most_common())})"
+    )
+    tok = Table(title="metadata markers seen across the run")
+    tok.add_column("count", justify="right")
+    tok.add_column("marker")
+    for name, count in gap_tokens.most_common():
+        tok.add_row(str(count), name)
+    console.print(tok)
+    for row in candidates[:_DISPLAY_CANDIDATE_LIMIT]:
+        detail = f"; markers={row['markers'].replace('|', ', ')}" if row["markers"] else ""
+        console.print(f"  {row['path']}  ->  {row['candidate_classes'].replace('|', ', ')}{detail}")
+    if (hidden := len(candidates) - _DISPLAY_CANDIDATE_LIMIT) > 0:
+        console.print(f"  ... {hidden} more candidate row(s); inspect the CSV for the complete set")
+
+
 @click.command()
 @click.option(
     "--corpus",
@@ -197,7 +306,9 @@ def _files(corpus: Path, since: date | None) -> list[Path]:
     help="Only scan files under YYYY-MM-DD corpus directories on or after this date.",
 )
 @click.option("--limit", type=int, default=0, help="Scan at most N files (0 = all).")
-def main(corpus: Path, report: Path | None, since, limit: int) -> None:  # noqa: ANN001
+@click.option("--workers", type=click.IntRange(min=1), default=max(1, (os.cpu_count() or 4) - 2), show_default=True)
+@click.option("--restart", is_flag=True, help="Discard this report's checkpoint and scan every selected file again.")
+def main(corpus: Path, report: Path | None, since, limit: int, workers: int, restart: bool) -> None:  # noqa: ANN001
     logging.basicConfig(level=logging.WARNING, format="%(message)s")
     report = report or corpus.parent / "detector_report.csv"
 
@@ -205,78 +316,59 @@ def main(corpus: Path, report: Path | None, since, limit: int) -> None:  # noqa:
     files = _files(corpus, since_date)
     if limit:
         files = files[:limit]
-    console.print(f"Scanning [bold]{len(files)}[/bold] files under {corpus} ...")
-
-    verdicts: Counter[str] = Counter()
-    platforms: Counter[str] = Counter()
-    shapes: Counter[str] = Counter()
-    rows: list[dict[str, str]] = []
-    errors = 0
+    checkpoint = report.with_suffix(report.suffix + ".progress.jsonl")
+    if restart:
+        checkpoint.unlink(missing_ok=True)
+        report.unlink(missing_ok=True)
+    _repair_checkpoint(checkpoint)
     lib_version = version("remove-ai-watermarks")
+    selected = [(path, str(path.relative_to(corpus))) for path in files]
+    selected_paths = {relative for _, relative in selected}
+    completed = {relative: row for relative, row in _read_checkpoint(checkpoint).items() if relative in selected_paths}
+    todo = [(path, relative) for path, relative in selected if relative not in completed]
+    console.print(
+        f"Scanning [bold]{len(files)}[/bold] files under {corpus}: "
+        f"[bold]{len(completed)}[/bold] checkpointed, [bold]{len(todo)}[/bold] remaining, workers {workers}"
+    )
 
-    with click.progressbar(files, label="identify") as bar:
-        for p in bar:
-            rel = str(p.relative_to(corpus))
-            shapes[p.suffix.lower() or "(none)"] += 1
-            try:
-                rep = identify(p)
-            except Exception as exc:
-                log.warning("identify failed on %s: %s", rel, exc)
-                errors += 1
-                row = _base_row(rel, p.suffix.lower(), lib_version)
-                row["candidate_classes"] = "identify_error"
-                row["error"] = f"{type(exc).__name__}: {exc}"[:300]
-                hits = _marker_hits_for_path(p)
-                row["markers"] = "|".join(hits)
-                rows.append(row)
-                continue
-            row = _row(rep, path=rel, suffix=p.suffix.lower(), lib_version=lib_version)
-            # identify() has already populated scan_head's bounded cache, so this
-            # reuses the exact metadata region rather than reading the file again.
-            hits = _marker_hits_for_path(p)
-            classes = _candidate_classes(rep, hits)
-            row["markers"] = "|".join(hits)
-            row["candidate_classes"] = "|".join(classes)
-            rows.append(row)
-            if rep.is_ai_generated:
-                verdicts["ai"] += 1
-                platforms[rep.platform or "?"] += 1
-                continue
-            verdicts["unknown"] += 1
+    checkpoint.parent.mkdir(parents=True, exist_ok=True)
+    with checkpoint.open("a", encoding="utf-8") as stream, ProcessPoolExecutor(max_workers=workers) as executor:
+        work = iter((str(path), relative, lib_version) for path, relative in todo)
 
+        def submit_batch() -> Future[list[dict[str, str]]] | None:
+            batch = tuple(itertools.islice(work, _WORKER_BATCH_SIZE))
+            return executor.submit(_scan_batch, batch) if batch else None
+
+        pending = {future for _ in range(workers * 2) if (future := submit_batch()) is not None}
+        processed = 0
+        next_progress = 500
+        while pending:
+            ready, pending = wait(pending, return_when=FIRST_COMPLETED)
+            for future in ready:
+                for row in future.result():
+                    completed[row["path"]] = row
+                    stream.write(json.dumps(row, ensure_ascii=False) + "\n")
+                    processed += 1
+                if replacement := submit_batch():
+                    pending.add(replacement)
+            stream.flush()
+            if processed >= next_progress:
+                console.print(
+                    f"  {processed}/{len(todo)} new; {len(completed)}/{len(files)} total",
+                    highlight=False,
+                )
+                next_progress += 500
+
+    rows = [completed[relative] for _, relative in selected]
     report.parent.mkdir(parents=True, exist_ok=True)
-    with report.open("w", newline="") as f:
-        writer = csv.DictWriter(f, fieldnames=REPORT_FIELDS)
+    temporary_report = report.with_suffix(report.suffix + ".tmp")
+    with temporary_report.open("w", newline="") as stream:
+        writer = csv.DictWriter(stream, fieldnames=REPORT_FIELDS)
         writer.writeheader()
         writer.writerows(rows)
+    temporary_report.replace(report)
     console.print(f"\nWrote [bold]{len(rows)}[/bold] rows -> {report}")
-
-    console.print(f"\n[bold]Verdicts:[/bold] AI {verdicts['ai']} | unknown {verdicts['unknown']} | errors {errors}")
-    console.print("[bold]Input shapes:[/bold] " + " | ".join(f"{name} {n}" for name, n in shapes.most_common()))
-    plat = Table(title="AI platforms", show_header=False)
-    for name, n in platforms.most_common():
-        plat.add_row(str(n), name)
-    console.print(plat)
-
-    candidates = [row for row in rows if row["candidate_classes"]]
-    if candidates:
-        candidate_counts = Counter(candidate for row in candidates for candidate in row["candidate_classes"].split("|"))
-        gap_tokens = Counter(marker for row in rows for marker in row["markers"].split("|") if marker)
-        console.print(
-            f"\n[bold red]Review candidates[/bold red]: {len(candidates)} file(s) "
-            f"({', '.join(f'{name}={n}' for name, n in candidate_counts.most_common())})"
-        )
-        tok = Table(title="metadata markers seen across the run")
-        tok.add_column("count", justify="right")
-        tok.add_column("marker")
-        for name, n in gap_tokens.most_common():
-            tok.add_row(str(n), name)
-        console.print(tok)
-        for row in candidates:
-            detail = f"; markers={row['markers'].replace('|', ', ')}" if row["markers"] else ""
-            console.print(f"  {row['path']}  ->  {row['candidate_classes'].replace('|', ', ')}{detail}")
-    else:
-        console.print("\n[green]No review candidates in this run.[/green]")
+    _summarize(rows)
 
 
 if __name__ == "__main__":

src/remove_ai_watermarks/__init__.py

@@ -33,7 +33,7 @@ _os.environ.setdefault("TRANSFORMERS_VERBOSITY", "error")
 _warnings.filterwarnings("ignore", message=r".*ImageProcessorFast.*")
 
