ComfyUI-Ethanfel-Prompt-Builder/docs/prompt-pool-routing-map.md

Prompt Pool Routing Map

This document maps the roads that can lead to prompt text. It intentionally does not duplicate every prompt phrase in the JSON pools. The useful debugging target is usually: which node path selected which pool, which seed axis controlled it, and whether the final wording was still raw builder text or rewritten by a formatter.

Mental Model

There are three layers:

1. Generation layer: selects category, subcategory, item/action, character descriptors, scene, expression, pose, composition, and camera directives.
2. Pair/adapter layer: optionally builds a softcore/hardcore pair, applies continuity, POV handling, camera-aware scene text, and clothing state.
3. Formatter layer: rewrites a metadata row into Krea2 prose, SDXL tags, or a naturalized caption.

When a result is wrong, first identify which layer owns the bad text:

• Raw builder prompt already wrong: edit prompt_builder.py or the relevant categories/*.json pool/template.
• Raw builder prompt acceptable, Krea2 output wrong: edit krea_formatter.py.
• Raw builder prompt acceptable, SDXL tags wrong: edit sdxl_formatter.py.
• Natural caption/training caption wrong: edit caption_naturalizer.py.
• UI/preview/loop behavior wrong: edit __init__.py, loop_nodes.py, or web/*.js.

High-Level Routes

flowchart TD
  A[SxCP Prompt Builder] --> B[prompt_builder.build_prompt]
  C[SxCP Prompt Builder From Configs] --> D[parse config nodes]
  D --> B
  E[SxCP Insta/OF Prompt Pair] --> F[prompt_builder.build_insta_of_pair]
  F --> B
  B --> R[metadata row + prompt + negative + caption]
  F --> P[pair metadata: soft row + hard row]
  R --> K[SxCP Krea2 Formatter]
  R --> S[SxCP SDXL Formatter]
  R --> N[SxCP Caption Naturalizer]
  P --> K
  P --> S
  P --> N

The config nodes mostly emit JSON. The final builder nodes parse that JSON and call the same core generation functions.

Main Entry Points

ComfyUI node	Python entry	What it owns
SxCP Prompt Builder	build_prompt	Direct single prompt generation. Can use built-in categories or JSON categories.
SxCP Prompt Builder From Configs	build_prompt_from_configs -> build_prompt	Same generator, but inputs come from category/cast/profile/filter helper nodes.
SxCP Insta/OF Prompt Pair	build_insta_of_pair	Builds a softcore row and hardcore row with shared cast/continuity options.
SxCP Krea2 Formatter	format_krea2_prompt	Converts metadata rows or pair metadata into Krea2-friendly prose.
SxCP SDXL Formatter	format_sdxl_prompt	Converts metadata rows or pair metadata into SDXL/tag style prompts.
SxCP Caption Naturalizer	naturalize_caption	Converts rows into more natural sentence captions.

Node IO Map

Use this when wiring or debugging a workflow. If the formatter can receive metadata_json, prefer wiring metadata instead of only prompt text. Metadata is what keeps cast, role graph, POV labels, camera config, and soft/hard pair state recoverable.

Node	Important inputs	Important outputs
SxCP Prompt Builder	category, subcategory, seed, optional config nodes	prompt, negative_prompt, caption, metadata_json, category, subcategory
SxCP Prompt Builder From Configs	category/cast/profile/filter/config node outputs	Same as SxCP Prompt Builder
SxCP Insta/OF Prompt Pair	options, seed_config, character_cast, location/composition/camera, hardcore_position_config	softcore_prompt, hardcore_prompt, both negatives, both captions, shared_descriptor, metadata_json
SxCP Krea2 Formatter	source_text, optional metadata_json, target	krea_prompt, both pair prompts if pair metadata exists, negative outputs, method
SxCP SDXL Formatter	source_text, optional metadata_json, target, style/quality preset	sdxl_prompt, both pair prompts if pair metadata exists, negative outputs, method
SxCP Caption Naturalizer	source_text, optional metadata_json	natural_caption, method

Practical Recipes

These recipes identify the intended road before editing prompt text.

