Extract caption naturalizer policy

caption_naturalizer.py

@@ -4,54 +4,26 @@ import re
 from typing import Any
 
 try:
+    from . import caption_policy
     from . import formatter_input as input_policy
-    from .hardcore_action_metadata import normalize_hardcore_action_family
     from . import krea_cast as cast_policy
     from .prompt_hygiene import sanitize_prose_text
 except ImportError:  # Allows local smoke tests with `python -c`.
+    import caption_policy
     import formatter_input as input_policy
-    from hardcore_action_metadata import normalize_hardcore_action_family
     import krea_cast as cast_policy
     from prompt_hygiene import sanitize_prose_text
 
 
-OLD_TRIGGER = "sxcpinup_coloredpencil"
+OLD_TRIGGER = caption_policy.OLD_TRIGGER
-DEFAULT_TRIGGER = "sxcppnl7"
+DEFAULT_TRIGGER = caption_policy.DEFAULT_TRIGGER
-
+STYLE_TAILS = caption_policy.STYLE_TAILS
-STYLE_TAILS = [
-    ", coloured pencil comic illustration, crisp linework, hatching, soft pastel palette, warm sensual lighting, textured parchment paper",
-    ", coloured pencil comic illustration, crisp linework, hatching, soft pastel palette, warm sensual lighting, textured paper",
-]
 
 PROMPT_FIELD_LABELS = input_policy.prompt_field_labels()
 
-ITEM_LABELS = (
+ITEM_LABELS = caption_policy.ITEM_LABELS
-    "Sexual pose",
+ACTION_FAMILY_CAPTION_LABELS = caption_policy.ACTION_FAMILY_CAPTION_LABELS
-    "Erotic outfit",
+POSITION_FAMILY_CAPTION_LABELS = caption_policy.POSITION_FAMILY_CAPTION_LABELS
-    "Clothing",
-)
-
-ACTION_FAMILY_CAPTION_LABELS = {
-    "foreplay": "foreplay action",
-    "outercourse": "non-penetrative action",
-    "oral": "oral action",
-    "penetration": "penetrative action",
-    "toy_double": "toy-assisted double-contact action",
-    "climax": "climax action",
-}
-
-POSITION_FAMILY_CAPTION_LABELS = {
-    "penetrative": "penetrative action",
-    "foreplay": "foreplay action",
-    "interaction": "interaction beat",
-    "manual": "manual action",
-    "oral": "oral action",
-    "outercourse": "non-penetrative action",
-    "anal": "anal action",
-    "climax": "climax action",
-    "threesome": "three-person action",
-    "group": "group action",
-}
 
 
 def _clean_text(value: Any) -> str:
@@ -105,13 +77,7 @@ def _human_join(parts: list[str]) -> str:
 
 
 def _metadata_action_label(row: dict[str, Any], default: str = "sexual pose") -> str:
-    position_family = _clean_text(row.get("position_family")).lower()
+    return caption_policy.metadata_action_label(row, default)
-    if position_family in POSITION_FAMILY_CAPTION_LABELS:
-        return POSITION_FAMILY_CAPTION_LABELS[position_family]
-    action_family = normalize_hardcore_action_family(row.get("action_family"))
-    if action_family in ACTION_FAMILY_CAPTION_LABELS:
-        return ACTION_FAMILY_CAPTION_LABELS[action_family]
-    return default
 
 
 def _prompt_cast_descriptors(text: str) -> str:
@@ -135,11 +101,7 @@ def _natural_label_text(text: Any, labels: list[str]) -> str:
 
 
 def _strip_style_tail(text: str) -> str:
-    text = _clean_text(text)
+    return caption_policy.strip_style_tail(text)
-    for tail in STYLE_TAILS:
-        if text.endswith(tail):
-            return text[: -len(tail)].strip(" ,")
-    return text
 
 
 def _remove_trigger(text: str, trigger: str) -> str:
@@ -185,14 +147,11 @@ def _field_from_any_prompt(text: str, labels: tuple[str, ...]) -> str:
 
 
 def _normalize_composition(text: str) -> str:
-    return re.sub(r"^vertical\s+", "", _clean_text(text), flags=re.IGNORECASE)
+    return caption_policy.normalize_composition(text)
 
 
 def _clean_clothing(text: str) -> str:
-    text = _clean_text(text)
+    return caption_policy.clean_clothing(text)
-    text = re.sub(r",?\s*fashion editorial styling$", "", text, flags=re.IGNORECASE)
-    text = re.sub(r",?\s*resort styling$", "", text, flags=re.IGNORECASE)
-    return text.strip(" ,")
 
 
 def _body_phrase(body: Any, figure_note: Any = "") -> str:
@@ -300,10 +259,7 @@ def _verb_for_row(row: dict[str, Any]) -> str:
 
 
 def _detail_allows(level: str, dense_only: bool = False) -> bool:
-    level = (level or "balanced").strip().lower()
+    return caption_policy.detail_allows(level, dense_only=dense_only)
-    if dense_only:
-        return level == "dense"
-    return level != "concise"
 