 
-__version__ = "0.40.1"
+__version__ = "0.40.2"
 
 __all__ = [
     "BatchItemResult",

src/remove_ai_watermarks/identify.py

@@ -66,6 +66,7 @@ from remove_ai_watermarks.metadata import (
     samsung_genai_in,
     scan_head,
     xai_signature,
+    xai_signature_in_metadata,
     xai_signature_pair,
 )
 from remove_ai_watermarks.watermark_registry import GEMINI_SPARKLE_TRUST_CONF
@@ -419,10 +420,16 @@ def evidence_from_metadata_record(
     values_by_key: dict[str, str] = {}
     for key, value in pairs:
         if isinstance(value, (bytes, str)):
-            values_by_key.setdefault(key.lower(), _external_text(value))
-    description = values_by_key.get("imagedescription", "")
-    artist = values_by_key.get("artist", "")
-    xai = xai_signature_pair(description, artist)
+            values_by_key.setdefault(key.lower().removeprefix("info:"), _external_text(value))
+    xai = any(
+        xai_signature_pair(values_by_key.get(description_key, ""), values_by_key.get(creator_key, ""))
+        for description_key, creator_key in (
+            ("imagedescription", "artist"),
+            ("description", "author"),
+            ("description", "creator"),
+            ("caption-abstract", "by-line"),
+        )
+    ) or xai_signature_in_metadata(scan)
 
     hf_job = next(
         (
@@ -446,7 +453,7 @@ def evidence_from_metadata_record(
             f"China AIGC label (TC260){f'; producer {producer}' if producer else ''}",
         )
     if xai:
-        ai_metadata.setdefault("xai_signature", "xAI/Grok EXIF signature (Artist UUID + Signature blob)")
+        ai_metadata.setdefault("xai_signature", "xAI/Grok signature (UUID + Signature blob)")
     if iptc_system:
         ai_metadata.setdefault("ai_system", f"IPTC 2025.1 AI disclosure ({iptc_system})")
     if hf_job:
@@ -1424,13 +1431,12 @@ def _identify_from_evidence(
         if v := _vendor_of(generator_tag):
             ai_vendor_claims["exif_generator"] = v
 
-    # ── xAI / Grok EXIF signature scheme (no C2PA/SynthID/IPTC) ──────
-    # Grok's only provenance signal: EXIF ImageDescription "Signature: <base64>"
-    # + a UUID Artist. Distinct from exif_generator (which matches generator
-    # tokens); verified stable across 3 generations. See CLAUDE.md.
+    # ── xAI / Grok signature scheme ───────────────────────────────────
+    # The original EXIF pair can move into equivalent XMP, PNG text, or IPTC
+    # fields during a metadata-preserving transcode. Neither shape alone is enough.
     if evidence.xai_signature:
-        signals.append(Signal("xai_signature", "EXIF Signature blob + UUID Artist", "high"))
-        watermarks.append("xAI/Grok EXIF signature")
+        signals.append(Signal("xai_signature", "Signature blob + UUID provenance pair", "high"))
+        watermarks.append("xAI/Grok provenance signature")
         if platform is None:
             platform = "xAI (Grok / Aurora)"
         ai_vendor_claims["xai"] = "xAI"

src/remove_ai_watermarks/metadata.py

@@ -527,7 +527,7 @@ def has_ai_metadata(image_path: Path) -> bool:
     # Hugging Face-hosted job marker (hf-job-id PNG text chunk).
     if huggingface_job(image_path):
         return True
-    # xAI / Grok: no C2PA/IPTC/XMP -- only the EXIF Signature + UUID-Artist pair.
+    # xAI / Grok signature + UUID pair, including container-native transcodes.
     return xai_signature(image_path)
 
 
@@ -1052,13 +1052,122 @@ def exif_generator(image_path: Path) -> str | None:
 # pair xAI-specific -- both required keeps the false-positive rate near zero.
 _XAI_SIGNATURE_RE = re.compile(r"Signature:\s*[A-Za-z0-9+/=]{64,}")
 _UUID_RE = re.compile(r"[0-9a-f]{8}-[0-9a-f]{4}-[0-9a-f]{4}-[0-9a-f]{4}-[0-9a-f]{12}", re.IGNORECASE)
+_RDF_DESCRIPTION_RE = re.compile(rb"<rdf:Description\b[^>]*>.*?</rdf:Description>", re.IGNORECASE | re.DOTALL)
+_XMP_DESCRIPTION_RE = re.compile(
+    rb"(?:<dc:description\b[^>]*>(.*?)</dc:description>|\bdc:description=[\"']([^\"']*)[\"'])",
+    re.IGNORECASE | re.DOTALL,
+)
+_XMP_CREATOR_RE = re.compile(
+    rb"(?:<dc:creator\b[^>]*>(.*?)</dc:creator>|\bdc:creator=[\"']([^\"']*)[\"'])",
+    re.IGNORECASE | re.DOTALL,
+)
 
 
 def xai_signature_pair(description: str, artist: str) -> bool:
-    """True if an EXIF (ImageDescription, Artist) pair is xAI/Grok's scheme."""
+    """True if a description/creator pair has xAI/Grok's two value shapes."""
     return _XAI_SIGNATURE_RE.match(description) is not None and _UUID_RE.fullmatch(artist) is not None
 
 
+def _decoded_xai_signature_pair(description: bytes, creator: bytes) -> bool:
+    description = re.sub(rb"<[^>]+>", b"", description).strip()
+    creator = re.sub(rb"<[^>]+>", b"", creator).strip()
+    return xai_signature_pair(
+        description.decode("utf-8", "replace").strip(),
+        creator.decode("utf-8", "replace").strip(),
+    )
+
+
+def _xmp_xai_signature_pair(data: bytes) -> bool:
+    """Match the two xAI values only inside the same RDF description."""
+    for block_match in _RDF_DESCRIPTION_RE.finditer(data):
+        block = block_match.group()
+        descriptions = [
+            next(value for value in match.groups() if value is not None)
+            for match in _XMP_DESCRIPTION_RE.finditer(block)
+        ]
+        creators = [
+            next(value for value in match.groups() if value is not None) for match in _XMP_CREATOR_RE.finditer(block)
+        ]
+        if any(
+            _decoded_xai_signature_pair(description, creator) for description in descriptions for creator in creators
+        ):
+            return True
+    return False
+
+
+def _iptc_dataset_values(data: bytes, dataset: int) -> list[bytes]:
+    """Return bounded IPTC application-record values for one dataset number."""
+    values: list[bytes] = []
+    marker = bytes((0x1C, 0x02, dataset))
+    start = 0
+    while (index := data.find(marker, start)) >= 0:
+        length_start = index + len(marker)
+        if length_start + 2 > len(data):
+            break
+        length = int.from_bytes(data[length_start : length_start + 2], "big")
+        value_start = length_start + 2
+        value_end = value_start + length
+        if value_end <= len(data) and length <= 4096:
+            values.append(data[value_start:value_end])
+        start = length_start
+    return values
+
+
+def _iptc_xai_signature_pair(data: bytes) -> bool:
+    descriptions = _iptc_dataset_values(data, 120)  # Caption-Abstract
+    creators = _iptc_dataset_values(data, 80)  # By-line
+    # Some metadata exports retain the field names instead of the binary dataset
+    # headers. Keep that observed representation precise by anchoring each value to
+    # its field label.
+    descriptions += re.findall(rb"Caption-Abstract\x00([^\x00]{1,4096})", data, re.IGNORECASE)
+    creators += re.findall(rb"By-line\x00([^\x00]{1,4096})", data, re.IGNORECASE)
+    return any(
+        _decoded_xai_signature_pair(description, creator) for description in descriptions for creator in creators
+    )
+
+
+def _jpeg_app_payloads(data: bytes) -> list[tuple[int, bytes]]:
+    """Extract complete XMP/IPTC APP payloads from a JPEG or concatenated regions."""
+    payloads: list[tuple[int, bytes]] = []
+    position = 2 if data.startswith(b"\xff\xd8") else 0
+    while position + 4 <= len(data):
+        if data[position] != 0xFF:
+            break
+        marker = data[position + 1]
+        if marker in (0xDA, 0xD9):  # SOS / EOI
+            break
+        if 0xD0 <= marker <= 0xD7 or marker == 0x01:
+            position += 2
+            continue
+        length = int.from_bytes(data[position + 2 : position + 4], "big")
+        end = position + 2 + length
+        if length >= 2 and end <= len(data):
+            if marker in (0xE1, 0xED):
+                payloads.append((marker, data[position + 4 : end]))
+            position = end
+        else:
+            break
+    return payloads
+
+
+def xai_signature_in_metadata(data: bytes) -> bool:
+    """True when one metadata record carries xAI's paired provenance fields.
+
+    Editors preserve the same two values under equivalent container-native fields:
+    XMP ``dc:description``/``dc:creator`` and IPTC
+    ``Caption-Abstract``/``By-line``. Values must be attached to the confirmed field
+    names inside one APP segment or RDF description; an unrelated XMP UUID must not
+    complete the pair.
+    """
+    payloads = _jpeg_app_payloads(data)
+    if payloads:
+        return any(
+            _xmp_xai_signature_pair(payload) if marker == 0xE1 else _iptc_xai_signature_pair(payload)
+            for marker, payload in payloads
+        )
+    return _xmp_xai_signature_pair(data) or _iptc_xai_signature_pair(data)
+
+
 def _exif_text(ifd: dict[int, Any], tag: int) -> str:
     """Decode a piexif 0th-IFD byte tag to a stripped string ('' if absent)."""
     value = ifd.get(tag)
@@ -1066,34 +1175,52 @@ def _exif_text(ifd: dict[int, Any], tag: int) -> str:
 