Request	Preferred node route	Critical settings	If wrong, inspect
Keep character/location but change only sexual pose	Global Seed or fixed seed config -> builder/pair	Keep person_seed and scene_seed fixed; change pose_seed and usually role_seed; for hardcore categories check content_seed_axis	sexual_poses.json, hardcore_position_config, krea_actions.hardcore_action_sentence
Generate a specific hardcore oral/blowjob scene	Hardcore Position Pool -> Hardcore Action Filter -> Insta/OF Prompt Pair or Prompt Builder	Use focus=oral_only or disable non-oral families; keep allow_oral=true; constrain position pool to kneeling/standing/oral variants when needed	sexual_poses.json oral subcategory/templates, _apply_hardcore_position_config_to_subcategory, krea_actions.hardcore_action_sentence
Generate POV oral or POV penetration	Man Slot with POV presence -> character_cast -> pair/builder -> Krea2 formatter	POV man must be in the cast; use metadata into Krea2; normal camera directive is suppressed by POV	krea_pov_actions.py, krea_pov.py, krea_cast.cast_prose omit-label handling
Generate porn-scene interaction beats	Hardcore Position Pool -> Hardcore Action Filter -> pair/builder	Use focus=interaction_only for kissing/body worship/transitions/guidance/camera/watching/aftercare, or focus=manual_only for fingering/clit/manual stimulation; constrain keys such as camera_showing, wrist_pinning, fingering, aftercare	sexual_poses.json interaction/manual subcategories, _role_graph, krea_action_context.is_foreplay_text / krea_actions.hardcore_action_sentence
Same woman, same room, softcore and hardcore outputs	Character Slot/Profile -> Insta/OF Options -> Insta/OF Prompt Pair	continuity=same_creator_same_room; set softcore_cast as needed; use pair metadata into formatter	build_insta_of_pair, softcore_row, hardcore_row, pair metadata fields
Same cast in softcore and hardcore	Character slot chain -> Insta/OF Options	softcore_cast=same_as_hardcore; configure partner slots/outfits if needed	_insta_of_partner_styling, character slot clothing, pair Krea branch
Change only outfit/clothing	Character clothing or category content route	Keep person_seed, scene_seed, pose_seed; change content_seed; slot softcore_outfit overrides Insta/OF outfit	SxCP Character Clothing, INSTA_OF_SOFTCORE_OUTFITS, category item templates
Force a custom location	SxCP Location Pool or SxCP Location Theme -> builder/pair	combine_mode=replace to force; add to mix with category scenes	_scene_pool, _apply_location_config_to_legacy_row, camera scene adapter
Force a custom frame/composition	SxCP Composition Pool or SxCP Location Theme -> builder/pair	combine_mode=replace to force; add to mix	_composition_pool, _apply_composition_config_to_legacy_row, Krea composition phrase
Use Qwen/orbit camera geometry	Qwen/orbit node -> camera_config -> builder/pair	For pair, use softcore_camera_config and/or hardcore_camera_config; set mode from config in options	_camera_config_with_mode, _camera_directive, _camera_scene_directive_for_context
Use Krea2 for only hard prompt from a pair	Pair metadata_json -> Krea2 Formatter	target=hardcore, input_hint=metadata_json or auto with metadata connected	_insta_pair_to_krea, hard row fields
Convert builder output to SDXL tags	Builder/pair metadata -> SDXL Formatter	Use metadata input; set target; select style and quality preset	_row_core_tags, _soft_tags, _hard_tags
Save/reuse character	Slot/profile nodes -> Profile Save/Load -> slot/builder	Save from the row/profile data you want, not a freshly randomized disconnected route	profile helpers, web/profile_buttons.js, profile JSON

Seed Axes

Seed routing is centralized around SEED_AXIS_SALTS, SEED_AXIS_ALIASES, and _axis_rng in prompt_builder.py.

Axis	Controls
category	Main category selection when random/auto.
subcategory	Subcategory selection.
content	Clothing item, category item, generated outfit, many softcore outfit choices.
person	Character appearance when not fully manual.
scene	Location/scene choice.
pose	Generic pose choice, and for pose-content categories, the hardcore position/action item.
role	Role graph / action choreography for sexual pose categories. Falls back through pose aliases.
expression	Expression choice and expression intensity randomization.
composition	Composition/framing choice.

SxCP Global Seed, SxCP Seed Control, and SxCP Seed Locker all feed seed_config. Values below zero mean the row's main seed still drives that axis. Fixed axis seeds allow changing only one road, for example changing pose/role while keeping person, scene, and category stable.

Seed Playbook

The seed system has two levels: the main row seed and optional per-axis seeds. If an axis seed is negative or absent, the main row seed plus row number drives that axis. If an axis seed is fixed, that axis is reproducible even while other axes change.

Goal	Seed setup
Exact full regeneration	Keep main seed, row_number, start_index, and every connected config identical.
Same person, new pose	Fix person_seed; change pose_seed and usually role_seed. For hardcore pose categories, changing content_seed may also matter if the selected category uses content for pose items.
Same scene, new character	Fix scene_seed; change person_seed.
Same action, new framing	Fix pose_seed, role_seed, and content_seed; change composition_seed and/or camera config.
Same outfit, new pose	Fix content_seed; change pose_seed/role_seed.
Same soft/hard pair but different hardcore action	In pair mode, keep person_seed, scene_seed, content_seed if clothing must stay; change pose_seed/role_seed.
Debug expression only	Fix everything except expression_seed or expression intensity.

Common trap: row_number participates in _axis_rng. If two workflows have the same seeds but different row_number, they are not expected to match.

Category Sources

There are two category systems.

Source	Files/functions	Notes
Built-in legacy generator	generate_prompt_batches.py, _build_direct_builtin_row, _build_auto_weighted_row	Handles legacy woman, man, couple, group_or_layout, auto_weighted, and auto_full.
JSON category library	categories/*.json, load_category_library, _build_custom_row	Handles expandable categories such as casual clothes, erotic clothes, and hardcore sexual poses.

JSON categories are the scalable system. Add new main categories or subcategories there unless the behavior needs Python logic.

