fix(nodes): fix metadata test

- make metadata a service
- rip out pydantic validation, implement metadata parsing as simple functions
- update tests
- address other minor feedback items

.github/stale.yaml

@@ -0,0 +1,19 @@
+# Number of days of inactivity before an issue becomes stale
+daysUntilStale: 28
+# Number of days of inactivity before a stale issue is closed
+daysUntilClose: 14
+# Issues with these labels will never be considered stale
+exemptLabels:
+  - pinned
+  - security
+# Label to use when marking an issue as stale
+staleLabel: stale
+# Comment to post when marking an issue as stale. Set to `false` to disable
+markComment: >
+  This issue has been automatically marked as stale because it has not had
+  recent activity. It will be closed if no further activity occurs. Please
+  update the ticket if this is still a problem on the latest release.
+# Comment to post when closing a stale issue. Set to `false` to disable
+closeComment: >
+  Due to inactivity, this issue has been automatically closed. If this is
+  still a problem on the latest release, please recreate the issue.

.github/workflows/build-container.yml

@@ -18,6 +18,7 @@ on:
 
 permissions:
   contents: write
+  packages: write
 
 jobs:
   docker:

README.md

@@ -84,7 +84,7 @@ installing lots of models.
 
 6. Wait while the installer does its thing. After installing the software,
 the installer will launch a script that lets you configure InvokeAI and
-select a set of starting image generaiton models.
+select a set of starting image generation models.
 
 7. Find the folder that InvokeAI was installed into (it is not the
 same as the unpacked zip file directory!) The default location of this
@@ -145,7 +145,7 @@ not supported.
     _For Linux with an AMD GPU:_
 
     ```sh
-    pip install InvokeAI --use-pep517 --extra-index-url https://download.pytorch.org/whl/rocm5.2
+    pip install InvokeAI --use-pep517 --extra-index-url https://download.pytorch.org/whl/rocm5.4.2
     ```
 
     _For Macintoshes, either Intel or M1/M2:_

docs/contributing/INVOCATIONS.md

@@ -1,10 +1,18 @@
 # Invocations
 
-Invocations represent a single operation, its inputs, and its outputs. These operations and their outputs can be chained together to generate and modify images.
+Invocations represent a single operation, its inputs, and its outputs. These
+operations and their outputs can be chained together to generate and modify
+images.
 
 ## Creating a new invocation
 
-To create a new invocation, either find the appropriate module file in `/ldm/invoke/app/invocations` to add your invocation to, or create a new one in that folder. All invocations in that folder will be discovered and made available to the CLI and API automatically. Invocations make use of [typing](https://docs.python.org/3/library/typing.html) and [pydantic](https://pydantic-docs.helpmanual.io/) for validation and integration into the CLI and API.
+To create a new invocation, either find the appropriate module file in
+`/ldm/invoke/app/invocations` to add your invocation to, or create a new one in
+that folder. All invocations in that folder will be discovered and made
+available to the CLI and API automatically. Invocations make use of
+[typing](https://docs.python.org/3/library/typing.html) and
+[pydantic](https://pydantic-docs.helpmanual.io/) for validation and integration
+into the CLI and API.
 
 An invocation looks like this:
 
@@ -41,34 +49,54 @@ class UpscaleInvocation(BaseInvocation):
 Each portion is important to implement correctly.
 
 ### Class definition and type
+
 ```py
 class UpscaleInvocation(BaseInvocation):
     """Upscales an image."""
     type: Literal['upscale'] = 'upscale'
 ```
-All invocations must derive from `BaseInvocation`. They should have a docstring that declares what they do in a single, short line. They should also have a `type` with a type hint that's `Literal["command_name"]`, where `command_name` is what the user will type on the CLI or use in the API to create this invocation. The `command_name` must be unique. The `type` must be assigned to the value of the literal in the type hint.
+
+All invocations must derive from `BaseInvocation`. They should have a docstring
+that declares what they do in a single, short line. They should also have a
+`type` with a type hint that's `Literal["command_name"]`, where `command_name`
+is what the user will type on the CLI or use in the API to create this
+invocation. The `command_name` must be unique. The `type` must be assigned to
+the value of the literal in the type hint.
 
 ### Inputs
+
 ```py
     # Inputs
     image: Union[ImageField,None] = Field(description="The input image")
     strength: float               = Field(default=0.75, gt=0, le=1, description="The strength")
     level: Literal[2,4]           = Field(default=2, description="The upscale level")
 ```
-Inputs consist of three parts: a name, a type hint, and a `Field` with default, description, and validation information. For example:
+
+Inputs consist of three parts: a name, a type hint, and a `Field` with default,
+description, and validation information. For example:
+
 | Part      | Value                                                         | Description                                                                                                               |
-| ---- | ----- | ----------- |
+| --------- | ------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------- |
 | Name      | `strength`                                                    | This field is referred to as `strength`                                                                                   |
 | Type Hint | `float`                                                       | This field must be of type `float`                                                                                        |
 | Field     | `Field(default=0.75, gt=0, le=1, description="The strength")` | The default value is `0.75`, the value must be in the range (0,1], and help text will show "The strength" for this field. |
 
-Notice that `image` has type `Union[ImageField,None]`. The `Union` allows this field to be parsed with `None` as a value, which enables linking to previous invocations. All fields should either provide a default value or allow `None` as a value, so that they can be overwritten with a linked output from another invocation.
+Notice that `image` has type `Union[ImageField,None]`. The `Union` allows this
+field to be parsed with `None` as a value, which enables linking to previous
+invocations. All fields should either provide a default value or allow `None` as
+a value, so that they can be overwritten with a linked output from another
+invocation.
 
-The special type `ImageField` is also used here. All images are passed as `ImageField`, which protects them from pydantic validation errors (since images only ever come from links).
+The special type `ImageField` is also used here. All images are passed as
+`ImageField`, which protects them from pydantic validation errors (since images
+only ever come from links).
 
-Finally, note that for all linking, the `type` of the linked fields must match. If the `name` also matches, then the field can be **automatically linked** to a previous invocation by name and matching.
+Finally, note that for all linking, the `type` of the linked fields must match.
+If the `name` also matches, then the field can be **automatically linked** to a
+previous invocation by name and matching.
 
 ### Invoke Function
+
 ```py
     def invoke(self, context: InvocationContext) -> ImageOutput:
         image = context.services.images.get(self.image.image_type, self.image.image_name)
@@ -88,13 +116,22 @@ Finally, note that for all linking, the `type` of the linked fields must match.
             image = ImageField(image_type = image_type, image_name = image_name)
         )
 ```
-The `invoke` function is the last portion of an invocation. It is provided an `InvocationContext` which contains services to perform work as well as a `session_id` for use as needed. It should return a class with output values that derives from `BaseInvocationOutput`.
 
-Before being called, the invocation will have all of its fields set from defaults, inputs, and finally links (overriding in that order).
+The `invoke` function is the last portion of an invocation. It is provided an
+`InvocationContext` which contains services to perform work as well as a
+`session_id` for use as needed. It should return a class with output values that
+derives from `BaseInvocationOutput`.
 
-Assume that this invocation may be running simultaneously with other invocations, may be running on another machine, or in other interesting scenarios. If you need functionality, please provide it as a service in the `InvocationServices` class, and make sure it can be overridden.
+Before being called, the invocation will have all of its fields set from
+defaults, inputs, and finally links (overriding in that order).
+
+Assume that this invocation may be running simultaneously with other
+invocations, may be running on another machine, or in other interesting
+scenarios. If you need functionality, please provide it as a service in the
+`InvocationServices` class, and make sure it can be overridden.
 