 
 def _xai_signature_impl(image_path: Path) -> bool:
-    """Detect xAI / Grok's EXIF provenance signature scheme.
+    """Detect xAI / Grok's provenance signature scheme.
 
-    Grok image downloads (Aurora model) carry no C2PA, XMP, SynthID, or IPTC --
-    their only provenance signal is a private EXIF pair: ``ImageDescription`` =
-    ``"Signature: <base64>"`` together with ``Artist`` = the image UUID. Verified
-    stable across three independent generations (2026-05-26; see CLAUDE.md). The
-    signature is xAI's and is not locally verifiable (no public key); detection
-    keys on this distinctive, low-false-positive shape, not on the signature's
-    validity. It survives only on the *original* JPEG download -- the web-UI
-    image is a re-encoded WebP that drops EXIF.
+    Legacy Grok/Aurora downloads use a private EXIF pair: ``ImageDescription`` =
+    ``"Signature: <base64>"`` together with ``Artist`` = the image UUID. Metadata-
+    preserving editors can move that exact pair into XMP, PNG text, or IPTC fields.
+    The signature is not locally verifiable (no public key); detection keys on both
+    distinctive shapes, not on the signature's validity.
     """
+    decoded_pair = False
     try:
         import piexif
         from PIL import Image
+        from PIL.IptcImagePlugin import getiptcinfo
 
         with Image.open(image_path) as img:
             exif_bytes = exif_bytes_from_image(img)
-        if not exif_bytes:
-            return False
-        tags = piexif.load(exif_bytes).get("0th", {})
+            descriptions = [img.info.get(key) for key in ("Description", "ImageDescription")]
+            creators = [img.info.get(key) for key in ("Author", "Artist", "Creator")]
+            decoded_pair = any(
+                xai_signature_pair(description, creator)
+                for description in descriptions
+                if isinstance(description, str)
+                for creator in creators
+                if isinstance(creator, str)
+            )
+            if iptc := getiptcinfo(img):
+                caption = iptc.get((2, 120), b"")
+                byline = iptc.get((2, 80), b"")
+                if isinstance(caption, bytes) and isinstance(byline, bytes):
+                    decoded_pair = decoded_pair or _decoded_xai_signature_pair(caption, byline)
+        if exif_bytes:
+            tags = piexif.load(exif_bytes).get("0th", {})
+            decoded_pair = decoded_pair or xai_signature_pair(
+                _exif_text(tags, piexif.ImageIFD.ImageDescription),
+                _exif_text(tags, piexif.ImageIFD.Artist),
+            )
     except Exception as exc:  # unopenable format / malformed EXIF
         logger.debug("xAI-signature EXIF read failed for %s: %s", image_path, exc)
+    if decoded_pair:
+        return True
+    try:
+        return xai_signature_in_metadata(scan_head(image_path))
+    except OSError as exc:
+        logger.debug("xAI-signature metadata scan failed for %s: %s", image_path, exc)
         return False
 
-    return xai_signature_pair(
-        _exif_text(tags, piexif.ImageIFD.ImageDescription), _exif_text(tags, piexif.ImageIFD.Artist)
-    )
-
 
 def _is_aigc_exif_value(raw: object) -> bool:
     """Whether an EXIF tag value carries a China TC260 AIGC producer/service block.
@@ -1276,9 +1403,9 @@ def get_ai_metadata(image_path: Path) -> dict[str, str]:
     if app_generator:
         result.setdefault("app_aigc", f"App AIGC disclosure ({app_generator})")
 
-    # xAI / Grok EXIF signature scheme (its only provenance signal).
+    # xAI / Grok signature pair in EXIF or an equivalent container-native field.
     if xai_signature(image_path):
-        result.setdefault("xai_signature", "xAI/Grok EXIF signature (Artist UUID + Signature blob)")
+        result.setdefault("xai_signature", "xAI/Grok signature (UUID + Signature blob)")
 
     # IPTC 2025.1 AI-disclosure XMP fields (Iptc4xmpExt:AISystemUsed etc.).
     if system := iptc_ai_system(image_path):
@@ -1359,12 +1486,15 @@ def _jpeg_app_carries_ai(marker: int, payload: bytes) -> bool:
             or any(m in payload for m in AIGC_MARKERS)
             or any(m in payload for m in IPTC_AI_MARKERS)
             or any(m in payload for m in IPTC_AI_FIELD_MARKERS)
+            or xai_signature_in_metadata(payload)
         )
     ):
         return True
     # IPTC "Made with AI" record (APP13).
     if marker == 0xED and (
-        any(m in payload for m in IPTC_AI_MARKERS) or any(m in payload for m in IPTC_AI_FIELD_MARKERS)
+        any(m in payload for m in IPTC_AI_MARKERS)
+        or any(m in payload for m in IPTC_AI_FIELD_MARKERS)
+        or xai_signature_in_metadata(payload)
     ):
         return True
     # A bare / wrapped China TC260 AIGC block (``AIGC{...}`` or ``{"AIGC":{...}}``) glued
@@ -1480,6 +1610,7 @@ def _strip_png_metadata_lossless(source_path: Path, output_path: Path, keep_stan
     data = source_path.read_bytes()
     if not data.startswith(PNG_SIGNATURE):
         return False
+    source_has_xai_pair = xai_signature(source_path)
     out = bytearray(PNG_SIGNATURE)
     view = memoryview(data)
     pos, n = 8, len(data)
@@ -1502,7 +1633,13 @@ def _strip_png_metadata_lossless(source_path: Path, output_path: Path, keep_stan
                     + scrubbed
                     + struct.pack(">I", zlib.crc32(b"eXIf" + scrubbed))
                 )
-        elif not _png_chunk_should_be_dropped(chunk_type, payload, keep_standard, png_text_decode):
+        elif not _png_chunk_should_be_dropped(
+            chunk_type,
+            payload,
+            keep_standard,
+            png_text_decode,
+            xai_pair_present=source_has_xai_pair,
+        ):
             out += view[pos:end]
         if chunk_type == b"IEND":
             break
@@ -1534,13 +1671,32 @@ def _scrubbed_png_exif(payload: bytes) -> bytes | None:
         return None
 
