feat: audit catalog media coverage

tests/test_media_assets.py

@@ -25,11 +25,21 @@ def test_does_not_mistake_seafishing_screenshot_for_fish_entity() -> None:
     assert items[0].entity_type == "reference"
 
 
+def test_gallery_page_provenance_classifies_generic_asset_urls() -> None:
+    fish = extract_media_candidates('<img src="/uploads/gallery/123.jpg" alt="Hecht">', source_page="https://rf4game.de/media/fische/")
+    water = extract_media_candidates('<img src="/uploads/gallery/456.jpg" alt="Kuori">', source_page="https://rf4game.de/media/levels/")
+    assert fish[0].entity_type == "fish"
+    assert water[0].entity_type == "waterbody"
+    chrome = extract_media_candidates('<img src="/wp-content/themes/rf4/img/banner_de.png">', source_page="https://rf4game.de/media/fische/")
+    assert chrome[0].entity_type == "reference"
+
+
 def test_manifest_merges_and_binary_store_is_content_addressed(tmp_path: Path) -> None:
     item = extract_media_candidates('<img src="/bait/worm.png" alt="Bait worm">', source_page="https://example.test")[0]
     first = merge_manifest(tmp_path / "manifest.json", [item])
     second = merge_manifest(tmp_path / "manifest.json", [item])
     assert len(first["assets"]) == len(second["assets"]) == 1
+    assert second["assets"][0]["source_pages"] == ["https://example.test"]
     image = io.BytesIO()
     Image.new("RGB", (3, 2), "green").save(image, format="PNG")
     digest, relative, width, height, mime = store_asset(tmp_path, image.getvalue(), content_type="image/png", source_url=item.asset_url)