feat: metadata refactor

- Refactor how metadata is handled to support a user-defined metadata in graphs
- Update workflow embed handling
- Update UI to work with these changes
- Update tests to support metadata/workflow changes

invokeai/app/api/routers/images.py

@@ -5,12 +5,13 @@ from fastapi import Body, HTTPException, Path, Query, Request, Response, UploadF
 from fastapi.responses import FileResponse
 from fastapi.routing import APIRouter
 from PIL import Image
-from pydantic import BaseModel, Field
+from pydantic import BaseModel, Field, ValidationError
 
-from invokeai.app.invocations.metadata import ImageMetadata
+from invokeai.app.invocations.baseinvocation import MetadataField, type_adapter_MetadataField
 from invokeai.app.services.image_records.image_records_common import ImageCategory, ImageRecordChanges, ResourceOrigin
 from invokeai.app.services.images.images_common import ImageDTO, ImageUrlsDTO
 from invokeai.app.services.shared.pagination import OffsetPaginatedResults
+from invokeai.app.services.workflow_records.workflow_records_common import type_adapter_WorkflowField
 
 from ..dependencies import ApiDependencies
 
@@ -45,8 +46,10 @@ async def upload_image(
     if not file.content_type or not file.content_type.startswith("image"):
         raise HTTPException(status_code=415, detail="Not an image")
 
-    contents = await file.read()
+    metadata = None
+    workflow = None
 
+    contents = await file.read()
     try:
         pil_image = Image.open(io.BytesIO(contents))
         if crop_visible:
@@ -56,6 +59,24 @@ async def upload_image(
         # Error opening the image
         raise HTTPException(status_code=415, detail="Failed to read image")
 
+    # attempt to parse metadata from image
+    metadata_raw = pil_image.info.get("invokeai_metadata", None)
+    if metadata_raw:
+        try:
+            metadata = type_adapter_MetadataField.validate_json(metadata_raw)
+        except ValidationError:
+            ApiDependencies.invoker.services.logger.warn("Failed to parse metadata for uploaded image")
+            pass
+
+    # attempt to parse workflow from image
+    workflow_raw = pil_image.info.get("invokeai_workflow", None)
+    if workflow_raw is not None:
+        try:
+            workflow = type_adapter_WorkflowField.validate_json(workflow_raw)
+        except ValidationError:
+            ApiDependencies.invoker.services.logger.warn("Failed to parse metadata for uploaded image")
+            pass
+
     try:
         image_dto = ApiDependencies.invoker.services.images.create(
             image=pil_image,
@@ -63,6 +84,8 @@ async def upload_image(
             image_category=image_category,
             session_id=session_id,
             board_id=board_id,
+            metadata=metadata,
+            workflow=workflow,
             is_intermediate=is_intermediate,
         )
 
@@ -146,11 +169,11 @@ async def get_image_dto(
 @images_router.get(
     "/i/{image_name}/metadata",
     operation_id="get_image_metadata",
-    response_model=ImageMetadata,
+    response_model=Optional[MetadataField],
 )
 async def get_image_metadata(
     image_name: str = Path(description="The name of image to get"),
-) -> ImageMetadata:
+) -> Optional[MetadataField]:
     """Gets an image's metadata"""
 
     try:

invokeai/app/api/routers/workflows.py

@@ -1,4 +1,4 @@
-from fastapi import APIRouter, Body, Path
+from fastapi import APIRouter, Path
 
 from invokeai.app.api.dependencies import ApiDependencies
 from invokeai.app.services.workflow_records.workflow_records_common import WorkflowField

invokeai/app/api_app.py

@@ -44,8 +44,8 @@ if True:  # hack to make flake8 happy with imports coming after setting up the c
         boards,
         images,
         models,
-        sessions,
         session_queue,
+        sessions,
         utilities,
         workflows,
     )

invokeai/app/invocations/baseinvocation.py

@@ -15,8 +15,8 @@ from pydantic.fields import _Unset
 from pydantic_core import PydanticUndefined
 
 from invokeai.app.services.config.config_default import InvokeAIAppConfig
-from invokeai.app.util.misc import uuid_string
 from invokeai.app.services.workflow_records.workflow_records_common import WorkflowField
+from invokeai.app.util.misc import uuid_string
 
 if TYPE_CHECKING:
     from ..services.invocation_services import InvocationServices
@@ -60,6 +60,11 @@ class FieldDescriptions:
     denoised_latents = "Denoised latents tensor"
     latents = "Latents tensor"
     strength = "Strength of denoising (proportional to steps)"
+    metadata = "Optional metadata to be saved with the image"
+    metadata_collection = "Collection of Metadata"
+    metadata_item_polymorphic = "A single metadata item or collection of metadata items"
+    metadata_item_label = "Label for this metadata item"
+    metadata_item_value = "The value for this metadata item (may be any type)"
     workflow = "Optional workflow to be saved with the image"
     interp_mode = "Interpolation mode"
     torch_antialias = "Whether or not to apply antialiasing (bilinear or bicubic only)"
@@ -167,8 +172,12 @@ class UIType(str, Enum):
     Scheduler = "Scheduler"
     WorkflowField = "WorkflowField"
     IsIntermediate = "IsIntermediate"
-    MetadataField = "MetadataField"
     BoardField = "BoardField"
+    Any = "Any"
+    MetadataItem = "MetadataItem"
+    MetadataItemCollection = "MetadataItemCollection"
+    MetadataItemPolymorphic = "MetadataItemPolymorphic"
+    MetadataDict = "MetadataDict"
     # endregion
 
 
@@ -807,3 +816,26 @@ def invocation_output(
 class WithWorkflow(BaseModel):
     workflow: Optional[WorkflowField] = InputField(default=None, description=FieldDescriptions.workflow)
 
+
+class MetadataItemField(BaseModel):
+    label: str = Field(description=FieldDescriptions.metadata_item_label)
+    value: Any = Field(description=FieldDescriptions.metadata_item_value)
+
+
+class MetadataField(RootModel):
+    """
+    Pydantic model for metadata with custom root of type dict[str, Any].
+    Metadata is stored without a strict schema.
+    """
+
+    root: dict[str, Any] = Field(description="A dictionary of metadata, shape of which is arbitrary")
+
+    def model_dump(self, *args, **kwargs) -> dict[str, Any]:
+        return super().model_dump(*args, **kwargs)["root"]
+
+
+type_adapter_MetadataField = TypeAdapter(MetadataField)
+
+
+class WithMetadata(BaseModel):
+    metadata: Optional[MetadataField] = InputField(default=None, description=FieldDescriptions.metadata)

invokeai/app/invocations/controlnet_image_processors.py

@@ -38,6 +38,8 @@ from .baseinvocation import (
     InputField,
     InvocationContext,
     OutputField,
+    WithMetadata,
+    WithWorkflow,
     invocation,
     invocation_output,
 )
@@ -127,12 +129,12 @@ class ControlNetInvocation(BaseInvocation):
 
 
 # This invocation exists for other invocations to subclass it - do not register with @invocation!
-class ImageProcessorInvocation(BaseInvocation):
+class ImageProcessorInvocation(BaseInvocation, WithMetadata, WithWorkflow):
     """Base class for invocations that preprocess images for ControlNet"""
 
     image: ImageField = InputField(description="The image to process")
 
-    def run_processor(self, image):
+    def run_processor(self, image: Image.Image) -> Image.Image:
         # superclass just passes through image without processing
         return image
 
@@ -150,6 +152,7 @@ class ImageProcessorInvocation(BaseInvocation):
             session_id=context.graph_execution_state_id,
             node_id=self.id,
             is_intermediate=self.is_intermediate,
+            metadata=self.metadata,
             workflow=self.workflow,
         )
 

invokeai/app/invocations/cv.py

@@ -8,11 +8,11 @@ from PIL import Image, ImageOps
 from invokeai.app.invocations.primitives import ImageField, ImageOutput
 from invokeai.app.services.image_records.image_records_common import ImageCategory, ResourceOrigin
 
-from .baseinvocation import BaseInvocation, InputField, InvocationContext, invocation
+from .baseinvocation import BaseInvocation, InputField, InvocationContext, WithMetadata, WithWorkflow, invocation
 
 
 @invocation("cv_inpaint", title="OpenCV Inpaint", tags=["opencv", "inpaint"], category="inpaint", version="1.0.0")
-class CvInpaintInvocation(BaseInvocation):
+class CvInpaintInvocation(BaseInvocation, WithMetadata, WithWorkflow):
     """Simple inpaint using opencv."""
 
     image: ImageField = InputField(description="The image to inpaint")

invokeai/app/invocations/facetools.py

@@ -16,6 +16,8 @@ from invokeai.app.invocations.baseinvocation import (
     InputField,
     InvocationContext,
     OutputField,
+    WithMetadata,
+    WithWorkflow,
     invocation,
     invocation_output,
 )
@@ -437,7 +439,7 @@ def get_faces_list(
 
 
 @invocation("face_off", title="FaceOff", tags=["image", "faceoff", "face", "mask"], category="image", version="1.0.2")
-class FaceOffInvocation(BaseInvocation):
+class FaceOffInvocation(BaseInvocation, WithWorkflow, WithMetadata):
     """Bound, extract, and mask a face from an image using MediaPipe detection"""
 
     image: ImageField = InputField(description="Image for face detection")
@@ -531,7 +533,7 @@ class FaceOffInvocation(BaseInvocation):
 
 
 @invocation("face_mask_detection", title="FaceMask", tags=["image", "face", "mask"], category="image", version="1.0.2")
-class FaceMaskInvocation(BaseInvocation):
+class FaceMaskInvocation(BaseInvocation, WithWorkflow, WithMetadata):
     """Face mask creation using mediapipe face detection"""
 
     image: ImageField = InputField(description="Image to face detect")
@@ -650,7 +652,7 @@ class FaceMaskInvocation(BaseInvocation):
 @invocation(
     "face_identifier", title="FaceIdentifier", tags=["image", "face", "identifier"], category="image", version="1.0.2"
 )
-class FaceIdentifierInvocation(BaseInvocation):
+class FaceIdentifierInvocation(BaseInvocation, WithWorkflow, WithMetadata):
     """Outputs an image with detected face IDs printed on each face. For use with other FaceTools."""
 
     image: ImageField = InputField(description="Image to face detect")

invokeai/app/invocations/image.py

@@ -7,13 +7,21 @@ import cv2
 import numpy
 from PIL import Image, ImageChops, ImageFilter, ImageOps
 
-from invokeai.app.invocations.metadata import CoreMetadata
 from invokeai.app.invocations.primitives import BoardField, ColorField, ImageField, ImageOutput
 from invokeai.app.services.image_records.image_records_common import ImageCategory, ResourceOrigin
 from invokeai.backend.image_util.invisible_watermark import InvisibleWatermark
 from invokeai.backend.image_util.safety_checker import SafetyChecker
 
-from .baseinvocation import BaseInvocation, FieldDescriptions, Input, InputField, InvocationContext, invocation
+from .baseinvocation import (
+    BaseInvocation,
+    FieldDescriptions,
+    Input,
+    InputField,
+    InvocationContext,
+    WithMetadata,
+    WithWorkflow,
+    invocation,
+)
 
 
 @invocation("show_image", title="Show Image", tags=["image"], category="image", version="1.0.0")
@@ -36,14 +44,8 @@ class ShowImageInvocation(BaseInvocation):
         )
 
 
-@invocation(
+@invocation("blank_image", title="Blank Image", tags=["image"], category="image", version="1.0.0")
-    "blank_image",
+class BlankImageInvocation(BaseInvocation, WithMetadata, WithWorkflow):
-    title="Blank Image",
-    tags=["image"],
-    category="image",
-    version="1.0.0",
-)
-class BlankImageInvocation(BaseInvocation):
     """Creates a blank image and forwards it to the pipeline"""
 
     width: int = InputField(default=512, description="The width of the image")
@@ -61,6 +63,7 @@ class BlankImageInvocation(BaseInvocation):
             node_id=self.id,
             session_id=context.graph_execution_state_id,
             is_intermediate=self.is_intermediate,
+            metadata=self.metadata,
             workflow=self.workflow,
         )
 
@@ -71,14 +74,8 @@ class BlankImageInvocation(BaseInvocation):
         )
 
 
-@invocation(
+@invocation("img_crop", title="Crop Image", tags=["image", "crop"], category="image", version="1.0.0")
-    "img_crop",
+class ImageCropInvocation(BaseInvocation, WithWorkflow, WithMetadata):
-    title="Crop Image",
-    tags=["image", "crop"],
-    category="image",
-    version="1.0.0",
-)
-class ImageCropInvocation(BaseInvocation):
     """Crops an image to a specified box. The box can be outside of the image."""
 
     image: ImageField = InputField(description="The image to crop")
@@ -100,6 +97,7 @@ class ImageCropInvocation(BaseInvocation):
             node_id=self.id,
             session_id=context.graph_execution_state_id,
             is_intermediate=self.is_intermediate,
+            metadata=self.metadata,
             workflow=self.workflow,
         )
 
@@ -110,14 +108,8 @@ class ImageCropInvocation(BaseInvocation):
         )
 
 
-@invocation(
+@invocation("img_paste", title="Paste Image", tags=["image", "paste"], category="image", version="1.0.1")
-    "img_paste",
+class ImagePasteInvocation(BaseInvocation, WithWorkflow, WithMetadata):
-    title="Paste Image",
-    tags=["image", "paste"],
-    category="image",
-    version="1.0.1",
-)
-class ImagePasteInvocation(BaseInvocation):
     """Pastes an image into another image."""
 
     base_image: ImageField = InputField(description="The base image")
@@ -159,6 +151,7 @@ class ImagePasteInvocation(BaseInvocation):
             node_id=self.id,
             session_id=context.graph_execution_state_id,
             is_intermediate=self.is_intermediate,
+            metadata=self.metadata,
             workflow=self.workflow,
         )
 
@@ -169,14 +162,8 @@ class ImagePasteInvocation(BaseInvocation):
         )
 
 
-@invocation(
+@invocation("tomask", title="Mask from Alpha", tags=["image", "mask"], category="image", version="1.0.0")
-    "tomask",
+class MaskFromAlphaInvocation(BaseInvocation, WithWorkflow, WithMetadata):
-    title="Mask from Alpha",
-    tags=["image", "mask"],
-    category="image",
-    version="1.0.0",
-)
-class MaskFromAlphaInvocation(BaseInvocation):
     """Extracts the alpha channel of an image as a mask."""
 
     image: ImageField = InputField(description="The image to create the mask from")
@@ -196,6 +183,7 @@ class MaskFromAlphaInvocation(BaseInvocation):
             node_id=self.id,
             session_id=context.graph_execution_state_id,
             is_intermediate=self.is_intermediate,
+            metadata=self.metadata,
             workflow=self.workflow,
         )
 
@@ -206,14 +194,8 @@ class MaskFromAlphaInvocation(BaseInvocation):
         )
 
 
-@invocation(
+@invocation("img_mul", title="Multiply Images", tags=["image", "multiply"], category="image", version="1.0.0")
-    "img_mul",
+class ImageMultiplyInvocation(BaseInvocation, WithWorkflow, WithMetadata):
-    title="Multiply Images",
-    tags=["image", "multiply"],
-    category="image",
-    version="1.0.0",
-)
-class ImageMultiplyInvocation(BaseInvocation):
     """Multiplies two images together using `PIL.ImageChops.multiply()`."""
 
     image1: ImageField = InputField(description="The first image to multiply")
@@ -232,6 +214,7 @@ class ImageMultiplyInvocation(BaseInvocation):
             node_id=self.id,
             session_id=context.graph_execution_state_id,
             is_intermediate=self.is_intermediate,
+            metadata=self.metadata,
             workflow=self.workflow,
         )
 
@@ -245,14 +228,8 @@ class ImageMultiplyInvocation(BaseInvocation):
 IMAGE_CHANNELS = Literal["A", "R", "G", "B"]
 
 
-@invocation(
+@invocation("img_chan", title="Extract Image Channel", tags=["image", "channel"], category="image", version="1.0.0")
-    "img_chan",
+class ImageChannelInvocation(BaseInvocation, WithWorkflow, WithMetadata):
-    title="Extract Image Channel",
-    tags=["image", "channel"],
-    category="image",
-    version="1.0.0",
-)
-class ImageChannelInvocation(BaseInvocation):
     """Gets a channel from an image."""
 
     image: ImageField = InputField(description="The image to get the channel from")
@@ -270,6 +247,7 @@ class ImageChannelInvocation(BaseInvocation):
             node_id=self.id,
             session_id=context.graph_execution_state_id,
             is_intermediate=self.is_intermediate,
+            metadata=self.metadata,
             workflow=self.workflow,
         )
 
@@ -283,14 +261,8 @@ class ImageChannelInvocation(BaseInvocation):
 IMAGE_MODES = Literal["L", "RGB", "RGBA", "CMYK", "YCbCr", "LAB", "HSV", "I", "F"]
 
 
-@invocation(
+@invocation("img_conv", title="Convert Image Mode", tags=["image", "convert"], category="image", version="1.0.0")
-    "img_conv",
+class ImageConvertInvocation(BaseInvocation, WithWorkflow, WithMetadata):
-    title="Convert Image Mode",
-    tags=["image", "convert"],
-    category="image",
-    version="1.0.0",
-)
-class ImageConvertInvocation(BaseInvocation):
     """Converts an image to a different mode."""
 
     image: ImageField = InputField(description="The image to convert")
@@ -308,6 +280,7 @@ class ImageConvertInvocation(BaseInvocation):
             node_id=self.id,
             session_id=context.graph_execution_state_id,
             is_intermediate=self.is_intermediate,
+            metadata=self.metadata,
             workflow=self.workflow,
         )
 
@@ -318,14 +291,8 @@ class ImageConvertInvocation(BaseInvocation):
         )
 
 
-@invocation(
+@invocation("img_blur", title="Blur Image", tags=["image", "blur"], category="image", version="1.0.0")
-    "img_blur",
+class ImageBlurInvocation(BaseInvocation, WithWorkflow, WithMetadata):
-    title="Blur Image",
-    tags=["image", "blur"],
-    category="image",
-    version="1.0.0",
-)
-class ImageBlurInvocation(BaseInvocation):
     """Blurs an image"""
 
     image: ImageField = InputField(description="The image to blur")
@@ -348,6 +315,7 @@ class ImageBlurInvocation(BaseInvocation):
             node_id=self.id,
             session_id=context.graph_execution_state_id,
             is_intermediate=self.is_intermediate,
+            metadata=self.metadata,
             workflow=self.workflow,
         )
 
@@ -378,23 +346,14 @@ PIL_RESAMPLING_MAP = {
 }
 
 
-@invocation(
+@invocation("img_resize", title="Resize Image", tags=["image", "resize"], category="image", version="1.0.0")
-    "img_resize",
+class ImageResizeInvocation(BaseInvocation, WithMetadata, WithWorkflow):
-    title="Resize Image",
-    tags=["image", "resize"],
-    category="image",
-    version="1.0.0",
-)
-class ImageResizeInvocation(BaseInvocation):
     """Resizes an image to specific dimensions"""
 
     image: ImageField = InputField(description="The image to resize")
     width: int = InputField(default=512, gt=0, description="The width to resize to (px)")
     height: int = InputField(default=512, gt=0, description="The height to resize to (px)")
     resample_mode: PIL_RESAMPLING_MODES = InputField(default="bicubic", description="The resampling mode")
-    metadata: Optional[CoreMetadata] = InputField(
-        default=None, description=FieldDescriptions.core_metadata, ui_hidden=True
-    )
 
     def invoke(self, context: InvocationContext) -> ImageOutput:
         image = context.services.images.get_pil_image(self.image.image_name)
@@ -413,7 +372,7 @@ class ImageResizeInvocation(BaseInvocation):
             node_id=self.id,
             session_id=context.graph_execution_state_id,
             is_intermediate=self.is_intermediate,
-            metadata=self.metadata.model_dump() if self.metadata else None,
+            metadata=self.metadata,
             workflow=self.workflow,
         )
 
@@ -424,14 +383,8 @@ class ImageResizeInvocation(BaseInvocation):
         )
 
 
-@invocation(
+@invocation("img_scale", title="Scale Image", tags=["image", "scale"], category="image", version="1.0.0")
-    "img_scale",
+class ImageScaleInvocation(BaseInvocation, WithMetadata, WithWorkflow):
-    title="Scale Image",
-    tags=["image", "scale"],
-    category="image",
-    version="1.0.0",
-)
-class ImageScaleInvocation(BaseInvocation):
     """Scales an image by a factor"""
 
     image: ImageField = InputField(description="The image to scale")
@@ -461,6 +414,7 @@ class ImageScaleInvocation(BaseInvocation):
             node_id=self.id,
             session_id=context.graph_execution_state_id,
             is_intermediate=self.is_intermediate,
+            metadata=self.metadata,
             workflow=self.workflow,
         )
 
@@ -471,14 +425,8 @@ class ImageScaleInvocation(BaseInvocation):
         )
 
 
-@invocation(
+@invocation("img_lerp", title="Lerp Image", tags=["image", "lerp"], category="image", version="1.0.0")
-    "img_lerp",
+class ImageLerpInvocation(BaseInvocation, WithWorkflow, WithMetadata):
-    title="Lerp Image",
-    tags=["image", "lerp"],
-    category="image",
-    version="1.0.0",
-)
-class ImageLerpInvocation(BaseInvocation):
     """Linear interpolation of all pixels of an image"""
 
     image: ImageField = InputField(description="The image to lerp")
@@ -500,6 +448,7 @@ class ImageLerpInvocation(BaseInvocation):
             node_id=self.id,
             session_id=context.graph_execution_state_id,
             is_intermediate=self.is_intermediate,
+            metadata=self.metadata,
             workflow=self.workflow,
         )
 
