# ISIS v2 Canonical Control Taxonomy

Date: 2026-02-23 Scope: WS01-002 canonical control taxonomy for deterministic,
guided, and lock-based generation controls.

## Purpose

Establish one canonical vocabulary for controls that must be exposed
consistently across workflow metadata, API/SDK contracts, RunPod handlers, and
Isis web UI.

## Source of Truth

- Machine-readable taxonomy:
  `docs/domains/isis/remediation-v2/ISIS_V2_CONTROL_TAXONOMY.json`
- Validation script: `scripts/isis/validate-v2-control-taxonomy.mjs`
- Renderer: `scripts/isis/render-v2-control-taxonomy.mjs`

## Required Canonical Controls

- `seed-lock`
- `controlnet-pose`
- `controlnet-depth`
- `controlnet-canny`
- `controlnet-lineart`
- `controlnet-segmentation`
- `controlnet-normal`
- `ipadapter-style`
- `ipadapter-face`
- `instantid`
- `faceid`
- `camera-path-guide`
- `temporal-lock`
- `style-lock`
- `identity-lock`

## Control Inventory by Category

### determinism

| Control                  | Scope      | Modalities                                        | Determinism | Required Inputs         | Node Families                           |
| ------------------------ | ---------- | ------------------------------------------------- | ----------- | ----------------------- | --------------------------------------- |
| `seed-lock`              | single-run | image, image_sequence, video, 3d, audio           | hard        | seed                    | KSampler                                |
| `sampler-scheduler-lock` | single-run | image, image_sequence, video, 3d                  | hard        | sampler_name, scheduler | KSampler                                |
| `model-lock`             | project    | image, image_sequence, video, 3d, audio           | hard        | checkpoint_id           | CheckpointLoader, VAELoader, LoraLoader |
| `prompt-template-lock`   | project    | image, image_sequence, video, 3d, audio, metadata | soft        | prompt_template_id      | CLIPTextEncode                          |

### spatial-guidance

| Control                   | Scope | Modalities                       | Determinism | Required Inputs  | Node Families                        |
| ------------------------- | ----- | -------------------------------- | ----------- | ---------------- | ------------------------------------ |
| `controlnet-pose`         | shot  | image, image_sequence, video     | guidance    | pose_map         | ControlNet, OpenPose                 |
| `controlnet-depth`        | shot  | image, image_sequence, video, 3d | guidance    | depth_map        | ControlNet, DepthAnything, MiDaS     |
| `controlnet-segmentation` | shot  | image, image_sequence, video     | guidance    | segmentation_map | ControlNet, SegmentationPreprocessor |
| `controlnet-normal`       | shot  | image, image_sequence, video, 3d | guidance    | normal_map       | ControlNet, NormalBae                |
| `controlnet-canny`        | shot  | image, image_sequence, video     | guidance    | edge_map         | ControlNet, Canny                    |
| `controlnet-lineart`      | shot  | image, image_sequence, video     | guidance    | lineart_map      | ControlNet, LineartPreprocessor      |
| `controlnet-softedge`     | shot  | image, image_sequence, video     | guidance    | softedge_map     | ControlNet, SoftEdgePreprocessor     |
| `controlnet-scribble`     | shot  | image, image_sequence, video     | guidance    | scribble_map     | ControlNet, ScribblePreprocessor     |
| `controlnet-tile`         | shot  | image, image_sequence, 3d        | guidance    | tile_source      | ControlNet, TileControl              |
| `controlnet-inpaint`      | shot  | image, image_sequence, video     | guidance    | mask, base_image | ControlNet, Inpaint                  |

### reference-guidance

| Control                 | Scope    | Modalities                   | Determinism | Required Inputs             | Node Families                      |
| ----------------------- | -------- | ---------------------------- | ----------- | --------------------------- | ---------------------------------- |
| `ipadapter-style`       | sequence | image, image_sequence, video | soft        | style_reference_image       | IPAdapter                          |
| `ipadapter-face`        | sequence | image, image_sequence, video | soft        | identity_reference_image    | IPAdapter, InsightFace             |
| `ipadapter-composition` | shot     | image, image_sequence, video | soft        | composition_reference_image | IPAdapter                          |
| `instantid`             | sequence | image, image_sequence, video | soft        | identity_reference_image    | InstantID, InsightFace, ControlNet |
| `faceid`                | sequence | image, image_sequence, video | soft        | identity_reference_image    | IPAdapter, FaceID, InsightFace     |

### camera-and-motion

| Control              | Scope    | Modalities                       | Determinism | Required Inputs   | Node Families              |
| -------------------- | -------- | -------------------------------- | ----------- | ----------------- | -------------------------- |
| `camera-path-guide`  | shot     | video, image_sequence, 3d        | guidance    | camera_path       | AnimateDiff, CameraControl |
| `lens-profile-lock`  | sequence | video, image, image_sequence, 3d | soft        | lens_profile      | CameraControl              |
| `motion-vector-lock` | sequence | video, image_sequence            | guidance    | motion_vector_map | AnimateDiff, FlowControl   |

### temporal-consistency

| Control                | Scope    | Modalities                | Determinism | Required Inputs               | Node Families                  |
| ---------------------- | -------- | ------------------------- | ----------- | ----------------------------- | ------------------------------ |
| `temporal-lock`        | sequence | video, image_sequence     | soft        | temporal_consistency_strength | AnimateDiff, FrameInterpolator |
| `keyframe-anchor-lock` | sequence | video, image_sequence, 3d | guidance    | keyframe_anchors              | AnimateDiff, KeyframeControl   |

### style-and-identity

| Control             | Scope    | Modalities                             | Determinism | Required Inputs     | Node Families                 |
| ------------------- | -------- | -------------------------------------- | ----------- | ------------------- | ----------------------------- |
| `style-lock`        | project  | image, image_sequence, video, 3d       | soft        | style_profile_id    | IPAdapter, PromptComposer     |
| `identity-lock`     | project  | image, image_sequence, video, 3d       | soft        | identity_profile_id | IPAdapter, FaceEmbedding      |
| `color-script-lock` | sequence | image, image_sequence, video, metadata | soft        | color_script        | PromptComposer, ColorTransfer |

## Governance Rules

1. Workflow definitions must only use canonical control IDs listed in this
   taxonomy.
2. Every control exposed in API/SDK must map to a canonical taxonomy entry.
3. UI labels may vary by domain, but underlying control IDs must remain
   canonical.
4. Any new control requires taxonomy update, validator pass, and remediation
   evidence entry.

## Traceability

- Checklist item: `WS01-002`
- Related asset taxonomy:
  `docs/domains/isis/remediation-v2/ISIS_V2_ASSET_TAXONOMY.json`