 
 def _single_from_row(row: dict[str, Any], detail_level: str, keep_style: bool) -> tuple[str, str] | None:
@@ -674,8 +630,8 @@ def naturalize_caption(
 ) -> tuple[str, str]:
     """Rewrite tag-style prompt/caption text into compact natural language."""
     input_hint = input_hint if input_hint in ("auto", "metadata_json", "caption_or_prompt") else "auto"
-    detail_level = detail_level if detail_level in ("concise", "balanced", "dense") else "balanced"
+    detail_level = caption_policy.normalize_detail_level(detail_level)
-    keep_style = style_policy == "keep_style_terms"
+    keep_style = caption_policy.keep_style_terms(style_policy)
     row, row_method = _row_from_inputs(source_text, metadata_json, input_hint)
     if row is not None:
         prose, method = _metadata_to_prose(row, detail_level, keep_style)

caption_policy.py

@@ -0,0 +1,99 @@
+from __future__ import annotations
+
+import re
+from typing import Any
+
+try:
+    from . import formatter_input as input_policy
+    from .hardcore_action_metadata import normalize_hardcore_action_family
+except ImportError:  # Allows local smoke tests with `python tools/prompt_smoke.py`.
+    import formatter_input as input_policy
+    from hardcore_action_metadata import normalize_hardcore_action_family
+
+
+OLD_TRIGGER = "sxcpinup_coloredpencil"
+DEFAULT_TRIGGER = "sxcppnl7"
+
+DETAIL_LEVELS = ("balanced", "concise", "dense")
+STYLE_POLICIES = ("drop_style_tail", "keep_style_terms")
+
+STYLE_TAILS = [
+    ", coloured pencil comic illustration, crisp linework, hatching, soft pastel palette, warm sensual lighting, textured parchment paper",
+    ", coloured pencil comic illustration, crisp linework, hatching, soft pastel palette, warm sensual lighting, textured paper",
+]
+
+ITEM_LABELS = (
+    "Sexual pose",
+    "Erotic outfit",
+    "Clothing",
+)
+
+ACTION_FAMILY_CAPTION_LABELS = {
+    "foreplay": "foreplay action",
+    "outercourse": "non-penetrative action",
+    "oral": "oral action",
+    "penetration": "penetrative action",
+    "toy_double": "toy-assisted double-contact action",
+    "climax": "climax action",
+}
+
+POSITION_FAMILY_CAPTION_LABELS = {
+    "penetrative": "penetrative action",
+    "foreplay": "foreplay action",
+    "interaction": "interaction beat",
+    "manual": "manual action",
+    "oral": "oral action",
+    "outercourse": "non-penetrative action",
+    "anal": "anal action",
+    "climax": "climax action",
+    "threesome": "three-person action",
+    "group": "group action",
+}
+
+
+def normalize_detail_level(value: str) -> str:
+    return value if value in DETAIL_LEVELS else "balanced"
+
+
+def normalize_style_policy(value: str) -> str:
+    return value if value in STYLE_POLICIES else "drop_style_tail"
+
+
+def keep_style_terms(style_policy: str) -> bool:
+    return normalize_style_policy(style_policy) == "keep_style_terms"
+
+
+def detail_allows(level: str, dense_only: bool = False) -> bool:
+    level = normalize_detail_level((level or "balanced").strip().lower())
+    if dense_only:
+        return level == "dense"
+    return level != "concise"
+
+
+def strip_style_tail(text: str) -> str:
+    text = input_policy.clean_text(text)
+    for tail in STYLE_TAILS:
+        if text.endswith(tail):
+            return text[: -len(tail)].strip(" ,")
+    return text
+
+
+def metadata_action_label(row: dict[str, Any], default: str = "sexual pose") -> str:
+    position_family = input_policy.clean_text(row.get("position_family")).lower()
+    if position_family in POSITION_FAMILY_CAPTION_LABELS:
+        return POSITION_FAMILY_CAPTION_LABELS[position_family]
+    action_family = normalize_hardcore_action_family(row.get("action_family"))
+    if action_family in ACTION_FAMILY_CAPTION_LABELS:
+        return ACTION_FAMILY_CAPTION_LABELS[action_family]
+    return default
+
+
+def normalize_composition(text: str) -> str:
+    return re.sub(r"^vertical\s+", "", input_policy.clean_text(text), flags=re.IGNORECASE)
+
+
+def clean_clothing(text: str) -> str:
+    text = input_policy.clean_text(text)
+    text = re.sub(r",?\s*fashion editorial styling$", "", text, flags=re.IGNORECASE)
+    text = re.sub(r",?\s*resort styling$", "", text, flags=re.IGNORECASE)
+    return text.strip(" ,")

docs/prompt-architecture-improvement-plan.md

@@ -302,10 +302,13 @@ Keep here:
 
 - natural sentence caption assembly;
 - training-caption trigger behavior;
-- style-tail policy.
+- style-tail policy from `caption_policy.py`.
-- metadata-family action labels from `action_family` and `position_family`.
+- metadata-family action labels from `action_family` and `position_family` via
+  `caption_policy.py`.
 - shared formatter input parsing from `formatter_input.py`.
 - shared cast descriptor parsing and label replacement from `krea_cast.py`.
+- caption detail-level/style-policy normalization, clothing cleanup, and
+  composition cleanup from `caption_policy.py`.
 