 
-def _keep_standard_text_metadata(key: str, value: object, keep_standard: bool) -> bool:
+def _xai_text_pair_member(key: str, value: object) -> bool:
+    """Whether one standard text item is a member of an already-proven xAI pair."""
+    if not isinstance(value, str):
+        return False
+    normalized = key.lower().replace("-", "").replace("_", "")
+    if normalized in {"description", "imagedescription", "captionabstract"}:
+        return _XAI_SIGNATURE_RE.search(value) is not None
+    if normalized in {"author", "artist", "creator", "byline"}:
+        return _UUID_RE.fullmatch(value.strip()) is not None
+    return False
+
+
+def _keep_standard_text_metadata(
+    key: str,
+    value: object,
+    keep_standard: bool,
+    *,
+    xai_pair_present: bool = False,
+) -> bool:
     """Whether a text metadata item is standard and free of AI provenance."""
     return (
         keep_standard
         and key in STANDARD_METADATA_KEYS
         and not _is_ai_key(key)
         and not (isinstance(value, str) and (_is_ai_value(value) or _is_aigc_exif_value(value)))
+        and not (xai_pair_present and _xai_text_pair_member(key, value))
     )
 
 
@@ -1549,6 +1705,8 @@ def _png_chunk_should_be_dropped(
     payload: bytes,
     keep_standard: bool,
     decode: Callable[[str, bytes], str],
+    *,
+    xai_pair_present: bool = False,
 ) -> bool:
     """True when a PNG chunk must be dropped by the lossless strip.
 
@@ -1565,7 +1723,12 @@ def _png_chunk_should_be_dropped(
     if chunk_type not in (b"tEXt", b"zTXt", b"iTXt"):
         return False
     keyword, _, text = decode(chunk_type.decode("ascii"), payload).partition("\x00")
-    return not _keep_standard_text_metadata(keyword, text, keep_standard)
+    return not _keep_standard_text_metadata(
+        keyword,
+        text,
+        keep_standard,
+        xai_pair_present=xai_pair_present,
+    )
 
 
 # Fallback extension -> PIL save format, used only when the content sniff is
@@ -1771,6 +1934,7 @@ def remove_ai_metadata(
     # Read image and filter metadata
     with Image.open(source_path) as source_image:
         source_exif_bytes = exif_bytes_from_image(source_image)
+        source_has_xai_pair = xai_signature(source_path)
         img = source_image.copy()
         # Pick the save format. Honor the caller's output extension (so a deliberate
         # source.png -> output.jpg conversion still works) UNLESS the SOURCE is misnamed
@@ -1815,7 +1979,12 @@ def remove_ai_metadata(
                 continue
             # NovelAI and TC260 can stamp AI provenance into standard Title/Source/
             # Description fields. Share this decision with the lossless PNG walker.
-            if _keep_standard_text_metadata(key, value, keep_standard):
+            if _keep_standard_text_metadata(
+                key,
+                value,
+                keep_standard,
+                xai_pair_present=source_has_xai_pair,
+            ):
                 kept_meta[key] = str(value) if not isinstance(value, str) else value
 
         if source_exif_bytes:

src/remove_ai_watermarks/metadata_record.py

@@ -281,8 +281,9 @@ def _exif_pairs(info: dict[str, Any]) -> dict[str, str]:
     """The 0th-IFD tags the verdict reads, under their tag NAMES.
 
     Not a convenience: two probes key on names rather than on the raw bytes already
-    in the regions. ``xai_signature_pair`` wants an (ImageDescription, Artist) pair,
-    and ``_external_exif_generator`` looks for Software / Make / Artist /
+    in the regions. The original xAI shape needs an (ImageDescription, Artist) pair
+    (container-native pairs arrive through ``pil`` or structured regions), and
+    ``_external_exif_generator`` looks for Software / Make / Artist /
     ImageDescription. Ship the bytes alone and both silently return nothing, which
     is how a collector can silently lose Grok and NovelAI verdicts.
     """

tests/test_metadata.py

@@ -1021,7 +1021,25 @@ def _grok_imagemagick_png(tmp_path: Path) -> Path:
 
 
 class TestXaiSignature:
-    """xAI / Grok's EXIF Signature + UUID-Artist provenance scheme."""
+    """xAI / Grok's Signature + UUID provenance scheme across metadata containers."""
+
+    @staticmethod
+    def _png_text_pair(tmp_path: Path) -> Path:
+        info = PngInfo()
+        info.add_text("Description", f"Signature: {_FAKE_SIG}")
+        info.add_text("Author", _FAKE_UUID)
+        path = tmp_path / "grok-text.png"
+        Image.new("RGB", (64, 64), (30, 40, 50)).save(path, pnginfo=info)
+        return path
+
+    @staticmethod
+    def _jpeg_app_pair(tmp_path: Path, marker: bytes, payload: bytes, name: str) -> Path:
+        path = tmp_path / name
+        Image.new("RGB", (64, 64), (30, 40, 50)).save(path, "JPEG")
+        data = path.read_bytes()
+        segment = marker + (len(payload) + 2).to_bytes(2, "big") + payload
+        path.write_bytes(data[:2] + segment + data[2:])
+        return path
 
     def test_signature_plus_uuid_detected(self, tmp_path: Path):
         assert xai_signature(_grok_jpeg(tmp_path)) is True
@@ -1036,6 +1054,79 @@ class TestXaiSignature:
         assert xai_signature(path) is True
         assert "xai_signature" in get_ai_metadata(path)
 
+    def test_png_description_and_author_pair_detected_and_stripped(self, tmp_path: Path):
+        path = self._png_text_pair(tmp_path)
+
+        assert xai_signature(path) is True
+        assert "xai_signature" in get_ai_metadata(path)
+
+        out = tmp_path / "clean.png"
+        remove_ai_metadata(path, out)
+
+        assert xai_signature(out) is False
+        with Image.open(out) as image:
+            assert "Description" not in image.info
+            assert "Author" not in image.info
+
+    def test_xmp_description_and_creator_pair_detected_and_stripped(self, tmp_path: Path):
+        payload = (
+            b"http://ns.adobe.com/xap/1.0/\x00"
+            b'<rdf:Description xmlns:dc="http://purl.org/dc/elements/1.1/">'
+            b"<dc:description>Signature: "
+            + _FAKE_SIG.encode()
+            + b"</dc:description><dc:creator>"
+            + _FAKE_UUID.encode()
+            + b"</dc:creator></rdf:Description>"
+        )
+        path = self._jpeg_app_pair(tmp_path, b"\xff\xe1", payload, "grok-xmp.jpg")
+
+        assert xai_signature(path) is True
+        out = tmp_path / "clean-xmp.jpg"
+        remove_ai_metadata(path, out)
+
+        assert xai_signature(out) is False
+        assert _FAKE_SIG.encode() not in out.read_bytes()
+        assert _FAKE_UUID.encode() not in out.read_bytes()
+
+    def test_xmp_signature_does_not_pair_with_an_unrelated_uuid(self, tmp_path: Path):
+        payload = (
+            b"http://ns.adobe.com/xap/1.0/\x00"
+            b'<rdf:Description xmlns:dc="http://purl.org/dc/elements/1.1/" '
+            b'xmp:InstanceID="xmp.iid:'
+            + _FAKE_UUID.encode()
+            + b'"><dc:description>Signature: '
+            + _FAKE_SIG.encode()
+            + b"</dc:description></rdf:Description>"
+        )
+        path = self._jpeg_app_pair(tmp_path, b"\xff\xe1", payload, "unrelated-xmp-uuid.jpg")
+
+        assert xai_signature(path) is False
+
+    def test_xmp_values_in_different_descriptions_do_not_form_a_pair(self, tmp_path: Path):
+        payload = (
+            b"http://ns.adobe.com/xap/1.0/\x00"
+            b"<rdf:Description><dc:description>Signature: "
+            + _FAKE_SIG.encode()
+            + b"</dc:description></rdf:Description>"
+            b"<rdf:Description><dc:creator>" + _FAKE_UUID.encode() + b"</dc:creator></rdf:Description>"
+        )
+        path = self._jpeg_app_pair(tmp_path, b"\xff\xe1", payload, "split-xmp-pair.jpg")
+
+        assert xai_signature(path) is False
+
+    def test_iptc_caption_and_byline_pair_detected_and_stripped(self, tmp_path: Path):
+        payload = b"Photoshop 3.0\x00Caption-Abstract\x00Signature: " + _FAKE_SIG.encode()
+        payload += b"\x00By-line\x00" + _FAKE_UUID.encode()
+        path = self._jpeg_app_pair(tmp_path, b"\xff\xed", payload, "grok-iptc.jpg")
+
+        assert xai_signature(path) is True
+        out = tmp_path / "clean-iptc.jpg"
+        remove_ai_metadata(path, out)
+
+        assert xai_signature(out) is False
+        assert _FAKE_SIG.encode() not in out.read_bytes()
+        assert _FAKE_UUID.encode() not in out.read_bytes()
+
     def test_real_grok_sample_detected(self):
         # Real committed Grok download (data/fixtures/provenance/grok-1.jpg); the EXIF
         # Signature + UUID-Artist pair is the only AI signal it carries.
@@ -1056,6 +1147,18 @@ class TestXaiSignature:
         # Incidental short "Signature: ..." text is below the 64-char base64 bar.
         assert xai_signature(_grok_jpeg(tmp_path, signature="ok")) is False
 