JSON Category Road

flowchart TD
  A[category/subcategory input] --> B[_find_subcategory]
  B --> C[item from subcategory.items]
  C --> D[_compose_item]
  D --> E[item_templates + axes]
  B --> F[_scene_pool]
  B --> G[_expression_pool]
  B --> H[_pose_pool]
  B --> I[_composition_pool]
  E --> J[role graph / item text / axis values]
  F --> R[scene_text]
  G --> R
  H --> R
  I --> R
  J --> R[metadata row]

Important JSON keys:

• categories: main category definitions.
• subcategories: selectable subcategories inside a category.
• items: item/action entries selected by the content or pose axis.
• item_templates: templates with axis placeholders.
• axes: values used to fill item_templates.
• scene_pool / scene_pools or direct scenes: location road.
• expression_pool / expression_pools or direct expressions: expression road.
• composition_pool / composition_pools or direct compositions: framing road.
• poses: category-specific pose fallback.
• prompt_template / caption_template: final prompt assembly for that category.
• inherit_scenes, inherit_expressions, inherit_compositions: stop or allow inheritance from category/subcategory/item levels.
• pool_extensions: patch legacy pools from JSON.

Current category/pool files:

File	Main ownership
categories/default_categories.json	Casual clothes, men/couple casual variants, normal JSON categories.
categories/erotic_clothes.json	Provocative/erotic clothing categories and their scene/expression/composition pools.
categories/sexual_poses.json	Hardcore sexual pose/action categories, role graphs, explicit scene/expression/composition pools.
categories/location_pools.json	Named scene pools and location pool extensions.
categories/expression_composition_pools.json	Named expression pools and composition pools.

Pool Ownership Matrix

This table is the first stop when the selected content is wrong.

File / pool area	Owns	Selection axis	Formatter risk
default_categories.json woman casual subcategories	Casual outfit items, casual scenes, casual expressions, casual compositions	category, subcategory, content, scene, expression, composition	Low unless Krea/SDXL needs shorter clothing tags
default_categories.json men casual subcategories	Male casual outfit/items and men-specific casual pools	Same as above	Medium if men are part of a mixed cast and clothing detail is too strong
default_categories.json couple casual subcategories	Couple outfit/action-ish soft poses and couple pools	Same as above	Medium because labels and partner styling can duplicate pair mode
erotic_clothes.json	Provocative/erotic clothing categories and softcore creator scenes	content, scene, expression, composition	Medium because nude/implied-nude wording can conflict with clothes
sexual_poses.json foreplay/interaction/manual/oral/outercourse/penetration/etc.	Hardcore action and porn-scene interaction templates, role graphs, axis values, hardcore pool references	pose for pose-content route, also role; sometimes content aliases matter	High because Krea2 rewrites action and POV position text
location_pools.json	Reusable scene pools and legacy scene extensions	scene	Medium when a camera-aware adapter changes scene/composition wording
expression_composition_pools.json	Reusable expressions and framing/composition pools	expression, composition	Medium because formatter may label or suppress expressions
generate_prompt_batches.py legacy pools	Built-in generator clothing, pose, expression, scene, composition lists	Main row seed plus axis config through legacy adapter	Medium because legacy prompt format is field-label heavy

When adding a new pool, choose JSON when the change is pure selectable wording. Choose Python only when selection logic, compatibility filters, camera adaptation, profile behavior, or formatter rewriting must change.

Pool Resolution

Scene / Location

Scene text is selected by _scene_pool.

Resolution order:

1. SxCP Location Pool / SxCP Location Theme with replace overrides the category scene pool.
2. Category/subcategory/item direct scenes and referenced scene pools are merged, unless inheritance is disabled.
3. SxCP Location Pool with add appends its entries.
4. Fallback is legacy g.SCENES or g.GROUP_SCENES.

Edit targets:

• Add reusable named locations: categories/location_pools.json.
• Add category-specific locations: the category JSON file.
• Add quick workflow-only locations: SxCP Location Pool custom locations.
• Add themed location packs: THEMATIC_LOCATION_PRESETS in prompt_builder.py.

Expression

Expression text is selected by _expression_pool, then filtered by _expression_entries_for_intensity.

Resolution order:

1. Category/subcategory/item direct expressions and named expression pools.
2. Inheritance can stop at item or subcategory level.
3. Fallback is legacy g.EXPRESSIONS.
4. Character slots can override intensity per softcore/hardcore phase.
5. Expression can be disabled globally or per character slot.

Edit targets:

• General expression pools: categories/expression_composition_pools.json.
• Hardcore-specific expressions: usually categories/sexual_poses.json or named hardcore expression pools.
• Character-level expression settings: slot config and _cast_expression_intensity_override.
• Formatter expression wording: krea_formatter.py or caption_naturalizer.py.

Pose / Action

Generic pose text is selected by _pose_pool. Hardcore sexual pose categories are different: the sexual position/action is an item/template selected by the content route, with content_seed_axis set to pose for pose-content categories.

Edit targets:

• Normal pose pools: legacy g.POSES, g.EVOCATIVE_POSES, or JSON poses.
• Hardcore positions/actions: categories/sexual_poses.json.
• Hardcore interaction beats: categories/sexual_poses.json subcategories foreplay_teasing, manual_stimulation, body_worship_touching, clothing_position_transitions, dominant_guidance, camera_performance, group_coordination, and aftercare_cleanup.
• Position filtering UI: build_hardcore_position_pool_json, build_hardcore_action_filter_json, _apply_hardcore_position_config_to_subcategory.
• Krea2 action rewrite orchestration: krea_formatter.py.
• Krea2 non-POV position anchors/arrangements: krea_action_positions.py.
• Krea2 non-climax item/detail cleanup: krea_action_details.py.
• Krea2 climax role/detail cleanup: krea_action_climax.py.
• Krea2 non-POV action-family routing: krea_action_dispatch.py.
• Krea2 non-POV action sentence assembly: krea_actions.py.
• Krea2 POV position rewrite: krea_pov_actions.py.