@@ -510,14 +459,8 @@ class ImageLerpInvocation(BaseInvocation):
         )
 
 
-@invocation(
+@invocation("img_ilerp", title="Inverse Lerp Image", tags=["image", "ilerp"], category="image", version="1.0.0")
-    "img_ilerp",
+class ImageInverseLerpInvocation(BaseInvocation, WithWorkflow, WithMetadata):
-    title="Inverse Lerp Image",
-    tags=["image", "ilerp"],
-    category="image",
-    version="1.0.0",
-)
-class ImageInverseLerpInvocation(BaseInvocation):
     """Inverse linear interpolation of all pixels of an image"""
 
     image: ImageField = InputField(description="The image to lerp")
@@ -539,6 +482,7 @@ class ImageInverseLerpInvocation(BaseInvocation):
             node_id=self.id,
             session_id=context.graph_execution_state_id,
             is_intermediate=self.is_intermediate,
+            metadata=self.metadata,
             workflow=self.workflow,
         )
 
@@ -549,20 +493,11 @@ class ImageInverseLerpInvocation(BaseInvocation):
         )
 
 
-@invocation(
+@invocation("img_nsfw", title="Blur NSFW Image", tags=["image", "nsfw"], category="image", version="1.0.0")
-    "img_nsfw",
+class ImageNSFWBlurInvocation(BaseInvocation, WithMetadata, WithWorkflow):
-    title="Blur NSFW Image",
-    tags=["image", "nsfw"],
-    category="image",
-    version="1.0.0",
-)
-class ImageNSFWBlurInvocation(BaseInvocation):
     """Add blur to NSFW-flagged images"""
 
     image: ImageField = InputField(description="The image to check")
-    metadata: Optional[CoreMetadata] = InputField(
-        default=None, description=FieldDescriptions.core_metadata, ui_hidden=True
-    )
 
     def invoke(self, context: InvocationContext) -> ImageOutput:
         image = context.services.images.get_pil_image(self.image.image_name)
@@ -583,7 +518,7 @@ class ImageNSFWBlurInvocation(BaseInvocation):
             node_id=self.id,
             session_id=context.graph_execution_state_id,
             is_intermediate=self.is_intermediate,
-            metadata=self.metadata.model_dump() if self.metadata else None,
+            metadata=self.metadata,
             workflow=self.workflow,
         )
 
@@ -607,14 +542,11 @@ class ImageNSFWBlurInvocation(BaseInvocation):
     category="image",
     version="1.0.0",
 )
-class ImageWatermarkInvocation(BaseInvocation):
+class ImageWatermarkInvocation(BaseInvocation, WithMetadata, WithWorkflow):
     """Add an invisible watermark to an image"""
 
     image: ImageField = InputField(description="The image to check")
     text: str = InputField(default="InvokeAI", description="Watermark text")
-    metadata: Optional[CoreMetadata] = InputField(
-        default=None, description=FieldDescriptions.core_metadata, ui_hidden=True
-    )
 
     def invoke(self, context: InvocationContext) -> ImageOutput:
         image = context.services.images.get_pil_image(self.image.image_name)
@@ -626,7 +558,7 @@ class ImageWatermarkInvocation(BaseInvocation):
             node_id=self.id,
             session_id=context.graph_execution_state_id,
             is_intermediate=self.is_intermediate,
-            metadata=self.metadata.model_dump() if self.metadata else None,
+            metadata=self.metadata,
             workflow=self.workflow,
         )
 
@@ -637,14 +569,8 @@ class ImageWatermarkInvocation(BaseInvocation):
         )
 
 
-@invocation(
+@invocation("mask_edge", title="Mask Edge", tags=["image", "mask", "inpaint"], category="image", version="1.0.0")
-    "mask_edge",
+class MaskEdgeInvocation(BaseInvocation, WithWorkflow, WithMetadata):
-    title="Mask Edge",
-    tags=["image", "mask", "inpaint"],
-    category="image",
-    version="1.0.0",
-)
-class MaskEdgeInvocation(BaseInvocation):
     """Applies an edge mask to an image"""
 
     image: ImageField = InputField(description="The image to apply the mask to")
@@ -678,6 +604,7 @@ class MaskEdgeInvocation(BaseInvocation):
             node_id=self.id,
             session_id=context.graph_execution_state_id,
             is_intermediate=self.is_intermediate,
+            metadata=self.metadata,
             workflow=self.workflow,
         )
 
@@ -695,7 +622,7 @@ class MaskEdgeInvocation(BaseInvocation):
     category="image",
     version="1.0.0",
 )
-class MaskCombineInvocation(BaseInvocation):
+class MaskCombineInvocation(BaseInvocation, WithWorkflow, WithMetadata):
     """Combine two masks together by multiplying them using `PIL.ImageChops.multiply()`."""
 
     mask1: ImageField = InputField(description="The first mask to combine")
@@ -714,6 +641,7 @@ class MaskCombineInvocation(BaseInvocation):
             node_id=self.id,
             session_id=context.graph_execution_state_id,
             is_intermediate=self.is_intermediate,
+            metadata=self.metadata,
             workflow=self.workflow,
         )
 
@@ -724,14 +652,8 @@ class MaskCombineInvocation(BaseInvocation):
         )
 
 
-@invocation(
+@invocation("color_correct", title="Color Correct", tags=["image", "color"], category="image", version="1.0.0")
-    "color_correct",
+class ColorCorrectInvocation(BaseInvocation, WithWorkflow, WithMetadata):
-    title="Color Correct",
-    tags=["image", "color"],
-    category="image",
-    version="1.0.0",
-)
-class ColorCorrectInvocation(BaseInvocation):
     """
     Shifts the colors of a target image to match the reference image, optionally
     using a mask to only color-correct certain regions of the target image.
@@ -830,6 +752,7 @@ class ColorCorrectInvocation(BaseInvocation):
             node_id=self.id,
             session_id=context.graph_execution_state_id,
             is_intermediate=self.is_intermediate,
+            metadata=self.metadata,
             workflow=self.workflow,
         )
 
@@ -840,14 +763,8 @@ class ColorCorrectInvocation(BaseInvocation):
         )
 
 
-@invocation(
+@invocation("img_hue_adjust", title="Adjust Image Hue", tags=["image", "hue"], category="image", version="1.0.0")
-    "img_hue_adjust",
+class ImageHueAdjustmentInvocation(BaseInvocation, WithWorkflow, WithMetadata):
-    title="Adjust Image Hue",
-    tags=["image", "hue"],
-    category="image",
-    version="1.0.0",
-)
-class ImageHueAdjustmentInvocation(BaseInvocation):
     """Adjusts the Hue of an image."""
 
     image: ImageField = InputField(description="The image to adjust")
@@ -875,6 +792,7 @@ class ImageHueAdjustmentInvocation(BaseInvocation):
             node_id=self.id,
             is_intermediate=self.is_intermediate,
             session_id=context.graph_execution_state_id,
+            metadata=self.metadata,
             workflow=self.workflow,
         )
 
@@ -950,7 +868,7 @@ CHANNEL_FORMATS = {
     category="image",
     version="1.0.0",
 )
-class ImageChannelOffsetInvocation(BaseInvocation):
+class ImageChannelOffsetInvocation(BaseInvocation, WithWorkflow, WithMetadata):
     """Add or subtract a value from a specific color channel of an image."""
 
     image: ImageField = InputField(description="The image to adjust")
@@ -984,6 +902,7 @@ class ImageChannelOffsetInvocation(BaseInvocation):
             node_id=self.id,
             is_intermediate=self.is_intermediate,
             session_id=context.graph_execution_state_id,
+            metadata=self.metadata,
             workflow=self.workflow,
         )
 
@@ -1020,7 +939,7 @@ class ImageChannelOffsetInvocation(BaseInvocation):
     category="image",
     version="1.0.0",
 )
-class ImageChannelMultiplyInvocation(BaseInvocation):
+class ImageChannelMultiplyInvocation(BaseInvocation, WithWorkflow, WithMetadata):
     """Scale a specific color channel of an image."""
 
     image: ImageField = InputField(description="The image to adjust")
@@ -1060,6 +979,7 @@ class ImageChannelMultiplyInvocation(BaseInvocation):
             is_intermediate=self.is_intermediate,
             session_id=context.graph_execution_state_id,
             workflow=self.workflow,
+            metadata=self.metadata,
         )
 
         return ImageOutput(
@@ -1079,16 +999,11 @@ class ImageChannelMultiplyInvocation(BaseInvocation):
     version="1.0.1",
     use_cache=False,
 )
-class SaveImageInvocation(BaseInvocation):
+class SaveImageInvocation(BaseInvocation, WithWorkflow, WithMetadata):
     """Saves an image. Unlike an image primitive, this invocation stores a copy of the image."""
 
     image: ImageField = InputField(description=FieldDescriptions.image)
-    board: Optional[BoardField] = InputField(default=None, description=FieldDescriptions.board, input=Input.Direct)
+    board: BoardField = InputField(default=None, description=FieldDescriptions.board, input=Input.Direct)
-    metadata: Optional[CoreMetadata] = InputField(
-        default=None,
-        description=FieldDescriptions.core_metadata,
-        ui_hidden=True,
-    )
 
     def invoke(self, context: InvocationContext) -> ImageOutput:
         image = context.services.images.get_pil_image(self.image.image_name)
@@ -1101,7 +1016,7 @@ class SaveImageInvocation(BaseInvocation):
             node_id=self.id,
             session_id=context.graph_execution_state_id,
             is_intermediate=self.is_intermediate,
-            metadata=self.metadata.model_dump() if self.metadata else None,
+            metadata=self.metadata,
             workflow=self.workflow,
         )
 

invokeai/app/invocations/infill.py

@@ -13,7 +13,7 @@ from invokeai.backend.image_util.cv2_inpaint import cv2_inpaint
 from invokeai.backend.image_util.lama import LaMA
 from invokeai.backend.image_util.patchmatch import PatchMatch
 
-from .baseinvocation import BaseInvocation, InputField, InvocationContext, invocation
+from .baseinvocation import BaseInvocation, InputField, InvocationContext, WithMetadata, WithWorkflow, invocation
 from .image import PIL_RESAMPLING_MAP, PIL_RESAMPLING_MODES
 
 
@@ -119,7 +119,7 @@ def tile_fill_missing(im: Image.Image, tile_size: int = 16, seed: Optional[int]
 
 
 @invocation("infill_rgba", title="Solid Color Infill", tags=["image", "inpaint"], category="inpaint", version="1.0.0")
-class InfillColorInvocation(BaseInvocation):
+class InfillColorInvocation(BaseInvocation, WithWorkflow, WithMetadata):
     """Infills transparent areas of an image with a solid color"""
 
     image: ImageField = InputField(description="The image to infill")
@@ -143,6 +143,7 @@ class InfillColorInvocation(BaseInvocation):
             node_id=self.id,
             session_id=context.graph_execution_state_id,
             is_intermediate=self.is_intermediate,
+            metadata=self.metadata,
             workflow=self.workflow,
         )
 
@@ -154,7 +155,7 @@ class InfillColorInvocation(BaseInvocation):
 
 
 @invocation("infill_tile", title="Tile Infill", tags=["image", "inpaint"], category="inpaint", version="1.0.0")
-class InfillTileInvocation(BaseInvocation):
+class InfillTileInvocation(BaseInvocation, WithWorkflow, WithMetadata):
     """Infills transparent areas of an image with tiles of the image"""
 
     image: ImageField = InputField(description="The image to infill")
@@ -179,6 +180,7 @@ class InfillTileInvocation(BaseInvocation):
             node_id=self.id,
             session_id=context.graph_execution_state_id,
             is_intermediate=self.is_intermediate,
+            metadata=self.metadata,
             workflow=self.workflow,
         )
 
@@ -192,7 +194,7 @@ class InfillTileInvocation(BaseInvocation):
 @invocation(
     "infill_patchmatch", title="PatchMatch Infill", tags=["image", "inpaint"], category="inpaint", version="1.0.0"
 )
-class InfillPatchMatchInvocation(BaseInvocation):
+class InfillPatchMatchInvocation(BaseInvocation, WithWorkflow, WithMetadata):
     """Infills transparent areas of an image using the PatchMatch algorithm"""
 
     image: ImageField = InputField(description="The image to infill")
@@ -232,6 +234,7 @@ class InfillPatchMatchInvocation(BaseInvocation):
             node_id=self.id,
             session_id=context.graph_execution_state_id,
             is_intermediate=self.is_intermediate,
+            metadata=self.metadata,
             workflow=self.workflow,
         )
 
@@ -243,7 +246,7 @@ class InfillPatchMatchInvocation(BaseInvocation):
 
 
 @invocation("infill_lama", title="LaMa Infill", tags=["image", "inpaint"], category="inpaint", version="1.0.0")
-class LaMaInfillInvocation(BaseInvocation):
+class LaMaInfillInvocation(BaseInvocation, WithWorkflow, WithMetadata):
     """Infills transparent areas of an image using the LaMa model"""
 
     image: ImageField = InputField(description="The image to infill")
@@ -260,6 +263,8 @@ class LaMaInfillInvocation(BaseInvocation):
             node_id=self.id,
             session_id=context.graph_execution_state_id,
             is_intermediate=self.is_intermediate,
+            metadata=self.metadata,
+            workflow=self.workflow,
         )
 
         return ImageOutput(
@@ -269,8 +274,8 @@ class LaMaInfillInvocation(BaseInvocation):
         )
 
 
-@invocation("infill_cv2", title="CV2 Infill", tags=["image", "inpaint"], category="inpaint", version="1.0.0")
+@invocation("infill_cv2", title="CV2 Infill", tags=["image", "inpaint"], category="inpaint")
-class CV2InfillInvocation(BaseInvocation):
+class CV2InfillInvocation(BaseInvocation, WithWorkflow, WithMetadata):
     """Infills transparent areas of an image using OpenCV Inpainting"""
 
     image: ImageField = InputField(description="The image to infill")
@@ -287,6 +292,8 @@ class CV2InfillInvocation(BaseInvocation):
             node_id=self.id,
             session_id=context.graph_execution_state_id,
             is_intermediate=self.is_intermediate,
+            metadata=self.metadata,
+            workflow=self.workflow,
         )
 
         return ImageOutput(

invokeai/app/invocations/latent.py

@@ -23,7 +23,6 @@ from pydantic import field_validator
 from torchvision.transforms.functional import resize as tv_resize
 
 from invokeai.app.invocations.ip_adapter import IPAdapterField
-from invokeai.app.invocations.metadata import CoreMetadata
 from invokeai.app.invocations.primitives import (
     DenoiseMaskField,
     DenoiseMaskOutput,
@@ -64,6 +63,8 @@ from .baseinvocation import (
     InvocationContext,
     OutputField,
     UIType,
+    WithMetadata,
+    WithWorkflow,
     invocation,
     invocation_output,
 )
@@ -792,7 +793,7 @@ class DenoiseLatentsInvocation(BaseInvocation):
     category="latents",
     version="1.0.0",
 )
-class LatentsToImageInvocation(BaseInvocation):
+class LatentsToImageInvocation(BaseInvocation, WithMetadata, WithWorkflow):
     """Generates an image from latents."""
 
     latents: LatentsField = InputField(
@@ -805,11 +806,6 @@ class LatentsToImageInvocation(BaseInvocation):
     )
     tiled: bool = InputField(default=False, description=FieldDescriptions.tiled)
     fp32: bool = InputField(default=DEFAULT_PRECISION == "float32", description=FieldDescriptions.fp32)
-    metadata: Optional[CoreMetadata] = InputField(
-        default=None,
-        description=FieldDescriptions.core_metadata,
-        ui_hidden=True,
-    )
 
     @torch.no_grad()
     def invoke(self, context: InvocationContext) -> ImageOutput:
@@ -878,7 +874,7 @@ class LatentsToImageInvocation(BaseInvocation):
             node_id=self.id,
             session_id=context.graph_execution_state_id,
             is_intermediate=self.is_intermediate,
-            metadata=self.metadata.model_dump() if self.metadata else None,
+            metadata=self.metadata,
             workflow=self.workflow,
         )
 

invokeai/app/invocations/metadata.py

@@ -1,13 +1,17 @@
-from typing import Optional
+from typing import Any, Literal, Optional, Union
 
-from pydantic import Field
+from pydantic import BaseModel, ConfigDict, Field
 
 from invokeai.app.invocations.baseinvocation import (
     BaseInvocation,
     BaseInvocationOutput,
+    FieldDescriptions,
     InputField,
     InvocationContext,
+    MetadataField,
+    MetadataItemField,
     OutputField,
+    UIType,
     invocation,
     invocation_output,
 )
@@ -16,116 +20,100 @@ from invokeai.app.invocations.ip_adapter import IPAdapterModelField
 from invokeai.app.invocations.model import LoRAModelField, MainModelField, VAEModelField
 from invokeai.app.invocations.primitives import ImageField
 from invokeai.app.invocations.t2i_adapter import T2IAdapterField
-from invokeai.app.util.model_exclude_null import BaseModelExcludeNull
 
 from ...version import __version__
 
 
-class LoRAMetadataField(BaseModelExcludeNull):
+class LoRAMetadataField(BaseModel):
-    """LoRA metadata for an image generated in InvokeAI."""
+    """LoRA Metadata Field"""
 
-    lora: LoRAModelField = Field(description="The LoRA model")
+    lora: LoRAModelField = Field(description=FieldDescriptions.lora_model)
-    weight: float = Field(description="The weight of the LoRA model")
+    weight: float = Field(description=FieldDescriptions.lora_weight)
 
 
-class IPAdapterMetadataField(BaseModelExcludeNull):
+class IPAdapterMetadataField(BaseModel):
+    """IP Adapter Field, minus the CLIP Vision Encoder model"""
+
     image: ImageField = Field(description="The IP-Adapter image prompt.")
-    ip_adapter_model: IPAdapterModelField = Field(description="The IP-Adapter model to use.")
+    ip_adapter_model: IPAdapterModelField = Field(
-    weight: float = Field(description="The weight of the IP-Adapter model")
+        description="The IP-Adapter model.",
+    )
+    weight: Union[float, list[float]] = Field(
+        default=1,
+        ge=0,
+        description="The weight given to the IP-Adapter",
+    )
     begin_step_percent: float = Field(
-        default=0, ge=0, le=1, description="When the IP-Adapter is first applied (% of total steps)"
+        default=0, ge=-1, le=2, description="When the IP-Adapter is first applied (% of total steps)"
     )
     end_step_percent: float = Field(
         default=1, ge=0, le=1, description="When the IP-Adapter is last applied (% of total steps)"
     )
 
 
-class CoreMetadata(BaseModelExcludeNull):
+@invocation_output("metadata_item_output")
-    """Core generation metadata for an image generated in InvokeAI."""
+class MetadataItemOutput(BaseInvocationOutput):
+    """Metadata Item Output"""
 
-    app_version: str = Field(default=__version__, description="The version of InvokeAI used to generate this image")
+    item: MetadataItemField = OutputField(description="Metadata Item")
-    generation_mode: Optional[str] = Field(
+
-        default=None,
+
-        description="The generation mode that output this image",
+@invocation("metadata_item", title="Metadata Item", tags=["metadata"], category="metadata", version="1.0.0")
-    )
+class MetadataItemInvocation(BaseInvocation):
-    created_by: Optional[str] = Field(default=None, description="The name of the creator of the image")
+    """Used to create an arbitrary metadata item. Provide "label" and make a connection to "value" to store that data as the value."""
-    positive_prompt: Optional[str] = Field(default=None, description="The positive prompt parameter")
+
-    negative_prompt: Optional[str] = Field(default=None, description="The negative prompt parameter")
+    label: str = InputField(description=FieldDescriptions.metadata_item_label)
-    width: Optional[int] = Field(default=None, description="The width parameter")
+    value: Any = InputField(description=FieldDescriptions.metadata_item_value, ui_type=UIType.Any)
-    height: Optional[int] = Field(default=None, description="The height parameter")
+
-    seed: Optional[int] = Field(default=None, description="The seed used for noise generation")
+    def invoke(self, context: InvocationContext) -> MetadataItemOutput:
-    rand_device: Optional[str] = Field(default=None, description="The device used for random number generation")
+        return MetadataItemOutput(item=MetadataItemField(label=self.label, value=self.value))
-    cfg_scale: Optional[float] = Field(default=None, description="The classifier-free guidance scale parameter")
+
-    steps: Optional[int] = Field(default=None, description="The number of steps used for inference")
+
-    scheduler: Optional[str] = Field(default=None, description="The scheduler used for inference")
+@invocation_output("metadata_output")
-    clip_skip: Optional[int] = Field(
+class MetadataOutput(BaseInvocationOutput):
-        default=None,
+    metadata: MetadataField = OutputField(description="Metadata Dict")
-        description="The number of skipped CLIP layers",
+
-    )
+
-    model: Optional[MainModelField] = Field(default=None, description="The main model used for inference")
+@invocation("metadata", title="Metadata", tags=["metadata"], category="metadata", version="1.0.0")
-    controlnets: Optional[list[ControlField]] = Field(default=None, description="The ControlNets used for inference")
+class MetadataInvocation(BaseInvocation):
-    ipAdapters: Optional[list[IPAdapterMetadataField]] = Field(
+    """Takes a MetadataItem or collection of MetadataItems and outputs a MetadataDict."""
-        default=None, description="The IP Adapters used for inference"
+
-    )
+    items: Union[list[MetadataItemField], MetadataItemField] = InputField(
-    t2iAdapters: Optional[list[T2IAdapterField]] = Field(default=None, description="The IP Adapters used for inference")
+        description=FieldDescriptions.metadata_item_polymorphic
-    loras: Optional[list[LoRAMetadataField]] = Field(default=None, description="The LoRAs used for inference")
-    vae: Optional[VAEModelField] = Field(
-        default=None,
-        description="The VAE used for decoding, if the main model's default was not used",
     )
 
-    # Latents-to-Latents
+    def invoke(self, context: InvocationContext) -> MetadataOutput:
-    strength: Optional[float] = Field(
+        if isinstance(self.items, MetadataItemField):
-        default=None,
+            # single metadata item
-        description="The strength used for latents-to-latents",
+            data = {self.items.label: self.items.value}
-    )
+        else:
-    init_image: Optional[str] = Field(default=None, description="The name of the initial image")
+            # collection of metadata items
+            data = {item.label: item.value for item in self.items}
 
-    # SDXL
+        # add app version
-    positive_style_prompt: Optional[str] = Field(default=None, description="The positive style prompt parameter")
+        data.update({"app_version": __version__})
-    negative_style_prompt: Optional[str] = Field(default=None, description="The negative style prompt parameter")
+        return MetadataOutput(metadata=MetadataField.model_validate(data))
-
-    # SDXL Refiner
-    refiner_model: Optional[MainModelField] = Field(default=None, description="The SDXL Refiner model used")
-    refiner_cfg_scale: Optional[float] = Field(
-        default=None,
-        description="The classifier-free guidance scale parameter used for the refiner",
-    )
-    refiner_steps: Optional[int] = Field(default=None, description="The number of steps used for the refiner")
-    refiner_scheduler: Optional[str] = Field(default=None, description="The scheduler used for the refiner")
-    refiner_positive_aesthetic_score: Optional[float] = Field(
-        default=None, description="The aesthetic score used for the refiner"
-    )
-    refiner_negative_aesthetic_score: Optional[float] = Field(
-        default=None, description="The aesthetic score used for the refiner"
-    )
-    refiner_start: Optional[float] = Field(default=None, description="The start value used for refiner denoising")
 
 
-class ImageMetadata(BaseModelExcludeNull):
+@invocation("merge_metadata", title="Metadata Merge", tags=["metadata"], category="metadata", version="1.0.0")
-    """An image's generation metadata"""
+class MergeMetadataInvocation(BaseInvocation):
+    """Merged a collection of MetadataDict into a single MetadataDict."""
 