+    def test_png_signature_without_uuid_is_preserved(self, tmp_path: Path):
+        info = PngInfo()
+        info.add_text("Description", f"Signature: {_FAKE_SIG}")
+        path = tmp_path / "signature-only.png"
+        Image.new("RGB", (64, 64)).save(path, pnginfo=info)
+
+        assert xai_signature(path) is False
+        out = tmp_path / "signature-only-clean.png"
+        remove_ai_metadata(path, out)
+        with Image.open(out) as image:
+            assert image.info["Description"] == f"Signature: {_FAKE_SIG}"
+
     def test_clean_image_is_false(self, tmp_clean_png: Path):
         assert xai_signature(tmp_clean_png) is False
 

tests/test_metadata_record.py

@@ -183,6 +183,17 @@ class TestPlacementsTheRecordCouldDrop:
         assert "xai_signature" in {s.name for s in identify(path, check_visible=False, check_invisible=False).signals}
         _assert_same_verdict(path)
 
+    def test_a_container_native_xai_pair_survives(self, tmp_path: Path):
+        """The portable record must preserve xAI pairs moved out of EXIF."""
+        info = PngInfo()
+        info.add_text("Description", "Signature: " + "A" * 80)
+        info.add_text("Author", "3f2504e0-4f89-11d3-9a0c-0305e82c3301")
+        path = tmp_path / "grok-text.png"
+        Image.fromarray(np.zeros((64, 64, 3), dtype=np.uint8)).save(path, "PNG", pnginfo=info)
+
+        assert "xai_signature" in {s.name for s in identify(path, check_visible=False, check_invisible=False).signals}
+        _assert_same_verdict(path)
+
     def test_isobmff_generation_tags_survive(self, tmp_path: Path):
         def box(kind: bytes, payload: bytes) -> bytes:
             return (8 + len(payload)).to_bytes(4, "big") + kind + payload

tests/test_signal_discovery_scripts.py

@@ -2,6 +2,7 @@
 
 from __future__ import annotations
 
+import json
 import sys
 from datetime import date
 from pathlib import Path
@@ -54,6 +55,49 @@ def test_since_filter_uses_corpus_date_directories(tmp_path):
     assert gap._files(tmp_path, date(2026, 9, 10)) == [current]
 
 
+def test_gap_checkpoint_keeps_complete_rows_and_ignores_a_torn_tail(tmp_path):
+    checkpoint = tmp_path / "report.csv.progress.jsonl"
+    row = gap._base_row("2026-09-10/example.png", ".png", "1.2.3")
+    checkpoint.write_text(json.dumps(row) + '\n{"path":"torn', encoding="utf-8")
+
+    assert gap._read_checkpoint(checkpoint) == {row["path"]: row}
+
+    gap._repair_checkpoint(checkpoint)
+    second = gap._base_row("2026-09-10/second.png", ".png", "1.2.3")
+    with checkpoint.open("a", encoding="utf-8") as stream:
+        stream.write(json.dumps(second) + "\n")
+
+    assert gap._read_checkpoint(checkpoint) == {row["path"]: row, second["path"]: second}
+
+
+def test_gap_worker_error_is_a_complete_resumable_row(monkeypatch, tmp_path):
+    path = tmp_path / "bad.bin"
+    path.write_bytes(b"not an image")
+    calls = []
+
+    def fail(_path, **kwargs):
+        calls.append(kwargs)
+        raise ValueError("broken\ninput")
+
+    monkeypatch.setattr(gap, "identify", fail)
+    row = gap._scan_one((str(path), "bad.bin", "1.2.3"))
+
+    assert set(row) == set(gap.REPORT_FIELDS)
+    assert row["candidate_classes"] == "identify_error"
+    assert row["error"] == "ValueError: broken input"
+    assert calls == [{"check_visible": False, "check_invisible": False}]
+
+
+def test_gap_summary_does_not_report_a_negative_hidden_count(capsys):
+    row = gap._base_row("candidate.png", ".png", "1.2.3")
+    row["candidate_classes"] = "blind_marker"
+
+    gap._summarize([row])
+
+    output = capsys.readouterr()
+    assert "more candidate row" not in output.out + output.err
+
+
 def test_sidecar_regression_prefers_stable_signal_names():
     sidecar = {"signals": ["visible_qwen"], "watermarks": ["wording may change"]}
     assert regression.sidecar_families(sidecar) == {"visible_qwen"}

uv.lock

@@ -800,7 +800,7 @@ name = "cuda-bindings"
 version = "13.3.1"
 source = { registry = "https://pypi.org/simple" }
 dependencies = [
-    { name = "cuda-pathfinder", marker = "sys_platform != 'darwin' and sys_platform != 'emscripten' and sys_platform != 'win32'" },
+    { name = "cuda-pathfinder" },
 ]
 wheels = [
     { url = "https://files.pythonhosted.org/packages/51/6b/457ca12dad3ee9bfcc9a545cfd6b64b359ba49de40f776f6e028e678f262/cuda_bindings-13.3.1-cp311-cp311-manylinux_2_24_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:c5879712accf6e14bb01aa5e67440eb84998b8d104b509cc7a6dc0b8f656a474", size = 6053539, upload-time = "2026-05-29T23:11:43.19Z" },
@@ -833,43 +833,43 @@ wheels = [
 