Composition

Composition text is selected by _composition_pool.

Resolution order:

1. SxCP Composition Pool / SxCP Location Theme with replace overrides the category composition pool.
2. Category/subcategory/item direct compositions and named composition pools are merged.
3. SxCP Composition Pool with add appends entries.
4. Fallback is legacy g.COMPOSITIONS or g.GROUP_COMPOSITIONS.

Edit targets:

• General composition pools: categories/expression_composition_pools.json.
• Category-specific composition pools: relevant category JSON.
• Workflow overrides: SxCP Composition Pool.
• Camera-aware composition replacements: _coworking_composition_prompt.

Character Route

flowchart TD
  A[Woman/Man/Character Slot chain] --> B[character_cast JSON]
  C[Profile Load] --> D[character_profile JSON]
  E[Manual Details / Hair / Body / Age / Eyes / Clothing nodes] --> A
  B --> F[_parse_character_cast]
  F --> G[_character_slot_label_map]
  G --> H[_context_from_character_slot]
  H --> R[cast descriptors + slot clothing + expression settings]
  D --> I[_apply_character_profile_to_context]
  I --> R

Important behavior:

• Slot chaining labels the closest slot to the final node as A for that gender, then upstream slots as B, C, and so on.
• character_cast is for multi-character configured casts. character_profile is for reusable primary-character details.
• SxCP Character Manual Details and characteristic pool nodes can feed slots.
• SxCP Character Clothing can feed softcore_outfit and hardcore_clothing.
• A man slot can be marked as POV. POV men are omitted from visible cast descriptors, and camera wording is adapted to first-person view.

Edit targets:

• Appearance field generation: _context_from_character_slot, _character_context_for_label, _cast_descriptor_entries.
• Profile save/load: SxCPCharacterProfileSave, SxCPCharacterProfileLoad, profile helpers in prompt_builder.py, and web/profile_buttons.js.
• Hair/body/ethnicity list behavior: characteristic config builders in prompt_builder.py.

Insta/OF Pair Route

flowchart TD
  O[SxCP Insta/OF Options] --> P[build_insta_of_pair]
  C[character_cast] --> P
  S[seed_config] --> P
  L[location_config] --> P
  M[composition_config] --> P
  H[hardcore_position_config] --> P
  P --> A[soft_row via build_prompt]
  P --> B[hard_row via build_prompt]
  A --> X[pair metadata]
  B --> X
  X --> K[Krea2/SDXL/Naturalizer]

Softcore row:

• Category comes from INSTA_OF_SOFTCORE_SUBCATEGORY_BY_LEVEL.
• Outfit comes from character slot softcore_outfit if present, otherwise INSTA_OF_SOFTCORE_OUTFITS.
• Soft pose comes from INSTA_OF_SOFTCORE_POSES.
• Partner styling comes from _insta_of_partner_styling when softcore cast is same_as_hardcore.

Hardcore row:

• Category is always Hardcore sexual poses.
• Cast count comes from SxCP Insta/OF Options.
• Position/action/interaction can be constrained by SxCP Hardcore Position Pool and SxCP Hardcore Action Filter.
• Clothing comes from character slot hardcore clothing first, then fallback hardcore_clothing_continuity.
• Men receive default hardcore clothing if visible and not configured.
• POV labels alter action, camera, composition, and visible cast descriptors.

Continuity:

• same_creator_same_room reuses the softcore scene for hardcore.
• same_creator_new_scene lets hardcore use its own scene.
• Shared cast descriptors are stored in pair metadata and consumed by formatters.

Metadata Field Dictionary

The builder outputs JSON metadata because downstream formatters need more than plain prompt text. When debugging, inspect these fields before editing pools.

