fix(tags): Title-Case operator-entered tags at create endpoint only (plan #701)

normalize_tag_name (per-word capitalize + whitespace collapse) is applied in
the POST /api/tags handler so operator-entered tags get clean display casing.
It is NOT applied in the shared find_or_create / rename paths — those are used
by the ML tagger and allowlist matching, which must preserve the booru
vocabulary's original casing (Title-Casing it broke apply_allowlist matching).

find_or_create / rename keep case-insensitive lookup + clash detection so a
differently-cased entry dedups onto the existing tag instead of forking.

Tests updated to expect Title-Cased create output (sunset→Sunset,
character:Saber→Character:saber, http://example.com→Http://example.com) and a
dedicated normalize_tag_name unit test.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

backend/app/api/tags.py

@@ -14,6 +14,7 @@ from ..services.tag_service import (
     TagMergeConflict,
     TagService,
     TagValidationError,
+    normalize_tag_name,
 )
 from ..utils.tag_prefix import parse_kind_prefix
 
@@ -141,6 +142,11 @@ async def create_tag():
 
     fandom_id = body.get("fandom_id")
 
+    # #701: Title-Case operator-entered tags. Only here (the explicit create
+    # endpoint), NOT in the shared find_or_create — the ML tagger uses that path
+    # and must keep the booru vocabulary's casing for allowlist matching.
+    name = normalize_tag_name(name)
+
     async with get_session() as session:
         svc = TagService(session)
         try:

backend/app/services/tag_service.py

@@ -84,7 +84,11 @@ class TagService:
         - if fandom_id is set, the referenced tag must exist and have
           kind == TagKind.fandom
         """
-        name = normalize_tag_name(name)
+        # NOTE: case is NOT normalized here — find_or_create is the shared path
+        # the ML tagger / allowlist also use, and Title-Casing the booru
+        # vocabulary breaks allowlist matching. Display-casing (Title Case) is
+        # applied at the user-entry layer (api/tags create_tag) only (#701).
+        name = name.strip()
         if not name:
             raise TagValidationError("Tag name cannot be empty")
 
@@ -259,14 +263,14 @@ class TagService:
         with an existing tag of the same (kind, fandom_id) — the API turns
         that into a 409 merge hint.
         """
-        new_name = normalize_tag_name(new_name)
+        new_name = new_name.strip()
         if not new_name:
             raise TagValidationError("Tag name cannot be empty")
         tag = await self.session.get(Tag, tag_id)
         if tag is None:
             raise TagValidationError(f"Tag {tag_id} not found")
 
-        # Case-insensitive clash (#701) — colliding with a differently-cased tag
+        # Case-insensitive clash (#701) — renaming onto a differently-cased tag
         # is still a merge, not a silent fork.
         clash_stmt = (
             select(Tag)

tests/test_api_tags.py

@@ -52,11 +52,12 @@ async def test_create_tag_missing_name_400(client):
 
 @pytest.mark.asyncio
 async def test_create_tag_name_only_defaults_to_general(client):
-    """IR-style: name without kind and without `kind:` prefix → general."""
+    """IR-style: name without kind and without `kind:` prefix → general.
+    #701: operator-entered names are Title-Cased at the create endpoint."""
     resp = await client.post("/api/tags", json={"name": "sunset"})
     assert resp.status_code == 201
     body = await resp.get_json()
-    assert body["name"] == "sunset"
+    assert body["name"] == "Sunset"
     assert body["kind"] == "general"
 
 
@@ -73,21 +74,22 @@ async def test_create_tag_with_kind_prefix(client):
 @pytest.mark.asyncio
 async def test_explicit_kind_overrides_prefix_parsing(client):
     """If caller passes explicit kind, don't re-parse the name —
-    colon and prefix stay literal."""
+    colon and prefix stay literal. #701: still Title-Cased as a single word."""
     resp = await client.post(
         "/api/tags", json={"name": "character:Saber", "kind": "general"}
     )
     assert resp.status_code == 201
     body = await resp.get_json()
-    assert body["name"] == "character:Saber"
+    assert body["name"] == "Character:saber"
     assert body["kind"] == "general"
 
 
 @pytest.mark.asyncio
 async def test_unknown_prefix_kept_literal(client):
-    """`http:example` — `http` not in KNOWN_KINDS → kind=general, literal name."""
+    """`http:example` — `http` not in KNOWN_KINDS → kind=general, literal name.
+    #701: Title-Cased as one word (no internal whitespace to split on)."""
     resp = await client.post("/api/tags", json={"name": "http://example.com"})
     assert resp.status_code == 201
     body = await resp.get_json()
-    assert body["name"] == "http://example.com"
+    assert body["name"] == "Http://example.com"
     assert body["kind"] == "general"

tests/test_tag_service.py

@@ -60,16 +60,24 @@ async def test_character_with_fandom(db):
 
 
 @pytest.mark.asyncio
-async def test_find_or_create_titlecases_and_dedups_case_insensitive(db):
-    """#701: tags normalize to Title Case + collapsed whitespace on create, and a
-    differently-cased entry finds the existing tag instead of forking."""
+async def test_find_or_create_dedups_case_insensitive(db):
+    """#701: find_or_create matches case-insensitively so a differently-cased
+    entry finds the existing tag instead of forking. (Display Title-Casing is
+    applied at the create API, not here — this path is shared with the ML
+    tagger, which keeps the booru vocabulary's casing.)"""
     svc = TagService(db)
     t1 = await svc.find_or_create("hatsune miku", TagKind.character)
-    assert t1.name == "Hatsune Miku"
-    t2 = await svc.find_or_create("HATSUNE   miku", TagKind.character)
-    assert t2.id == t1.id  # case + whitespace variant → same tag, no fork
-    g = await svc.find_or_create("  looking   at  viewer ", TagKind.general)
-    assert g.name == "Looking At Viewer"
+    assert t1.name == "hatsune miku"  # stored as given; not recased here
+    t2 = await svc.find_or_create("HATSUNE MIKU", TagKind.character)
+    assert t2.id == t1.id  # case variant → same tag, no fork
+
+
+@pytest.mark.asyncio
+async def test_normalize_tag_name():
+    from backend.app.services.tag_service import normalize_tag_name
+    assert normalize_tag_name("hatsune miku") == "Hatsune Miku"
+    assert normalize_tag_name("  looking   at  viewer ") == "Looking At Viewer"
+    assert normalize_tag_name("HATSUNE MIKU") == "Hatsune Miku"
 
 
 @pytest.mark.asyncio