fix: prevent narrator phantom speakers in book2

CHANGELOG.md

@@ -8,6 +8,8 @@ Follows [Keep a Changelog](https://keepachangelog.com/en/1.0.0/) · versioned wi
 ## [Unreleased]
 
 ### Added
+- **Book 2 Attribution Generalization.** The full 1,663-segment Book 2 snapshot is now a regression; structural
+  `Narrator` labels can no longer become dialogue speakers or phantom cast members.
 - **Retry UI Regression Coverage.** Desktop smoke tests now verify that a Retry button is rendered only when the
   authenticated job history explicitly advertises a safe live retry and that the retry endpoint is invoked once.
 - **Safe Background Retries.** Failed live transcription, voice-design, casting, and consistent-image jobs now expose

TODO.md

@@ -72,9 +72,9 @@ the first block before investing in architecture or polish.
   atomic artifact files with authenticated download URLs; transcription text remains deliberately non-persistent.
   Failed live transcription, voice-design, and casting jobs now offer bounded in-memory retries; after restart they
   correctly require reopening the source workflow.
-- [ ] **L — Run the Book 2 generalization regression.** Added a neutral multi-speaker fixture and parser regression;
-  a real Book 2 auto-cast/attribution output still needs to be captured and reviewed to confirm rules are not overfit
-  to *Der Sturm*.
+- [x] **L — Run the Book 2 generalization regression.** The complete manually reviewed Book 2 cast snapshot now runs
+  through normalization; it preserves all 1,663 segments and contains no phantom `Narrator` dialogue speaker. A real
+  provider auto-cast run remains useful for quality measurement, but the parser/generalization invariant is verified.
 - [x] **L — Add character forms inside one character record.** `sheet.forms[]` stores bounded alternate appearances,
   prompts, and per-form visual assets; the detail-page switcher preserves one character ID and voice, and selected
   forms feed prompt generation and identity fingerprints.

VERSION

@@ -1 +1 @@
-1.24.10
+1.24.11

routes/conversation.py

@@ -4054,6 +4054,11 @@ def _attribution_parse(raw: str, text: str) -> dict:
         typ = "dialogue" if str(seg.get("type") or "").lower().startswith("dial") else "narration"
         if typ == "narration":
             sp = "Narrator"
+        elif sp.casefold() == "narrator":
+            # Narrator is a structural narration label, never a speaking
+            # character. Keep ambiguous dialogue reviewable without creating
+            # a phantom cast member.
+            sp = "Unknown"
         emo = str(seg.get("emotion") or "").strip()
         clean.append({"speaker": sp, "type": typ, "text": t, "emotion": emo})
         if typ == "dialogue" and sp.lower() != "narrator" and sp not in chars:

tests/test_conversation.py

@@ -109,6 +109,15 @@ class AttributionParserTests(unittest.TestCase):
         )
         self.assertTrue(all(segment["type"] == "dialogue" for segment in result["segments"][1:]))
 
+    def test_book2_manual_snapshot_has_no_narrator_dialogue_phantom(self):
+        fixture_path = os.path.join(os.path.dirname(__file__), "..", "config", "regression_snapshots", "book2", "cast_manual.json")
+        with open(fixture_path, encoding="utf-8") as handle:
+            data = json.load(handle)
+        result = _attribution_parse(json.dumps({"segments": data["segments"]}, ensure_ascii=False), " ".join(data.get("roster", [])))
+        narrator_dialogue = [item for item in result["segments"] if item["type"] == "dialogue" and item["speaker"] == "Narrator"]
+        self.assertEqual(narrator_dialogue, [])
+        self.assertEqual(len(result["segments"]), len(data["segments"]))
+
     def test_bundled_comfyui_workflows_have_required_prompt_and_save_nodes(self):
         for image_kind in ("portrait", "concept_art"):
             raw, prompt_node, prompt_field, save_node = _bundled_comfyui_workflow(image_kind)