-    metadata: Optional[dict] = Field(
+    collection: list[MetadataField] = InputField(description=FieldDescriptions.metadata_collection)
-        default=None,
+
-        description="The image's core metadata, if it was created in the Linear or Canvas UI",
+    def invoke(self, context: InvocationContext) -> MetadataOutput:
-    )
+        data = {}
-    graph: Optional[dict] = Field(default=None, description="The graph that created the image")
+        for item in self.collection:
+            data.update(item.model_dump())
+
+        return MetadataOutput(metadata=MetadataField.model_validate(data))
 
 
-@invocation_output("metadata_accumulator_output")
+@invocation("core_metadata", title="Core Metadata", tags=["metadata"], category="metadata", version="1.0.0")
-class MetadataAccumulatorOutput(BaseInvocationOutput):
+class CoreMetadataInvocation(BaseInvocation):
-    """The output of the MetadataAccumulator node"""
+    """Collects core generation metadata into a MetadataField"""
 
-    metadata: CoreMetadata = OutputField(description="The core metadata for the image")
+    generation_mode: Literal["txt2img", "img2img", "inpaint", "outpaint"] = InputField(
-
-
-@invocation(
-    "metadata_accumulator", title="Metadata Accumulator", tags=["metadata"], category="metadata", version="1.0.0"
-)
-class MetadataAccumulatorInvocation(BaseInvocation):
-    """Outputs a Core Metadata Object"""
-
-    generation_mode: Optional[str] = InputField(
         default=None,
         description="The generation mode that output this image",
     )
@@ -138,6 +126,8 @@ class MetadataAccumulatorInvocation(BaseInvocation):
     cfg_scale: Optional[float] = InputField(default=None, description="The classifier-free guidance scale parameter")
     steps: Optional[int] = InputField(default=None, description="The number of steps used for inference")
     scheduler: Optional[str] = InputField(default=None, description="The scheduler used for inference")
+    seamless_x: Optional[bool] = InputField(default=None, description="Whether seamless tiling was used on the X axis")
+    seamless_y: Optional[bool] = InputField(default=None, description="Whether seamless tiling was used on the Y axis")
     clip_skip: Optional[int] = InputField(
         default=None,
         description="The number of skipped CLIP layers",
@@ -220,7 +210,13 @@ class MetadataAccumulatorInvocation(BaseInvocation):
         description="The start value used for refiner denoising",
     )
 
-    def invoke(self, context: InvocationContext) -> MetadataAccumulatorOutput:
+    def invoke(self, context: InvocationContext) -> MetadataOutput:
         """Collects and outputs a CoreMetadata object"""
 
-        return MetadataAccumulatorOutput(metadata=CoreMetadata(**self.model_dump()))
+        return MetadataOutput(
+            metadata=MetadataField.model_validate(
+                self.model_dump(exclude_none=True, exclude={"id", "type", "is_intermediate", "use_cache"})
+            )
+        )
+
+    model_config = ConfigDict(extra="allow")

invokeai/app/invocations/onnx.py

@@ -4,7 +4,7 @@ import inspect
 import re
 
 # from contextlib import ExitStack
-from typing import List, Literal, Optional, Union
+from typing import List, Literal, Union
 
 import numpy as np
 import torch
@@ -12,7 +12,6 @@ from diffusers.image_processor import VaeImageProcessor
 from pydantic import BaseModel, ConfigDict, Field, field_validator
 from tqdm import tqdm
 
-from invokeai.app.invocations.metadata import CoreMetadata
 from invokeai.app.invocations.primitives import ConditioningField, ConditioningOutput, ImageField, ImageOutput
 from invokeai.app.services.image_records.image_records_common import ImageCategory, ResourceOrigin
 from invokeai.app.util.step_callback import stable_diffusion_step_callback
@@ -31,6 +30,8 @@ from .baseinvocation import (
     OutputField,
     UIComponent,
     UIType,
+    WithMetadata,
+    WithWorkflow,
     invocation,
     invocation_output,
 )
@@ -327,7 +328,7 @@ class ONNXTextToLatentsInvocation(BaseInvocation):
     category="image",
     version="1.0.0",
 )
-class ONNXLatentsToImageInvocation(BaseInvocation):
+class ONNXLatentsToImageInvocation(BaseInvocation, WithMetadata, WithWorkflow):
     """Generates an image from latents."""
 
     latents: LatentsField = InputField(
@@ -338,11 +339,6 @@ class ONNXLatentsToImageInvocation(BaseInvocation):
         description=FieldDescriptions.vae,
         input=Input.Connection,
     )
-    metadata: Optional[CoreMetadata] = InputField(
-        default=None,
-        description=FieldDescriptions.core_metadata,
-        ui_hidden=True,
-    )
     # tiled: bool = InputField(default=False, description="Decode latents by overlaping tiles(less memory consumption)")
 
     def invoke(self, context: InvocationContext) -> ImageOutput:
@@ -381,7 +377,7 @@ class ONNXLatentsToImageInvocation(BaseInvocation):
             node_id=self.id,
             session_id=context.graph_execution_state_id,
             is_intermediate=self.is_intermediate,
-            metadata=self.metadata.model_dump() if self.metadata else None,
+            metadata=self.metadata,
             workflow=self.workflow,
         )
 

invokeai/app/invocations/primitives.py

@@ -251,7 +251,9 @@ class ImageCollectionOutput(BaseInvocationOutput):
 
 
 @invocation("image", title="Image Primitive", tags=["primitives", "image"], category="primitives", version="1.0.0")
-class ImageInvocation(BaseInvocation):
+class ImageInvocation(
+    BaseInvocation,
+):
     """An image primitive value"""
 
     image: ImageField = InputField(description="The image to load")

invokeai/app/invocations/upscale.py

@@ -14,7 +14,7 @@ from invokeai.app.invocations.primitives import ImageField, ImageOutput
 from invokeai.app.services.image_records.image_records_common import ImageCategory, ResourceOrigin
 from invokeai.backend.util.devices import choose_torch_device
 
-from .baseinvocation import BaseInvocation, InputField, InvocationContext, invocation
+from .baseinvocation import BaseInvocation, InputField, InvocationContext, WithMetadata, WithWorkflow, invocation
 
 # TODO: Populate this from disk?
 # TODO: Use model manager to load?
@@ -30,7 +30,7 @@ if choose_torch_device() == torch.device("mps"):
 
 
 @invocation("esrgan", title="Upscale (RealESRGAN)", tags=["esrgan", "upscale"], category="esrgan", version="1.1.0")
-class ESRGANInvocation(BaseInvocation):
+class ESRGANInvocation(BaseInvocation, WithWorkflow, WithMetadata):
     """Upscales an image using RealESRGAN."""
 
     image: ImageField = InputField(description="The input image")
@@ -123,6 +123,7 @@ class ESRGANInvocation(BaseInvocation):
             node_id=self.id,
             session_id=context.graph_execution_state_id,
             is_intermediate=self.is_intermediate,
+            metadata=self.metadata,
             workflow=self.workflow,
         )
 

invokeai/app/services/image_files/image_files_base.py

@@ -4,6 +4,9 @@ from typing import Optional
 
 from PIL.Image import Image as PILImageType
 
+from invokeai.app.invocations.metadata import MetadataField
+from invokeai.app.services.workflow_records.workflow_records_common import WorkflowField
+
 
 class ImageFileStorageBase(ABC):
     """Low-level service responsible for storing and retrieving image files."""
@@ -30,8 +33,8 @@ class ImageFileStorageBase(ABC):
         self,
         image: PILImageType,
         image_name: str,
-        metadata: Optional[dict] = None,
+        metadata: Optional[MetadataField] = None,
-        workflow: Optional[str] = None,
+        workflow: Optional[WorkflowField] = None,
         thumbnail_size: int = 256,
     ) -> None:
         """Saves an image and a 256x256 WEBP thumbnail. Returns a tuple of the image name, thumbnail name, and created timestamp."""

invokeai/app/services/image_files/image_files_disk.py

@@ -1,5 +1,4 @@
 # Copyright (c) 2022 Kyle Schouviller (https://github.com/kyle0654) and the InvokeAI Team
-import json
 from pathlib import Path
 from queue import Queue
 from typing import Dict, Optional, Union
@@ -8,7 +7,9 @@ from PIL import Image, PngImagePlugin
 from PIL.Image import Image as PILImageType
 from send2trash import send2trash
 
+from invokeai.app.invocations.metadata import MetadataField
 from invokeai.app.services.invoker import Invoker
+from invokeai.app.services.workflow_records.workflow_records_common import WorkflowField
 from invokeai.app.util.thumbnails import get_thumbnail_name, make_thumbnail
 
 from .image_files_base import ImageFileStorageBase
@@ -55,8 +56,8 @@ class DiskImageFileStorage(ImageFileStorageBase):
         self,
         image: PILImageType,
         image_name: str,
-        metadata: Optional[dict] = None,
+        metadata: Optional[MetadataField] = None,
-        workflow: Optional[str] = None,
+        workflow: Optional[WorkflowField] = None,
         thumbnail_size: int = 256,
     ) -> None:
         try:
@@ -67,9 +68,9 @@ class DiskImageFileStorage(ImageFileStorageBase):
 
             if metadata is not None or workflow is not None:
                 if metadata is not None:
-                    pnginfo.add_text("invokeai_metadata", json.dumps(metadata))
+                    pnginfo.add_text("invokeai_metadata", metadata.model_dump_json())
                 if workflow is not None:
-                    pnginfo.add_text("invokeai_workflow", workflow)
+                    pnginfo.add_text("invokeai_workflow", workflow.model_dump_json())
             else:
                 # For uploaded images, we want to retain metadata. PIL strips it on save; manually add it back
                 # TODO: retain non-invokeai metadata on save...

invokeai/app/services/image_records/image_records_base.py

@@ -2,6 +2,7 @@ from abc import ABC, abstractmethod
 from datetime import datetime
 from typing import Optional
 
+from invokeai.app.invocations.metadata import MetadataField
 from invokeai.app.services.shared.pagination import OffsetPaginatedResults
 
 from .image_records_common import ImageCategory, ImageRecord, ImageRecordChanges, ResourceOrigin
@@ -18,7 +19,7 @@ class ImageRecordStorageBase(ABC):
         pass
 
     @abstractmethod
-    def get_metadata(self, image_name: str) -> Optional[dict]:
+    def get_metadata(self, image_name: str) -> Optional[MetadataField]:
         """Gets an image's metadata'."""
         pass
 
@@ -78,7 +79,8 @@ class ImageRecordStorageBase(ABC):
         starred: Optional[bool] = False,
         session_id: Optional[str] = None,
         node_id: Optional[str] = None,
-        metadata: Optional[dict] = None,
+        metadata: Optional[MetadataField] = None,
+        workflow_id: Optional[str] = None,
     ) -> datetime:
         """Saves an image record."""
         pass

invokeai/app/services/image_records/image_records_common.py

@@ -100,6 +100,7 @@ IMAGE_DTO_COLS = ", ".join(
                 "width",
                 "height",
                 "session_id",
+                "workflow_id",
                 "node_id",
                 "is_intermediate",
                 "created_at",
@@ -140,6 +141,11 @@ class ImageRecord(BaseModelExcludeNull):
         description="The session ID that generated this image, if it is a generated image.",
     )
     """The session ID that generated this image, if it is a generated image."""
+    workflow_id: Optional[str] = Field(
+        default=None,
+        description="The workflow that generated this image.",
+    )
+    """The workflow that generated this image."""
     node_id: Optional[str] = Field(
         default=None,
         description="The node ID that generated this image, if it is a generated image.",
@@ -184,6 +190,7 @@ def deserialize_image_record(image_dict: dict) -> ImageRecord:
     width = image_dict.get("width", 0)
     height = image_dict.get("height", 0)
     session_id = image_dict.get("session_id", None)
+    workflow_id = image_dict.get("workflow_id", None)
     node_id = image_dict.get("node_id", None)
     created_at = image_dict.get("created_at", get_iso_timestamp())
     updated_at = image_dict.get("updated_at", get_iso_timestamp())
@@ -198,6 +205,7 @@ def deserialize_image_record(image_dict: dict) -> ImageRecord:
         width=width,
         height=height,
         session_id=session_id,
+        workflow_id=workflow_id,
         node_id=node_id,
         created_at=created_at,
         updated_at=updated_at,

invokeai/app/services/image_records/image_records_sqlite.py

@@ -1,9 +1,9 @@
-import json
 import sqlite3
 import threading
 from datetime import datetime
 from typing import Optional, Union, cast
 
+from invokeai.app.invocations.baseinvocation import MetadataField, type_adapter_MetadataField
 from invokeai.app.services.shared.pagination import OffsetPaginatedResults
 from invokeai.app.services.shared.sqlite import SqliteDatabase
 
@@ -76,6 +76,16 @@ class SqliteImageRecordStorage(ImageRecordStorageBase):
                 """
             )
 
+        if "workflow_id" not in columns:
+            self._cursor.execute(
+                """--sql
+                ALTER TABLE images
+                ADD COLUMN workflow_id TEXT;
+                -- TODO: This requires a migration:
+                -- FOREIGN KEY (workflow_id) REFERENCES workflows (workflow_id) ON DELETE SET NULL;
+                """
+            )
+
         # Create the `images` table indices.
         self._cursor.execute(
             """--sql
@@ -141,22 +151,26 @@ class SqliteImageRecordStorage(ImageRecordStorageBase):
 
         return deserialize_image_record(dict(result))
 
-    def get_metadata(self, image_name: str) -> Optional[dict]:
+    def get_metadata(self, image_name: str) -> Optional[MetadataField]:
         try:
             self._lock.acquire()
 
             self._cursor.execute(
                 """--sql
-                SELECT images.metadata FROM images
+                SELECT metadata FROM images
                 WHERE image_name = ?;
                 """,
                 (image_name,),
             )
 
             result = cast(Optional[sqlite3.Row], self._cursor.fetchone())
-            if not result or not result[0]:
+
-                return None
+            if not result:
-            return json.loads(result[0])
+                raise ImageRecordNotFoundException
+
+            as_dict = dict(result)
+            metadata_raw = cast(Optional[str], as_dict.get("metadata", None))
+            return type_adapter_MetadataField.validate_json(metadata_raw) if metadata_raw is not None else None
         except sqlite3.Error as e:
             self._conn.rollback()
             raise ImageRecordNotFoundException from e
@@ -408,10 +422,11 @@ class SqliteImageRecordStorage(ImageRecordStorageBase):
         starred: Optional[bool] = False,
         session_id: Optional[str] = None,
         node_id: Optional[str] = None,
-        metadata: Optional[dict] = None,
+        metadata: Optional[MetadataField] = None,
+        workflow_id: Optional[str] = None,
     ) -> datetime:
         try:
-            metadata_json = None if metadata is None else json.dumps(metadata)
+            metadata_json = metadata.model_dump_json() if metadata is not None else None
             self._lock.acquire()
             self._cursor.execute(
                 """--sql
@@ -424,10 +439,11 @@ class SqliteImageRecordStorage(ImageRecordStorageBase):
                     node_id,
                     session_id,
                     metadata,
+                    workflow_id,
                     is_intermediate,
                     starred
                     )
-                VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?);
+                VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?);
                 """,
                 (
                     image_name,
@@ -438,6 +454,7 @@ class SqliteImageRecordStorage(ImageRecordStorageBase):
                     node_id,
                     session_id,
                     metadata_json,
+                    workflow_id,
                     is_intermediate,
                     starred,
                 ),

invokeai/app/services/images/images_base.py

@@ -3,7 +3,7 @@ from typing import Callable, Optional
 
 from PIL.Image import Image as PILImageType
 
-from invokeai.app.invocations.metadata import ImageMetadata
+from invokeai.app.invocations.metadata import MetadataField
 from invokeai.app.services.image_records.image_records_common import (
     ImageCategory,
     ImageRecord,
@@ -12,6 +12,7 @@ from invokeai.app.services.image_records.image_records_common import (
 )
 from invokeai.app.services.images.images_common import ImageDTO
 from invokeai.app.services.shared.pagination import OffsetPaginatedResults
+from invokeai.app.services.workflow_records.workflow_records_common import WorkflowField
 
 
 class ImageServiceABC(ABC):
@@ -50,8 +51,8 @@ class ImageServiceABC(ABC):
         session_id: Optional[str] = None,
         board_id: Optional[str] = None,
         is_intermediate: Optional[bool] = False,
-        metadata: Optional[dict] = None,
+        metadata: Optional[MetadataField] = None,
-        workflow: Optional[str] = None,
+        workflow: Optional[WorkflowField] = None,
     ) -> ImageDTO:
         """Creates an image, storing the file and its metadata."""
         pass
@@ -81,7 +82,7 @@ class ImageServiceABC(ABC):
         pass
 
     @abstractmethod
-    def get_metadata(self, image_name: str) -> ImageMetadata:
+    def get_metadata(self, image_name: str) -> Optional[MetadataField]:
         """Gets an image's metadata."""
         pass
 

invokeai/app/services/images/images_common.py

@@ -25,8 +25,6 @@ class ImageDTO(ImageRecord, ImageUrlsDTO):
     )
     """The id of the board the image belongs to, if one exists."""
 