 Improve later:
 

docs/prompt-pool-routing-map.md

@@ -97,6 +97,7 @@ Core helper ownership:
 | `row_normalization.py` | Final prompt-row and pair metadata normalization: trigger prepending, extra-positive append, negative merge/dedupe, caption-part joining, and embedded soft/hard row sanitation. |
 | `formatter_input.py` | Shared formatter input parsing: text cleanup, metadata/source JSON detection, trigger-prefix stripping, shared prompt field-label inventory, `Avoid:` splitting, prompt-field extraction, and metadata row-value fallback. |
 | `sdxl_presets.py` | SDXL style presets, quality presets, default negative prompt, and metadata-family tag hints used by the SDXL formatter and node choice lists. |
+| `caption_policy.py` | Caption naturalizer policy data and helpers: style tails, item labels, metadata-family caption labels, detail/style-policy normalization, clothing cleanup, and composition cleanup. |
 
 ## Node IO Map
 

tools/prompt_smoke.py

@@ -24,6 +24,7 @@ if str(ROOT) not in sys.path:
     sys.path.insert(0, str(ROOT))
 
 import caption_naturalizer  # noqa: E402
+import caption_policy  # noqa: E402
 import character_config  # noqa: E402
 import character_profile  # noqa: E402
 import category_cast_config  # noqa: E402
@@ -951,6 +952,47 @@ def smoke_formatter_cast_policy() -> None:
     )
 
 
+def smoke_caption_policy() -> None:
+    _expect(
+        caption_naturalizer.STYLE_TAILS is caption_policy.STYLE_TAILS,
+        "Caption naturalizer style tails should delegate to caption_policy",
+    )
+    _expect(
+        caption_naturalizer.ITEM_LABELS is caption_policy.ITEM_LABELS,
+        "Caption naturalizer item labels should delegate to caption_policy",
+    )
+    _expect(
+        caption_naturalizer.ACTION_FAMILY_CAPTION_LABELS is caption_policy.ACTION_FAMILY_CAPTION_LABELS,
+        "Caption naturalizer action labels should delegate to caption_policy",
+    )
+    _expect(caption_policy.normalize_detail_level("bad") == "balanced", "Caption invalid detail fallback changed")
+    _expect(caption_policy.keep_style_terms("keep_style_terms") is True, "Caption style policy keep flag changed")
+    _expect(caption_policy.detail_allows("concise") is False, "Caption concise detail gate changed")
+    _expect(caption_policy.detail_allows("dense", dense_only=True) is True, "Caption dense-only gate changed")
+
+    style_tail = caption_policy.STYLE_TAILS[0]
+    _expect(
+        caption_policy.strip_style_tail(f"caption body{style_tail}") == "caption body",
+        "Caption style-tail stripping changed",
+    )
+    _expect(
+        caption_naturalizer._strip_style_tail(f"caption body{style_tail}") == "caption body",
+        "Caption naturalizer style-tail wrapper should delegate",
+    )
+    _expect(
+        caption_policy.normalize_composition("vertical centered body frame") == "centered body frame",
+        "Caption composition normalization changed",
+    )
+    _expect(
+        caption_policy.clean_clothing("silk dress, fashion editorial styling") == "silk dress",
+        "Caption clothing cleanup changed",
+    )
+    row = {"action_family": "oral", "position_family": ""}
+    _expect(caption_policy.metadata_action_label(row) == "oral action", "Caption action-family label changed")
+    row = {"action_family": "oral", "position_family": "anal"}
+    _expect(caption_naturalizer._metadata_action_label(row) == "anal action", "Caption position-family label priority changed")
+
+
 def smoke_sdxl_presets_policy() -> None:
     _expect(
         sdxl_formatter.SDXL_STYLE_PRESETS is sdxl_presets.SDXL_STYLE_PRESETS,
@@ -2964,6 +3006,7 @@ SMOKE_CASES: list[tuple[str, Callable[[], None]]] = [
     ("row_normalization_policy", smoke_row_normalization_policy),
     ("formatter_input_policy", smoke_formatter_input_policy),
     ("formatter_cast_policy", smoke_formatter_cast_policy),
+    ("caption_policy", smoke_caption_policy),
     ("sdxl_presets_policy", smoke_sdxl_presets_policy),
     ("hardcore_position_config_policy", smoke_hardcore_position_config_policy),
     ("category_library_route", smoke_category_library_route),