 [package.optional-dependencies]
 cublas = [
-    { name = "nvidia-cublas", marker = "(platform_machine == 'aarch64' and sys_platform == 'linux') or (platform_machine == 'x86_64' and sys_platform == 'linux')" },
-    { name = "nvidia-cuda-nvrtc", marker = "(platform_machine == 'aarch64' and sys_platform == 'linux') or (platform_machine == 'x86_64' and sys_platform == 'linux')" },
+    { name = "nvidia-cublas", marker = "platform_machine == 'aarch64' or platform_machine == 'x86_64'" },
+    { name = "nvidia-cuda-nvrtc", marker = "platform_machine == 'aarch64' or platform_machine == 'x86_64'" },
 ]
 cudart = [
-    { name = "nvidia-cuda-runtime", marker = "(platform_machine == 'aarch64' and sys_platform == 'linux') or (platform_machine == 'x86_64' and sys_platform == 'linux')" },
+    { name = "nvidia-cuda-runtime", marker = "platform_machine == 'aarch64' or platform_machine == 'x86_64'" },
 ]
 cufft = [
-    { name = "nvidia-cufft", marker = "(platform_machine == 'aarch64' and sys_platform == 'linux') or (platform_machine == 'x86_64' and sys_platform == 'linux')" },
-    { name = "nvidia-nvjitlink", marker = "(platform_machine == 'aarch64' and sys_platform == 'linux') or (platform_machine == 'x86_64' and sys_platform == 'linux')" },
+    { name = "nvidia-cufft", marker = "platform_machine == 'aarch64' or platform_machine == 'x86_64'" },
+    { name = "nvidia-nvjitlink", marker = "platform_machine == 'aarch64' or platform_machine == 'x86_64'" },
 ]
 cufile = [
-    { name = "nvidia-cufile", marker = "(platform_machine == 'aarch64' and sys_platform == 'linux') or (platform_machine == 'x86_64' and sys_platform == 'linux')" },
+    { name = "nvidia-cufile", marker = "platform_machine == 'aarch64' or platform_machine == 'x86_64'" },
 ]
 cupti = [
-    { name = "nvidia-cuda-cupti", marker = "(platform_machine == 'aarch64' and sys_platform == 'linux') or (platform_machine == 'x86_64' and sys_platform == 'linux')" },
+    { name = "nvidia-cuda-cupti", marker = "platform_machine == 'aarch64' or platform_machine == 'x86_64'" },
 ]
 curand = [
-    { name = "nvidia-curand", marker = "(platform_machine == 'aarch64' and sys_platform == 'linux') or (platform_machine == 'x86_64' and sys_platform == 'linux')" },
+    { name = "nvidia-curand", marker = "platform_machine == 'aarch64' or platform_machine == 'x86_64'" },
 ]
 cusolver = [
-    { name = "nvidia-cublas", marker = "(platform_machine == 'aarch64' and sys_platform == 'linux') or (platform_machine == 'x86_64' and sys_platform == 'linux')" },
-    { name = "nvidia-cusolver", marker = "(platform_machine == 'aarch64' and sys_platform == 'linux') or (platform_machine == 'x86_64' and sys_platform == 'linux')" },
-    { name = "nvidia-cusparse", marker = "(platform_machine == 'aarch64' and sys_platform == 'linux') or (platform_machine == 'x86_64' and sys_platform == 'linux')" },
-    { name = "nvidia-nvjitlink", marker = "(platform_machine == 'aarch64' and sys_platform == 'linux') or (platform_machine == 'x86_64' and sys_platform == 'linux')" },
+    { name = "nvidia-cublas", marker = "platform_machine == 'aarch64' or platform_machine == 'x86_64'" },
+    { name = "nvidia-cusolver", marker = "platform_machine == 'aarch64' or platform_machine == 'x86_64'" },
+    { name = "nvidia-cusparse", marker = "platform_machine == 'aarch64' or platform_machine == 'x86_64'" },
+    { name = "nvidia-nvjitlink", marker = "platform_machine == 'aarch64' or platform_machine == 'x86_64'" },
 ]
 cusparse = [
-    { name = "nvidia-cusparse", marker = "(platform_machine == 'aarch64' and sys_platform == 'linux') or (platform_machine == 'x86_64' and sys_platform == 'linux')" },
-    { name = "nvidia-nvjitlink", marker = "(platform_machine == 'aarch64' and sys_platform == 'linux') or (platform_machine == 'x86_64' and sys_platform == 'linux')" },
+    { name = "nvidia-cusparse", marker = "platform_machine == 'aarch64' or platform_machine == 'x86_64'" },
+    { name = "nvidia-nvjitlink", marker = "platform_machine == 'aarch64' or platform_machine == 'x86_64'" },
 ]
 nvjitlink = [
-    { name = "nvidia-nvjitlink", marker = "(platform_machine == 'aarch64' and sys_platform == 'linux') or (platform_machine == 'x86_64' and sys_platform == 'linux')" },
+    { name = "nvidia-nvjitlink", marker = "platform_machine == 'aarch64' or platform_machine == 'x86_64'" },
 ]
 nvrtc = [
-    { name = "nvidia-cuda-nvrtc", marker = "(platform_machine == 'aarch64' and sys_platform == 'linux') or (platform_machine == 'x86_64' and sys_platform == 'linux')" },
+    { name = "nvidia-cuda-nvrtc", marker = "platform_machine == 'aarch64' or platform_machine == 'x86_64'" },
 ]
 nvtx = [
-    { name = "nvidia-nvtx", marker = "(platform_machine == 'aarch64' and sys_platform == 'linux') or (platform_machine == 'x86_64' and sys_platform == 'linux')" },
+    { name = "nvidia-nvtx", marker = "platform_machine == 'aarch64' or platform_machine == 'x86_64'" },
 ]
 
 [[package]]
@@ -1066,7 +1066,7 @@ wheels = [
 
 [package.optional-dependencies]
 http = [
-    { name = "aiohttp", marker = "python_full_version < '3.13'" },
+    { name = "aiohttp" },
 ]
 