-    pass
-
 
 def image_record_to_dto(
     image_record: ImageRecord,

invokeai/app/services/images/images_default.py

@@ -2,10 +2,10 @@ from typing import Optional
 
 from PIL.Image import Image as PILImageType
 
-from invokeai.app.invocations.metadata import ImageMetadata
+from invokeai.app.invocations.metadata import MetadataField
 from invokeai.app.services.invoker import Invoker
 from invokeai.app.services.shared.pagination import OffsetPaginatedResults
-from invokeai.app.util.metadata import get_metadata_graph_from_raw_session
+from invokeai.app.services.workflow_records.workflow_records_common import WorkflowField
 
 from ..image_files.image_files_common import (
     ImageFileDeleteException,
@@ -42,8 +42,8 @@ class ImageService(ImageServiceABC):
         session_id: Optional[str] = None,
         board_id: Optional[str] = None,
         is_intermediate: Optional[bool] = False,
-        metadata: Optional[dict] = None,
+        metadata: Optional[MetadataField] = None,
-        workflow: Optional[str] = None,
+        workflow: Optional[WorkflowField] = None,
     ) -> ImageDTO:
         if image_origin not in ResourceOrigin:
             raise InvalidOriginException
@@ -56,6 +56,12 @@ class ImageService(ImageServiceABC):
         (width, height) = image.size
 
         try:
+            if workflow is not None:
+                created_workflow = self.__invoker.services.workflow_records.create(workflow)
+                workflow_id = created_workflow.model_dump()["id"]
+            else:
+                workflow_id = None
+
             # TODO: Consider using a transaction here to ensure consistency between storage and database
             self.__invoker.services.image_records.save(
                 # Non-nullable fields
@@ -69,6 +75,7 @@ class ImageService(ImageServiceABC):
                 # Nullable fields
                 node_id=node_id,
                 metadata=metadata,
+                workflow_id=workflow_id,
                 session_id=session_id,
             )
             if board_id is not None:
@@ -146,25 +153,9 @@ class ImageService(ImageServiceABC):
             self.__invoker.services.logger.error("Problem getting image DTO")
             raise e
 
-    def get_metadata(self, image_name: str) -> ImageMetadata:
+    def get_metadata(self, image_name: str) -> Optional[MetadataField]:
         try:
-            image_record = self.__invoker.services.image_records.get(image_name)
+            return self.__invoker.services.image_records.get_metadata(image_name)
-            metadata = self.__invoker.services.image_records.get_metadata(image_name)
-
-            if not image_record.session_id:
-                return ImageMetadata(metadata=metadata)
-
-            session_raw = self.__invoker.services.graph_execution_manager.get_raw(image_record.session_id)
-            graph = None
-
-            if session_raw:
-                try:
-                    graph = get_metadata_graph_from_raw_session(session_raw)
-                except Exception as e:
-                    self.__invoker.services.logger.warn(f"Failed to parse session graph: {e}")
-                    graph = None
-
-            return ImageMetadata(graph=graph, metadata=metadata)
         except ImageRecordNotFoundException:
             self.__invoker.services.logger.error("Image record not found")
             raise

invokeai/app/services/shared/graph.py

@@ -439,6 +439,14 @@ class Graph(BaseModel):
         except Exception as e:
             raise UnknownGraphValidationError(f"Problem validating graph {e}") from e
 
+    def _is_destination_field_Any(self, edge: Edge) -> bool:
+        """Checks if the destination field for an edge is of type typing.Any"""
+        return get_input_field(self.get_node(edge.destination.node_id), edge.destination.field) == Any
+
+    def _is_destination_field_list_of_Any(self, edge: Edge) -> bool:
+        """Checks if the destination field for an edge is of type typing.Any"""
+        return get_input_field(self.get_node(edge.destination.node_id), edge.destination.field) == list[Any]
+
     def _validate_edge(self, edge: Edge):
         """Validates that a new edge doesn't create a cycle in the graph"""
 
@@ -491,8 +499,19 @@ class Graph(BaseModel):
                     f"Collector output type does not match collector input type: {edge.source.node_id}.{edge.source.field} to {edge.destination.node_id}.{edge.destination.field}"
                 )
 
-        # Validate if collector output type matches input type (if this edge results in both being set)
+        # Validate that we are not connecting collector to iterator (currently unsupported)
-        if isinstance(from_node, CollectInvocation) and edge.source.field == "collection":
+        if isinstance(from_node, CollectInvocation) and isinstance(to_node, IterateInvocation):
+            raise InvalidEdgeError(
+                f"Cannot connect collector to iterator: {edge.source.node_id}.{edge.source.field} to {edge.destination.node_id}.{edge.destination.field}"
+            )
+
+        # Validate if collector output type matches input type (if this edge results in both being set) - skip if the destination field is not Any or list[Any]
+        if (
+            isinstance(from_node, CollectInvocation)
+            and edge.source.field == "collection"
+            and not self._is_destination_field_list_of_Any(edge)
+            and not self._is_destination_field_Any(edge)
+        ):
             if not self._is_collector_connection_valid(edge.source.node_id, new_output=edge.destination):
                 raise InvalidEdgeError(
                     f"Collector input type does not match collector output type: {edge.source.node_id}.{edge.source.field} to {edge.destination.node_id}.{edge.destination.field}"
@@ -725,16 +744,15 @@ class Graph(BaseModel):
         # Get the input root type
         input_root_type = next(t[0] for t in type_degrees if t[1] == 0)  # type: ignore
 
-        # Verify that all outputs are lists
-        # if not all((get_origin(f) == list for f in output_fields)):
-        #     return False
-
         # Verify that all outputs are lists
         if not all(is_list_or_contains_list(f) for f in output_fields):
             return False
 
         # Verify that all outputs match the input type (are a base class or the same class)
-        if not all((issubclass(input_root_type, get_args(f)[0]) for f in output_fields)):
+        if not all(
+            is_union_subtype(input_root_type, get_args(f)[0]) or issubclass(input_root_type, get_args(f)[0])
+            for f in output_fields
+        ):
             return False
 
         return True

invokeai/frontend/web/src/features/gallery/components/CurrentImage/CurrentImageButtons.tsx

@@ -27,7 +27,7 @@ import {
   setShouldShowImageDetails,
   setShouldShowProgressInViewer,
 } from 'features/ui/store/uiSlice';
-import { memo, useCallback, useMemo } from 'react';
+import { memo, useCallback } from 'react';
 import { useHotkeys } from 'react-hotkeys-hook';
 import { useTranslation } from 'react-i18next';
 import {
@@ -40,11 +40,13 @@ import {
 import { FaCircleNodes, FaEllipsis } from 'react-icons/fa6';
 import {
   useGetImageDTOQuery,
-  useGetImageMetadataFromFileQuery,
+  useGetImageMetadataQuery,
 } from 'services/api/endpoints/images';
 import { menuListMotionProps } from 'theme/components/menu';
+import { useDebounce } from 'use-debounce';
 import { sentImageToImg2Img } from '../../store/actions';
 import SingleSelectionMenuItems from '../ImageContextMenu/SingleSelectionMenuItems';
+import { useGetWorkflowQuery } from 'services/api/endpoints/workflows';
 
 const currentImageButtonsSelector = createSelector(
   [stateSelector, activeTabNameSelector],
@@ -89,7 +91,6 @@ const CurrentImageButtons = () => {
     shouldShowImageDetails,
     lastSelectedImage,
     shouldShowProgressInViewer,
-    shouldFetchMetadataFromApi,
   } = useAppSelector(currentImageButtonsSelector);
 
   const isUpscalingEnabled = useFeatureStatus('upscaling').isFeatureEnabled;
@@ -104,23 +105,17 @@ const CurrentImageButtons = () => {
     lastSelectedImage?.image_name ?? skipToken
   );
 
-  const getMetadataArg = useMemo(() => {
+  const [debouncedImageName] = useDebounce(lastSelectedImage?.image_name, 300);
-    if (lastSelectedImage) {
+  const [debouncedWorkflowId] = useDebounce(
-      return { image: lastSelectedImage, shouldFetchMetadataFromApi };
+    lastSelectedImage?.workflow_id,
-    } else {
+    300
-      return skipToken;
+  );
-    }
-  }, [lastSelectedImage, shouldFetchMetadataFromApi]);
 
-  const { metadata, workflow, isLoading } = useGetImageMetadataFromFileQuery(
+  const { data: metadata, isLoading: isLoadingMetadata } =
-    getMetadataArg,
+    useGetImageMetadataQuery(debouncedImageName ?? skipToken);
-    {
+
-      selectFromResult: (res) => ({
+  const { data: workflow, isLoading: isLoadingWorkflow } = useGetWorkflowQuery(
-        isLoading: res.isFetching,
+    debouncedWorkflowId ?? skipToken
-        metadata: res?.currentData?.metadata,
-        workflow: res?.currentData?.workflow,
-      }),
-    }
   );
 
   const handleLoadWorkflow = useCallback(() => {
@@ -257,7 +252,7 @@ const CurrentImageButtons = () => {
 
         <ButtonGroup isAttached={true} isDisabled={shouldDisableToolbarButtons}>
           <IAIIconButton
-            isLoading={isLoading}
+            isLoading={isLoadingWorkflow}
             icon={<FaCircleNodes />}
             tooltip={`${t('nodes.loadWorkflow')} (W)`}
             aria-label={`${t('nodes.loadWorkflow')} (W)`}
@@ -265,7 +260,7 @@ const CurrentImageButtons = () => {
             onClick={handleLoadWorkflow}
           />
           <IAIIconButton
-            isLoading={isLoading}
+            isLoading={isLoadingMetadata}
             icon={<FaQuoteRight />}
             tooltip={`${t('parameters.usePrompt')} (P)`}
             aria-label={`${t('parameters.usePrompt')} (P)`}
@@ -273,7 +268,7 @@ const CurrentImageButtons = () => {
             onClick={handleUsePrompt}
           />
           <IAIIconButton
-            isLoading={isLoading}
+            isLoading={isLoadingMetadata}
             icon={<FaSeedling />}
             tooltip={`${t('parameters.useSeed')} (S)`}
             aria-label={`${t('parameters.useSeed')} (S)`}
@@ -281,7 +276,7 @@ const CurrentImageButtons = () => {
             onClick={handleUseSeed}
           />
           <IAIIconButton
-            isLoading={isLoading}
+            isLoading={isLoadingMetadata}
             icon={<FaAsterisk />}
             tooltip={`${t('parameters.useAll')} (A)`}
             aria-label={`${t('parameters.useAll')} (A)`}

invokeai/frontend/web/src/features/gallery/components/ImageContextMenu/SingleSelectionMenuItems.tsx

@@ -1,8 +1,9 @@
 import { Flex, MenuItem, Spinner } from '@chakra-ui/react';
 import { useStore } from '@nanostores/react';
+import { skipToken } from '@reduxjs/toolkit/dist/query';
 import { useAppToaster } from 'app/components/Toaster';
 import { $customStarUI } from 'app/store/nanostores/customStarUI';
-import { useAppDispatch, useAppSelector } from 'app/store/storeHooks';
+import { useAppDispatch } from 'app/store/storeHooks';
 import { setInitialCanvasImage } from 'features/canvas/store/canvasSlice';
 import {
   imagesToChangeSelected,
@@ -32,12 +33,13 @@ import {
 import { FaCircleNodes } from 'react-icons/fa6';
 import { MdStar, MdStarBorder } from 'react-icons/md';
 import {
-  useGetImageMetadataFromFileQuery,
+  useGetImageMetadataQuery,
   useStarImagesMutation,
   useUnstarImagesMutation,
 } from 'services/api/endpoints/images';
+import { useGetWorkflowQuery } from 'services/api/endpoints/workflows';
 import { ImageDTO } from 'services/api/types';
-import { configSelector } from '../../../system/store/configSelectors';
+import { useDebounce } from 'use-debounce';
 import { sentImageToCanvas, sentImageToImg2Img } from '../../store/actions';
 
 type SingleSelectionMenuItemsProps = {
@@ -53,18 +55,16 @@ const SingleSelectionMenuItems = (props: SingleSelectionMenuItemsProps) => {
   const toaster = useAppToaster();
 
   const isCanvasEnabled = useFeatureStatus('unifiedCanvas').isFeatureEnabled;
-  const { shouldFetchMetadataFromApi } = useAppSelector(configSelector);
   const customStarUi = useStore($customStarUI);
 
-  const { metadata, workflow, isLoading } = useGetImageMetadataFromFileQuery(
+  const [debouncedImageName] = useDebounce(imageDTO?.image_name, 300);
-    { image: imageDTO, shouldFetchMetadataFromApi },
+  const [debouncedWorkflowId] = useDebounce(imageDTO?.workflow_id, 300);
-    {
+
-      selectFromResult: (res) => ({
+  const { data: metadata, isLoading: isLoadingMetadata } =
-        isLoading: res.isFetching,
+    useGetImageMetadataQuery(debouncedImageName ?? skipToken);
-        metadata: res?.currentData?.metadata,
+
-        workflow: res?.currentData?.workflow,
+  const { data: workflow, isLoading: isLoadingWorkflow } = useGetWorkflowQuery(
-      }),
+    debouncedWorkflowId ?? skipToken
-    }
   );
 
   const [starImages] = useStarImagesMutation();
@@ -181,17 +181,17 @@ const SingleSelectionMenuItems = (props: SingleSelectionMenuItemsProps) => {
         {t('parameters.downloadImage')}
       </MenuItem>
       <MenuItem
-        icon={isLoading ? <SpinnerIcon /> : <FaCircleNodes />}
+        icon={isLoadingWorkflow ? <SpinnerIcon /> : <FaCircleNodes />}
         onClickCapture={handleLoadWorkflow}
-        isDisabled={isLoading || !workflow}
+        isDisabled={isLoadingWorkflow || !workflow}
       >
         {t('nodes.loadWorkflow')}
       </MenuItem>
       <MenuItem
-        icon={isLoading ? <SpinnerIcon /> : <FaQuoteRight />}
+        icon={isLoadingMetadata ? <SpinnerIcon /> : <FaQuoteRight />}
         onClickCapture={handleRecallPrompt}
         isDisabled={
-          isLoading ||
+          isLoadingMetadata ||
           (metadata?.positive_prompt === undefined &&
             metadata?.negative_prompt === undefined)
         }
@@ -199,16 +199,16 @@ const SingleSelectionMenuItems = (props: SingleSelectionMenuItemsProps) => {
         {t('parameters.usePrompt')}
       </MenuItem>
       <MenuItem
-        icon={isLoading ? <SpinnerIcon /> : <FaSeedling />}
+        icon={isLoadingMetadata ? <SpinnerIcon /> : <FaSeedling />}
         onClickCapture={handleRecallSeed}
-        isDisabled={isLoading || metadata?.seed === undefined}
+        isDisabled={isLoadingMetadata || metadata?.seed === undefined}
       >
         {t('parameters.useSeed')}
       </MenuItem>
       <MenuItem
-        icon={isLoading ? <SpinnerIcon /> : <FaAsterisk />}
+        icon={isLoadingMetadata ? <SpinnerIcon /> : <FaAsterisk />}
         onClickCapture={handleUseAllParameters}
-        isDisabled={isLoading || !metadata}
+        isDisabled={isLoadingMetadata || !metadata}
       >
         {t('parameters.useAll')}
       </MenuItem>

invokeai/frontend/web/src/features/gallery/components/ImageMetadataViewer/ImageMetadataViewer.tsx

@@ -9,16 +9,17 @@ import {
   Tabs,
   Text,
 } from '@chakra-ui/react';
+import { skipToken } from '@reduxjs/toolkit/dist/query';
 import { IAINoContentFallback } from 'common/components/IAIImageFallback';
+import ScrollableContent from 'features/nodes/components/sidePanel/ScrollableContent';
 import { memo } from 'react';
-import { useGetImageMetadataFromFileQuery } from 'services/api/endpoints/images';
+import { useTranslation } from 'react-i18next';
+import { useGetImageMetadataQuery } from 'services/api/endpoints/images';
+import { useGetWorkflowQuery } from 'services/api/endpoints/workflows';
 import { ImageDTO } from 'services/api/types';
+import { useDebounce } from 'use-debounce';
 import DataViewer from './DataViewer';
 import ImageMetadataActions from './ImageMetadataActions';
-import { useAppSelector } from '../../../../app/store/storeHooks';
-import { configSelector } from '../../../system/store/configSelectors';
-import { useTranslation } from 'react-i18next';
-import ScrollableContent from 'features/nodes/components/sidePanel/ScrollableContent';
 
 type ImageMetadataViewerProps = {
   image: ImageDTO;
@@ -32,16 +33,15 @@ const ImageMetadataViewer = ({ image }: ImageMetadataViewerProps) => {
   // });
   const { t } = useTranslation();
 
-  const { shouldFetchMetadataFromApi } = useAppSelector(configSelector);
+  const [debouncedImageName] = useDebounce(image.image_name, 300);
+  const [debouncedWorkflowId] = useDebounce(image.workflow_id, 300);
 
-  const { metadata, workflow } = useGetImageMetadataFromFileQuery(
+  const { data: metadata } = useGetImageMetadataQuery(
-    { image, shouldFetchMetadataFromApi },
+    debouncedImageName ?? skipToken
-    {
+  );
-      selectFromResult: (res) => ({
+
-        metadata: res?.currentData?.metadata,
+  const { data: workflow } = useGetWorkflowQuery(
-        workflow: res?.currentData?.workflow,
+    debouncedWorkflowId ?? skipToken
-      }),
-    }
   );
 
   return (

invokeai/frontend/web/src/features/nodes/components/flow/nodes/Invocation/EmbedWorkflowCheckbox.tsx

@@ -1,13 +1,13 @@
 import { Checkbox, Flex, FormControl, FormLabel } from '@chakra-ui/react';
 import { useAppDispatch } from 'app/store/storeHooks';
 import { useEmbedWorkflow } from 'features/nodes/hooks/useEmbedWorkflow';
-import { useHasImageOutput } from 'features/nodes/hooks/useHasImageOutput';
+import { useWithWorkflow } from 'features/nodes/hooks/useWithWorkflow';
 import { nodeEmbedWorkflowChanged } from 'features/nodes/store/nodesSlice';
 import { ChangeEvent, memo, useCallback } from 'react';
 
 const EmbedWorkflowCheckbox = ({ nodeId }: { nodeId: string }) => {
   const dispatch = useAppDispatch();
-  const hasImageOutput = useHasImageOutput(nodeId);
+  const withWorkflow = useWithWorkflow(nodeId);
   const embedWorkflow = useEmbedWorkflow(nodeId);
   const handleChange = useCallback(
     (e: ChangeEvent<HTMLInputElement>) => {
@@ -21,7 +21,7 @@ const EmbedWorkflowCheckbox = ({ nodeId }: { nodeId: string }) => {
     [dispatch, nodeId]
   );
 
-  if (!hasImageOutput) {
+  if (!withWorkflow) {
     return null;
   }
 

invokeai/frontend/web/src/features/nodes/components/flow/nodes/Invocation/InvocationNodeFooter.tsx

@@ -1,11 +1,11 @@
 import { Flex } from '@chakra-ui/react';
+import { useHasImageOutput } from 'features/nodes/hooks/useHasImageOutput';
 import { DRAG_HANDLE_CLASSNAME } from 'features/nodes/types/constants';
 import { memo } from 'react';
+import { useFeatureStatus } from '../../../../../system/hooks/useFeatureStatus';
 import EmbedWorkflowCheckbox from './EmbedWorkflowCheckbox';
 import SaveToGalleryCheckbox from './SaveToGalleryCheckbox';
 import UseCacheCheckbox from './UseCacheCheckbox';
-import { useHasImageOutput } from 'features/nodes/hooks/useHasImageOutput';
-import { useFeatureStatus } from '../../../../../system/hooks/useFeatureStatus';
 
 type Props = {
   nodeId: string;

invokeai/frontend/web/src/features/nodes/hooks/useWithWorkflow.ts

@@ -0,0 +1,31 @@
+import { createSelector } from '@reduxjs/toolkit';
+import { stateSelector } from 'app/store/store';
+import { useAppSelector } from 'app/store/storeHooks';
+import { defaultSelectorOptions } from 'app/store/util/defaultMemoizeOptions';
+import { useMemo } from 'react';
+import { isInvocationNode } from '../types/types';
+
+export const useWithWorkflow = (nodeId: string) => {
+  const selector = useMemo(
+    () =>
+      createSelector(
+        stateSelector,
+        ({ nodes }) => {
+          const node = nodes.nodes.find((node) => node.id === nodeId);
+          if (!isInvocationNode(node)) {
+            return false;
+          }
+          const nodeTemplate = nodes.nodeTemplates[node?.data.type ?? ''];
+          if (!nodeTemplate) {
+            return false;
+          }
+          return nodeTemplate.withWorkflow;
+        },
+        defaultSelectorOptions
+      ),
+    [nodeId]
+  );
+
+  const withWorkflow = useAppSelector(selector);
+  return withWorkflow;
+};

invokeai/frontend/web/src/features/nodes/store/util/validateSourceAndTargetTypes.ts

@@ -69,6 +69,8 @@ export const validateSourceAndTargetTypes = (
     (sourceType === 'integer' || sourceType === 'float') &&
     targetType === 'string';
 
+  const isTargetAnyType = targetType === 'Any';
+
   return (
     isCollectionItemToNonCollection ||
     isNonCollectionToCollectionItem ||
@@ -76,6 +78,7 @@ export const validateSourceAndTargetTypes = (
     isGenericCollectionToAnyCollectionOrPolymorphic ||
     isCollectionToGenericCollection ||
     isIntToFloat ||
-    isIntOrFloatToString
+    isIntOrFloatToString ||
+    isTargetAnyType
   );
 };

invokeai/frontend/web/src/features/nodes/types/constants.ts

@@ -33,6 +33,8 @@ export const COLLECTION_TYPES: FieldType[] = [
   'ColorCollection',
   'T2IAdapterCollection',
   'IPAdapterCollection',
+  'MetadataItemCollection',
+  'MetadataCollection',
 ];
 
 export const POLYMORPHIC_TYPES: FieldType[] = [
@@ -47,6 +49,7 @@ export const POLYMORPHIC_TYPES: FieldType[] = [
   'ColorPolymorphic',
   'T2IAdapterPolymorphic',
   'IPAdapterPolymorphic',
+  'MetadataItemPolymorphic',
 ];
 
 export const MODEL_TYPES: FieldType[] = [
@@ -78,6 +81,8 @@ export const COLLECTION_MAP: FieldTypeMapWithNumber = {
   ColorField: 'ColorCollection',
   T2IAdapterField: 'T2IAdapterCollection',
   IPAdapterField: 'IPAdapterCollection',
+  MetadataItemField: 'MetadataItemCollection',
+  MetadataField: 'MetadataCollection',
 };
 export const isCollectionItemType = (
   itemType: string | undefined
@@ -97,6 +102,7 @@ export const SINGLE_TO_POLYMORPHIC_MAP: FieldTypeMapWithNumber = {
   ColorField: 'ColorPolymorphic',
   T2IAdapterField: 'T2IAdapterPolymorphic',
   IPAdapterField: 'IPAdapterPolymorphic',
+  MetadataItemField: 'MetadataItemPolymorphic',
 };
 
 export const POLYMORPHIC_TO_SINGLE_MAP: FieldTypeMap = {
@@ -111,6 +117,7 @@ export const POLYMORPHIC_TO_SINGLE_MAP: FieldTypeMap = {
   ColorPolymorphic: 'ColorField',
   T2IAdapterPolymorphic: 'T2IAdapterField',
   IPAdapterPolymorphic: 'IPAdapterField',
+  MetadataItemPolymorphic: 'MetadataItemField',
 };
 
 export const TYPES_WITH_INPUT_COMPONENTS: FieldType[] = [
@@ -144,6 +151,37 @@ export const isPolymorphicItemType = (
   Boolean(itemType && itemType in SINGLE_TO_POLYMORPHIC_MAP);
 
 export const FIELDS: Record<FieldType, FieldUIConfig> = {
+  Any: {
+    color: 'gray.500',
+    description: 'Any field type is accepted.',
+    title: 'Any',
+  },
+  MetadataField: {
+    color: 'gray.500',
+    description: 'A metadata dict.',
+    title: 'Metadata Dict',
+  },
+  MetadataCollection: {
+    color: 'gray.500',
+    description: 'A collection of metadata dicts.',
+    title: 'Metadata Dict Collection',
+  },
+  MetadataItemField: {
+    color: 'gray.500',
+    description: 'A metadata item.',
+    title: 'Metadata Item',
+  },
+  MetadataItemCollection: {
+    color: 'gray.500',
+    description: 'Any field type is accepted.',
+    title: 'Metadata Item Collection',
+  },
+  MetadataItemPolymorphic: {
+    color: 'gray.500',
+    description:
+      'MetadataItem or MetadataItemCollection field types are accepted.',
+    title: 'Metadata Item Polymorphic',
+  },
   boolean: {
     color: 'green.500',
     description: t('nodes.booleanDescription'),

invokeai/frontend/web/src/features/nodes/types/types.ts

@@ -54,6 +54,10 @@ export type InvocationTemplate = {
    * The type of this node's output
    */
   outputType: string; // TODO: generate a union of output types
+  /**
+   * Whether or not this invocation supports workflows
+   */
+  withWorkflow: boolean;
   /**
    * The invocation's version.
    */
@@ -72,6 +76,7 @@ export type FieldUIConfig = {
 