Normal Row Metadata

Field	Owner	Consumed by	Meaning
source	build_prompt / row builder	All formatters	Usually json_category or built_in_generator; tells which route created the row.
main_category, subcategory	Category selection	All formatters and debug	Human-readable selected category route.
category_slug, subcategory_slug	JSON category normalization	Debug/filtering	Stable-ish machine labels for selected category route.
content_seed_axis	_build_custom_row	Debug	Shows whether the item/action was driven by content or pose. Critical for hardcore pose categories.
item	_compose_item or Insta override	Krea/SDXL/Naturalizer	Clothing item, category item, or sexual scene/action text.
item_axis_values	_compose_item	Krea hardcore rewrite, SDXL tags	Filled template axes such as position/action/detail values.
custom_item, item_label	Category/pair route	Formatters and debug	Label/name for item route.
role_graph	_role_graph, POV adapter	Krea/Naturalizer	Choreography/action relationship text after POV adaptation.
source_role_graph	_role_graph before POV rewrite	Krea hardcore rewrite	Raw action graph used to infer position and contact.
scene_text	_scene_pool or location config	All formatters	Final location text.
source_scene_text	location/body-exposure/camera adapters	Debug/continuity	Previous scene text before an override.
location_config	Location config parser	Debug	Active location pool config, if connected.
pose	_pose_pool or category item route	Formatters	Generic pose text. Less important for hardcore action categories than item/role_graph.
expression	_expression_pool and intensity filter	All formatters	Final expression text unless disabled.
shared_expression	Expression selection	Debug	Expression before character-specific expansion.
character_expression_text	Character slot expression route	Krea/Naturalizer	Per-character expression clauses.
expression_enabled, expression_disabled	Builder/slot override	All formatters	Hard gate for whether expression text should appear.
expression_intensity_source	Builder/slot override	Debug	Explains whether intensity came from input, random, slot, or disabled state.
composition	_composition_pool, POV/camera adapter	All formatters	Final framing phrase.
source_composition	Composition adapter	Krea hardcore rewrite	Previous/raw composition, often better for action inference.
composition_config	Composition config parser	Debug	Active composition pool config, if connected.
camera_config	Camera nodes/parser	Krea/SDXL/debug	Structured camera settings.
camera_directive	_camera_directive	Krea/Naturalizer/prompt text	Human camera sentence. Suppressed for POV.
camera_scene_directive	scene-camera adapter	Krea/Naturalizer/prompt text	Location-aware camera layout sentence.
subject_type, subject_phrase	Subject/context builder	Formatters	Single/couple/group/configured cast route.
women_count, men_count, person_count	Cast route	Pair/formatters/debug	Effective cast counts.
cast_descriptors, cast_descriptor_text	Character/cast route	Krea/SDXL/Naturalizer	Visible cast descriptors.
character_cast_slots	Character slot chain	POV/camera/formatters	Raw configured slots.
character_slot_status, character_profile_status	Character/profile application	Debug	Explains whether slot/profile was applied or skipped.
pov_character_labels	Character slot presence mode	Krea/prompt/camera	Labels omitted from visible cast and rewritten as first-person POV.
hardcore_position_config	Hardcore position/filter nodes	Debug	Active hardcore family/position/action/interaction constraints, including interaction_only and manual_only.
negative_prompt	Category/pair/default negative route	Formatter output	Base negative text before formatter extras.
trigger	Builder input	Formatter/fallback/debug	Active trigger after fallback to default.

Insta/OF Pair Metadata

Field	Owner	Consumed by	Meaning
mode	build_insta_of_pair	Formatters	Insta/OF selects pair formatter branches.
options	SxCP Insta/OF Options	Formatters/debug	Soft/hard level, cast mode, continuity, camera modes, expression settings.
shared_descriptor	Soft row descriptor	Pair formatters	Primary creator descriptor.
shared_cast_descriptors	Cast descriptor builder	Pair formatters	Full cast descriptor list.
softcore_row, hardcore_row	Pair route	Pair formatters	Full normal metadata rows for each side.
softcore_prompt, hardcore_prompt	Pair assembly	Direct output/fallback	Raw pair prompts before formatter rewrite.
softcore_negative_prompt, hardcore_negative_prompt	Pair assembly	Formatter negatives	Separate negatives for each side.
softcore_partner_styling	_insta_of_partner_styling	Krea/SDXL pair branch	Partner softcore clothing and pose when same-cast softcore is enabled.
character_hardcore_clothing	Character slots	Krea pair branch	Explicit per-character hardcore clothing state.
default_man_hardcore_clothing	Pair fallback	Krea pair branch	Auto clothing for visible men without configured clothing.
hardcore_clothing_state	Pair clothing continuity	Krea/SDXL pair branch	Final hard clothing/body exposure sentence before Krea cleanup.
hardcore_detail_density	Insta/OF options	Krea hardcore action rewrite	Controls compact/balanced/dense action detail.
softcore_camera_config, hardcore_camera_config	Pair camera route	Krea/SDXL pair branch	Separate camera configs after option mode resolution.
softcore_camera_directive, hardcore_camera_directive	Pair camera route	Krea pair branch	Separate plain camera sentences, suppressed for POV.
softcore_camera_scene_directive, hardcore_camera_scene_directive	Scene-camera adapter	Krea/Naturalizer pair branch	Separate location-aware camera layout text.

Hardcore Position Route

SxCP Hardcore Position Pool and SxCP Hardcore Action Filter both emit hardcore_position_config.

flowchart TD
  A[Hardcore Position Pool] --> C[hardcore_position_config]
  B[Hardcore Action Filter] --> C
  C --> D[_filter_hardcore_categories_for_position]
  C --> E[_apply_hardcore_position_config_to_subcategory]
  E --> F[item_templates / axes in sexual_poses.json]
  F --> G[role_graph + item + axis_values]
  G --> H[Krea2 action sentence and POV rewrite]

What each part owns:

• sexual_poses.json: available positions, families, action templates, role graph templates, interaction templates, and action-specific pool references.
• prompt_builder.py: filters which templates/axes remain available.
• krea_formatter.py: orchestrates the selected action rewrite into model-readable prose.
• krea_action_positions.py: resolves non-POV pose anchors, body-arrangement text, duplicate arrangement checks, and action-position phrases.
• krea_action_details.py: normalizes non-climax item/detail text and dedupes foreplay, outercourse, oral, penetration, toy/double-contact, and anchor details.
• krea_action_climax.py: rewrites climax role graphs and dedupes aftermath detail/view clauses.
• krea_action_dispatch.py: normalizes non-POV role graphs, classifies action families, and applies the matching detail cleanup.
• krea_actions.py: assembles the final non-POV hardcore action sentence.
• krea_pov_actions.py: rewrites POV variants with first-person geometry.