 [[package]]
@@ -1198,8 +1198,8 @@ name = "httpcore2"
 version = "2.12.0"
 source = { registry = "https://pypi.org/simple" }
 dependencies = [
-    { name = "h11", marker = "sys_platform != 'emscripten'" },
-    { name = "truststore", marker = "sys_platform != 'emscripten'" },
+    { name = "h11" },
+    { name = "truststore" },
 ]
 sdist = { url = "https://files.pythonhosted.org/packages/be/ad/f4f0e57345f1870f3e8cb624e058d7eca6e5a27d33bcc3311d9b618734cd/httpcore2-2.12.0.tar.gz", hash = "sha256:9293522bba0aa7c4c8e9e3f040c16575bd8868e155a77fa30c7a9085a5eae648", size = 67548, upload-time = "2026-08-18T13:22:08.211Z" }
 wheels = [
@@ -1461,15 +1461,15 @@ name = "lightning"
 version = "2.6.6"
 source = { registry = "https://pypi.org/simple" }
 dependencies = [
-    { name = "fsspec", extra = ["http"], marker = "python_full_version < '3.13'" },
-    { name = "lightning-utilities", marker = "python_full_version < '3.13'" },
-    { name = "packaging", marker = "python_full_version < '3.13'" },
-    { name = "pytorch-lightning", marker = "python_full_version < '3.13'" },
-    { name = "pyyaml", marker = "python_full_version < '3.13'" },
-    { name = "torch", marker = "python_full_version < '3.13'" },
-    { name = "torchmetrics", marker = "python_full_version < '3.13'" },
-    { name = "tqdm", marker = "python_full_version < '3.13'" },
-    { name = "typing-extensions", marker = "python_full_version < '3.13'" },
+    { name = "fsspec", extra = ["http"] },
+    { name = "lightning-utilities" },
+    { name = "packaging" },
+    { name = "pytorch-lightning" },
+    { name = "pyyaml" },
+    { name = "torch" },
+    { name = "torchmetrics" },
+    { name = "tqdm" },
+    { name = "typing-extensions" },
 ]
 sdist = { url = "https://files.pythonhosted.org/packages/39/b8/fbf1469d917c3ed3b22e00ab3a852fb9a5f85b9dc7f7060eedc3449098e7/lightning-2.6.6.tar.gz", hash = "sha256:1c6329e241f8a85f29fedead89c1ee138b36caa08de60c260c6a7eaee81c8295", size = 659902, upload-time = "2026-09-10T09:41:03.009Z" }
 wheels = [
@@ -1481,8 +1481,8 @@ name = "lightning-utilities"
 version = "0.15.3"
 source = { registry = "https://pypi.org/simple" }
 dependencies = [
-    { name = "packaging", marker = "python_full_version < '3.13'" },
-    { name = "typing-extensions", marker = "python_full_version < '3.13'" },
+    { name = "packaging" },
+    { name = "typing-extensions" },
 ]
 sdist = { url = "https://files.pythonhosted.org/packages/f1/45/7fa8f56b17dc0f0a41ec70dd307ecd6787254483549843bef4c30ab5adce/lightning_utilities-0.15.3.tar.gz", hash = "sha256:792ae0204c79f6859721ac7f386c237a33b0ed06ba775009cb894e010a842033", size = 33553, upload-time = "2026-02-22T14:48:53.348Z" }
 wheels = [
@@ -1896,7 +1896,7 @@ name = "nvidia-cublas"
 version = "13.1.1.3"
 source = { registry = "https://pypi.org/simple" }
 dependencies = [
-    { name = "nvidia-cuda-nvrtc", marker = "sys_platform != 'darwin' and sys_platform != 'emscripten' and sys_platform != 'win32'" },
+    { name = "nvidia-cuda-nvrtc" },
 ]
 wheels = [
     { url = "https://files.pythonhosted.org/packages/a7/a1/0bd24ee8c8d03adac032fd2909426a00c88f8c57961b1277ded97f91119f/nvidia_cublas-13.1.1.3-py3-none-manylinux_2_27_aarch64.whl", hash = "sha256:b7a210458267ac818974c53038fbec2e969d5c99f305ab15c72522fa9f001dd5", size = 542848918, upload-time = "2026-04-08T18:46:22.985Z" },
@@ -1935,7 +1935,7 @@ name = "nvidia-cudnn-cu13"
 version = "9.24.0.43"
 source = { registry = "https://pypi.org/simple" }
 dependencies = [
-    { name = "nvidia-cublas", marker = "sys_platform != 'darwin' and sys_platform != 'emscripten' and sys_platform != 'win32'" },
+    { name = "nvidia-cublas" },
 ]
 wheels = [
     { url = "https://files.pythonhosted.org/packages/ca/30/7c257e3d5cb4fecb147b93895c66e29c93f8e76d74b45bb418ff0587c4ec/nvidia_cudnn_cu13-9.24.0.43-py3-none-manylinux_2_27_aarch64.whl", hash = "sha256:a6812a554a1ff0413e9c52b84c26c050380649ab9615f9c16bded368ce9f421f", size = 650976863, upload-time = "2026-07-02T16:23:39.248Z" },
@@ -1947,7 +1947,7 @@ name = "nvidia-cufft"
 version = "12.0.0.61"
 source = { registry = "https://pypi.org/simple" }
 dependencies = [
-    { name = "nvidia-nvjitlink", marker = "sys_platform != 'darwin' and sys_platform != 'emscripten' and sys_platform != 'win32'" },
+    { name = "nvidia-nvjitlink" },
 ]
 wheels = [
     { url = "https://files.pythonhosted.org/packages/8b/ae/f417a75c0259e85c1d2f83ca4e960289a5f814ed0cea74d18c353d3e989d/nvidia_cufft-12.0.0.61-py3-none-manylinux2014_aarch64.manylinux_2_17_aarch64.whl", hash = "sha256:2708c852ef8cd89d1d2068bdbece0aa188813a0c934db3779b9b1faa8442e5f5", size = 214053554, upload-time = "2025-09-04T08:31:38.196Z" },
@@ -1977,9 +1977,9 @@ name = "nvidia-cusolver"
 version = "12.0.4.66"
 source = { registry = "https://pypi.org/simple" }
 dependencies = [
-    { name = "nvidia-cublas", marker = "sys_platform != 'darwin' and sys_platform != 'emscripten' and sys_platform != 'win32'" },
-    { name = "nvidia-cusparse", marker = "sys_platform != 'darwin' and sys_platform != 'emscripten' and sys_platform != 'win32'" },
-    { name = "nvidia-nvjitlink", marker = "sys_platform != 'darwin' and sys_platform != 'emscripten' and sys_platform != 'win32'" },
+    { name = "nvidia-cublas" },
+    { name = "nvidia-cusparse" },
+    { name = "nvidia-nvjitlink" },
 ]
 wheels = [
     { url = "https://files.pythonhosted.org/packages/c8/c3/b30c9e935fc01e3da443ec0116ed1b2a009bb867f5324d3f2d7e533e776b/nvidia_cusolver-12.0.4.66-py3-none-manylinux_2_27_aarch64.whl", hash = "sha256:02c2457eaa9e39de20f880f4bd8820e6a1cfb9f9a34f820eb12a155aa5bc92d2", size = 223467760, upload-time = "2025-09-04T08:33:04.222Z" },
@@ -1991,7 +1991,7 @@ name = "nvidia-cusparse"
 version = "12.6.3.3"
 source = { registry = "https://pypi.org/simple" }
 dependencies = [
-    { name = "nvidia-nvjitlink", marker = "sys_platform != 'darwin' and sys_platform != 'emscripten' and sys_platform != 'win32'" },
+    { name = "nvidia-nvjitlink" },
 ]
 wheels = [
     { url = "https://files.pythonhosted.org/packages/f8/94/5c26f33738ae35276672f12615a64bd008ed5be6d1ebcb23579285d960a9/nvidia_cusparse-12.6.3.3-py3-none-manylinux2014_aarch64.manylinux_2_17_aarch64.whl", hash = "sha256:80bcc4662f23f1054ee334a15c72b8940402975e0eab63178fc7e670aa59472c", size = 162155568, upload-time = "2025-09-04T08:33:42.864Z" },
@@ -2146,7 +2146,7 @@ resolution-markers = [
     "(python_full_version < '3.12' and platform_machine != 'aarch64' and sys_platform == 'linux') or (python_full_version < '3.12' and sys_platform != 'darwin' and sys_platform != 'emscripten' and sys_platform != 'linux' and sys_platform != 'win32')",
 ]
 dependencies = [
-    { name = "numpy", version = "1.26.4", source = { registry = "https://pypi.org/simple" }, marker = "python_full_version < '3.13'" },
+    { name = "numpy", version = "1.26.4", source = { registry = "https://pypi.org/simple" } },
 ]
 sdist = { url = "https://files.pythonhosted.org/packages/17/06/68c27a523103dad5837dc5b87e71285280c4f098c60e4fe8a8db6486ab09/opencv-python-4.11.0.86.tar.gz", hash = "sha256:03d60ccae62304860d232272e4a4fda93c39d595780cb40b161b310244b736a4", size = 95171956, upload-time = "2025-01-16T13:52:24.737Z" }
 wheels = [
@@ -2175,7 +2175,7 @@ resolution-markers = [
     "(python_full_version == '3.13.*' and platform_machine != 'aarch64' and sys_platform == 'linux') or (python_full_version == '3.13.*' and sys_platform != 'darwin' and sys_platform != 'emscripten' and sys_platform != 'linux' and sys_platform != 'win32')",
 ]
 dependencies = [
-    { name = "numpy", version = "2.3.5", source = { registry = "https://pypi.org/simple" }, marker = "python_full_version >= '3.13'" },
+    { name = "numpy", version = "2.3.5", source = { registry = "https://pypi.org/simple" } },
 ]
 sdist = { url = "https://files.pythonhosted.org/packages/ed/e8/c2959c0f969ea4ae96b55ad10c108c0410de4e8b3550d9941a4cfdd52703/opencv_python-4.14.0.94.tar.gz", hash = "sha256:afadc76a38a07713a4c2b4702da917cf0ed489be57e2a40a34e670189731de7d", size = 96400225, upload-time = "2026-07-28T19:23:09.546Z" }
 wheels = [
@@ -2206,7 +2206,7 @@ resolution-markers = [
     "(python_full_version < '3.12' and platform_machine != 'aarch64' and sys_platform == 'linux') or (python_full_version < '3.12' and sys_platform != 'darwin' and sys_platform != 'emscripten' and sys_platform != 'linux' and sys_platform != 'win32')",
 ]
 dependencies = [
-    { name = "numpy", version = "1.26.4", source = { registry = "https://pypi.org/simple" }, marker = "python_full_version < '3.13'" },
+    { name = "numpy", version = "1.26.4", source = { registry = "https://pypi.org/simple" } },
 ]
 sdist = { url = "https://files.pythonhosted.org/packages/36/2f/5b2b3ba52c864848885ba988f24b7f105052f68da9ab0e693cc7c25b0b30/opencv-python-headless-4.11.0.86.tar.gz", hash = "sha256:996eb282ca4b43ec6a3972414de0e2331f5d9cda2b41091a49739c19fb843798", size = 95177929, upload-time = "2025-01-16T13:53:40.22Z" }
 wheels = [
@@ -2235,7 +2235,7 @@ resolution-markers = [
     "(python_full_version == '3.13.*' and platform_machine != 'aarch64' and sys_platform == 'linux') or (python_full_version == '3.13.*' and sys_platform != 'darwin' and sys_platform != 'emscripten' and sys_platform != 'linux' and sys_platform != 'win32')",
 ]
 dependencies = [
-    { name = "numpy", version = "2.3.5", source = { registry = "https://pypi.org/simple" }, marker = "python_full_version >= '3.13'" },
+    { name = "numpy", version = "2.3.5", source = { registry = "https://pypi.org/simple" } },
 ]
 sdist = { url = "https://files.pythonhosted.org/packages/67/8a/51ff1c6287be0b6cd3070cf095875da4ff7d6eda38507e6ada5608786a78/opencv_python_headless-4.14.0.94.tar.gz", hash = "sha256:4afa2ea1214453648be88259f035712454faa9039b686de7753569ba8eec1577", size = 96405664, upload-time = "2026-07-28T19:23:48.758Z" }
 wheels = [
@@ -3179,14 +3179,14 @@ name = "pytorch-lightning"
 version = "2.6.6"
 source = { registry = "https://pypi.org/simple" }
 dependencies = [
-    { name = "fsspec", extra = ["http"], marker = "python_full_version < '3.13'" },
-    { name = "lightning-utilities", marker = "python_full_version < '3.13'" },
-    { name = "packaging", marker = "python_full_version < '3.13'" },
-    { name = "pyyaml", marker = "python_full_version < '3.13'" },
-    { name = "torch", marker = "python_full_version < '3.13'" },
-    { name = "torchmetrics", marker = "python_full_version < '3.13'" },
-    { name = "tqdm", marker = "python_full_version < '3.13'" },
-    { name = "typing-extensions", marker = "python_full_version < '3.13'" },
+    { name = "fsspec", extra = ["http"] },
+    { name = "lightning-utilities" },
+    { name = "packaging" },
+    { name = "pyyaml" },
+    { name = "torch" },
+    { name = "torchmetrics" },
+    { name = "tqdm" },
+    { name = "typing-extensions" },
 ]
 sdist = { url = "https://files.pythonhosted.org/packages/97/e4/5025cb42311b4f8cc6d1b12cd81399ca433159f8e3831ab229c5a6c5177e/pytorch_lightning-2.6.6.tar.gz", hash = "sha256:52c52b20522dd3d3feb3bb3aae2dbe3f291e8e836d4111a74cfd72e9df018834", size = 662507, upload-time = "2026-09-10T09:41:16.684Z" }
 wheels = [
@@ -3205,7 +3205,7 @@ resolution-markers = [
     "(python_full_version < '3.12' and platform_machine != 'aarch64' and sys_platform == 'linux') or (python_full_version < '3.12' and sys_platform != 'darwin' and sys_platform != 'emscripten' and sys_platform != 'linux' and sys_platform != 'win32')",
 ]
 dependencies = [
-    { name = "numpy", version = "1.26.4", source = { registry = "https://pypi.org/simple" }, marker = "python_full_version < '3.12'" },
+    { name = "numpy", version = "1.26.4", source = { registry = "https://pypi.org/simple" } },
 ]
 sdist = { url = "https://files.pythonhosted.org/packages/5a/75/50581633d199812205ea8cdd0f6d52f12a624886b74bf1486335b67f01ff/pywavelets-1.9.0.tar.gz", hash = "sha256:148d12203377772bea452a59211d98649c8ee4a05eff019a9021853a36babdc8", size = 3938340, upload-time = "2025-08-04T16:20:04.978Z" }
 wheels = [
@@ -3281,7 +3281,7 @@ resolution-markers = [
     "(python_full_version == '3.12.*' and platform_machine != 'aarch64' and sys_platform == 'linux') or (python_full_version == '3.12.*' and sys_platform != 'darwin' and sys_platform != 'emscripten' and sys_platform != 'linux' and sys_platform != 'win32')",
 ]
 dependencies = [
-    { name = "numpy", version = "1.26.4", source = { registry = "https://pypi.org/simple" }, marker = "python_full_version == '3.12.*'" },
+    { name = "numpy", version = "1.26.4", source = { registry = "https://pypi.org/simple" }, marker = "python_full_version < '3.13'" },
     { name = "numpy", version = "2.3.5", source = { registry = "https://pypi.org/simple" }, marker = "python_full_version >= '3.13'" },
 ]
 sdist = { url = "https://files.pythonhosted.org/packages/25/84/a9e80e33eb114e78bd2f1829a90d72d8d4d660109c721a8aa2f0d58d7592/pywavelets-1.10.0.tar.gz", hash = "sha256:f3cb8640225a8f3fdcbc3f6f3bd112c78927c9407cbfa65a5ff303990a6a3e32", size = 3948163, upload-time = "2026-09-07T17:15:15.637Z" }
@@ -3449,7 +3449,7 @@ wheels = [
 