 ### Outputs
+
 ```py
 class ImageOutput(BaseInvocationOutput):
     """Base class for invocations that output an image"""
@@ -102,4 +139,64 @@ class ImageOutput(BaseInvocationOutput):
 
     image: ImageField = Field(default=None, description="The output image")
 ```
-Output classes look like an invocation class without the invoke method. Prefer to use an existing output class if available, and prefer to name inputs the same as outputs when possible, to promote automatic invocation linking.
+
+Output classes look like an invocation class without the invoke method. Prefer
+to use an existing output class if available, and prefer to name inputs the same
+as outputs when possible, to promote automatic invocation linking.
+
+## Schema Generation
+
+Invocation, output and related classes are used to generate an OpenAPI schema.
+
+### Required Properties
+
+The schema generation treat all properties with default values as optional. This
+makes sense internally, but when when using these classes via the generated
+schema, we end up with e.g. the `ImageOutput` class having its `image` property
+marked as optional.
+
+We know that this property will always be present, so the additional logic
+needed to always check if the property exists adds a lot of extraneous cruft.
+
+To fix this, we can leverage `pydantic`'s
+[schema customisation](https://docs.pydantic.dev/usage/schema/#schema-customization)
+to mark properties that we know will always be present as required.
+
+Here's that `ImageOutput` class, without the needed schema customisation:
+
+```python
+class ImageOutput(BaseInvocationOutput):
+    """Base class for invocations that output an image"""
+
+    type: Literal["image"] = "image"
+    image:      ImageField = Field(default=None, description="The output image")
+```
+
+The generated OpenAPI schema, and all clients/types generated from it, will have
+the `type` and `image` properties marked as optional, even though we know they
+will always have a value by the time we can interact with them via the API.
+
+Here's the same class, but with the schema customisation added:
+
+```python
+class ImageOutput(BaseInvocationOutput):
+    """Base class for invocations that output an image"""
+
+    type: Literal["image"] = "image"
+    image:      ImageField = Field(default=None, description="The output image")
+
+    class Config:
+        schema_extra = {
+            'required': [
+                'type',
+                'image',
+            ]
+        }
+```
+
+The resultant schema (and any API client or types generated from it) will now
+have see `type` as string literal `"image"` and `image` as an `ImageField`
+object.
+
+See this `pydantic` issue for discussion on this solution:
+<https://github.com/pydantic/pydantic/discussions/4577>