 // TODO: Get this from the OpenAPI schema? may be tricky...
 export const zFieldType = z.enum([
+  'Any',
   'BoardField',
   'boolean',
   'BooleanCollection',
@@ -109,6 +114,11 @@ export const zFieldType = z.enum([
   'LatentsPolymorphic',
   'LoRAModelField',
   'MainModelField',
+  'MetadataField',
+  'MetadataCollection',
+  'MetadataItemField',
+  'MetadataItemCollection',
+  'MetadataItemPolymorphic',
   'ONNXModelField',
   'Scheduler',
   'SDXLMainModelField',
@@ -685,6 +695,57 @@ export type CollectionItemInputFieldValue = z.infer<
   typeof zCollectionItemInputFieldValue
 >;
 
+export const zMetadataItemField = z.object({
+  label: z.string(),
+  value: z.any(),
+});
+export type MetadataItemField = z.infer<typeof zMetadataItemField>;
+
+export const zMetadataItemInputFieldValue = zInputFieldValueBase.extend({
+  type: z.literal('MetadataItemField'),
+  value: zMetadataItemField.optional(),
+});
+export type MetadataItemInputFieldValue = z.infer<
+  typeof zMetadataItemInputFieldValue
+>;
+
+export const zMetadataItemCollectionInputFieldValue =
+  zInputFieldValueBase.extend({
+    type: z.literal('MetadataItemCollection'),
+    value: z.array(zMetadataItemField).optional(),
+  });
+export type MetadataItemCollectionInputFieldValue = z.infer<
+  typeof zMetadataItemCollectionInputFieldValue
+>;
+
+export const zMetadataItemPolymorphicInputFieldValue =
+  zInputFieldValueBase.extend({
+    type: z.literal('MetadataItemPolymorphic'),
+    value: z
+      .union([zMetadataItemField, z.array(zMetadataItemField)])
+      .optional(),
+  });
+export type MetadataItemPolymorphicInputFieldValue = z.infer<
+  typeof zMetadataItemPolymorphicInputFieldValue
+>;
+
+export const zMetadataField = z.record(z.any());
+export type MetadataField = z.infer<typeof zMetadataField>;
+
+export const zMetadataInputFieldValue = zInputFieldValueBase.extend({
+  type: z.literal('MetadataField'),
+  value: zMetadataField.optional(),
+});
+export type MetadataInputFieldValue = z.infer<typeof zMetadataInputFieldValue>;
+
+export const zMetadataCollectionInputFieldValue = zInputFieldValueBase.extend({
+  type: z.literal('MetadataCollection'),
+  value: z.array(zMetadataField).optional(),
+});
+export type MetadataCollectionInputFieldValue = z.infer<
+  typeof zMetadataCollectionInputFieldValue
+>;
+
 export const zColorField = z.object({
   r: z.number().int().min(0).max(255),
   g: z.number().int().min(0).max(255),
@@ -723,7 +784,13 @@ export type SchedulerInputFieldValue = z.infer<
   typeof zSchedulerInputFieldValue
 >;
 
+export const zAnyInputFieldValue = zInputFieldValueBase.extend({
+  type: z.literal('Any'),
+  value: z.any().optional(),
+});
+
 export const zInputFieldValue = z.discriminatedUnion('type', [
+  zAnyInputFieldValue,
   zBoardInputFieldValue,
   zBooleanCollectionInputFieldValue,
   zBooleanInputFieldValue,
@@ -774,6 +841,11 @@ export const zInputFieldValue = z.discriminatedUnion('type', [
   zUNetInputFieldValue,
   zVaeInputFieldValue,
   zVaeModelInputFieldValue,
+  zMetadataItemInputFieldValue,
+  zMetadataItemCollectionInputFieldValue,
+  zMetadataItemPolymorphicInputFieldValue,
+  zMetadataInputFieldValue,
+  zMetadataCollectionInputFieldValue,
 ]);
 
 export type InputFieldValue = z.infer<typeof zInputFieldValue>;
@@ -786,6 +858,11 @@ export type InputFieldTemplateBase = {
   fieldKind: 'input';
 } & _InputField;
 
+export type AnyInputFieldTemplate = InputFieldTemplateBase & {
+  type: 'Any';
+  default: undefined;
+};
+
 export type IntegerInputFieldTemplate = InputFieldTemplateBase & {
   type: 'integer';
   default: number;
@@ -939,6 +1016,11 @@ export type UNetInputFieldTemplate = InputFieldTemplateBase & {
   type: 'UNetField';
 };
 
+export type MetadataItemFieldTemplate = InputFieldTemplateBase & {
+  default: undefined;
+  type: 'MetadataItemField';
+};
+
 export type ClipInputFieldTemplate = InputFieldTemplateBase & {
   default: undefined;
   type: 'ClipField';
@@ -1087,6 +1169,34 @@ export type WorkflowInputFieldTemplate = InputFieldTemplateBase & {
   type: 'WorkflowField';
 };
 
+export type MetadataItemInputFieldTemplate = InputFieldTemplateBase & {
+  default: undefined;
+  type: 'MetadataItemField';
+};
+
+export type MetadataItemCollectionInputFieldTemplate =
+  InputFieldTemplateBase & {
+    default: undefined;
+    type: 'MetadataItemCollection';
+  };
+
+export type MetadataItemPolymorphicInputFieldTemplate = Omit<
+  MetadataItemInputFieldTemplate,
+  'type'
+> & {
+  type: 'MetadataItemPolymorphic';
+};
+
+export type MetadataInputFieldTemplate = InputFieldTemplateBase & {
+  default: undefined;
+  type: 'MetadataField';
+};
+
+export type MetadataCollectionInputFieldTemplate = InputFieldTemplateBase & {
+  default: undefined;
+  type: 'MetadataCollection';
+};
+
 /**
  * An input field template is generated on each page load from the OpenAPI schema.
  *
@@ -1094,6 +1204,7 @@ export type WorkflowInputFieldTemplate = InputFieldTemplateBase & {
  * maximum length, pattern to match, etc).
  */
 export type InputFieldTemplate =
+  | AnyInputFieldTemplate
   | BoardInputFieldTemplate
   | BooleanCollectionInputFieldTemplate
   | BooleanPolymorphicInputFieldTemplate
@@ -1143,7 +1254,12 @@ export type InputFieldTemplate =
   | T2IAdapterPolymorphicInputFieldTemplate
   | UNetInputFieldTemplate
   | VaeInputFieldTemplate
-  | VaeModelInputFieldTemplate;
+  | VaeModelInputFieldTemplate
+  | MetadataItemInputFieldTemplate
+  | MetadataItemCollectionInputFieldTemplate
+  | MetadataInputFieldTemplate
+  | MetadataItemPolymorphicInputFieldTemplate
+  | MetadataCollectionInputFieldTemplate;
 
 export const isInputFieldValue = (
   field?: InputFieldValue | OutputFieldValue
@@ -1264,7 +1380,7 @@ export const isInvocationFieldSchema = (
 
 export type InvocationEdgeExtra = { type: 'default' | 'collapsed' };
 
-const zLoRAMetadataItem = z.object({
+export const zLoRAMetadataItem = z.object({
   lora: zLoRAModelField.deepPartial(),
   weight: z.number(),
 });

invokeai/frontend/web/src/features/nodes/util/fieldTemplateBuilders.ts

@@ -7,6 +7,7 @@ import {
   startCase,
 } from 'lodash-es';
 import { OpenAPIV3_1 } from 'openapi-types';
+import { ControlField } from 'services/api/types';
 import {
   COLLECTION_MAP,
   POLYMORPHIC_TYPES,
@@ -15,36 +16,70 @@ import {
   isPolymorphicItemType,
 } from '../types/constants';
 import {
+  AnyInputFieldTemplate,
+  BoardInputFieldTemplate,
   BooleanCollectionInputFieldTemplate,
   BooleanInputFieldTemplate,
+  BooleanPolymorphicInputFieldTemplate,
   ClipInputFieldTemplate,
   CollectionInputFieldTemplate,
   CollectionItemInputFieldTemplate,
+  ColorCollectionInputFieldTemplate,
   ColorInputFieldTemplate,
+  ColorPolymorphicInputFieldTemplate,
+  ConditioningCollectionInputFieldTemplate,
+  ConditioningField,
   ConditioningInputFieldTemplate,
+  ConditioningPolymorphicInputFieldTemplate,
+  ControlCollectionInputFieldTemplate,
   ControlInputFieldTemplate,
   ControlNetModelInputFieldTemplate,
+  ControlPolymorphicInputFieldTemplate,
   DenoiseMaskInputFieldTemplate,
   EnumInputFieldTemplate,
   FieldType,
   FloatCollectionInputFieldTemplate,
-  FloatPolymorphicInputFieldTemplate,
   FloatInputFieldTemplate,
+  FloatPolymorphicInputFieldTemplate,
+  IPAdapterCollectionInputFieldTemplate,
+  IPAdapterField,
+  IPAdapterInputFieldTemplate,
+  IPAdapterModelInputFieldTemplate,
+  IPAdapterPolymorphicInputFieldTemplate,
   ImageCollectionInputFieldTemplate,
+  ImageField,
   ImageInputFieldTemplate,
+  ImagePolymorphicInputFieldTemplate,
+  InputFieldTemplate,
   InputFieldTemplateBase,
   IntegerCollectionInputFieldTemplate,
   IntegerInputFieldTemplate,
+  IntegerPolymorphicInputFieldTemplate,
   InvocationFieldSchema,
   InvocationSchemaObject,
+  LatentsCollectionInputFieldTemplate,
+  LatentsField,
   LatentsInputFieldTemplate,
+  LatentsPolymorphicInputFieldTemplate,
   LoRAModelInputFieldTemplate,
   MainModelInputFieldTemplate,
+  MetadataCollectionInputFieldTemplate,
+  MetadataInputFieldTemplate,
+  MetadataItemCollectionInputFieldTemplate,
+  MetadataItemInputFieldTemplate,
+  MetadataItemPolymorphicInputFieldTemplate,
+  OpenAPIV3_1SchemaOrRef,
   SDXLMainModelInputFieldTemplate,
   SDXLRefinerModelInputFieldTemplate,
   SchedulerInputFieldTemplate,
   StringCollectionInputFieldTemplate,
   StringInputFieldTemplate,
+  StringPolymorphicInputFieldTemplate,
+  T2IAdapterCollectionInputFieldTemplate,
+  T2IAdapterField,
+  T2IAdapterInputFieldTemplate,
+  T2IAdapterModelInputFieldTemplate,
+  T2IAdapterPolymorphicInputFieldTemplate,
   UNetInputFieldTemplate,
   VaeInputFieldTemplate,
   VaeModelInputFieldTemplate,
@@ -52,36 +87,7 @@ import {
   isNonArraySchemaObject,
   isRefObject,
   isSchemaObject,
-  ControlPolymorphicInputFieldTemplate,
-  ColorPolymorphicInputFieldTemplate,
-  ColorCollectionInputFieldTemplate,
-  IntegerPolymorphicInputFieldTemplate,
-  StringPolymorphicInputFieldTemplate,
-  BooleanPolymorphicInputFieldTemplate,
-  ImagePolymorphicInputFieldTemplate,
-  LatentsPolymorphicInputFieldTemplate,
-  LatentsCollectionInputFieldTemplate,
-  ConditioningPolymorphicInputFieldTemplate,
-  ConditioningCollectionInputFieldTemplate,
-  ControlCollectionInputFieldTemplate,
-  ImageField,
-  LatentsField,
-  ConditioningField,
-  IPAdapterField,
-  IPAdapterInputFieldTemplate,
-  IPAdapterModelInputFieldTemplate,
-  IPAdapterPolymorphicInputFieldTemplate,
-  IPAdapterCollectionInputFieldTemplate,
-  T2IAdapterField,
-  T2IAdapterInputFieldTemplate,
-  T2IAdapterModelInputFieldTemplate,
-  T2IAdapterPolymorphicInputFieldTemplate,
-  T2IAdapterCollectionInputFieldTemplate,
-  BoardInputFieldTemplate,
-  InputFieldTemplate,
-  OpenAPIV3_1SchemaOrRef,
 } from '../types/types';
-import { ControlField } from 'services/api/types';
 
 export type BaseFieldProperties = 'name' | 'title' | 'description';
 
@@ -851,6 +857,78 @@ const buildCollectionItemInputFieldTemplate = ({
   return template;
 };
 
+const buildAnyInputFieldTemplate = ({
+  baseField,
+}: BuildInputFieldArg): AnyInputFieldTemplate => {
+  const template: AnyInputFieldTemplate = {
+    ...baseField,
+    type: 'Any',
+    default: undefined,
+  };
+
+  return template;
+};
+
+const buildMetadataItemInputFieldTemplate = ({
+  baseField,
+}: BuildInputFieldArg): MetadataItemInputFieldTemplate => {
+  const template: MetadataItemInputFieldTemplate = {
+    ...baseField,
+    type: 'MetadataItemField',
+    default: undefined,
+  };
+
+  return template;
+};
+
+const buildMetadataItemCollectionInputFieldTemplate = ({
+  baseField,
+}: BuildInputFieldArg): MetadataItemCollectionInputFieldTemplate => {
+  const template: MetadataItemCollectionInputFieldTemplate = {
+    ...baseField,
+    type: 'MetadataItemCollection',
+    default: undefined,
+  };
+
+  return template;
+};
+
+const buildMetadataItemPolymorphicInputFieldTemplate = ({
+  baseField,
+}: BuildInputFieldArg): MetadataItemPolymorphicInputFieldTemplate => {
+  const template: MetadataItemPolymorphicInputFieldTemplate = {
+    ...baseField,
+    type: 'MetadataItemPolymorphic',
+    default: undefined,
+  };
+
+  return template;
+};
+
+const buildMetadataDictInputFieldTemplate = ({
+  baseField,
+}: BuildInputFieldArg): MetadataInputFieldTemplate => {
+  const template: MetadataInputFieldTemplate = {
+    ...baseField,
+    type: 'MetadataField',
+    default: undefined,
+  };
+
+  return template;
+};
+
+const buildMetadataCollectionInputFieldTemplate = ({
+  baseField,
+}: BuildInputFieldArg): MetadataCollectionInputFieldTemplate => {
+  const template: MetadataCollectionInputFieldTemplate = {
+    ...baseField,
+    type: 'MetadataCollection',
+    default: undefined,
+  };
+
+  return template;
+};
+
 const buildColorInputFieldTemplate = ({
   schemaObject,
   baseField,
@@ -1012,6 +1090,7 @@ const TEMPLATE_BUILDER_MAP: {
   [key in FieldType]?: (arg: BuildInputFieldArg) => InputFieldTemplate;
 } = {
   BoardField: buildBoardInputFieldTemplate,
+  Any: buildAnyInputFieldTemplate,
   boolean: buildBooleanInputFieldTemplate,
   BooleanCollection: buildBooleanCollectionInputFieldTemplate,
   BooleanPolymorphic: buildBooleanPolymorphicInputFieldTemplate,
@@ -1047,6 +1126,11 @@ const TEMPLATE_BUILDER_MAP: {
   LatentsField: buildLatentsInputFieldTemplate,
   LatentsPolymorphic: buildLatentsPolymorphicInputFieldTemplate,
   LoRAModelField: buildLoRAModelInputFieldTemplate,
+  MetadataItemField: buildMetadataItemInputFieldTemplate,
+  MetadataItemCollection: buildMetadataItemCollectionInputFieldTemplate,
+  MetadataItemPolymorphic: buildMetadataItemPolymorphicInputFieldTemplate,
+  MetadataField: buildMetadataDictInputFieldTemplate,
+  MetadataCollection: buildMetadataCollectionInputFieldTemplate,
   MainModelField: buildMainModelInputFieldTemplate,
   Scheduler: buildSchedulerInputFieldTemplate,
   SDXLMainModelField: buildSDXLMainModelInputFieldTemplate,

invokeai/frontend/web/src/features/nodes/util/fieldValueBuilders.ts

@@ -3,6 +3,7 @@ import { FieldType, InputFieldTemplate, InputFieldValue } from '../types/types';
 const FIELD_VALUE_FALLBACK_MAP: {
   [key in FieldType]: InputFieldValue['value'];
 } = {
+  Any: undefined,
   enum: '',
   BoardField: undefined,
   boolean: false,
@@ -38,6 +39,11 @@ const FIELD_VALUE_FALLBACK_MAP: {
   LatentsCollection: [],
   LatentsField: undefined,
   LatentsPolymorphic: undefined,
+  MetadataItemField: undefined,
+  MetadataItemCollection: [],
+  MetadataItemPolymorphic: undefined,
+  MetadataField: undefined,
+  MetadataCollection: [],
   LoRAModelField: undefined,
   MainModelField: undefined,
   ONNXModelField: undefined,

invokeai/frontend/web/src/features/nodes/util/graphBuilders/addControlNetToLinearGraph.ts

@@ -5,14 +5,14 @@ import {
   CollectInvocation,
   ControlField,
   ControlNetInvocation,
-  MetadataAccumulatorInvocation,
+  CoreMetadataInvocation,
 } from 'services/api/types';
 import { NonNullableGraph } from '../../types/types';
 import {
   CANVAS_COHERENCE_DENOISE_LATENTS,
   CONTROL_NET_COLLECT,
-  METADATA_ACCUMULATOR,
 } from './constants';
+import { upsertMetadata } from './metadata';
 
 export const addControlNetToLinearGraph = (
   state: RootState,
@@ -23,9 +23,11 @@ export const addControlNetToLinearGraph = (
     (ca) => ca.model?.base_model === state.generation.model?.base_model
   );
 
-  const metadataAccumulator = graph.nodes[METADATA_ACCUMULATOR] as
+  // const metadataAccumulator = graph.nodes[METADATA_ACCUMULATOR] as
-    | MetadataAccumulatorInvocation
+  //   | MetadataAccumulatorInvocation
-    | undefined;
+  //   | undefined;
+
+  const controlNetMetadata: CoreMetadataInvocation['controlnets'] = [];
 
   if (validControlNets.length) {
     // Even though denoise_latents' control input is polymorphic, keep it simple and always use a collect
@@ -99,15 +101,9 @@ export const addControlNetToLinearGraph = (
 
       graph.nodes[controlNetNode.id] = controlNetNode as ControlNetInvocation;
 
-      if (metadataAccumulator?.controlnets) {
+      controlNetMetadata.push(
-        // metadata accumulator only needs a control field - not the whole node
+        omit(controlNetNode, ['id', 'type', 'is_intermediate']) as ControlField
-        // extract what we need and add to the accumulator
+      );
-        const controlField = omit(controlNetNode, [
-          'id',
-          'type',
-        ]) as ControlField;
-        metadataAccumulator.controlnets.push(controlField);
-      }
 
       graph.edges.push({
         source: { node_id: controlNetNode.id, field: 'control' },
@@ -117,5 +113,6 @@ export const addControlNetToLinearGraph = (
         },
       });
     });
+    upsertMetadata(graph, { controlnets: controlNetMetadata });
   }
 };

invokeai/frontend/web/src/features/nodes/util/graphBuilders/addHrfToGraph.ts

@@ -1,25 +1,25 @@
+import { logger } from 'app/logging/logger';
 import { RootState } from 'app/store/store';
 import { NonNullableGraph } from 'features/nodes/types/types';
 import {
   DenoiseLatentsInvocation,
-  ResizeLatentsInvocation,
-  NoiseInvocation,
-  LatentsToImageInvocation,
   Edge,
+  LatentsToImageInvocation,
+  NoiseInvocation,
+  ResizeLatentsInvocation,
 } from 'services/api/types';
 import {
-  LATENTS_TO_IMAGE,
   DENOISE_LATENTS,
-  NOISE,
-  MAIN_MODEL_LOADER,
-  METADATA_ACCUMULATOR,
-  LATENTS_TO_IMAGE_HRF,
   DENOISE_LATENTS_HRF,
-  RESCALE_LATENTS,
+  LATENTS_TO_IMAGE,
+  LATENTS_TO_IMAGE_HRF,
+  MAIN_MODEL_LOADER,
+  NOISE,
   NOISE_HRF,
+  RESCALE_LATENTS,
   VAE_LOADER,
 } from './constants';
-import { logger } from 'app/logging/logger';
+import { upsertMetadata } from './metadata';
 
 // Copy certain connections from previous DENOISE_LATENTS to new DENOISE_LATENTS_HRF.
 function copyConnectionsToDenoiseLatentsHrf(graph: NonNullableGraph): void {
@@ -71,10 +71,8 @@ export const addHrfToGraph = (
   }
   const log = logger('txt2img');
 
-  const { vae } = state.generation;
+  const { vae, hrfWidth, hrfHeight, hrfStrength } = state.generation;
   const isAutoVae = !vae;
-  const hrfWidth = state.generation.hrfWidth;
-  const hrfHeight = state.generation.hrfHeight;
 