Current broad hardcore families:

Family / focus	Subcategories
penetrative / penetration_only	penetrative_sex
foreplay / foreplay_only	foreplay_teasing
interaction / interaction_only	foreplay_teasing, body_worship_touching, clothing_position_transitions, dominant_guidance, camera_performance, group_coordination, aftercare_cleanup
manual / manual_only	manual_stimulation
oral / oral_only	oral_sex
outercourse / outercourse_only	outercourse_sex, manual_stimulation
anal / anal_only	anal_double_penetration
climax / climax_only	cumshot_climax
threesome / threesome_only	threesomes
group / group_only	group_sex_orgy

The action filter also has independent gates for toys, double-contact, penetration, foreplay, interaction, manual stimulation, oral, outercourse, anal, and climax. Keep allow_interaction=true when using the broader interaction family; keep allow_manual=true when manual stimulation should remain possible. allow_anal=false blocks anal-sex wording, not ordinary ass-touching interaction phrases such as ass grabbing or body worship.

Interaction selector keys include kissing/caressing/undressing, body worship, nipple play, ass grabbing, thigh kissing, hair holding, wrist pinning, dirty talk, position transitions, guided positioning, camera presentation, watching, aftercare, cleanup, fingering, clit rubbing, and mutual masturbation. These keys are still routed through hardcore_position_config.positions, so they are controlled by the same pose/role debug path as other hardcore action keys.

If one action keeps recurring, inspect:

1. The enabled position family/action flags.
2. weight values in sexual_poses.json.
3. Whether the selected subcategory has only a small compatible set after filtering.
4. Whether the formatter collapses several raw actions into the same final wording.

Camera And Scene Route

Camera config nodes:

• SxCP Camera Control: direct camera settings.
• SxCP Camera Orbit Control: orbit-like camera settings.
• SxCP Qwen Camera Translator: converts qwen multi-angle info into camera_config.

Camera handling:

1. Camera nodes emit camera_config.
2. build_prompt calls _apply_camera_config.
3. _camera_directive creates a plain camera sentence unless disabled/off.
4. _camera_scene_directive_for_context can add location-aware camera text.
5. POV rows suppress the normal camera directive and use first-person camera wording instead.

Current camera-aware scene adapter:

• Coworking/business-cafe/office scenes are detected by _is_coworking_scene.
• Location profile comes from _coworking_location_profile.
• Direction, distance, and elevation details come from _coworking_direction_detail, _coworking_distance_detail, and _coworking_elevation_detail.
• Composition cleanup for coworking outfit-check wording happens in _coworking_composition_prompt.

Important POV rule:

• In POV rows, location anchors must stay behind, beside, or at the frame edges. The foreground belongs to the POV body/hands and visible partner/action.

Formatter Routes

Krea2

format_krea2_prompt chooses between three roads:

• Pair metadata: _insta_pair_to_krea.
• Normal metadata row: _normal_row_to_krea.
• Plain text fallback: _fallback_text_to_krea.

Key Krea2 ownership:

• Cast descriptor naturalization: krea_cast.cast_prose, krea_cast.natural_label_text.
• Action context and family predicates: krea_action_context.py.
• Non-POV pose anchors and arrangements: krea_action_positions.py.
• Non-climax item/detail cleanup: krea_action_details.py.
• Climax role/detail cleanup: krea_action_climax.py.
• Non-POV action-family routing: krea_action_dispatch.py.
• Non-POV hardcore action sentence: krea_actions.hardcore_action_sentence.
• POV labels, filtering, and camera/composition support: krea_pov.py.
• Detail clause splitting and density limiting: krea_detail.py.
• POV hardcore sentence: krea_pov_actions.pov_action_phrase.
• Clothing state cleanup: krea_clothing.natural_clothing_state.
• Camera scene preservation: _camera_scene_phrase.

Krea2 field consumption:

Branch	Reads most from	Key functions
Normal single row	subject_type, item, pose, scene_text, expression, composition, camera_*, style fields	_normal_row_to_krea
Normal configured cast/hardcore row	cast_descriptor_text, women_count, men_count, source_role_graph, role_graph, item, item_axis_values, source_composition, pov_character_labels	_normal_row_to_krea, krea_actions.hardcore_action_sentence, krea_pov_actions.pov_action_phrase
Insta/OF pair softcore	shared_descriptor, softcore_row, softcore_partner_styling, options, soft camera fields	_insta_pair_to_krea
Insta/OF pair hardcore	hardcore_row, shared_cast_descriptors, hardcore_clothing_state, hardcore_detail_density, hard camera fields, POV labels	_insta_pair_to_krea, krea_actions.hardcore_action_sentence, krea_pov_actions.pov_action_phrase, krea_clothing.natural_clothing_state
Plain text fallback	source_text only	_fallback_text_to_krea

If metadata is connected and method says text(fallback), the formatter did not parse metadata. That is a wiring/input-hint issue, not a prompt pool issue.

SDXL

format_sdxl_prompt chooses between:

• Pair metadata: _soft_tags and _hard_tags.
• Normal metadata row: _row_core_tags.
• Plain text fallback: _fallback_text_to_sdxl.

Use this route for style triggers, weighted tag style, nude weighting, and Pony / SDXL quality/style presets.

SDXL field consumption:

Branch	Reads most from	Key functions
Normal metadata	cast descriptors, age/body/skin/hair/eyes, item, role graph, scene, camera config/directive	_row_core_tags, _appearance_tags, _camera_tags
Pair softcore	softcore_row, pair partner styling, root soft camera config	_soft_tags
Pair hardcore	hardcore_row, hardcore_clothing_state, hard camera fields, hard prompt text	_hard_tags
Text fallback	source_text, preserve-trigger setting	_fallback_text_to_sdxl

SDXL is the right place for model trigger handling, tag ordering, weight syntax, quality/style preset changes, and nude-weight defaults. Do not solve those in JSON category pools unless the raw builder text is also wrong.