docs/features/PROMPTS.md

@@ -268,7 +268,7 @@ model is so good at inpainting, a good substitute is to use the `clipseg` text
 masking option:
 
 ```bash
-invoke> a fluffy cat eating a hotdot
+invoke> a fluffy cat eating a hotdog
 Outputs:
 [1010] outputs/000025.2182095108.png: a fluffy cat eating a hotdog
 invoke> a smiling dog eating a hotdog -I 000025.2182095108.png -tm cat

docs/installation/010_INSTALL_AUTOMATED.md

@@ -417,7 +417,7 @@ Then type the following commands:
 
 === "AMD System"
     ```bash
-    pip install torch torchvision --force-reinstall --extra-index-url https://download.pytorch.org/whl/rocm5.2
+    pip install torch torchvision --force-reinstall --extra-index-url https://download.pytorch.org/whl/rocm5.4.2
     ```
 
 ### Corrupted configuration file

docs/installation/020_INSTALL_MANUAL.md

@@ -154,7 +154,7 @@ manager, please follow these steps:
     === "ROCm (AMD)"
 
         ```bash
-        pip install InvokeAI --use-pep517 --extra-index-url https://download.pytorch.org/whl/rocm5.2
+        pip install InvokeAI --use-pep517 --extra-index-url https://download.pytorch.org/whl/rocm5.4.2
         ```
 
     === "CPU (Intel Macs & non-GPU systems)"
@@ -315,7 +315,7 @@ installation protocol (important!)
 
     === "ROCm (AMD)"
         ```bash
-        pip install -e . --use-pep517 --extra-index-url https://download.pytorch.org/whl/rocm5.2
+        pip install -e . --use-pep517 --extra-index-url https://download.pytorch.org/whl/rocm5.4.2
         ```
 
     === "CPU (Intel Macs & non-GPU systems)"

docs/installation/030_INSTALL_CUDA_AND_ROCM.md

@@ -110,7 +110,7 @@ recipes are available
 
 When installing torch and torchvision manually with `pip`, remember to provide
 the argument `--extra-index-url
-https://download.pytorch.org/whl/rocm5.2` as described in the [Manual
+https://download.pytorch.org/whl/rocm5.4.2` as described in the [Manual
 Installation Guide](020_INSTALL_MANUAL.md).
 
 This will be done automatically for you if you use the installer

docs/installation/050_INSTALLING_MODELS.md

@@ -50,7 +50,7 @@ subset that are currently installed are found in
 |stable-diffusion-1.5|runwayml/stable-diffusion-v1-5|Stable Diffusion version 1.5 diffusers model (4.27 GB)|https://huggingface.co/runwayml/stable-diffusion-v1-5 |
 |sd-inpainting-1.5|runwayml/stable-diffusion-inpainting|RunwayML SD 1.5 model optimized for inpainting, diffusers version (4.27 GB)|https://huggingface.co/runwayml/stable-diffusion-inpainting |
 |stable-diffusion-2.1|stabilityai/stable-diffusion-2-1|Stable Diffusion version 2.1 diffusers model, trained on 768 pixel images (5.21 GB)|https://huggingface.co/stabilityai/stable-diffusion-2-1 |
-|sd-inpainting-2.0|stabilityai/stable-diffusion-2-1|Stable Diffusion version 2.0 inpainting model (5.21 GB)|https://huggingface.co/stabilityai/stable-diffusion-2-1 |
+|sd-inpainting-2.0|stabilityai/stable-diffusion-2-inpainting|Stable Diffusion version 2.0 inpainting model (5.21 GB)|https://huggingface.co/stabilityai/stable-diffusion-2-inpainting |
 |analog-diffusion-1.0|wavymulder/Analog-Diffusion|An SD-1.5 model trained on diverse analog photographs (2.13 GB)|https://huggingface.co/wavymulder/Analog-Diffusion |
 |deliberate-1.0|XpucT/Deliberate|Versatile model that produces detailed images up to 768px (4.27 GB)|https://huggingface.co/XpucT/Deliberate |
 |d&d-diffusion-1.0|0xJustin/Dungeons-and-Diffusion|Dungeons & Dragons characters (2.13 GB)|https://huggingface.co/0xJustin/Dungeons-and-Diffusion |

installer/lib/installer.py

@@ -456,7 +456,7 @@ def get_torch_source() -> (Union[str, None],str):
     optional_modules = None
     if OS == "Linux":
         if device == "rocm":
-            url = "https://download.pytorch.org/whl/rocm5.2"
+            url = "https://download.pytorch.org/whl/rocm5.4.2"
         elif device == "cpu":
             url = "https://download.pytorch.org/whl/cpu"
 

invokeai/app/api/dependencies.py

@@ -3,10 +3,16 @@
 import os
 from argparse import Namespace
 
+from invokeai.app.services.metadata import PngMetadataService, MetadataServiceBase
+
+from ..services.default_graphs import create_system_graphs
+
+from ..services.latent_storage import DiskLatentsStorage, ForwardCacheLatentsStorage
+
 from ...backend import Globals
 from ..services.model_manager_initializer import get_model_manager
 from ..services.restoration_services import RestorationServices
-from ..services.graph import GraphExecutionState
+from ..services.graph import GraphExecutionState, LibraryGraph
 from ..services.image_storage import DiskImageStorage
 from ..services.invocation_queue import MemoryInvocationQueue
 from ..services.invocation_services import InvocationServices
@@ -54,7 +60,11 @@ class ApiDependencies:
             os.path.join(os.path.dirname(__file__), "../../../../outputs")
         )
 
-        images = DiskImageStorage(output_folder)
+        latents = ForwardCacheLatentsStorage(DiskLatentsStorage(f'{output_folder}/latents'))
+
+        metadata = PngMetadataService()
+
+        images = DiskImageStorage(f'{output_folder}/images', metadata_service=metadata)
 
         # TODO: build a file/path manager?
         db_location = os.path.join(output_folder, "invokeai.db")
@@ -62,8 +72,13 @@ class ApiDependencies:
         services = InvocationServices(
             model_manager=get_model_manager(config),
             events=events,
+            latents=latents,
             images=images,
+            metadata=metadata,
             queue=MemoryInvocationQueue(),
+            graph_library=SqliteItemStorage[LibraryGraph](
+                filename=db_location, table_name="graphs"
+            ),
             graph_execution_manager=SqliteItemStorage[GraphExecutionState](
                 filename=db_location, table_name="graph_executions"
             ),
@@ -71,6 +86,8 @@ class ApiDependencies:
             restoration=RestorationServices(config),
         )
 
+        create_system_graphs(services.graph_library)
+
         ApiDependencies.invoker = Invoker(services)
 
     @staticmethod

invokeai/app/api/models/images.py

@@ -0,0 +1,34 @@
+from typing import Optional
+from pydantic import BaseModel, Field
+
+from invokeai.app.models.image import ImageType
+from invokeai.app.services.metadata import InvokeAIMetadata
+
+
+class ImageResponseMetadata(BaseModel):
+    """An image's metadata. Used only in HTTP responses."""
+
+    created: int = Field(description="The creation timestamp of the image")
+    width: int = Field(description="The width of the image in pixels")
+    height: int = Field(description="The height of the image in pixels")
+    invokeai: Optional[InvokeAIMetadata] = Field(
+        description="The image's InvokeAI-specific metadata"
+    )
+
+
+class ImageResponse(BaseModel):
+    """The response type for images"""
+
+    image_type: ImageType = Field(description="The type of the image")
+    image_name: str = Field(description="The name of the image")
+    image_url: str = Field(description="The url of the image")
+    thumbnail_url: str = Field(description="The url of the image's thumbnail")
+    metadata: ImageResponseMetadata = Field(description="The image's metadata")
+
+
+class ProgressImage(BaseModel):
+    """The progress image sent intermittently during processing"""
+
+    width: int = Field(description="The effective width of the image in pixels")
+    height: int = Field(description="The effective height of the image in pixels")
+    dataURL: str = Field(description="The image data as a b64 data URL")

invokeai/app/api/routers/images.py

@@ -1,11 +1,18 @@
 # Copyright (c) 2022 Kyle Schouviller (https://github.com/kyle0654)
-
+import io
 from datetime import datetime, timezone
+import json
+import os
+from typing import Any
+import uuid
 
-from fastapi import Path, Request, UploadFile
+from fastapi import HTTPException, Path, Query, Request, UploadFile
 from fastapi.responses import FileResponse, Response
 from fastapi.routing import APIRouter
 from PIL import Image
+from invokeai.app.api.models.images import ImageResponse, ImageResponseMetadata
+from invokeai.app.services.metadata import InvokeAIMetadata
+from invokeai.app.services.item_storage import PaginatedResults
 
 from ...services.image_storage import ImageType
 from ..dependencies import ApiDependencies
@@ -17,40 +24,105 @@ images_router = APIRouter(prefix="/v1/images", tags=["images"])
 async def get_image(
     image_type: ImageType = Path(description="The type of image to get"),
     image_name: str = Path(description="The name of the image to get"),
-):
+) -> FileResponse | Response:
     """Gets a result"""
-    # TODO: This is not really secure at all. At least make sure only output results are served
-    filename = ApiDependencies.invoker.services.images.get_path(image_type, image_name)
-    return FileResponse(filename)
+
+    path = ApiDependencies.invoker.services.images.get_path(
+        image_type=image_type, image_name=image_name
+    )
+
+    if ApiDependencies.invoker.services.images.validate_path(path):
+        return FileResponse(path)
+    else:
+        raise HTTPException(status_code=404)
+
+
+@images_router.get(
+    "/{image_type}/thumbnails/{image_name}", operation_id="get_thumbnail"
+)
+async def get_thumbnail(
+    image_type: ImageType = Path(description="The type of image to get"),
+    image_name: str = Path(description="The name of the image to get"),
+) -> FileResponse | Response:
+    """Gets a thumbnail"""
+
+    path = ApiDependencies.invoker.services.images.get_path(
+        image_type=image_type, image_name=image_name, is_thumbnail=True
+    )
+
+    if ApiDependencies.invoker.services.images.validate_path(path):
+        return FileResponse(path)
+    else:
+        raise HTTPException(status_code=404)
 
 
 @images_router.post(
     "/uploads/",
     operation_id="upload_image",
     responses={
-        201: {"description": "The image was uploaded successfully"},
-        404: {"description": "Session not found"},
+        201: {
+            "description": "The image was uploaded successfully",
+            "model": ImageResponse,
         },
+        415: {"description": "Image upload failed"},
+    },
+    status_code=201,
 )
-async def upload_image(file: UploadFile, request: Request):
+async def upload_image(
+    file: UploadFile, request: Request, response: Response
+) -> ImageResponse:
     if not file.content_type.startswith("image"):
-        return Response(status_code=415)
+        raise HTTPException(status_code=415, detail="Not an image")
 
     contents = await file.read()
+
     try:
-        im = Image.open(contents)
+        img = Image.open(io.BytesIO(contents))
     except:
         # Error opening the image
-        return Response(status_code=415)
+        raise HTTPException(status_code=415, detail="Failed to read image")
 
-    filename = f"{str(int(datetime.now(timezone.utc).timestamp()))}.png"
-    ApiDependencies.invoker.services.images.save(ImageType.UPLOAD, filename, im)
+    filename = f"{uuid.uuid4()}_{str(int(datetime.now(timezone.utc).timestamp()))}.png"
 
-    return Response(
-        status_code=201,
-        headers={
-            "Location": request.url_for(
-                "get_image", image_type=ImageType.UPLOAD, image_name=filename
+    (image_path, thumbnail_path, ctime) = ApiDependencies.invoker.services.images.save(
+        ImageType.UPLOAD, filename, img
     )
-        },
+
+    invokeai_metadata = ApiDependencies.invoker.services.metadata.get_metadata(img)
+
+    res = ImageResponse(
+        image_type=ImageType.UPLOAD,
+        image_name=filename,
+        image_url=f"api/v1/images/{ImageType.UPLOAD.value}/{filename}",
+        thumbnail_url=f"api/v1/images/{ImageType.UPLOAD.value}/thumbnails/{os.path.splitext(filename)[0]}.webp",
+        metadata=ImageResponseMetadata(
+            created=ctime,
+            width=img.width,
+            height=img.height,
+            invokeai=invokeai_metadata,
+        ),
     )
+
+    response.status_code = 201
+    response.headers["Location"] = request.url_for(
+        "get_image", image_type=ImageType.UPLOAD.value, image_name=filename
+    )
+
+    return res
+
+
+@images_router.get(
+    "/",
+    operation_id="list_images",
+    responses={200: {"model": PaginatedResults[ImageResponse]}},
+)
+async def list_images(
+    image_type: ImageType = Query(
+        default=ImageType.RESULT, description="The type of images to get"
+    ),
+    page: int = Query(default=0, description="The page of images to get"),
+    per_page: int = Query(default=10, description="The number of images per page"),
+) -> PaginatedResults[ImageResponse]:
+    """Gets a list of images"""
+    result = ApiDependencies.invoker.services.images.list(image_type, page, per_page)
+    return result

invokeai/app/api/routers/models.py

@@ -0,0 +1,251 @@
+# Copyright (c) 2023 Kyle Schouviller (https://github.com/kyle0654) and 2023 Kent Keirsey (https://github.com/hipsterusername)
+
+import shutil
+import asyncio
+from typing import Annotated, Any, List, Literal, Optional, Union
+
+from fastapi.routing import APIRouter, HTTPException
+from pydantic import BaseModel, Field, parse_obj_as
+from pathlib import Path
+from ..dependencies import ApiDependencies
+from invokeai.backend.globals import Globals, global_converted_ckpts_dir
+from invokeai.backend.args import Args
+
+
+
+models_router = APIRouter(prefix="/v1/models", tags=["models"])
+
+
+class VaeRepo(BaseModel):
+    repo_id: str = Field(description="The repo ID to use for this VAE")
+    path: Optional[str] = Field(description="The path to the VAE")
+    subfolder: Optional[str] = Field(description="The subfolder to use for this VAE")
+
+class ModelInfo(BaseModel):
+    description: Optional[str] = Field(description="A description of the model")
+    
+class CkptModelInfo(ModelInfo):
+    format: Literal['ckpt'] = 'ckpt'
+
+    config: str = Field(description="The path to the model config")
+    weights: str = Field(description="The path to the model weights")
+    vae: str = Field(description="The path to the model VAE")
+    width: Optional[int] = Field(description="The width of the model")
+    height: Optional[int] = Field(description="The height of the model")
+
+class DiffusersModelInfo(ModelInfo):
+    format: Literal['diffusers'] = 'diffusers'
+
+    vae: Optional[VaeRepo] = Field(description="The VAE repo to use for this model")
+    repo_id: Optional[str] = Field(description="The repo ID to use for this model")
+    path: Optional[str] = Field(description="The path to the model")
+
+class CreateModelRequest(BaseModel):
+    name: str = Field(description="The name of the model")
+    info: Union[CkptModelInfo, DiffusersModelInfo] = Field(discriminator="format", description="The model info")
+
+class CreateModelResponse(BaseModel):
+    name: str = Field(description="The name of the new model")
+    info: Union[CkptModelInfo, DiffusersModelInfo] = Field(discriminator="format", description="The model info")
+    status: str = Field(description="The status of the API response")
+
+class ConversionRequest(BaseModel):
+    name: str = Field(description="The name of the new model")
+    info: CkptModelInfo = Field(description="The converted model info")
+    save_location: str = Field(description="The path to save the converted model weights")
+    
+
+class ConvertedModelResponse(BaseModel):
+    name: str = Field(description="The name of the new model")
+    info: DiffusersModelInfo = Field(description="The converted model info")
+
+class ModelsList(BaseModel):
+    models: dict[str, Annotated[Union[(CkptModelInfo,DiffusersModelInfo)], Field(discriminator="format")]]
+
+
+@models_router.get(
+    "/",
+    operation_id="list_models",
+    responses={200: {"model": ModelsList }},
+)
+async def list_models() -> ModelsList:
+    """Gets a list of models"""
+    models_raw = ApiDependencies.invoker.services.model_manager.list_models()
+    models = parse_obj_as(ModelsList, { "models": models_raw })
+    return models
+
+
+@models_router.post(
+    "/",
+    operation_id="update_model",
+    responses={200: {"status": "success"}},
+)
+async def update_model(
+    model_request: CreateModelRequest
+) -> CreateModelResponse:
+    """ Add Model """
+    model_request_info = model_request.info
+    info_dict = model_request_info.dict()
+    model_response = CreateModelResponse(name=model_request.name, info=model_request.info, status="success")
+
+    ApiDependencies.invoker.services.model_manager.add_model(
+        model_name=model_request.name,
+        model_attributes=info_dict,
+        clobber=True,
+    )
+
+    return model_response
+
+
+@models_router.delete(
+    "/{model_name}",
+    operation_id="del_model",
+    responses={
+        204: {
+        "description": "Model deleted successfully"
+        }, 
+        404: {
+        "description": "Model not found"
+        }
+    },
+)
+async def delete_model(model_name: str) -> None:
+    """Delete Model"""
+    model_names = ApiDependencies.invoker.services.model_manager.model_names()
+    model_exists = model_name in model_names
+
+    # check if model exists
+    print(f">> Checking for model {model_name}...")
+           
+    if model_exists:
+        print(f">> Deleting Model: {model_name}")
+        ApiDependencies.invoker.services.model_manager.del_model(model_name, delete_files=True)
+        print(f">> Model Deleted: {model_name}")
+        raise HTTPException(status_code=204, detail=f"Model '{model_name}' deleted successfully")
+    
+    else:
+        print(f">> Model not found")
+        raise HTTPException(status_code=404, detail=f"Model '{model_name}' not found")
+    
+
+            # @socketio.on("convertToDiffusers")
+        # def convert_to_diffusers(model_to_convert: dict):
+        #     try:
+        #         if model_info := self.generate.model_manager.model_info(
+        #             model_name=model_to_convert["model_name"]
+        #         ):
+        #             if "weights" in model_info:
+        #                 ckpt_path = Path(model_info["weights"])
+        #                 original_config_file = Path(model_info["config"])
+        #                 model_name = model_to_convert["model_name"]
+        #                 model_description = model_info["description"]
+        #             else:
+        #                 self.socketio.emit(
+        #                     "error", {"message": "Model is not a valid checkpoint file"}
+        #                 )
+        #         else:
+        #             self.socketio.emit(
+        #                 "error", {"message": "Could not retrieve model info."}
+        #             )
+
+        #         if not ckpt_path.is_absolute():
+        #             ckpt_path = Path(Globals.root, ckpt_path)
+
+        #         if original_config_file and not original_config_file.is_absolute():
+        #             original_config_file = Path(Globals.root, original_config_file)
+
+        #         diffusers_path = Path(
+        #             ckpt_path.parent.absolute(), f"{model_name}_diffusers"
+        #         )
+
+        #         if model_to_convert["save_location"] == "root":
+        #             diffusers_path = Path(
+        #                 global_converted_ckpts_dir(), f"{model_name}_diffusers"
+        #             )
+
+        #         if (
+        #             model_to_convert["save_location"] == "custom"
+        #             and model_to_convert["custom_location"] is not None
+        #         ):
+        #             diffusers_path = Path(
+        #                 model_to_convert["custom_location"], f"{model_name}_diffusers"
+        #             )
+
+        #         if diffusers_path.exists():
+        #             shutil.rmtree(diffusers_path)
+
+        #         self.generate.model_manager.convert_and_import(
+        #             ckpt_path,
+        #             diffusers_path,
+        #             model_name=model_name,
+        #             model_description=model_description,
+        #             vae=None,
+        #             original_config_file=original_config_file,
+        #             commit_to_conf=opt.conf,
+        #         )
+
+        #         new_model_list = self.generate.model_manager.list_models()
+        #         socketio.emit(
+        #             "modelConverted",
+        #             {
+        #                 "new_model_name": model_name,
+        #                 "model_list": new_model_list,
+        #                 "update": True,
+        #             },
+        #         )
+        #         print(f">> Model Converted: {model_name}")
+        #     except Exception as e:
+        #         self.handle_exceptions(e)
+
+        # @socketio.on("mergeDiffusersModels")
+        # def merge_diffusers_models(model_merge_info: dict):
+        #     try:
+        #         models_to_merge = model_merge_info["models_to_merge"]
+        #         model_ids_or_paths = [
+        #             self.generate.model_manager.model_name_or_path(x)
+        #             for x in models_to_merge
+        #         ]
+        #         merged_pipe = merge_diffusion_models(
+        #             model_ids_or_paths,
+        #             model_merge_info["alpha"],
+        #             model_merge_info["interp"],