 [[package]]
 name = "remove-ai-watermarks"
-version = "0.40.1"
+version = "0.40.2"
 source = { editable = "." }
 dependencies = [
     { name = "c2pa-python" },
@@ -4067,10 +4067,10 @@ name = "torchmetrics"
 version = "1.9.0"
 source = { registry = "https://pypi.org/simple" }
 dependencies = [
-    { name = "lightning-utilities", marker = "python_full_version < '3.13'" },
-    { name = "numpy", version = "1.26.4", source = { registry = "https://pypi.org/simple" }, marker = "python_full_version < '3.13'" },
-    { name = "packaging", marker = "python_full_version < '3.13'" },
-    { name = "torch", marker = "python_full_version < '3.13'" },
+    { name = "lightning-utilities" },
+    { name = "numpy", version = "1.26.4", source = { registry = "https://pypi.org/simple" } },
+    { name = "packaging" },
+    { name = "torch" },
 ]
 sdist = { url = "https://files.pythonhosted.org/packages/81/34/39b8b749333db56c0585d7a11fa62a283c087bb1dfc897d69fb8cedbefb1/torchmetrics-1.9.0.tar.gz", hash = "sha256:a488609948600df52d3db4fcdab02e62aab2a85ef34da67037dc3e65b8512faa", size = 581765, upload-time = "2026-03-09T17:41:22.443Z" }
 wheels = [
@@ -4165,13 +4165,13 @@ name = "trustmark"
 version = "0.9.2"
 source = { registry = "https://pypi.org/simple" }
 dependencies = [
-    { name = "einops", marker = "python_full_version < '3.13'" },
-    { name = "lightning", marker = "python_full_version < '3.13'" },
-    { name = "numpy", version = "1.26.4", source = { registry = "https://pypi.org/simple" }, marker = "python_full_version < '3.13'" },
-    { name = "omegaconf", marker = "python_full_version < '3.13'" },
-    { name = "six", marker = "python_full_version < '3.13'" },
-    { name = "torch", marker = "python_full_version < '3.13'" },
-    { name = "torchvision", marker = "python_full_version < '3.13'" },
+    { name = "einops" },
+    { name = "lightning" },
+    { name = "numpy", version = "1.26.4", source = { registry = "https://pypi.org/simple" } },
+    { name = "omegaconf" },
+    { name = "six" },
+    { name = "torch" },
+    { name = "torchvision" },
 ]
 sdist = { url = "https://files.pythonhosted.org/packages/53/41/32e84596af94e8696c55060d55eb9eed9d99722a673d12212a1c0c4ef1d2/trustmark-0.9.2.tar.gz", hash = "sha256:bcaa3bc38ead1c0558df81932672f34e306efaed31389da57dbed100e236b28a", size = 63780, upload-time = "2026-09-08T10:58:11.601Z" }