   // Pre-existing (original) graph nodes.
   const originalDenoiseLatentsNode = graph.nodes[DENOISE_LATENTS] as
@@ -116,7 +114,7 @@ export const addHrfToGraph = (
     cfg_scale: originalDenoiseLatentsNode?.cfg_scale,
     scheduler: originalDenoiseLatentsNode?.scheduler,
     steps: originalDenoiseLatentsNode?.steps,
-    denoising_start: 1 - state.generation.hrfStrength,
+    denoising_start: 1 - hrfStrength,
     denoising_end: 1,
   };
 
@@ -221,16 +219,6 @@ export const addHrfToGraph = (
         field: 'latents',
       },
     },
-    {
-      source: {
-        node_id: METADATA_ACCUMULATOR,
-        field: 'metadata',
-      },
-      destination: {
-        node_id: LATENTS_TO_IMAGE_HRF,
-        field: 'metadata',
-      },
-    },
     {
       source: {
         node_id: isAutoVae ? MAIN_MODEL_LOADER : VAE_LOADER,
@@ -243,5 +231,11 @@ export const addHrfToGraph = (
     }
   );
 
+  upsertMetadata(graph, {
+    hrf_height: hrfHeight,
+    hrf_width: hrfWidth,
+    hrf_strength: hrfStrength,
+  });
+
   copyConnectionsToDenoiseLatentsHrf(graph);
 };

invokeai/frontend/web/src/features/nodes/util/graphBuilders/addIPAdapterToLinearGraph.ts

@@ -1,16 +1,18 @@
 import { RootState } from 'app/store/store';
 import { selectValidIPAdapters } from 'features/controlAdapters/store/controlAdaptersSlice';
+import { omit } from 'lodash-es';
 import {
   CollectInvocation,
+  CoreMetadataInvocation,
   IPAdapterInvocation,
-  MetadataAccumulatorInvocation,
+  IPAdapterMetadataField,
 } from 'services/api/types';
 import { NonNullableGraph } from '../../types/types';
 import {
   CANVAS_COHERENCE_DENOISE_LATENTS,
   IP_ADAPTER_COLLECT,
-  METADATA_ACCUMULATOR,
 } from './constants';
+import { upsertMetadata } from './metadata';
 
 export const addIPAdapterToLinearGraph = (
   state: RootState,
@@ -21,10 +23,6 @@ export const addIPAdapterToLinearGraph = (
     (ca) => ca.model?.base_model === state.generation.model?.base_model
   );
 
-  const metadataAccumulator = graph.nodes[METADATA_ACCUMULATOR] as
-    | MetadataAccumulatorInvocation
-    | undefined;
-
   if (validIPAdapters.length) {
     // Even though denoise_latents' control input is polymorphic, keep it simple and always use a collect
     const ipAdapterCollectNode: CollectInvocation = {
@@ -50,6 +48,7 @@ export const addIPAdapterToLinearGraph = (
         },
       });
     }
+    const ipAdapterMetdata: CoreMetadataInvocation['ipAdapters'] = [];
 
     validIPAdapters.forEach((ipAdapter) => {
       if (!ipAdapter.model) {
@@ -76,19 +75,13 @@ export const addIPAdapterToLinearGraph = (
 
       graph.nodes[ipAdapterNode.id] = ipAdapterNode as IPAdapterInvocation;
 
-      if (metadataAccumulator?.ipAdapters) {
+      ipAdapterMetdata.push(
-        const ipAdapterField = {
+        omit(ipAdapterNode, [
-          image: {
+          'id',
-            image_name: ipAdapter.controlImage,
+          'type',
-          },
+          'is_intermediate',
-          weight,
+        ]) as IPAdapterMetadataField
-          ip_adapter_model: model,
+      );
-          begin_step_percent: beginStepPct,
-          end_step_percent: endStepPct,
-        };
-
-        metadataAccumulator.ipAdapters.push(ipAdapterField);
-      }
 
       graph.edges.push({
         source: { node_id: ipAdapterNode.id, field: 'ip_adapter' },
@@ -98,5 +91,7 @@ export const addIPAdapterToLinearGraph = (
         },
       });
     });
+
+    upsertMetadata(graph, { ipAdapters: ipAdapterMetdata });
   }
 };

invokeai/frontend/web/src/features/nodes/util/graphBuilders/addLoRAsToGraph.ts

@@ -2,20 +2,20 @@ import { RootState } from 'app/store/store';
 import { NonNullableGraph } from 'features/nodes/types/types';
 import { forEach, size } from 'lodash-es';
 import {
+  CoreMetadataInvocation,
   LoraLoaderInvocation,
-  MetadataAccumulatorInvocation,
 } from 'services/api/types';
 import {
+  CANVAS_COHERENCE_DENOISE_LATENTS,
   CANVAS_INPAINT_GRAPH,
   CANVAS_OUTPAINT_GRAPH,
-  CANVAS_COHERENCE_DENOISE_LATENTS,
   CLIP_SKIP,
   LORA_LOADER,
   MAIN_MODEL_LOADER,
-  METADATA_ACCUMULATOR,
   NEGATIVE_CONDITIONING,
   POSITIVE_CONDITIONING,
 } from './constants';
+import { upsertMetadata } from './metadata';
 
 export const addLoRAsToGraph = (
   state: RootState,
@@ -33,29 +33,29 @@ export const addLoRAsToGraph = (
 
   const { loras } = state.lora;
   const loraCount = size(loras);
-  const metadataAccumulator = graph.nodes[METADATA_ACCUMULATOR] as
-    | MetadataAccumulatorInvocation
-    | undefined;
 
-  if (loraCount > 0) {
+  if (loraCount === 0) {
-    // Remove modelLoaderNodeId unet connection to feed it to LoRAs
+    return;
-    graph.edges = graph.edges.filter(
-      (e) =>
-        !(
-          e.source.node_id === modelLoaderNodeId &&
-          ['unet'].includes(e.source.field)
-        )
-    );
-    // Remove CLIP_SKIP connections to conditionings to feed it through LoRAs
-    graph.edges = graph.edges.filter(
-      (e) =>
-        !(e.source.node_id === CLIP_SKIP && ['clip'].includes(e.source.field))
-    );
   }
 
+  // Remove modelLoaderNodeId unet connection to feed it to LoRAs
+  graph.edges = graph.edges.filter(
+    (e) =>
+      !(
+        e.source.node_id === modelLoaderNodeId &&
+        ['unet'].includes(e.source.field)
+      )
+  );
+  // Remove CLIP_SKIP connections to conditionings to feed it through LoRAs
+  graph.edges = graph.edges.filter(
+    (e) =>
+      !(e.source.node_id === CLIP_SKIP && ['clip'].includes(e.source.field))
+  );
+
   // we need to remember the last lora so we can chain from it
   let lastLoraNodeId = '';
   let currentLoraIndex = 0;
+  const loraMetadata: CoreMetadataInvocation['loras'] = [];
 
   forEach(loras, (lora) => {
     const { model_name, base_model, weight } = lora;
@@ -69,13 +69,10 @@ export const addLoRAsToGraph = (
       weight,
     };
 
-    // add the lora to the metadata accumulator
+    loraMetadata.push({
-    if (metadataAccumulator?.loras) {
+      lora: { model_name, base_model },
-      metadataAccumulator.loras.push({
+      weight,
-        lora: { model_name, base_model },
+    });
-        weight,
-      });
-    }
 
     // add to graph
     graph.nodes[currentLoraNodeId] = loraLoaderNode;
@@ -182,4 +179,6 @@ export const addLoRAsToGraph = (
     lastLoraNodeId = currentLoraNodeId;
     currentLoraIndex += 1;
   });
+
+  upsertMetadata(graph, { loras: loraMetadata });
 };

invokeai/frontend/web/src/features/nodes/util/graphBuilders/addSDXLLoRAstoGraph.ts

@@ -1,14 +1,14 @@
 import { RootState } from 'app/store/store';
-import { NonNullableGraph } from 'features/nodes/types/types';
-import { forEach, size } from 'lodash-es';
 import {
-  MetadataAccumulatorInvocation,
+  LoRAMetadataItem,
-  SDXLLoraLoaderInvocation,
+  NonNullableGraph,
-} from 'services/api/types';
+  zLoRAMetadataItem,
+} from 'features/nodes/types/types';
+import { forEach, size } from 'lodash-es';
+import { SDXLLoraLoaderInvocation } from 'services/api/types';
 import {
   CANVAS_COHERENCE_DENOISE_LATENTS,
   LORA_LOADER,
-  METADATA_ACCUMULATOR,
   NEGATIVE_CONDITIONING,
   POSITIVE_CONDITIONING,
   SDXL_CANVAS_INPAINT_GRAPH,
@@ -17,6 +17,7 @@ import {
   SDXL_REFINER_INPAINT_CREATE_MASK,
   SEAMLESS,
 } from './constants';
+import { upsertMetadata } from './metadata';
 
 export const addSDXLLoRAsToGraph = (
   state: RootState,
@@ -34,9 +35,12 @@ export const addSDXLLoRAsToGraph = (
 
   const { loras } = state.lora;
   const loraCount = size(loras);
-  const metadataAccumulator = graph.nodes[METADATA_ACCUMULATOR] as
+
-    | MetadataAccumulatorInvocation
+  if (loraCount === 0) {
-    | undefined;
+    return;
+  }
+
+  const loraMetadata: LoRAMetadataItem[] = [];
 
   // Handle Seamless Plugs
   const unetLoaderId = modelLoaderNodeId;
@@ -47,22 +51,17 @@ export const addSDXLLoRAsToGraph = (
     clipLoaderId = SDXL_MODEL_LOADER;
   }
 
-  if (loraCount > 0) {
+  // Remove modelLoaderNodeId unet/clip/clip2 connections to feed it to LoRAs
-    // Remove modelLoaderNodeId unet/clip/clip2 connections to feed it to LoRAs
+  graph.edges = graph.edges.filter(
-    graph.edges = graph.edges.filter(
+    (e) =>
-      (e) =>
+      !(
-        !(
+        e.source.node_id === unetLoaderId && ['unet'].includes(e.source.field)
-          e.source.node_id === unetLoaderId && ['unet'].includes(e.source.field)
+      ) &&
-        ) &&
+      !(
-        !(
+        e.source.node_id === clipLoaderId && ['clip'].includes(e.source.field)
-          e.source.node_id === clipLoaderId && ['clip'].includes(e.source.field)
+      ) &&
-        ) &&
+      !(e.source.node_id === clipLoaderId && ['clip2'].includes(e.source.field))
-        !(
+  );
-          e.source.node_id === clipLoaderId &&
-          ['clip2'].includes(e.source.field)
-        )
-    );
-  }
 
   // we need to remember the last lora so we can chain from it
   let lastLoraNodeId = '';
@@ -80,16 +79,12 @@ export const addSDXLLoRAsToGraph = (
       weight,
     };
 
-    // add the lora to the metadata accumulator
+    loraMetadata.push(
-    if (metadataAccumulator) {
+      zLoRAMetadataItem.parse({
-      if (!metadataAccumulator.loras) {
-        metadataAccumulator.loras = [];
-      }
-      metadataAccumulator.loras.push({
         lora: { model_name, base_model },
         weight,
-      });
+      })
-    }
+    );
 
     // add to graph
     graph.nodes[currentLoraNodeId] = loraLoaderNode;
@@ -242,4 +237,6 @@ export const addSDXLLoRAsToGraph = (
     lastLoraNodeId = currentLoraNodeId;
     currentLoraIndex += 1;
   });
+
+  upsertMetadata(graph, { loras: loraMetadata });
 };

invokeai/frontend/web/src/features/nodes/util/graphBuilders/addSDXLRefinerToGraph.ts

@@ -2,7 +2,6 @@ import { RootState } from 'app/store/store';
 import {
   CreateDenoiseMaskInvocation,
   ImageDTO,
-  MetadataAccumulatorInvocation,
   SeamlessModeInvocation,
 } from 'services/api/types';
 import { NonNullableGraph } from '../../types/types';
@@ -12,7 +11,6 @@ import {
   LATENTS_TO_IMAGE,
   MASK_COMBINE,
   MASK_RESIZE_UP,
-  METADATA_ACCUMULATOR,
   SDXL_CANVAS_IMAGE_TO_IMAGE_GRAPH,
   SDXL_CANVAS_INPAINT_GRAPH,
   SDXL_CANVAS_OUTPAINT_GRAPH,
@@ -26,6 +24,7 @@ import {
   SDXL_REFINER_SEAMLESS,
 } from './constants';
 import { buildSDXLStylePrompts } from './helpers/craftSDXLStylePrompt';
+import { upsertMetadata } from './metadata';
 
 export const addSDXLRefinerToGraph = (
   state: RootState,
@@ -58,21 +57,15 @@ export const addSDXLRefinerToGraph = (
     return;
   }
 
-  const metadataAccumulator = graph.nodes[METADATA_ACCUMULATOR] as
+  upsertMetadata(graph, {
-    | MetadataAccumulatorInvocation
+    refiner_model: refinerModel,
-    | undefined;
+    refiner_positive_aesthetic_score: refinerPositiveAestheticScore,
-
+    refiner_negative_aesthetic_score: refinerNegativeAestheticScore,
-  if (metadataAccumulator) {
+    refiner_cfg_scale: refinerCFGScale,
-    metadataAccumulator.refiner_model = refinerModel;
+    refiner_scheduler: refinerScheduler,
-    metadataAccumulator.refiner_positive_aesthetic_score =
+    refiner_start: refinerStart,
-      refinerPositiveAestheticScore;
+    refiner_steps: refinerSteps,
-    metadataAccumulator.refiner_negative_aesthetic_score =
+  });
-      refinerNegativeAestheticScore;
-    metadataAccumulator.refiner_cfg_scale = refinerCFGScale;
-    metadataAccumulator.refiner_scheduler = refinerScheduler;
-    metadataAccumulator.refiner_start = refinerStart;
-    metadataAccumulator.refiner_steps = refinerSteps;
-  }
 
   const modelLoaderId = modelLoaderNodeId
     ? modelLoaderNodeId

invokeai/frontend/web/src/features/nodes/util/graphBuilders/addSaveImageNode.ts

@@ -1,19 +1,15 @@
+import { RootState } from 'app/store/store';
 import { NonNullableGraph } from 'features/nodes/types/types';
+import { activeTabNameSelector } from 'features/ui/store/uiSelectors';
+import { SaveImageInvocation } from 'services/api/types';
 import {
   CANVAS_OUTPUT,
   LATENTS_TO_IMAGE,
   LATENTS_TO_IMAGE_HRF,
-  METADATA_ACCUMULATOR,
   NSFW_CHECKER,
   SAVE_IMAGE,
   WATERMARKER,
 } from './constants';
-import {
-  MetadataAccumulatorInvocation,
-  SaveImageInvocation,
-} from 'services/api/types';
-import { RootState } from 'app/store/store';
-import { activeTabNameSelector } from 'features/ui/store/uiSelectors';
 
 /**
  * Set the `use_cache` field on the linear/canvas graph's final image output node to False.
@@ -37,23 +33,6 @@ export const addSaveImageNode = (
 
   graph.nodes[SAVE_IMAGE] = saveImageNode;
 
-  const metadataAccumulator = graph.nodes[METADATA_ACCUMULATOR] as
-    | MetadataAccumulatorInvocation
-    | undefined;
-
-  if (metadataAccumulator) {
-    graph.edges.push({
-      source: {
-        node_id: METADATA_ACCUMULATOR,
-        field: 'metadata',
-      },
-      destination: {
-        node_id: SAVE_IMAGE,
-        field: 'metadata',
-      },
-    });
-  }
-
   const destination = {
     node_id: SAVE_IMAGE,
     field: 'image',

invokeai/frontend/web/src/features/nodes/util/graphBuilders/addSeamlessToLinearGraph.ts

@@ -1,6 +1,7 @@
 import { RootState } from 'app/store/store';
 import { SeamlessModeInvocation } from 'services/api/types';
 import { NonNullableGraph } from '../../types/types';
+import { upsertMetadata } from './metadata';
 import {
   CANVAS_COHERENCE_DENOISE_LATENTS,
   CANVAS_INPAINT_GRAPH,
@@ -31,6 +32,17 @@ export const addSeamlessToLinearGraph = (
     seamless_y: seamlessYAxis,
   } as SeamlessModeInvocation;
 
+  if (seamlessXAxis) {
+    upsertMetadata(graph, {
+      seamless_x: seamlessXAxis,
+    });
+  }
+  if (seamlessYAxis) {
+    upsertMetadata(graph, {
+      seamless_y: seamlessYAxis,
+    });
+  }
+
   let denoisingNodeId = DENOISE_LATENTS;
 
   if (

invokeai/frontend/web/src/features/nodes/util/graphBuilders/addT2IAdapterToLinearGraph.ts

@@ -3,15 +3,15 @@ import { selectValidT2IAdapters } from 'features/controlAdapters/store/controlAd
 import { omit } from 'lodash-es';
 import {
   CollectInvocation,
-  MetadataAccumulatorInvocation,
+  CoreMetadataInvocation,
   T2IAdapterInvocation,
 } from 'services/api/types';
 import { NonNullableGraph, T2IAdapterField } from '../../types/types';
 import {
   CANVAS_COHERENCE_DENOISE_LATENTS,
-  METADATA_ACCUMULATOR,
   T2I_ADAPTER_COLLECT,
 } from './constants';
+import { upsertMetadata } from './metadata';
 
 export const addT2IAdaptersToLinearGraph = (
   state: RootState,
@@ -22,10 +22,6 @@ export const addT2IAdaptersToLinearGraph = (
     (ca) => ca.model?.base_model === state.generation.model?.base_model
   );
 
-  const metadataAccumulator = graph.nodes[METADATA_ACCUMULATOR] as
-    | MetadataAccumulatorInvocation
-    | undefined;
-
   if (validT2IAdapters.length) {
     // Even though denoise_latents' control input is polymorphic, keep it simple and always use a collect
     const t2iAdapterCollectNode: CollectInvocation = {
@@ -51,6 +47,7 @@ export const addT2IAdaptersToLinearGraph = (
         },
       });
     }
+    const t2iAdapterMetdata: CoreMetadataInvocation['t2iAdapters'] = [];
 
     validT2IAdapters.forEach((t2iAdapter) => {
       if (!t2iAdapter.model) {
@@ -96,15 +93,13 @@ export const addT2IAdaptersToLinearGraph = (
 
       graph.nodes[t2iAdapterNode.id] = t2iAdapterNode as T2IAdapterInvocation;
 
-      if (metadataAccumulator?.t2iAdapters) {
+      t2iAdapterMetdata.push(
-        // metadata accumulator only needs a control field - not the whole node
+        omit(t2iAdapterNode, [
-        // extract what we need and add to the accumulator
-        const t2iAdapterField = omit(t2iAdapterNode, [
           'id',
           'type',
-        ]) as T2IAdapterField;
+          'is_intermediate',
-        metadataAccumulator.t2iAdapters.push(t2iAdapterField);
+        ]) as T2IAdapterField
-      }
+      );
 
       graph.edges.push({
         source: { node_id: t2iAdapterNode.id, field: 't2i_adapter' },
@@ -114,5 +109,7 @@ export const addT2IAdaptersToLinearGraph = (
         },
       });
     });
+
+    upsertMetadata(graph, { t2iAdapters: t2iAdapterMetdata });
   }
 };

invokeai/frontend/web/src/features/nodes/util/graphBuilders/addVAEToGraph.ts

@@ -1,6 +1,5 @@
 import { RootState } from 'app/store/store';
 import { NonNullableGraph } from 'features/nodes/types/types';
-import { MetadataAccumulatorInvocation } from 'services/api/types';
 import {
   CANVAS_COHERENCE_INPAINT_CREATE_MASK,
   CANVAS_IMAGE_TO_IMAGE_GRAPH,
@@ -14,7 +13,6 @@ import {
   INPAINT_IMAGE,
   LATENTS_TO_IMAGE,
   MAIN_MODEL_LOADER,
-  METADATA_ACCUMULATOR,
   ONNX_MODEL_LOADER,
   SDXL_CANVAS_IMAGE_TO_IMAGE_GRAPH,
   SDXL_CANVAS_INPAINT_GRAPH,
@@ -26,6 +24,7 @@ import {
   TEXT_TO_IMAGE_GRAPH,
   VAE_LOADER,
 } from './constants';
+import { upsertMetadata } from './metadata';
 
 export const addVAEToGraph = (
   state: RootState,
@@ -41,9 +40,6 @@ export const addVAEToGraph = (
   );
 
   const isAutoVae = !vae;
-  const metadataAccumulator = graph.nodes[METADATA_ACCUMULATOR] as
-    | MetadataAccumulatorInvocation
-    | undefined;
 
   if (!isAutoVae) {
     graph.nodes[VAE_LOADER] = {
@@ -181,7 +177,7 @@ export const addVAEToGraph = (
     }
   }
 
-  if (vae && metadataAccumulator) {
+  if (vae) {
-    metadataAccumulator.vae = vae;
+    upsertMetadata(graph, { vae });
   }
 };

invokeai/frontend/web/src/features/nodes/util/graphBuilders/addWatermarkerToGraph.ts

@@ -5,14 +5,8 @@ import {
   ImageNSFWBlurInvocation,
   ImageWatermarkInvocation,
   LatentsToImageInvocation,
-  MetadataAccumulatorInvocation,
 } from 'services/api/types';
-import {
+import { LATENTS_TO_IMAGE, NSFW_CHECKER, WATERMARKER } from './constants';
-  LATENTS_TO_IMAGE,
-  METADATA_ACCUMULATOR,
-  NSFW_CHECKER,
-  WATERMARKER,
-} from './constants';
 
 export const addWatermarkerToGraph = (
   state: RootState,
@@ -32,10 +26,6 @@ export const addWatermarkerToGraph = (
     | ImageNSFWBlurInvocation
     | undefined;
 