+        #             model_merge_info["force"],
+        #         )
+
+        #         dump_path = global_models_dir() / "merged_models"
+        #         if model_merge_info["model_merge_save_path"] is not None:
+        #             dump_path = Path(model_merge_info["model_merge_save_path"])
+
+        #         os.makedirs(dump_path, exist_ok=True)
+        #         dump_path = dump_path / model_merge_info["merged_model_name"]
+        #         merged_pipe.save_pretrained(dump_path, safe_serialization=1)
+
+        #         merged_model_config = dict(
+        #             model_name=model_merge_info["merged_model_name"],
+        #             description=f'Merge of models {", ".join(models_to_merge)}',
+        #             commit_to_conf=opt.conf,
+        #         )
+
+        #         if vae := self.generate.model_manager.config[models_to_merge[0]].get(
+        #             "vae", None
+        #         ):
+        #             print(f">> Using configured VAE assigned to {models_to_merge[0]}")
+        #             merged_model_config.update(vae=vae)
+
+        #         self.generate.model_manager.import_diffuser_model(
+        #             dump_path, **merged_model_config
+        #         )
+        #         new_model_list = self.generate.model_manager.list_models()
+
+        #         socketio.emit(
+        #             "modelsMerged",
+        #             {
+        #                 "merged_models": models_to_merge,
+        #                 "merged_model_name": model_merge_info["merged_model_name"],
+        #                 "model_list": new_model_list,
+        #                 "update": True,
+        #             },
+        #         )
+        #         print(f">> Models Merged: {models_to_merge}")
+        #         print(f">> New Model Added: {model_merge_info['merged_model_name']}")
+        #     except Exception as e:

invokeai/app/api/routers/sessions.py

@@ -51,7 +51,7 @@ async def list_sessions(
     query: str = Query(default="", description="The query string to search for"),
 ) -> PaginatedResults[GraphExecutionState]:
     """Gets a list of sessions, optionally searching"""
-    if filter == "":
+    if query == "":
         result = ApiDependencies.invoker.services.graph_execution_manager.list(
             page, per_page
         )
@@ -270,3 +270,18 @@ async def invoke_session(
 
     ApiDependencies.invoker.invoke(session, invoke_all=all)
     return Response(status_code=202)
+
+
+@session_router.delete(
+    "/{session_id}/invoke",
+    operation_id="cancel_session_invoke",
+    responses={
+        202: {"description": "The invocation is canceled"}
+    },
+)
+async def cancel_session_invoke(
+    session_id: str = Path(description="The id of the session to cancel"),
+) -> None:
+    """Invokes a session"""
+    ApiDependencies.invoker.cancel(session_id)
+    return Response(status_code=202)

invokeai/app/api_app.py

@@ -14,7 +14,7 @@ from pydantic.schema import schema
 
 from ..backend import Args
 from .api.dependencies import ApiDependencies
-from .api.routers import images, sessions
+from .api.routers import images, sessions, models
 from .api.sockets import SocketIO
 from .invocations import *
 from .invocations.baseinvocation import BaseInvocation
@@ -76,6 +76,8 @@ app.include_router(sessions.session_router, prefix="/api")
 
 app.include_router(images.images_router, prefix="/api")
 
+app.include_router(models.models_router, prefix="/api")
+
 
 # Build a custom OpenAPI to include all outputs
 # TODO: can outputs be included on metadata of invocation schemas somehow?

invokeai/app/cli/commands.py

@@ -4,12 +4,43 @@ from abc import ABC, abstractmethod
 import argparse
 from typing import Any, Callable, Iterable, Literal, get_args, get_origin, get_type_hints
 from pydantic import BaseModel, Field
+import networkx as nx
+import matplotlib.pyplot as plt
 
+from ..invocations.baseinvocation import BaseInvocation
 from ..invocations.image import ImageField
-from ..services.graph import GraphExecutionState
+from ..services.graph import GraphExecutionState, LibraryGraph, GraphInvocation, Edge
 from ..services.invoker import Invoker
 
 
+def add_field_argument(command_parser, name: str, field, default_override = None):
+    default = default_override if default_override is not None else field.default if field.default_factory is None else field.default_factory()
+    if get_origin(field.type_) == Literal:
+        allowed_values = get_args(field.type_)
+        allowed_types = set()
+        for val in allowed_values:
+            allowed_types.add(type(val))
+        allowed_types_list = list(allowed_types)
+        field_type = allowed_types_list[0] if len(allowed_types) == 1 else Union[allowed_types_list]  # type: ignore
+
+        command_parser.add_argument(
+            f"--{name}",
+            dest=name,
+            type=field_type,
+            default=default,
+            choices=allowed_values,
+            help=field.field_info.description,
+        )
+    else:
+        command_parser.add_argument(
+            f"--{name}",
+            dest=name,
+            type=field.type_,
+            default=default,
+            help=field.field_info.description,
+        )
+
+
 def add_parsers(
     subparsers,
     commands: list[type],
@@ -34,30 +65,26 @@ def add_parsers(
             if name in exclude_fields:
                 continue
 
-            if get_origin(field.type_) == Literal:
-                allowed_values = get_args(field.type_)
-                allowed_types = set()
-                for val in allowed_values:
-                    allowed_types.add(type(val))
-                allowed_types_list = list(allowed_types)
-                field_type = allowed_types_list[0] if len(allowed_types) == 1 else Union[allowed_types_list]  # type: ignore
+            add_field_argument(command_parser, name, field)
 
-                command_parser.add_argument(
-                    f"--{name}",
-                    dest=name,
-                    type=field_type,
-                    default=field.default,
-                    choices=allowed_values,
-                    help=field.field_info.description,
-                )
-            else:
-                command_parser.add_argument(
-                    f"--{name}",
-                    dest=name,
-                    type=field.type_,
-                    default=field.default,
-                    help=field.field_info.description,
-                )
+
+def add_graph_parsers(
+    subparsers,
+    graphs: list[LibraryGraph],
+    add_arguments: Callable[[argparse.ArgumentParser], None]|None = None
+):
+    for graph in graphs:
+        command_parser = subparsers.add_parser(graph.name, help=graph.description)
+        
+        if add_arguments is not None:
+            add_arguments(command_parser)
+
+        # Add arguments for inputs
+        for exposed_input in graph.exposed_inputs:
+            node = graph.graph.get_node(exposed_input.node_path)
+            field = node.__fields__[exposed_input.field]
+            default_override = getattr(node, exposed_input.field)
+            add_field_argument(command_parser, exposed_input.alias, field, default_override)
 
 
 class CliContext:
@@ -65,17 +92,38 @@ class CliContext:
     session: GraphExecutionState
     parser: argparse.ArgumentParser
     defaults: dict[str, Any]
+    graph_nodes: dict[str, str]
+    nodes_added: list[str]
 
     def __init__(self, invoker: Invoker, session: GraphExecutionState, parser: argparse.ArgumentParser):
         self.invoker = invoker
         self.session = session
         self.parser = parser
         self.defaults = dict()
+        self.graph_nodes = dict()
+        self.nodes_added = list()
 
     def get_session(self):
         self.session = self.invoker.services.graph_execution_manager.get(self.session.id)
         return self.session
 
+    def reset(self):
+        self.session = self.invoker.create_execution_state()
+        self.graph_nodes = dict()
+        self.nodes_added = list()
+        # Leave defaults unchanged
+
+    def add_node(self, node: BaseInvocation):
+        self.get_session()
+        self.session.graph.add_node(node)
+        self.nodes_added.append(node.id)
+        self.invoker.services.graph_execution_manager.set(self.session)
+
+    def add_edge(self, edge: Edge):
+        self.get_session()
+        self.session.add_edge(edge)
+        self.invoker.services.graph_execution_manager.set(self.session)
+
 
 class ExitCli(Exception):
     """Exception to exit the CLI"""
@@ -200,3 +248,39 @@ class SetDefaultCommand(BaseCommand):
                 del context.defaults[self.field]
         else:
             context.defaults[self.field] = self.value
+
+
+class DrawGraphCommand(BaseCommand):
+    """Debugs a graph"""
+    type: Literal['draw_graph'] = 'draw_graph'
+
+    def run(self, context: CliContext) -> None:
+        session: GraphExecutionState = context.invoker.services.graph_execution_manager.get(context.session.id)
+        nxgraph = session.graph.nx_graph_flat()
+
+        # Draw the networkx graph
+        plt.figure(figsize=(20, 20))
+        pos = nx.spectral_layout(nxgraph)
+        nx.draw_networkx_nodes(nxgraph, pos, node_size=1000)
+        nx.draw_networkx_edges(nxgraph, pos, width=2)
+        nx.draw_networkx_labels(nxgraph, pos, font_size=20, font_family="sans-serif")
+        plt.axis("off")
+        plt.show()
+
+
+class DrawExecutionGraphCommand(BaseCommand):
+    """Debugs an execution graph"""
+    type: Literal['draw_xgraph'] = 'draw_xgraph'
+
+    def run(self, context: CliContext) -> None:
+        session: GraphExecutionState = context.invoker.services.graph_execution_manager.get(context.session.id)
+        nxgraph = session.execution_graph.nx_graph_flat()
+
+        # Draw the networkx graph
+        plt.figure(figsize=(20, 20))
+        pos = nx.spectral_layout(nxgraph)
+        nx.draw_networkx_nodes(nxgraph, pos, node_size=1000)
+        nx.draw_networkx_edges(nxgraph, pos, width=2)
+        nx.draw_networkx_labels(nxgraph, pos, font_size=20, font_family="sans-serif")
+        plt.axis("off")
+        plt.show()

invokeai/app/cli/completer.py

@@ -0,0 +1,167 @@
+"""
+Readline helper functions for cli_app.py
+You may import the global singleton `completer` to get access to the
+completer object.
+"""
+import atexit
+import readline
+import shlex
+
+from pathlib import Path
+from typing import List, Dict, Literal, get_args, get_type_hints, get_origin
+
+from ...backend import ModelManager, Globals
+from ..invocations.baseinvocation import BaseInvocation
+from .commands import BaseCommand
+
