Durable curation: persisted splits + pre-resolve title edits, catalog A→Z

Wiz-War had no split button: can_split required a matches row, but fresh
extractions leave multi-photo titles rowless until resolve runs. Splits
are now a title-level decision persisted in data/title_splits.json,
honored by extract's dedupe and resolve's dedupe on every rebuild, with
the button on any multi-photo line — resolved or not.

Same mechanism carries human corrections: data/title_edits.json stores
fixed misreads and known cues (publisher/edition/year/language), applied
before dedupe on every titles.json rebuild, editable from a new inline
form on every catalog line. An edit drops the title's stale matches rows
so resolve re-queries with the corrected data.

The catalog page now sorts alphabetically (case-insensitive; split
copies stay adjacent) instead of extraction order.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016jXZFSTZQKzAC8fqpWSz9g

tests/test_extract.py

@@ -12,9 +12,16 @@ import typer
 
 from bggpipe.config import Config
 from bggpipe.extract import (
+    apply_title_edits,
     dedupe_entries,
+    load_title_edits,
+    load_title_splits,
     parse_vision_response,
     prepare_image,
+    rebuild_artifacts,
+    record_title_edit,
+    record_title_split,
+    replay_titles,
     run_extract,
 )
 
@@ -167,6 +174,90 @@ def test_dedupe_conflicting_years_stay_separate():
     assert len(deduped) == 2
 
 
+def test_dedupe_split_titles_never_merge():
+    deduped = dedupe_entries(
+        [_entry("Wiz-War", "a.jpg"), _entry("Wiz-War", "b.jpg")],
+        split_titles={"wiz war"},
+    )
+    assert len(deduped) == 2  # the human said: separate physical copies
+
+
+def test_edits_fix_misreads_before_dedupe():
+    # a corrected misspelling merges with the correctly-read sighting
+    edits = [{"match": "Hebarceos", "title_raw": "Herbaceous"}]
+    deduped = dedupe_entries(
+        apply_title_edits(
+            [_entry("Hebarceos", "a.jpg"), _entry("Herbaceous", "b.jpg")], edits
+        )
+    )
+    assert len(deduped) == 1
+    assert deduped[0]["title_raw"] == "Herbaceous"
+    assert deduped[0]["source_photos"] == ["a.jpg", "b.jpg"]
+
+
+def test_edits_chain_and_target_photos():
+    edits = [
+        {"match": "Wiz-War", "photos": ["a.jpg"], "edition_hint": "7th Edition"},
+        {"match": "Wiz-War", "title_raw": "Wiz-War!", "photos": ["a.jpg"]},
+        # made later, against the renamed title — must chain onto the result
+        {"match": "Wiz-War!", "photos": ["a.jpg"], "year_hint": 1997},
+    ]
+    entries = apply_title_edits(
+        [_entry("Wiz-War", "a.jpg"), _entry("Wiz-War", "b.jpg")], edits
+    )
+    assert entries[0]["title_raw"] == "Wiz-War!"
+    assert entries[0]["edition_hint"] == "7th Edition"
+    assert entries[0]["year_hint"] == 1997
+    assert entries[1] == _entry("Wiz-War", "b.jpg")  # untargeted copy untouched
+
+
+def test_stores_roundtrip_and_replay_from_raw(tmp_path):
+    cfg = Config(data_dir=tmp_path / "data", photos_dir=tmp_path / "photos")
+    raw = cfg.extract_raw_dir
+    raw.mkdir(parents=True)
+    for photo in ("a.jpg", "b.jpg"):
+        (raw / f"{photo}.json").write_text(
+            json.dumps({"titles": [_entry("Wiz-War", photo)], "unidentified": []})
+        )
+    record_title_split(cfg.title_splits_path, "Wiz-War")
+    record_title_split(cfg.title_splits_path, "wiz war")  # dupe, normalized away
+    record_title_edit(
+        cfg.title_edits_path,
+        {"match": "Wiz-War", "photos": ["a.jpg"], "edition_hint": "7th Edition"},
+    )
+    replay_titles(cfg)
+    titles = json.loads(cfg.titles_path.read_text())
+    assert [e["source_photos"] for e in titles] == [["a.jpg"], ["b.jpg"]]
+    assert titles[0]["edition_hint"] == "7th Edition"
+    assert len(load_title_splits(cfg.title_splits_path)) == 1
+    assert len(load_title_edits(cfg.title_edits_path)) == 1
+    # a later full rebuild (a real extract run) honors the same stores
+    rebuild_artifacts(
+        raw,
+        cfg.titles_path,
+        cfg.unidentified_path,
+        load_title_splits(cfg.title_splits_path),
+        load_title_edits(cfg.title_edits_path),
+    )
+    assert len(json.loads(cfg.titles_path.read_text())) == 2
+
+
+def test_replay_without_raw_caches_explodes_merged_entries(tmp_path):
+    cfg = Config(data_dir=tmp_path / "data", photos_dir=tmp_path / "photos")
+    cfg.data_dir.mkdir(parents=True)
+    merged = _entry("Wiz-War", "a.jpg")
+    merged["source_photos"] = ["a.jpg", "b.jpg", "c.jpg"]
+    cfg.titles_path.write_text(json.dumps([merged, _entry("Catan", "a.jpg")]))
+    record_title_split(cfg.title_splits_path, "Wiz-War")
+    replay_titles(cfg)
+    titles = json.loads(cfg.titles_path.read_text())
+    by_title = {}
+    for e in titles:
+        by_title.setdefault(e["title_raw"], []).append(e["source_photos"])
+    assert by_title["Wiz-War"] == [["a.jpg"], ["b.jpg"], ["c.jpg"]]
+    assert by_title["Catan"] == [["a.jpg"]]  # non-split entries survive intact
+
+
 def test_dedupe_upgrades_confidence():
     deduped = dedupe_entries(
         [