-  const metadataAccumulator = graph.nodes[METADATA_ACCUMULATOR] as
-    | MetadataAccumulatorInvocation
-    | undefined;
-
   if (!nodeToAddTo) {
     // something has gone terribly awry
     return;
@@ -80,17 +70,4 @@ export const addWatermarkerToGraph = (
       },
     });
   }
-
-  if (metadataAccumulator) {
-    graph.edges.push({
-      source: {
-        node_id: METADATA_ACCUMULATOR,
-        field: 'metadata',
-      },
-      destination: {
-        node_id: WATERMARKER,
-        field: 'metadata',
-      },
-    });
-  }
 };

invokeai/frontend/web/src/features/nodes/util/graphBuilders/buildAdHocUpscaleGraph.ts

@@ -1,12 +1,13 @@
+import { BoardId } from 'features/gallery/store/types';
 import { NonNullableGraph } from 'features/nodes/types/types';
 import { ESRGANModelName } from 'features/parameters/store/postprocessingSlice';
 import {
-  Graph,
   ESRGANInvocation,
+  Graph,
   SaveImageInvocation,
 } from 'services/api/types';
 import { REALESRGAN as ESRGAN, SAVE_IMAGE } from './constants';
-import { BoardId } from 'features/gallery/store/types';
+import { addCoreMetadataNode } from './metadata';
 
 type Arg = {
   image_name: string;
@@ -55,5 +56,9 @@ export const buildAdHocUpscaleGraph = ({
     ],
   };
 
+  addCoreMetadataNode(graph, {
+    esrgan_model: esrganModelName,
+  });
+
   return graph;
 };

invokeai/frontend/web/src/features/nodes/util/graphBuilders/buildCanvasImageToImageGraph.ts

@@ -20,12 +20,12 @@ import {
   IMG2IMG_RESIZE,
   LATENTS_TO_IMAGE,
   MAIN_MODEL_LOADER,
-  METADATA_ACCUMULATOR,
   NEGATIVE_CONDITIONING,
   NOISE,
   POSITIVE_CONDITIONING,
   SEAMLESS,
 } from './constants';
+import { addCoreMetadataNode } from './metadata';
 
 /**
  * Builds the Canvas tab's Image to Image graph.
@@ -308,10 +308,7 @@ export const buildCanvasImageToImageGraph = (
     });
   }
 
-  // add metadata accumulator, which is only mostly populated - some fields are added later
+  addCoreMetadataNode(graph, {
-  graph.nodes[METADATA_ACCUMULATOR] = {
-    id: METADATA_ACCUMULATOR,
-    type: 'metadata_accumulator',
     generation_mode: 'img2img',
     cfg_scale,
     width: !isUsingScaledDimensions ? width : scaledBoundingBoxDimensions.width,
@@ -325,15 +322,10 @@ export const buildCanvasImageToImageGraph = (
     steps,
     rand_device: use_cpu ? 'cpu' : 'cuda',
     scheduler,
-    vae: undefined, // option; set in addVAEToGraph
-    controlnets: [], // populated in addControlNetToLinearGraph
-    loras: [], // populated in addLoRAsToGraph
-    ipAdapters: [], // populated in addIPAdapterToLinearGraph
-    t2iAdapters: [],
     clip_skip: clipSkip,
     strength,
     init_image: initialImage.image_name,
-  };
+  });
 
   // Add Seamless To Graph
   if (seamlessXAxis || seamlessYAxis) {

invokeai/frontend/web/src/features/nodes/util/graphBuilders/buildCanvasSDXLImageToImageGraph.ts

@@ -16,7 +16,6 @@ import {
   IMAGE_TO_LATENTS,
   IMG2IMG_RESIZE,
   LATENTS_TO_IMAGE,
-  METADATA_ACCUMULATOR,
   NEGATIVE_CONDITIONING,
   NOISE,
   POSITIVE_CONDITIONING,
@@ -28,6 +27,7 @@ import {
 } from './constants';
 import { buildSDXLStylePrompts } from './helpers/craftSDXLStylePrompt';
 import { addT2IAdaptersToLinearGraph } from './addT2IAdapterToLinearGraph';
+import { addCoreMetadataNode } from './metadata';
 
 /**
  * Builds the Canvas tab's Image to Image graph.
@@ -319,10 +319,7 @@ export const buildCanvasSDXLImageToImageGraph = (
     });
   }
 
-  // add metadata accumulator, which is only mostly populated - some fields are added later
+  addCoreMetadataNode(graph, {
-  graph.nodes[METADATA_ACCUMULATOR] = {
-    id: METADATA_ACCUMULATOR,
-    type: 'metadata_accumulator',
     generation_mode: 'img2img',
     cfg_scale,
     width: !isUsingScaledDimensions ? width : scaledBoundingBoxDimensions.width,
@@ -336,24 +333,8 @@ export const buildCanvasSDXLImageToImageGraph = (
     steps,
     rand_device: use_cpu ? 'cpu' : 'cuda',
     scheduler,
-    vae: undefined, // option; set in addVAEToGraph
-    controlnets: [], // populated in addControlNetToLinearGraph
-    loras: [], // populated in addLoRAsToGraph
-    ipAdapters: [], // populated in addIPAdapterToLinearGraph
-    t2iAdapters: [],
     strength,
     init_image: initialImage.image_name,
-  };
-
-  graph.edges.push({
-    source: {
-      node_id: METADATA_ACCUMULATOR,
-      field: 'metadata',
-    },
-    destination: {
-      node_id: CANVAS_OUTPUT,
-      field: 'metadata',
-    },
   });
 
   // Add Seamless To Graph

invokeai/frontend/web/src/features/nodes/util/graphBuilders/buildCanvasSDXLTextToImageGraph.ts

@@ -18,7 +18,6 @@ import { addWatermarkerToGraph } from './addWatermarkerToGraph';
 import {
   CANVAS_OUTPUT,
   LATENTS_TO_IMAGE,
-  METADATA_ACCUMULATOR,
   NEGATIVE_CONDITIONING,
   NOISE,
   ONNX_MODEL_LOADER,
@@ -30,6 +29,7 @@ import {
   SEAMLESS,
 } from './constants';
 import { buildSDXLStylePrompts } from './helpers/craftSDXLStylePrompt';
+import { addCoreMetadataNode } from './metadata';
 
 /**
  * Builds the Canvas tab's Text to Image graph.
@@ -301,10 +301,7 @@ export const buildCanvasSDXLTextToImageGraph = (
     });
   }
 
-  // add metadata accumulator, which is only mostly populated - some fields are added later
+  addCoreMetadataNode(graph, {
-  graph.nodes[METADATA_ACCUMULATOR] = {
-    id: METADATA_ACCUMULATOR,
-    type: 'metadata_accumulator',
     generation_mode: 'txt2img',
     cfg_scale,
     width: !isUsingScaledDimensions ? width : scaledBoundingBoxDimensions.width,
@@ -318,22 +315,6 @@ export const buildCanvasSDXLTextToImageGraph = (
     steps,
     rand_device: use_cpu ? 'cpu' : 'cuda',
     scheduler,
-    vae: undefined, // option; set in addVAEToGraph
-    controlnets: [], // populated in addControlNetToLinearGraph
-    loras: [], // populated in addLoRAsToGraph
-    ipAdapters: [], // populated in addIPAdapterToLinearGraph
-    t2iAdapters: [],
-  };
-
-  graph.edges.push({
-    source: {
-      node_id: METADATA_ACCUMULATOR,
-      field: 'metadata',
-    },
-    destination: {
-      node_id: CANVAS_OUTPUT,
-      field: 'metadata',
-    },
   });
 
   // Add Seamless To Graph

invokeai/frontend/web/src/features/nodes/util/graphBuilders/buildCanvasTextToImageGraph.ts

@@ -21,13 +21,13 @@ import {
   DENOISE_LATENTS,
   LATENTS_TO_IMAGE,
   MAIN_MODEL_LOADER,
-  METADATA_ACCUMULATOR,
   NEGATIVE_CONDITIONING,
   NOISE,
   ONNX_MODEL_LOADER,
   POSITIVE_CONDITIONING,
   SEAMLESS,
 } from './constants';
+import { addCoreMetadataNode } from './metadata';
 
 /**
  * Builds the Canvas tab's Text to Image graph.
@@ -289,10 +289,7 @@ export const buildCanvasTextToImageGraph = (
     });
   }
 
-  // add metadata accumulator, which is only mostly populated - some fields are added later
+  addCoreMetadataNode(graph, {
-  graph.nodes[METADATA_ACCUMULATOR] = {
-    id: METADATA_ACCUMULATOR,
-    type: 'metadata_accumulator',
     generation_mode: 'txt2img',
     cfg_scale,
     width: !isUsingScaledDimensions ? width : scaledBoundingBoxDimensions.width,
@@ -306,23 +303,7 @@ export const buildCanvasTextToImageGraph = (
     steps,
     rand_device: use_cpu ? 'cpu' : 'cuda',
     scheduler,
-    vae: undefined, // option; set in addVAEToGraph
-    controlnets: [], // populated in addControlNetToLinearGraph
-    loras: [], // populated in addLoRAsToGraph
-    ipAdapters: [], // populated in addIPAdapterToLinearGraph
-    t2iAdapters: [],
     clip_skip: clipSkip,
-  };
-
-  graph.edges.push({
-    source: {
-      node_id: METADATA_ACCUMULATOR,
-      field: 'metadata',
-    },
-    destination: {
-      node_id: CANVAS_OUTPUT,
-      field: 'metadata',
-    },
   });
 
   // Add Seamless To Graph

invokeai/frontend/web/src/features/nodes/util/graphBuilders/buildLinearBatchConfig.ts

@@ -7,10 +7,12 @@ import { components } from 'services/api/schema';
 import { Batch, BatchConfig } from 'services/api/types';
 import {
   CANVAS_COHERENCE_NOISE,
+  METADATA,
   METADATA_ACCUMULATOR,
   NOISE,
   POSITIVE_CONDITIONING,
 } from './constants';
+import { removeMetadata } from './metadata';
 
 export const prepareLinearUIBatch = (
   state: RootState,
@@ -24,7 +26,6 @@ export const prepareLinearUIBatch = (
   const data: Batch['data'] = [];
 
   if (prompts.length === 1) {
-    unset(graph.nodes[METADATA_ACCUMULATOR], 'seed');
     const seeds = generateSeeds({
       count: iterations,
       start: shouldRandomizeSeed ? undefined : seed,
@@ -40,13 +41,13 @@ export const prepareLinearUIBatch = (
       });
     }
 
-    if (graph.nodes[METADATA_ACCUMULATOR]) {
+    // add to metadata
-      zipped.push({
+    removeMetadata(graph, 'seed');
-        node_path: METADATA_ACCUMULATOR,
+    zipped.push({
-        field_name: 'seed',
+      node_path: METADATA,
-        items: seeds,
+      field_name: 'seed',
-      });
+      items: seeds,
-    }
+    });
 
     if (graph.nodes[CANVAS_COHERENCE_NOISE]) {
       zipped.push({
@@ -77,13 +78,13 @@ export const prepareLinearUIBatch = (
         });
       }
 
-      if (graph.nodes[METADATA_ACCUMULATOR]) {
+      // add to metadata
-        firstBatchDatumList.push({
+      removeMetadata(graph, 'seed');
-          node_path: METADATA_ACCUMULATOR,
+      firstBatchDatumList.push({
-          field_name: 'seed',
+        node_path: METADATA,
-          items: seeds,
+        field_name: 'seed',
-        });
+        items: seeds,
-      }
+      });
 
       if (graph.nodes[CANVAS_COHERENCE_NOISE]) {
         firstBatchDatumList.push({
@@ -106,13 +107,15 @@ export const prepareLinearUIBatch = (
           items: seeds,
         });
       }
-      if (graph.nodes[METADATA_ACCUMULATOR]) {
+
-        secondBatchDatumList.push({
+      // add to metadata
-          node_path: METADATA_ACCUMULATOR,
+      removeMetadata(graph, 'seed');
-          field_name: 'seed',
+      secondBatchDatumList.push({
-          items: seeds,
+        node_path: METADATA,
-        });
+        field_name: 'seed',
-      }
+        items: seeds,
+      });
+
       if (graph.nodes[CANVAS_COHERENCE_NOISE]) {
         secondBatchDatumList.push({
           node_path: CANVAS_COHERENCE_NOISE,
@@ -137,13 +140,13 @@ export const prepareLinearUIBatch = (
       });
     }
 
-    if (graph.nodes[METADATA_ACCUMULATOR]) {
+    // add to metadata
-      firstBatchDatumList.push({
+    removeMetadata(graph, 'positive_prompt');
-        node_path: METADATA_ACCUMULATOR,
+    firstBatchDatumList.push({
-        field_name: 'positive_prompt',
+      node_path: METADATA,
-        items: extendedPrompts,
+      field_name: 'positive_prompt',
-      });
+      items: extendedPrompts,
-    }
+    });
 
     if (shouldConcatSDXLStylePrompt && model?.base_model === 'sdxl') {
       unset(graph.nodes[METADATA_ACCUMULATOR], 'positive_style_prompt');
@@ -160,13 +163,13 @@ export const prepareLinearUIBatch = (
         });
       }
 
-      if (graph.nodes[METADATA_ACCUMULATOR]) {
+      // add to metadata
-        firstBatchDatumList.push({
+      removeMetadata(graph, 'positive_style_prompt');
-          node_path: METADATA_ACCUMULATOR,
+      firstBatchDatumList.push({
-          field_name: 'positive_style_prompt',
+        node_path: METADATA,
-          items: stylePrompts,
+        field_name: 'positive_style_prompt',
-        });
+        items: extendedPrompts,
-      }
+      });
     }
 
     data.push(firstBatchDatumList);

invokeai/frontend/web/src/features/nodes/util/graphBuilders/buildLinearImageToImageGraph.ts

@@ -21,13 +21,13 @@ import {
   IMAGE_TO_LATENTS,
   LATENTS_TO_IMAGE,
   MAIN_MODEL_LOADER,
-  METADATA_ACCUMULATOR,
   NEGATIVE_CONDITIONING,
   NOISE,
   POSITIVE_CONDITIONING,
   RESIZE,
   SEAMLESS,
 } from './constants';
+import { addCoreMetadataNode } from './metadata';
 
 /**
  * Builds the Image to Image tab graph.
@@ -311,10 +311,7 @@ export const buildLinearImageToImageGraph = (
     });
   }
 
-  // add metadata accumulator, which is only mostly populated - some fields are added later
+  addCoreMetadataNode(graph, {
-  graph.nodes[METADATA_ACCUMULATOR] = {
-    id: METADATA_ACCUMULATOR,
-    type: 'metadata_accumulator',
     generation_mode: 'img2img',
     cfg_scale,
     height,
@@ -326,25 +323,9 @@ export const buildLinearImageToImageGraph = (
     steps,
     rand_device: use_cpu ? 'cpu' : 'cuda',
     scheduler,
-    vae: undefined, // option; set in addVAEToGraph
-    controlnets: [], // populated in addControlNetToLinearGraph
-    loras: [], // populated in addLoRAsToGraph
-    ipAdapters: [], // populated in addIPAdapterToLinearGraph
-    t2iAdapters: [], // populated in addT2IAdapterToLinearGraph
     clip_skip: clipSkip,
     strength,
     init_image: initialImage.imageName,
-  };
-
-  graph.edges.push({
-    source: {
-      node_id: METADATA_ACCUMULATOR,
-      field: 'metadata',
-    },
-    destination: {
-      node_id: LATENTS_TO_IMAGE,
-      field: 'metadata',
-    },
   });
 
   // Add Seamless To Graph

invokeai/frontend/web/src/features/nodes/util/graphBuilders/buildLinearSDXLImageToImageGraph.ts

@@ -18,7 +18,6 @@ import { addWatermarkerToGraph } from './addWatermarkerToGraph';
 import {
   IMAGE_TO_LATENTS,
   LATENTS_TO_IMAGE,
-  METADATA_ACCUMULATOR,
   NEGATIVE_CONDITIONING,
   NOISE,
   POSITIVE_CONDITIONING,
@@ -30,6 +29,7 @@ import {
   SEAMLESS,
 } from './constants';
 import { buildSDXLStylePrompts } from './helpers/craftSDXLStylePrompt';
+import { addCoreMetadataNode } from './metadata';
 
 /**
  * Builds the Image to Image tab graph.
@@ -331,10 +331,7 @@ export const buildLinearSDXLImageToImageGraph = (
     });
   }
 
-  // add metadata accumulator, which is only mostly populated - some fields are added later
+  addCoreMetadataNode(graph, {
-  graph.nodes[METADATA_ACCUMULATOR] = {
-    id: METADATA_ACCUMULATOR,
-    type: 'metadata_accumulator',
     generation_mode: 'sdxl_img2img',
     cfg_scale,
     height,
@@ -346,26 +343,10 @@ export const buildLinearSDXLImageToImageGraph = (
     steps,
     rand_device: use_cpu ? 'cpu' : 'cuda',
     scheduler,
-    vae: undefined,
+    strength,
-    controlnets: [],
-    loras: [],
-    ipAdapters: [],
-    t2iAdapters: [],
-    strength: strength,
     init_image: initialImage.imageName,
     positive_style_prompt: positiveStylePrompt,
     negative_style_prompt: negativeStylePrompt,
-  };
-
-  graph.edges.push({
-    source: {
-      node_id: METADATA_ACCUMULATOR,
-      field: 'metadata',
-    },
-    destination: {
-      node_id: LATENTS_TO_IMAGE,
-      field: 'metadata',
-    },
   });
 
   // Add Seamless To Graph

invokeai/frontend/web/src/features/nodes/util/graphBuilders/buildLinearSDXLTextToImageGraph.ts

@@ -11,9 +11,9 @@ import { addSeamlessToLinearGraph } from './addSeamlessToLinearGraph';
 import { addT2IAdaptersToLinearGraph } from './addT2IAdapterToLinearGraph';
 import { addVAEToGraph } from './addVAEToGraph';
 import { addWatermarkerToGraph } from './addWatermarkerToGraph';
+import { addCoreMetadataNode } from './metadata';
 import {
   LATENTS_TO_IMAGE,
-  METADATA_ACCUMULATOR,
   NEGATIVE_CONDITIONING,
   NOISE,
   POSITIVE_CONDITIONING,
@@ -225,10 +225,7 @@ export const buildLinearSDXLTextToImageGraph = (
     ],
   };
 
-  // add metadata accumulator, which is only mostly populated - some fields are added later
+  addCoreMetadataNode(graph, {
-  graph.nodes[METADATA_ACCUMULATOR] = {
-    id: METADATA_ACCUMULATOR,
-    type: 'metadata_accumulator',
     generation_mode: 'sdxl_txt2img',
     cfg_scale,
     height,
@@ -240,24 +237,8 @@ export const buildLinearSDXLTextToImageGraph = (
     steps,
     rand_device: use_cpu ? 'cpu' : 'cuda',
     scheduler,
-    vae: undefined,
-    controlnets: [],
-    loras: [],
-    ipAdapters: [],
-    t2iAdapters: [],
     positive_style_prompt: positiveStylePrompt,
     negative_style_prompt: negativeStylePrompt,
-  };
-
-  graph.edges.push({
-    source: {
-      node_id: METADATA_ACCUMULATOR,
-      field: 'metadata',
-    },
-    destination: {
-      node_id: LATENTS_TO_IMAGE,
-      field: 'metadata',
-    },
   });
 
   // Add Seamless To Graph

invokeai/frontend/web/src/features/nodes/util/graphBuilders/buildLinearTextToImageGraph.ts

@@ -15,12 +15,12 @@ import { addSeamlessToLinearGraph } from './addSeamlessToLinearGraph';
 import { addT2IAdaptersToLinearGraph } from './addT2IAdapterToLinearGraph';
 import { addVAEToGraph } from './addVAEToGraph';
 import { addWatermarkerToGraph } from './addWatermarkerToGraph';
+import { addCoreMetadataNode } from './metadata';
 import {
   CLIP_SKIP,
   DENOISE_LATENTS,
   LATENTS_TO_IMAGE,
   MAIN_MODEL_LOADER,
-  METADATA_ACCUMULATOR,
   NEGATIVE_CONDITIONING,
   NOISE,
   ONNX_MODEL_LOADER,
@@ -48,10 +48,6 @@ export const buildLinearTextToImageGraph = (
     seamlessXAxis,
     seamlessYAxis,
     seed,
-    hrfWidth,
-    hrfHeight,
-    hrfStrength,
-    hrfEnabled: hrfEnabled,
   } = state.generation;
 
   const use_cpu = shouldUseCpuNoise;
@@ -238,10 +234,7 @@ export const buildLinearTextToImageGraph = (
     ],
   };
 
-  // add metadata accumulator, which is only mostly populated - some fields are added later
+  addCoreMetadataNode(graph, {
-  graph.nodes[METADATA_ACCUMULATOR] = {
-    id: METADATA_ACCUMULATOR,
-    type: 'metadata_accumulator',
     generation_mode: 'txt2img',
     cfg_scale,
     height,
@@ -253,26 +246,7 @@ export const buildLinearTextToImageGraph = (
     steps,
     rand_device: use_cpu ? 'cpu' : 'cuda',
     scheduler,
-    vae: undefined, // option; set in addVAEToGraph
-    controlnets: [], // populated in addControlNetToLinearGraph
-    loras: [], // populated in addLoRAsToGraph
-    ipAdapters: [], // populated in addIPAdapterToLinearGraph
-    t2iAdapters: [], // populated in addT2IAdapterToLinearGraph
     clip_skip: clipSkip,
-    hrf_width: hrfEnabled ? hrfWidth : undefined,
-    hrf_height: hrfEnabled ? hrfHeight : undefined,
-    hrf_strength: hrfEnabled ? hrfStrength : undefined,
-  };
-
-  graph.edges.push({
-    source: {
-      node_id: METADATA_ACCUMULATOR,
-      field: 'metadata',
-    },
-    destination: {
-      node_id: LATENTS_TO_IMAGE,
-      field: 'metadata',
-    },
   });
 