+# singleton object, class variable
+completer = None
+
+class Completer(object):
+    
+    def __init__(self, model_manager: ModelManager):
+        self.commands = self.get_commands()
+        self.matches = None
+        self.linebuffer = None
+        self.manager = model_manager
+        return
+
+    def complete(self, text, state):
+        """
+        Complete commands and switches fromm the node CLI command line.
+        Switches are determined in a context-specific manner.
+        """
+
+        buffer = readline.get_line_buffer()
+        if state == 0:
+            options = None
+            try:
+                current_command, current_switch = self.get_current_command(buffer)
+                options = self.get_command_options(current_command, current_switch)
+            except IndexError:
+                pass
+            options = options or list(self.parse_commands().keys())
+            
+            if not text:  # first time
+                self.matches = options
+            else:
+                self.matches = [s for s in options if s and s.startswith(text)]
+
+        try:
+            match = self.matches[state]
+        except IndexError:
+            match = None
+        return match
+
+    @classmethod
+    def get_commands(self)->List[object]:
+        """
+        Return a list of all the client commands and invocations.
+        """
+        return BaseCommand.get_commands() + BaseInvocation.get_invocations()
+
+    def get_current_command(self, buffer: str)->tuple[str, str]:
+        """
+        Parse the readline buffer to find the most recent command and its switch.
+        """
+        if len(buffer)==0:
+            return None, None
+        tokens = shlex.split(buffer)
+        command = None
+        switch = None
+        for t in tokens:
+            if t[0].isalpha():
+                if switch is None:
+                    command = t
+            else:
+                switch = t
+        # don't try to autocomplete switches that are already complete
+        if switch and buffer.endswith(' '):
+            switch=None
+        return command or '', switch or ''
+
+    def parse_commands(self)->Dict[str, List[str]]:
+        """
+        Return a dict in which the keys are the command name
+        and the values are the parameters the command takes.
+        """
+        result = dict()
+        for command in self.commands:
+            hints = get_type_hints(command)
+            name = get_args(hints['type'])[0]
+            result.update({name:hints})
+        return result
+
+    def get_command_options(self, command: str, switch: str)->List[str]:
+        """
+        Return all the parameters that can be passed to the command as
+        command-line switches. Returns None if the command is unrecognized.
+        """
+        parsed_commands = self.parse_commands()
+        if command not in parsed_commands:
+            return None
+        
+        # handle switches in the format "-foo=bar"
+        argument = None
+        if switch and '=' in switch:
+            switch, argument = switch.split('=')
+            
+        parameter = switch.strip('-')
+        if parameter in parsed_commands[command]:
+            if argument is None:
+                return self.get_parameter_options(parameter, parsed_commands[command][parameter])
+            else:
+                return [f"--{parameter}={x}" for x in self.get_parameter_options(parameter, parsed_commands[command][parameter])]
+        else:
+            return [f"--{x}" for x in parsed_commands[command].keys()]
+
+    def get_parameter_options(self, parameter: str, typehint)->List[str]:
+        """
+        Given a parameter type (such as Literal), offers autocompletions.
+        """
+        if get_origin(typehint) == Literal:
+            return get_args(typehint)
+        if parameter == 'model':
+            return self.manager.model_names()
+        
+    def _pre_input_hook(self):
+        if self.linebuffer:
+            readline.insert_text(self.linebuffer)
+            readline.redisplay()
+            self.linebuffer = None
+    
+def set_autocompleter(model_manager: ModelManager) -> Completer:
+    global completer
+    
+    if completer:
+        return completer
+    
+    completer = Completer(model_manager)
+
+    readline.set_completer(completer.complete)
+    # pyreadline3 does not have a set_auto_history() method
+    try:
+        readline.set_auto_history(True)
+    except:
+        pass
+    readline.set_pre_input_hook(completer._pre_input_hook)
+    readline.set_completer_delims(" ")
+    readline.parse_and_bind("tab: complete")
+    readline.parse_and_bind("set print-completions-horizontally off")
+    readline.parse_and_bind("set page-completions on")
+    readline.parse_and_bind("set skip-completed-text on")
+    readline.parse_and_bind("set show-all-if-ambiguous on")
+
+    histfile = Path(Globals.root, ".invoke_history")
+    try:
+        readline.read_history_file(histfile)
+        readline.set_history_length(1000)
+    except FileNotFoundError:
+        pass
+    except OSError:  # file likely corrupted
+        newname = f"{histfile}.old"
+        print(
+            f"## Your history file {histfile} couldn't be loaded and may be corrupted. Renaming it to {newname}"
+        )
+        histfile.replace(Path(newname))
+    atexit.register(readline.write_history_file, histfile)

invokeai/app/cli_app.py

@@ -2,6 +2,7 @@
 
 import argparse
 import os
+import re
 import shlex
 import time
 from typing import (
@@ -12,14 +13,22 @@ from typing import (
 from pydantic import BaseModel
 from pydantic.fields import Field
 
+from invokeai.app.services.metadata import PngMetadataService
+
+from .services.default_graphs import create_system_graphs
+
+from .services.latent_storage import DiskLatentsStorage, ForwardCacheLatentsStorage
+
 from ..backend import Args
-from .cli.commands import BaseCommand, CliContext, ExitCli, add_parsers, get_graph_execution_history
+from .cli.commands import BaseCommand, CliContext, ExitCli, add_graph_parsers, add_parsers, get_graph_execution_history
+from .cli.completer import set_autocompleter
 from .invocations import *
 from .invocations.baseinvocation import BaseInvocation
 from .services.events import EventServiceBase
 from .services.model_manager_initializer import get_model_manager
 from .services.restoration_services import RestorationServices
-from .services.graph import Edge, EdgeConnection, GraphExecutionState
+from .services.graph import Edge, EdgeConnection, ExposedNodeInput, GraphExecutionState, GraphInvocation, LibraryGraph, are_connection_types_compatible
+from .services.default_graphs import default_text_to_image_graph_id
 from .services.image_storage import DiskImageStorage
 from .services.invocation_queue import MemoryInvocationQueue
 from .services.invocation_services import InvocationServices
@@ -43,7 +52,7 @@ def add_invocation_args(command_parser):
         "-l",
         action="append",
         nargs=3,
-        help="A link in the format 'dest_field source_node source_field'. source_node can be relative to history (e.g. -1)",
+        help="A link in the format 'source_node source_field dest_field'. source_node can be relative to history (e.g. -1)",
     )
 
     command_parser.add_argument(
@@ -54,7 +63,7 @@ def add_invocation_args(command_parser):
     )
 
 
-def get_command_parser() -> argparse.ArgumentParser:
+def get_command_parser(services: InvocationServices) -> argparse.ArgumentParser:
     # Create invocation parser
     parser = argparse.ArgumentParser()
 
@@ -72,20 +81,72 @@ def get_command_parser() -> argparse.ArgumentParser:
     commands = BaseCommand.get_all_subclasses()
     add_parsers(subparsers, commands, exclude_fields=["type"])
 
+    # Create subparsers for exposed CLI graphs
+    # TODO: add a way to identify these graphs
+    text_to_image = services.graph_library.get(default_text_to_image_graph_id)
+    add_graph_parsers(subparsers, [text_to_image], add_arguments=add_invocation_args)
+
     return parser
 
 
+class NodeField():
+    alias: str
+    node_path: str
+    field: str
+    field_type: type
+
+    def __init__(self, alias: str, node_path: str, field: str, field_type: type):
+        self.alias = alias
+        self.node_path = node_path
+        self.field = field
+        self.field_type = field_type
+
+
+def fields_from_type_hints(hints: dict[str, type], node_path: str) -> dict[str,NodeField]:
+    return {k:NodeField(alias=k, node_path=node_path, field=k, field_type=v) for k, v in hints.items()}
+
+
+def get_node_input_field(graph: LibraryGraph, field_alias: str, node_id: str) -> NodeField:
+    """Gets the node field for the specified field alias"""
+    exposed_input = next(e for e in graph.exposed_inputs if e.alias == field_alias)
+    node_type = type(graph.graph.get_node(exposed_input.node_path))
+    return NodeField(alias=exposed_input.alias, node_path=f'{node_id}.{exposed_input.node_path}', field=exposed_input.field, field_type=get_type_hints(node_type)[exposed_input.field])
+
+
+def get_node_output_field(graph: LibraryGraph, field_alias: str, node_id: str) -> NodeField:
+    """Gets the node field for the specified field alias"""
+    exposed_output = next(e for e in graph.exposed_outputs if e.alias == field_alias)
+    node_type = type(graph.graph.get_node(exposed_output.node_path))
+    node_output_type = node_type.get_output_type()
+    return NodeField(alias=exposed_output.alias, node_path=f'{node_id}.{exposed_output.node_path}', field=exposed_output.field, field_type=get_type_hints(node_output_type)[exposed_output.field])
+
+
+def get_node_inputs(invocation: BaseInvocation, context: CliContext) -> dict[str, NodeField]:
+    """Gets the inputs for the specified invocation from the context"""
+    node_type = type(invocation)
+    if node_type is not GraphInvocation:
+        return fields_from_type_hints(get_type_hints(node_type), invocation.id)
+    else:
+        graph: LibraryGraph = context.invoker.services.graph_library.get(context.graph_nodes[invocation.id])
+        return {e.alias: get_node_input_field(graph, e.alias, invocation.id) for e in graph.exposed_inputs}
+
+
+def get_node_outputs(invocation: BaseInvocation, context: CliContext) -> dict[str, NodeField]:
+    """Gets the outputs for the specified invocation from the context"""
+    node_type = type(invocation)
+    if node_type is not GraphInvocation:
+        return fields_from_type_hints(get_type_hints(node_type.get_output_type()), invocation.id)
+    else:
+        graph: LibraryGraph = context.invoker.services.graph_library.get(context.graph_nodes[invocation.id])