Naturalizer

naturalize_caption chooses metadata-specific renderers such as _configured_cast_from_row, _couple_from_row, and single/group renderers.

Use this route when the row metadata is correct but the sentence-style caption is too mechanical or unsuitable for training captions.

Naturalizer field consumption:

Branch	Reads most from	Key functions
Normal single/couple/group	subject fields, age/body, item, scene, expression, composition, camera scene	_single_from_row, _couple_from_row, _group_or_layout_from_row
Configured cast/hardcore	cast_descriptor_text, role_graph, item, scene_text, expression, composition	_configured_cast_from_row
Insta/OF pair	softcore_row, hardcore_row, pair options and continuity	_insta_pair_from_row
Text fallback	caption or prompt text	_text_to_prose

Final Text Hygiene

prompt_hygiene.py owns route-agnostic final cleanup. It is intentionally small: whitespace, punctuation, empty field labels, adjacent duplicate sentences, repeated trigger prefixes, duplicate comma-list items, and dangling connectors.

It is called from:

• prompt_builder.build_prompt
• prompt_builder.build_insta_of_pair
• krea_formatter.format_krea2_prompt
• sdxl_formatter.format_sdxl_prompt
• caption_naturalizer.naturalize_caption

Do not put semantic fixes in prompt_hygiene.py. Sexual action readability, POV geometry, clothing state, Krea prose, SDXL weighting, and training-caption policy still belong to their route-specific owner.

Utility / Workflow Nodes

These do not own prompt pool wording, but they affect execution and review:

Node family	Files	Purpose
Loop nodes	loop_nodes.py, web/loop_slots.js	While/for loop execution and carry values.
Index switch	loop_nodes.py, web/index_switch_slots.js	Multi-input to selected output, and selected input to multi-output routing.
Accumulator	loop_nodes.py, web/accumulator_preview.js	Stores generated values/images during workflow execution and previews/reorders/deletes them.
Persistent text preview	loop_nodes.py, web/preview_any_text.js	Stores any value as text and keeps it after workflow reload.
SDXL bucket size	SxCPSDXLBucketSize in __init__.py	Random/fixed SDXL bucket width and height selection.
Krea2 resolution selector	SxCPKrea2ResolutionSelector in __init__.py	Krea-compatible width/height and API aspect/resolution helper.

Drift Audit Helper

The map should be checked when adding nodes, category files, or named pools. Run:

python tools/prompt_map_audit.py

The script does not import ComfyUI. It parses the repo and prints:

• registered display node names and known return names;
• per-JSON category counts;
• named scene/expression/composition pool inventory.
• JSON reference validation for every scene_pools, expression_pools, and composition_pools reference;
• item template validation so {placeholder} names resolve to item_axes.

Use its output to spot doc drift after adding a new node or pool. If a new node or pool appears there but not in this map, update the relevant route table. The script exits nonzero when JSON pool references or item template axes do not resolve.

Behavioral Smoke Helper

Route behavior should be checked when changing prompt generation, pair assembly, formatter metadata parsing, trigger handling, expression disabling, or scene continuity. Run:

python tools/prompt_smoke.py

The script does not import ComfyUI. It builds representative metadata rows and pair metadata through the core Python APIs, then verifies:

• generated rows keep prompt, negative prompt, scene, composition, action item, and role graph metadata populated;
• Krea2, SDXL, and natural caption routes use metadata instead of text fallback;
• SDXL and caption trigger handling keeps one trigger;
• negative prompts do not duplicate comma-list items;
• SxCP Prompt Builder From Configs-style wiring preserves category, cast, generation profile, seed lock, camera, location theme, and composition config;
• same-room Insta/OF continuity keeps prompt text and hardcore_row.scene_text synchronized;
• camera-aware coworking scene text survives single-row Krea formatting;
• softcore and hardcore pair rows can carry different camera configs without collapsing to the same camera phrase;
• POV camera-scene directives suppress normal third-person camera text while preserving first-person spatial layout;
• Krea close-interaction routes keep rewritten action wording, avoid raw builder labels, and catch invalid surface joins such as on against a wall;
• Krea POV penetration routes keep first-person position anchors, suppress normal camera text, and preserve composition punctuation before the style suffix;
• expression-disabled rows do not fall back to generated expression text.