   // Add Seamless To Graph

invokeai/frontend/web/src/features/nodes/util/graphBuilders/buildNodesGraph.ts

@@ -35,7 +35,7 @@ export const buildNodesGraph = (nodesState: NodesState): Graph => {
   const { nodes, edges } = nodesState;
 
   const filteredNodes = nodes.filter(isInvocationNode);
-  const workflowJSON = JSON.stringify(buildWorkflow(nodesState));
+  // const workflowJSON = JSON.stringify(buildWorkflow(nodesState));
 
   // Reduce the node editor nodes into invocation graph nodes
   const parsedNodes = filteredNodes.reduce<NonNullable<Graph['nodes']>>(
@@ -68,7 +68,8 @@ export const buildNodesGraph = (nodesState: NodesState): Graph => {
 
       if (embedWorkflow) {
         // add the workflow to the node
-        Object.assign(graphNode, { workflow: workflowJSON });
+        // Object.assign(graphNode, { workflow: workflowJSON });
+        Object.assign(graphNode, { workflow: buildWorkflow(nodesState) });
       }
 
       // Add it to the nodes object

invokeai/frontend/web/src/features/nodes/util/graphBuilders/constants.ts

@@ -56,7 +56,15 @@ export const IP_ADAPTER = 'ip_adapter';
 export const DYNAMIC_PROMPT = 'dynamic_prompt';
 export const IMAGE_COLLECTION = 'image_collection';
 export const IMAGE_COLLECTION_ITERATE = 'image_collection_iterate';
+export const METADATA = 'core_metadata';
+export const BATCH_METADATA = 'batch_metadata';
+export const BATCH_METADATA_COLLECT = 'batch_metadata_collect';
+export const BATCH_SEED = 'batch_seed';
+export const BATCH_PROMPT = 'batch_prompt';
+export const BATCH_STYLE_PROMPT = 'batch_style_prompt';
+export const METADATA_COLLECT = 'metadata_collect';
 export const METADATA_ACCUMULATOR = 'metadata_accumulator';
+export const MERGE_METADATA = 'merge_metadata';
 export const REALESRGAN = 'esrgan';
 export const DIVIDE = 'divide';
 export const SCALE = 'scale_image';

invokeai/frontend/web/src/features/nodes/util/graphBuilders/metadata.ts

@@ -0,0 +1,58 @@
+import { NonNullableGraph } from 'features/nodes/types/types';
+import { CoreMetadataInvocation } from 'services/api/types';
+import { JsonObject } from 'type-fest';
+import { METADATA, SAVE_IMAGE } from './constants';
+
+export const addCoreMetadataNode = (
+  graph: NonNullableGraph,
+  metadata: Partial<CoreMetadataInvocation> | JsonObject
+): void => {
+  graph.nodes[METADATA] = {
+    id: METADATA,
+    type: 'core_metadata',
+    ...metadata,
+  };
+
+  graph.edges.push({
+    source: {
+      node_id: METADATA,
+      field: 'metadata',
+    },
+    destination: {
+      node_id: SAVE_IMAGE,
+      field: 'metadata',
+    },
+  });
+
+  return;
+};
+
+export const upsertMetadata = (
+  graph: NonNullableGraph,
+  metadata: Partial<CoreMetadataInvocation> | JsonObject
+): void => {
+  const metadataNode = graph.nodes[METADATA] as
+    | CoreMetadataInvocation
+    | undefined;
+
+  if (!metadataNode) {
+    return;
+  }
+
+  Object.assign(metadataNode, metadata);
+};
+
+export const removeMetadata = (
+  graph: NonNullableGraph,
+  key: keyof CoreMetadataInvocation
+): void => {
+  const metadataNode = graph.nodes[METADATA] as
+    | CoreMetadataInvocation
+    | undefined;
+
+  if (!metadataNode) {
+    return;
+  }
+
+  delete metadataNode[key];
+};

invokeai/frontend/web/src/features/nodes/util/parseSchema.ts

@@ -4,7 +4,6 @@ import { reduce, startCase } from 'lodash-es';
 import { OpenAPIV3_1 } from 'openapi-types';
 import { AnyInvocationType } from 'services/events/types';
 import {
-  FieldType,
   InputFieldTemplate,
   InvocationSchemaObject,
   InvocationTemplate,
@@ -16,18 +15,11 @@ import {
 } from '../types/types';
 import { buildInputFieldTemplate, getFieldType } from './fieldTemplateBuilders';
 
-const RESERVED_INPUT_FIELD_NAMES = ['id', 'type', 'metadata', 'use_cache'];
+const RESERVED_INPUT_FIELD_NAMES = ['id', 'type', 'use_cache'];
 const RESERVED_OUTPUT_FIELD_NAMES = ['type'];
-const RESERVED_FIELD_TYPES = [
+const RESERVED_FIELD_TYPES = ['IsIntermediate'];
-  'WorkflowField',
-  'MetadataField',
-  'IsIntermediate',
-];
 
-const invocationDenylist: AnyInvocationType[] = [
+const invocationDenylist: AnyInvocationType[] = ['graph'];
-  'graph',
-  'metadata_accumulator',
-];
 
 const isReservedInputField = (nodeType: string, fieldName: string) => {
   if (RESERVED_INPUT_FIELD_NAMES.includes(fieldName)) {
@@ -42,7 +34,7 @@ const isReservedInputField = (nodeType: string, fieldName: string) => {
   return false;
 };
 
-const isReservedFieldType = (fieldType: FieldType) => {
+const isReservedFieldType = (fieldType: string) => {
   if (RESERVED_FIELD_TYPES.includes(fieldType)) {
     return true;
   }
@@ -86,6 +78,7 @@ export const parseSchema = (
     const tags = schema.tags ?? [];
     const description = schema.description ?? '';
     const version = schema.version;
+    let withWorkflow = false;
 
     const inputs = reduce(
       schema.properties,
@@ -112,7 +105,7 @@ export const parseSchema = (
 
         const fieldType = property.ui_type ?? getFieldType(property);
 
-        if (!isFieldType(fieldType)) {
+        if (!fieldType) {
           logger('nodes').warn(
             {
               node: type,
@@ -120,11 +113,16 @@ export const parseSchema = (
               fieldType,
               field: parseify(property),
             },
-            'Skipping unknown input field type'
+            'Missing input field type'
           );
           return inputsAccumulator;
         }
 
+        if (fieldType === 'WorkflowField') {
+          withWorkflow = true;
+          return inputsAccumulator;
+        }
+
         if (isReservedFieldType(fieldType)) {
           logger('nodes').trace(
             {
@@ -133,7 +131,20 @@ export const parseSchema = (
               fieldType,
               field: parseify(property),
             },
-            'Skipping reserved field type'
+            `Skipping reserved input field type: ${fieldType}`
+          );
+          return inputsAccumulator;
+        }
+
+        if (!isFieldType(fieldType)) {
+          logger('nodes').warn(
+            {
+              node: type,
+              fieldName: propertyName,
+              fieldType,
+              field: parseify(property),
+            },
+            `Skipping unknown input field type: ${fieldType}`
           );
           return inputsAccumulator;
         }
@@ -146,7 +157,7 @@ export const parseSchema = (
         );
 
         if (!field) {
-          logger('nodes').debug(
+          logger('nodes').warn(
             {
               node: type,
               fieldName: propertyName,
@@ -248,6 +259,7 @@ export const parseSchema = (
       inputs,
       outputs,
       useCache,
+      withWorkflow,
     };
 
     Object.assign(invocationsAccumulator, { [type]: invocation });

invokeai/frontend/web/src/services/api/endpoints/images.ts

@@ -1,6 +1,7 @@
 import { EntityState, Update } from '@reduxjs/toolkit';
 import { fetchBaseQuery } from '@reduxjs/toolkit/dist/query';
 import { PatchCollection } from '@reduxjs/toolkit/dist/query/core/buildThunks';
+import { logger } from 'app/logging/logger';
 import {
   ASSETS_CATEGORIES,
   BoardId,
@@ -8,6 +9,7 @@ import {
   IMAGE_LIMIT,
 } from 'features/gallery/store/types';
 import {
+  CoreMetadata,
   ImageMetadataAndWorkflow,
   zCoreMetadata,
 } from 'features/nodes/types/types';
@@ -23,7 +25,6 @@ import {
   ListImagesArgs,
   OffsetPaginatedResults_ImageDTO_,
   PostUploadAction,
-  UnsafeImageMetadata,
 } from '../types';
 import {
   getCategories,
@@ -114,11 +115,24 @@ export const imagesApi = api.injectEndpoints({
       ],
       keepUnusedDataFor: 86400, // 24 hours
     }),
-    getImageMetadata: build.query<UnsafeImageMetadata, string>({
+    getImageMetadata: build.query<CoreMetadata | undefined, string>({
       query: (image_name) => ({ url: `images/i/${image_name}/metadata` }),
       providesTags: (result, error, image_name) => [
         { type: 'ImageMetadata', id: image_name },
       ],
+      transformResponse: (
+        response: paths['/api/v1/images/i/{image_name}/metadata']['get']['responses']['200']['content']['application/json']
+      ) => {
+        if (response) {
+          const result = zCoreMetadata.safeParse(response);
+          if (result.success) {
+            return result.data;
+          } else {
+            logger('images').warn('Problem parsing metadata');
+          }
+        }
+        return;
+      },
       keepUnusedDataFor: 86400, // 24 hours
     }),
     getImageMetadataFromFile: build.query<

invokeai/frontend/web/src/services/api/endpoints/workflows.ts

@@ -0,0 +1,31 @@
+import { logger } from 'app/logging/logger';
+import { Workflow, zWorkflow } from 'features/nodes/types/types';
+import { api } from '..';
+import { paths } from '../schema';
+
+export const workflowsApi = api.injectEndpoints({
+  endpoints: (build) => ({
+    getWorkflow: build.query<Workflow | undefined, string>({
+      query: (workflow_id) => `workflows/i/${workflow_id}`,
+      keepUnusedDataFor: 86400, // 24 hours
+      providesTags: (result, error, workflow_id) => [
+        { type: 'Workflow', id: workflow_id },
+      ],
+      transformResponse: (
+        response: paths['/api/v1/workflows/i/{workflow_id}']['get']['responses']['200']['content']['application/json']
+      ) => {
+        if (response) {
+          const result = zWorkflow.safeParse(response);
+          if (result.success) {
+            return result.data;
+          } else {
+            logger('images').warn('Problem parsing metadata');
+          }
+        }
+        return;
+      },
+    }),
+  }),
+});
+
+export const { useGetWorkflowQuery } = workflowsApi;

invokeai/frontend/web/src/services/api/index.ts

@@ -37,6 +37,7 @@ export const tagTypes = [
   'ControlNetModel',
   'LoRAModel',
   'SDXLRefinerModel',
+  'Workflow',
 ] as const;
 export type ApiTagDescription = TagDescription<(typeof tagTypes)[number]>;
 export const LIST_TAG = 'LIST';

invokeai/frontend/web/src/services/api/types.ts

@@ -27,14 +27,6 @@ export type BatchConfig =
 
 export type EnqueueBatchResult = components['schemas']['EnqueueBatchResult'];
 
-/**
- * This is an unsafe type; the object inside is not guaranteed to be valid.
- */
-export type UnsafeImageMetadata = {
-  metadata: s['CoreMetadata'];
-  graph: NonNullable<s['Graph']>;
-};
-
 export type _InputField = s['_InputField'];
 export type _OutputField = s['_OutputField'];
 
@@ -50,7 +42,6 @@ export type ImageChanges = s['ImageRecordChanges'];
 export type ImageCategory = s['ImageCategory'];
 export type ResourceOrigin = s['ResourceOrigin'];
 export type ImageField = s['ImageField'];
-export type ImageMetadata = s['ImageMetadata'];
 export type OffsetPaginatedResults_BoardDTO_ =
   s['OffsetPaginatedResults_BoardDTO_'];
 export type OffsetPaginatedResults_ImageDTO_ =
@@ -145,13 +136,19 @@ export type ImageCollectionInvocation = s['ImageCollectionInvocation'];
 export type MainModelLoaderInvocation = s['MainModelLoaderInvocation'];
 export type OnnxModelLoaderInvocation = s['OnnxModelLoaderInvocation'];
 export type LoraLoaderInvocation = s['LoraLoaderInvocation'];
-export type MetadataAccumulatorInvocation = s['MetadataAccumulatorInvocation'];
 export type ESRGANInvocation = s['ESRGANInvocation'];
 export type DivideInvocation = s['DivideInvocation'];
 export type ImageNSFWBlurInvocation = s['ImageNSFWBlurInvocation'];
 export type ImageWatermarkInvocation = s['ImageWatermarkInvocation'];
 export type SeamlessModeInvocation = s['SeamlessModeInvocation'];
 export type SaveImageInvocation = s['SaveImageInvocation'];
+export type MetadataInvocation = s['MetadataInvocation'];
+export type CoreMetadataInvocation = s['CoreMetadataInvocation'];
+export type MetadataItemInvocation = s['MetadataItemInvocation'];
+export type MergeMetadataInvocation = s['MergeMetadataInvocation'];
+export type IPAdapterMetadataField = s['IPAdapterMetadataField'];
+export type T2IAdapterField = s['T2IAdapterField'];
+export type LoRAMetadataField = s['LoRAMetadataField'];
 
 // ControlNet Nodes
 export type ControlNetInvocation = s['ControlNetInvocation'];

tests/nodes/test_node_graph.py

@@ -10,7 +10,12 @@ from invokeai.app.invocations.baseinvocation import (
 )
 from invokeai.app.invocations.image import ShowImageInvocation
 from invokeai.app.invocations.math import AddInvocation, SubtractInvocation
-from invokeai.app.invocations.primitives import FloatInvocation, IntegerInvocation
+from invokeai.app.invocations.primitives import (
+    FloatCollectionInvocation,
+    FloatInvocation,
+    IntegerInvocation,
+    StringInvocation,
+)
 from invokeai.app.invocations.upscale import ESRGANInvocation
 from invokeai.app.services.shared.default_graphs import create_text_to_image
 from invokeai.app.services.shared.graph import (
@@ -27,8 +32,11 @@ from invokeai.app.services.shared.graph import (
 )
 
 from .test_nodes import (
+    AnyTypeTestInvocation,
     ImageToImageTestInvocation,
     ListPassThroughInvocation,
+    PolymorphicStringTestInvocation,
+    PromptCollectionTestInvocation,
     PromptTestInvocation,
     TextToImageTestInvocation,
 )
@@ -692,6 +700,144 @@ def test_ints_do_not_accept_floats():
         g.add_edge(e)
 
 
+def test_polymorphic_accepts_single():
+    g = Graph()
+    n1 = StringInvocation(id="1", value="banana")
+    n2 = PolymorphicStringTestInvocation(id="2")
+    g.add_node(n1)
+    g.add_node(n2)
+    e1 = create_edge(n1.id, "value", n2.id, "value")
+    # Not throwing on this line is sufficient
+    g.add_edge(e1)
+
+
+def test_polymorphic_accepts_collection_of_same_base_type():
+    g = Graph()
+    n1 = PromptCollectionTestInvocation(id="1", collection=["banana", "sundae"])
+    n2 = PolymorphicStringTestInvocation(id="2")
+    g.add_node(n1)
+    g.add_node(n2)
+    e1 = create_edge(n1.id, "collection", n2.id, "value")
+    # Not throwing on this line is sufficient
+    g.add_edge(e1)
+
+
+def test_polymorphic_does_not_accept_collection_of_different_base_type():
+    g = Graph()
+    n1 = FloatCollectionInvocation(id="1", collection=[1.0, 2.0, 3.0])
+    n2 = PolymorphicStringTestInvocation(id="2")
+    g.add_node(n1)
+    g.add_node(n2)
+    e1 = create_edge(n1.id, "collection", n2.id, "value")
+    with pytest.raises(InvalidEdgeError):
+        g.add_edge(e1)
+
+
+def test_polymorphic_does_not_accept_generic_collection():
+    g = Graph()
+    n1 = IntegerInvocation(id="1", value=1)
+    n2 = IntegerInvocation(id="2", value=2)
+    n3 = CollectInvocation(id="3")
+    n4 = PolymorphicStringTestInvocation(id="4")
+    g.add_node(n1)
+    g.add_node(n2)
+    g.add_node(n3)
+    g.add_node(n4)
+    e1 = create_edge(n1.id, "value", n3.id, "item")
+    e2 = create_edge(n2.id, "value", n3.id, "item")
+    e3 = create_edge(n3.id, "collection", n4.id, "value")
+    g.add_edge(e1)
+    g.add_edge(e2)
+    with pytest.raises(InvalidEdgeError):
+        g.add_edge(e3)
+
+
+def test_any_accepts_integer():
+    g = Graph()
+    n1 = IntegerInvocation(id="1", value=1)
+    n2 = AnyTypeTestInvocation(id="2")
+    g.add_node(n1)
+    g.add_node(n2)
+    e = create_edge(n1.id, "value", n2.id, "value")
+    # Not throwing on this line is sufficient
+    g.add_edge(e)
+
+
+def test_any_accepts_string():
+    g = Graph()
+    n1 = StringInvocation(id="1", value="banana sundae")
+    n2 = AnyTypeTestInvocation(id="2")
+    g.add_node(n1)
+    g.add_node(n2)
+    e = create_edge(n1.id, "value", n2.id, "value")
+    # Not throwing on this line is sufficient
+    g.add_edge(e)
+
+
+def test_any_accepts_generic_collection():
+    g = Graph()
+    n1 = IntegerInvocation(id="1", value=1)
+    n2 = IntegerInvocation(id="2", value=2)
+    n3 = CollectInvocation(id="3")
+    n4 = AnyTypeTestInvocation(id="4")
+    g.add_node(n1)
+    g.add_node(n2)
+    g.add_node(n3)
+    g.add_node(n4)
+    e1 = create_edge(n1.id, "value", n3.id, "item")
+    e2 = create_edge(n2.id, "value", n3.id, "item")
+    e3 = create_edge(n3.id, "collection", n4.id, "value")
+    g.add_edge(e1)
+    g.add_edge(e2)
+    # Not throwing on this line is sufficient
+    g.add_edge(e3)
+
+
+def test_any_accepts_prompt_collection():
+    g = Graph()
+    n1 = PromptCollectionTestInvocation(id="1", collection=["banana", "sundae"])
+    n2 = AnyTypeTestInvocation(id="2")
+    g.add_node(n1)
+    g.add_node(n2)
+    e = create_edge(n1.id, "collection", n2.id, "value")
+    # Not throwing on this line is sufficient
+    g.add_edge(e)
+
+
+def test_any_accepts_any():
+    g = Graph()
+    n1 = AnyTypeTestInvocation(id="1")
+    n2 = AnyTypeTestInvocation(id="2")
+    g.add_node(n1)
+    g.add_node(n2)
+    e = create_edge(n1.id, "value", n2.id, "value")
+    # Not throwing on this line is sufficient
+    g.add_edge(e)
+
+
+def test_iterate_accepts_collection():
+    """We need to update the validation for Collect -> Iterate to traverse to the Iterate
+    node's output and compare that against the item type of the Collect node's collection. Until
+    then, Collect nodes may not output into Iterate nodes."""
+    g = Graph()
+    n1 = IntegerInvocation(id="1", value=1)
+    n2 = IntegerInvocation(id="2", value=2)
+    n3 = CollectInvocation(id="3")
+    n4 = IterateInvocation(id="4")
+    g.add_node(n1)
+    g.add_node(n2)
+    g.add_node(n3)
+    g.add_node(n4)
+    e1 = create_edge(n1.id, "value", n3.id, "item")
+    e2 = create_edge(n2.id, "value", n3.id, "item")
+    e3 = create_edge(n3.id, "collection", n4.id, "collection")
+    g.add_edge(e1)
+    g.add_edge(e2)
+    # Once we fix the validation logic as described, this should should not raise an error
+    with pytest.raises(InvalidEdgeError, match="Cannot connect collector to iterator"):
+        g.add_edge(e3)
+
+
 def test_graph_can_generate_schema():
     # Not throwing on this line is sufficient
     # NOTE: if this test fails, it's PROBABLY because a new invocation type is breaking schema generation

tests/nodes/test_nodes.py

@@ -81,6 +81,29 @@ class PromptCollectionTestInvocation(BaseInvocation):
         return PromptCollectionTestInvocationOutput(collection=self.collection.copy())
 
 
+@invocation_output("test_any_output")
+class AnyTypeTestInvocationOutput(BaseInvocationOutput):
+    value: Any = Field()
+
+
+@invocation("test_any")
+class AnyTypeTestInvocation(BaseInvocation):
+    value: Any = Field(default=None)
+
+    def invoke(self, context: InvocationContext) -> AnyTypeTestInvocationOutput:
+        return AnyTypeTestInvocationOutput(value=self.value)
+
+
+@invocation("test_polymorphic")
+class PolymorphicStringTestInvocation(BaseInvocation):
+    value: Union[str, list[str]] = Field(default="")
+
+    def invoke(self, context: InvocationContext) -> PromptCollectionTestInvocationOutput:
+        if isinstance(self.value, str):
+            return PromptCollectionTestInvocationOutput(collection=[self.value])
+        return PromptCollectionTestInvocationOutput(collection=self.value)
+
+
 # Importing these must happen after test invocations are defined or they won't register
 from invokeai.app.services.events.events_base import EventServiceBase  # noqa: E402
 from invokeai.app.services.shared.graph import Edge, EdgeConnection  # noqa: E402