+        return {e.alias: get_node_output_field(graph, e.alias, invocation.id) for e in graph.exposed_outputs}
+
+
 def generate_matching_edges(
-    a: BaseInvocation, b: BaseInvocation
+    a: BaseInvocation, b: BaseInvocation, context: CliContext
 ) -> list[Edge]:
     """Generates all possible edges between two invocations"""
-    atype = type(a)
-    btype = type(b)
-
-    aoutputtype = atype.get_output_type()
-
-    afields = get_type_hints(aoutputtype)
-    bfields = get_type_hints(btype)
+    afields = get_node_outputs(a, context)
+    bfields = get_node_inputs(b, context)
 
     matching_fields = set(afields.keys()).intersection(bfields.keys())
 
@@ -93,12 +154,15 @@ def generate_matching_edges(
     invalid_fields = set(["type", "id"])
     matching_fields = matching_fields.difference(invalid_fields)
 
+    # Validate types
+    matching_fields = [f for f in matching_fields if are_connection_types_compatible(afields[f].field_type, bfields[f].field_type)]
+
     edges = [
         Edge(
-            source=EdgeConnection(node_id=a.id, field=field),
-            destination=EdgeConnection(node_id=b.id, field=field)
+            source=EdgeConnection(node_id=afields[alias].node_path, field=afields[alias].field),
+            destination=EdgeConnection(node_id=bfields[alias].node_path, field=bfields[alias].field)
         )
-        for field in matching_fields
+        for alias in matching_fields
     ]
     return edges
 
@@ -130,8 +194,16 @@ def invoke_cli():
     config.parse_args()
     model_manager = get_model_manager(config)
 
+    # This initializes the autocompleter and returns it.
+    # Currently nothing is done with the returned Completer
+    # object, but the object can be used to change autocompletion
+    # behavior on the fly, if desired.
+    completer = set_autocompleter(model_manager)
+
     events = EventServiceBase()
 
+    metadata = PngMetadataService()
+
     output_folder = os.path.abspath(
         os.path.join(os.path.dirname(__file__), "../../../outputs")
     )
@@ -142,8 +214,13 @@ def invoke_cli():
     services = InvocationServices(
         model_manager=model_manager,
         events=events,
-        images=DiskImageStorage(output_folder),
+        latents = ForwardCacheLatentsStorage(DiskLatentsStorage(f'{output_folder}/latents')),
+        images=DiskImageStorage(f'{output_folder}/images', metadata_service=metadata),
+        metadata=metadata,
         queue=MemoryInvocationQueue(),
+        graph_library=SqliteItemStorage[LibraryGraph](
+            filename=db_location, table_name="graphs"
+        ),
         graph_execution_manager=SqliteItemStorage[GraphExecutionState](
             filename=db_location, table_name="graph_executions"
         ),
@@ -151,9 +228,14 @@ def invoke_cli():
         restoration=RestorationServices(config),
     )
 
+    system_graphs = create_system_graphs(services.graph_library)
+    system_graph_names = set([g.name for g in system_graphs])
+
     invoker = Invoker(services)
     session: GraphExecutionState = invoker.create_execution_state()
-    parser = get_command_parser()
+    parser = get_command_parser(services)
+
+    re_negid = re.compile('^-[0-9]+$')
 
     # Uncomment to print out previous sessions at startup
     # print(services.session_manager.list())
@@ -162,18 +244,19 @@ def invoke_cli():
 
     while True:
         try:
-            cmd_input = input("> ")
-        except KeyboardInterrupt:
+            cmd_input = input("invoke> ")
+        except (KeyboardInterrupt, EOFError):
             # Ctrl-c exits
             break
 
         try:
             # Refresh the state of the session
-            history = list(get_graph_execution_history(context.session))
+            #history = list(get_graph_execution_history(context.session))
+            history = list(reversed(context.nodes_added))
 
             # Split the command for piping
             cmds = cmd_input.split("|")
-            start_id = len(history)
+            start_id = len(context.nodes_added)
             current_id = start_id
             new_invocations = list()
             for cmd in cmds:
@@ -189,9 +272,25 @@ def invoke_cli():
                         args[field_name] = field_default
 
                 # Parse invocation
+                command: CliCommand = None # type:ignore
+                system_graph: LibraryGraph|None = None
+                if args['type'] in system_graph_names:
+                    system_graph = next(filter(lambda g: g.name == args['type'], system_graphs))
+                    invocation = GraphInvocation(graph=system_graph.graph, id=str(current_id))
+                    for exposed_input in system_graph.exposed_inputs:
+                        if exposed_input.alias in args:
+                            node = invocation.graph.get_node(exposed_input.node_path)
+                            field = exposed_input.field
+                            setattr(node, field, args[exposed_input.alias])
+                    command = CliCommand(command = invocation)
+                    context.graph_nodes[invocation.id] = system_graph.id
+                else:
                     args["id"] = current_id
                     command = CliCommand(command=args)
 
+                if command is None:
+                    continue
+
                 # Run any CLI commands immediately
                 if isinstance(command.command, BaseCommand):
                     # Invoke all current nodes to preserve operation order
@@ -201,6 +300,7 @@ def invoke_cli():
                     command.command.run(context)
                     continue
 
+                # TODO: handle linking with library graphs
                 # Pipe previous command output (if there was a previous command)
                 edges: list[Edge] = list()
                 if len(history) > 0 or current_id != start_id:
@@ -213,16 +313,20 @@ def invoke_cli():
                         else context.session.graph.get_node(from_id)
                     )
                     matching_edges = generate_matching_edges(
-                        from_node, command.command
+                        from_node, command.command, context
                     )
                     edges.extend(matching_edges)
 
                 # Parse provided links
                 if "link_node" in args and args["link_node"]:
                     for link in args["link_node"]:
-                        link_node = context.session.graph.get_node(link)
+                        node_id = link
+                        if re_negid.match(node_id):
+                            node_id = str(current_id + int(node_id))
+
+                        link_node = context.session.graph.get_node(node_id)
                         matching_edges = generate_matching_edges(
-                            link_node, command.command
+                            link_node, command.command, context
                         )
                         matching_destinations = [e.destination for e in matching_edges]
                         edges = [e for e in edges if e.destination not in matching_destinations]
@@ -230,13 +334,20 @@ def invoke_cli():
 
                 if "link" in args and args["link"]:
                     for link in args["link"]:
-                        edges = [e for e in edges if e.destination.node_id != command.command.id and e.destination.field != link[2]]
+                        edges = [e for e in edges if e.destination.node_id != command.command.id or e.destination.field != link[2]]
+
+                        node_id = link[0]
+                        if re_negid.match(node_id):
+                            node_id = str(current_id + int(node_id))
+
+                        # TODO: handle missing input/output
+                        node_output = get_node_outputs(context.session.graph.get_node(node_id), context)[link[1]]
+                        node_input = get_node_inputs(command.command, context)[link[2]]
+
                         edges.append(
                             Edge(
-                                source=EdgeConnection(node_id=link[1], field=link[0]),
-                                destination=EdgeConnection(
-                                    node_id=command.command.id, field=link[2]
-                                )
+                                source=EdgeConnection(node_id=node_output.node_path, field=node_output.field),
+                                destination=EdgeConnection(node_id=node_input.node_path, field=node_input.field)
                             )
                         )
 
@@ -245,10 +356,10 @@ def invoke_cli():
                 current_id = current_id + 1
 
                 # Add the node to the session
-                context.session.add_node(command.command)
+                context.add_node(command.command)
                 for edge in edges:
                     print(edge)
-                    context.session.add_edge(edge)
+                    context.add_edge(edge)
 
             # Execute all remaining nodes
             invoke_all(context)
@@ -260,7 +371,7 @@ def invoke_cli():
         except SessionError:
             # Start a new session
             print("Session error: creating a new session")
-            context.session = context.invoker.create_execution_state()
+            context.reset()
 
         except ExitCli:
             break

invokeai/app/invocations/baseinvocation.py

@@ -2,7 +2,7 @@
 
 from abc import ABC, abstractmethod
 from inspect import signature
-from typing import get_args, get_type_hints
+from typing import get_args, get_type_hints, Dict, List, Literal, TypedDict
 
 from pydantic import BaseModel, Field
 
@@ -76,3 +76,56 @@ class BaseInvocation(ABC, BaseModel):
     #fmt: off
     id: str = Field(description="The id of this node. Must be unique among all nodes.")
     #fmt: on
+
+
+# TODO: figure out a better way to provide these hints
+# TODO: when we can upgrade to python 3.11, we can use the`NotRequired` type instead of `total=False`
+class UIConfig(TypedDict, total=False):
+    type_hints: Dict[
+        str,
+        Literal[
+            "integer",
+            "float",
+            "boolean",
+            "string",
+            "enum",
+            "image",
+            "latents",
+            "model",
+        ],
+    ]
+    tags: List[str]
+    title: str
+
+class CustomisedSchemaExtra(TypedDict):
+    ui: UIConfig
+
+
+class InvocationConfig(BaseModel.Config):
+    """Customizes pydantic's BaseModel.Config class for use by Invocations.
+
+    Provide `schema_extra` a `ui` dict to add hints for generated UIs.
+
+    `tags`