Editing Cheatsheet

Symptom	First file/function to inspect
Wrong main category/subcategory frequency	Category node config, load_category_library, category JSON weights.
Wrong outfit/clothing item	Relevant category JSON, INSTA_OF_SOFTCORE_OUTFITS, SxCP Character Clothing.
Nude/clothing state confusing Krea2	build_insta_of_pair clothing state helpers, then krea_clothing.natural_clothing_state.
Wrong location	categories/location_pools.json, category scene_pool, _scene_pool.
Location good but camera/location layout wrong	_camera_scene_directive_for_context, coworking adapter functions.
Repeated desk/anchor in POV foreground	Coworking direction/distance/elevation helpers.
Wrong expression intensity	Character slot expression settings, _expression_entries_for_intensity, expression pools.
Expression appears when disabled	_disable_row_expression, formatter expression extraction.
Same hardcore action repeats	Hardcore filter config, sexual_poses.json weights, _apply_hardcore_position_config_to_subcategory.
Hardcore interaction beat falls back to penetration/oral	sexual_poses.json interaction subcategory, _role_graph, and krea_action_context.is_foreplay_text / krea_action_positions.hardcore_pose_anchor.
Raw hardcore prompt position is vague	sexual_poses.json item templates and role graph templates.
Krea2 hardcore prompt position is vague	krea_actions.hardcore_action_sentence or krea_pov_actions.py.
Man appears described in POV	POV labels, krea_cast.cast_prose omit labels, krea_pov_actions.pov_action_phrase.
Camera prompt missing from Krea2	Row camera_directive / camera_scene_directive, then Krea _camera_phrase.
Trigger missing in Krea2 fallback	format_krea2_prompt preserve-trigger fallback behavior.
SDXL tags too weak/wrong style	sdxl_formatter.py presets and _row_core_tags / _soft_tags / _hard_tags.
Duplicate punctuation, empty labels, repeated trigger, repeated tag item	prompt_hygiene.py, then the route-specific formatter if the repeated content is semantic.
Bed/sheet/couch or malformed surface wording leaks into hardcore prompts	hardcore_text_cleanup.py, then the relevant category pool/template.
Saved profile does not match liked character	Profile save/load path and whether the saved input is row metadata or regenerated slot config.
Accumulator preview behavior wrong	loop_nodes.py accumulator methods and web/accumulator_preview.js.

Debug Route Traces

Use these traces to narrow a problem in one pass.

Hardcore action keeps selecting the same family

1. Check metadata main_category, subcategory, content_seed_axis, hardcore_position_config, item, role_graph, and item_axis_values.
2. If hardcore_position_config disabled most families, the repeated action may be the only compatible pool left.
3. Inspect categories/sexual_poses.json for the selected subcategory, item_templates, axes, and weight.
4. If raw item differs but Krea output looks identical, inspect krea_action_context.py family predicates first, then krea_action_positions.py pose anchors/arrangements, krea_action_details.py item/detail cleanup, krea_action_climax.py climax cleanup, krea_action_dispatch.py family routing, and krea_actions.py action sentence assembly.

POV position is spatially wrong

1. Confirm pov_character_labels includes the intended man label.
2. Confirm Krea input uses metadata, not plain prompt fallback.
3. Inspect source_role_graph, item, source_composition, and item_axis_values.
4. Inspect krea_pov.py if the label omission, camera phrase, or POV composition cleanup is wrong.
5. Edit krea_pov_actions.py if the first-person body geometry is wrong.
6. Edit sexual_poses.json if the raw action lacks enough body-position anchor for any formatter to infer a good POV prompt.

Camera disappears or becomes too generic

1. Check row camera_config, camera_directive, and camera_scene_directive.
2. If camera_detail=off or camera_mode=disabled, missing camera text is expected.
3. If POV labels exist, plain camera_directive is intentionally suppressed.
4. If a location-aware sentence is missing, inspect _camera_scene_directive_for_context and the scene detector for that location family.
5. If raw metadata has camera text but Krea omits it, inspect _camera_phrase, _pair_camera_phrase, and _camera_scene_phrase.

Nude/clothing wording conflicts

1. Check pair root hardcore_clothing_state.
2. Check hard row item and source_role_graph for access flags.
3. Character slot hardcore_clothing overrides pair fallback clothing.
4. For Krea wording, inspect krea_clothing.natural_clothing_state.
5. For generation wording, inspect _insta_of_hardcore_clothing_state, _hardcore_row_access_flags, and character_hardcore_clothing_values.

Softcore contains strange no-contact or bed/action leakage

1. Check whether the prompt came from pair softcore or normal category builder.
2. In pair softcore, inspect softcore_partner_styling, softcore_row.item, softcore_row.pose, and options softcore_cast.
3. If the raw soft prompt contains awkward defensive clauses, fix build_insta_of_pair soft prompt assembly.
4. If Krea adds the awkwardness, inspect _insta_pair_to_krea.

Location composition mentions irrelevant props

1. Check scene_text and composition separately.
2. If scene is good and composition is bad, edit composition pools, not location pools.
3. If a scene-camera adapter rewrote composition, inspect scene_camera_adapters.py.
4. If the issue comes from Location Theme, edit THEMATIC_LOCATION_PRESETS.

Trigger missing after formatting

1. For builder raw prompts, check trigger and prepend_trigger_to_prompt.
2. For Krea fallback, check preserve_trigger; metadata route usually rebuilds prose and does not use prompt text as a raw string.
3. For SDXL, trigger handling belongs to format_sdxl_prompt style assembly: trigger, prepend_trigger_to_prompt, preserve_trigger, and style preset.

Safe Edit Workflow

Before changing prompt behavior:

1. Capture one raw builder output and the formatter output.
2. Check metadata JSON for source, main_category, subcategory, content_seed_axis, scene_text, item, role_graph, source_role_graph, item_axis_values, composition, camera_scene_directive, and pov_character_labels.
3. Decide whether the bug is selection, raw wording, camera adaptation, or formatter rewrite.
4. Edit the smallest owning pool/template/function.
5. Re-run a small simulation with fixed person/scene/category seeds and only the target axis varied.

The repo may have unrelated dirty files during interactive prompt work. Always stage only the intended files for commits.