+    - A list of strings, used to categorise invocations.
+
+    `type_hints`
+    - A dict of field types which override the types in the invocation definition.
+    - Each key should be the name of one of the invocation's fields.
+    - Each value should be one of the valid types:
+      - `integer`, `float`, `boolean`, `string`, `enum`, `image`, `latents`, `model`
+
+    ```python
+    class Config(InvocationConfig):
+      schema_extra = {
+          "ui": {
+              "tags": ["stable-diffusion", "image"],
+              "type_hints": {
+                  "initial_image": "image",
+              },
+          },
+      }
+    ```
+    """
+
+    schema_extra: CustomisedSchemaExtra

invokeai/app/invocations/collections.py

@@ -0,0 +1,64 @@
+# Copyright (c) 2023 Kyle Schouviller (https://github.com/kyle0654)
+
+from typing import Literal, Optional
+
+import numpy as np
+import numpy.random
+from pydantic import Field
+
+from .baseinvocation import (
+    BaseInvocation,
+    InvocationConfig,
+    InvocationContext,
+    BaseInvocationOutput,
+)
+
+
+class IntCollectionOutput(BaseInvocationOutput):
+    """A collection of integers"""
+
+    type: Literal["int_collection"] = "int_collection"
+
+    # Outputs
+    collection: list[int] = Field(default=[], description="The int collection")
+
+
+class RangeInvocation(BaseInvocation):
+    """Creates a range"""
+
+    type: Literal["range"] = "range"
+
+    # Inputs
+    start: int = Field(default=0, description="The start of the range")
+    stop: int = Field(default=10, description="The stop of the range")
+    step: int = Field(default=1, description="The step of the range")
+
+    def invoke(self, context: InvocationContext) -> IntCollectionOutput:
+        return IntCollectionOutput(
+            collection=list(range(self.start, self.stop, self.step))
+        )
+
+
+class RandomRangeInvocation(BaseInvocation):
+    """Creates a collection of random numbers"""
+
+    type: Literal["random_range"] = "random_range"
+
+    # Inputs
+    low: int = Field(default=0, description="The inclusive low value")
+    high: int = Field(
+        default=np.iinfo(np.int32).max, description="The exclusive high value"
+    )
+    size: int = Field(default=1, description="The number of values to generate")
+    seed: Optional[int] = Field(
+        ge=0,
+        le=np.iinfo(np.int32).max,
+        description="The seed for the RNG",
+        default_factory=lambda: numpy.random.randint(0, np.iinfo(np.int32).max),
+    )
+
+    def invoke(self, context: InvocationContext) -> IntCollectionOutput:
+        rng = np.random.default_rng(self.seed)
+        return IntCollectionOutput(
+            collection=list(rng.integers(low=self.low, high=self.high, size=self.size))
+        )

invokeai/app/invocations/cv.py

@@ -5,14 +5,26 @@ from typing import Literal
 import cv2 as cv
 import numpy
 from PIL import Image, ImageOps
-from pydantic import Field
+from pydantic import BaseModel, Field
 
-from ..services.image_storage import ImageType
-from .baseinvocation import BaseInvocation, InvocationContext
-from .image import ImageField, ImageOutput
+from invokeai.app.models.image import ImageField, ImageType
+from .baseinvocation import BaseInvocation, InvocationContext, InvocationConfig
+from .image import ImageOutput, build_image_output
 
 
-class CvInpaintInvocation(BaseInvocation):
+class CvInvocationConfig(BaseModel):
+    """Helper class to provide all OpenCV invocations with additional config"""
+
+    # Schema customisation
+    class Config(InvocationConfig):
+        schema_extra = {
+            "ui": {
+                "tags": ["cv", "image"],
+            },
+        }
+
+
+class CvInpaintInvocation(BaseInvocation, CvInvocationConfig):
     """Simple inpaint using opencv."""
     #fmt: off
     type: Literal["cv_inpaint"] = "cv_inpaint"
@@ -44,7 +56,14 @@ class CvInpaintInvocation(BaseInvocation):
         image_name = context.services.images.create_name(
             context.graph_execution_state_id, self.id
         )
-        context.services.images.save(image_type, image_name, image_inpainted)
-        return ImageOutput(
-            image=ImageField(image_type=image_type, image_name=image_name)
+
+        metadata = context.services.metadata.build_metadata(
+            session_id=context.graph_execution_state_id, node=self
+        )
+        
+        context.services.images.save(image_type, image_name, image_inpainted, metadata)
+        return build_image_output(
+            image_type=image_type,
+            image_name=image_name,
+            image=image_inpainted,
         )

invokeai/app/invocations/generate.py

@@ -1,29 +1,41 @@
 # Copyright (c) 2022 Kyle Schouviller (https://github.com/kyle0654)
 
-from datetime import datetime, timezone
-from typing import Any, Literal, Optional, Union
+from functools import partial
+from typing import Literal, Optional, Union
 
 import numpy as np
-
 from torch import Tensor
-from PIL import Image
-from pydantic import Field
-from skimage.exposure.histogram_matching import match_histograms
 
-from ..services.image_storage import ImageType
-from ..services.invocation_services import InvocationServices
-from .baseinvocation import BaseInvocation, InvocationContext
-from .image import ImageField, ImageOutput
-from ...backend.generator import Txt2Img, Img2Img, Inpaint, InvokeAIGenerator, Generator
+from pydantic import BaseModel, Field
+
+from invokeai.app.models.image import ImageField, ImageType
+from invokeai.app.invocations.util.choose_model import choose_model
+from .baseinvocation import BaseInvocation, InvocationContext, InvocationConfig
+from .image import ImageOutput, build_image_output
+from ...backend.generator import Txt2Img, Img2Img, Inpaint, InvokeAIGenerator
 from ...backend.stable_diffusion import PipelineIntermediateState
-from ...backend.util.util import image_to_dataURL
+from ..util.step_callback import stable_diffusion_step_callback
+
+SAMPLER_NAME_VALUES = Literal[tuple(InvokeAIGenerator.schedulers())]
+
+
+class SDImageInvocation(BaseModel):
+    """Helper class to provide all Stable Diffusion raster image invocations with additional config"""
+
+    # Schema customisation
+    class Config(InvocationConfig):
+        schema_extra = {
+            "ui": {
+                "tags": ["stable-diffusion", "image"],
+                "type_hints": {
+                    "model": "model",
+                },
+            },
+        }
 
-SAMPLER_NAME_VALUES = Literal[
-    tuple(InvokeAIGenerator.schedulers())
-]
 
 # Text to image
-class TextToImageInvocation(BaseInvocation):
+class TextToImageInvocation(BaseInvocation, SDImageInvocation):
     """Generates an image using text2img."""
 
     type: Literal["txt2img"] = "txt2img"
@@ -37,7 +49,7 @@ class TextToImageInvocation(BaseInvocation):
     width:       int = Field(default=512, multiple_of=64, gt=0, description="The width of the resulting image", )
     height:      int = Field(default=512, multiple_of=64, gt=0, description="The height of the resulting image", )
     cfg_scale: float = Field(default=7.5, gt=0, description="The Classifier-Free Guidance, higher values may result in a result closer to the prompt", )
-    sampler_name: SAMPLER_NAME_VALUES = Field(default="k_lms", description="The sampler to use" )
+    scheduler: SAMPLER_NAME_VALUES = Field(default="k_lms", description="The scheduler to use" )
     seamless:   bool = Field(default=False, description="Whether or not to generate an image that can tile without seams", )
     model:       str = Field(default="", description="The model to use (currently ignored)")
     progress_images: bool = Field(default=False, description="Whether or not to produce progress images during generation",  )
@@ -45,41 +57,31 @@ class TextToImageInvocation(BaseInvocation):
 
     # TODO: pass this an emitter method or something? or a session for dispatching?
     def dispatch_progress(
-        self, context: InvocationContext, sample: Tensor, step: int
+        self,
+        context: InvocationContext,
+        source_node_id: str,
+        intermediate_state: PipelineIntermediateState,
     ) -> None:
-        # TODO: only output a preview image when requested
-        image = Generator.sample_to_lowres_estimated_image(sample)
-
-        (width, height) = image.size
-        width *= 8
-        height *= 8
-
-        dataURL = image_to_dataURL(image, image_format="JPEG")
-
-        context.services.events.emit_generator_progress(
-            context.graph_execution_state_id,
-            self.id,
-            {
-                "width": width,
-                "height": height,
-                "dataURL": dataURL
-            },
-            step,
-            self.steps,
+        stable_diffusion_step_callback(
+            context=context,
+            intermediate_state=intermediate_state,
+            node=self.dict(),
+            source_node_id=source_node_id,
         )
 
     def invoke(self, context: InvocationContext) -> ImageOutput:
-        def step_callback(state: PipelineIntermediateState):
-            self.dispatch_progress(context, state.latents, state.step)
-
         # Handle invalid model parameter
-        # TODO: figure out if this can be done via a validator that uses the model_cache
-        # TODO: How to get the default model name now?
-        #       (right now uses whatever current model is set in model manager)
-        model= context.services.model_manager.get_model()
+        model = choose_model(context.services.model_manager, self.model)
+
+        # Get the source node id (we are invoking the prepared node)
+        graph_execution_state = context.services.graph_execution_manager.get(
+            context.graph_execution_state_id
+        )
+        source_node_id = graph_execution_state.prepared_source_mapping[self.id]
+
         outputs = Txt2Img(model).generate(
             prompt=self.prompt,
-            step_callback=step_callback,
+            step_callback=partial(self.dispatch_progress, context, source_node_id),
             **self.dict(
                 exclude={"prompt"}
             ),  # Shorthand for passing all of the parameters above manually
@@ -95,9 +97,18 @@ class TextToImageInvocation(BaseInvocation):
         image_name = context.services.images.create_name(
             context.graph_execution_state_id, self.id
         )
-        context.services.images.save(image_type, image_name, generate_output.image)
-        return ImageOutput(
-            image=ImageField(image_type=image_type, image_name=image_name)
+
+        metadata = context.services.metadata.build_metadata(
+            session_id=context.graph_execution_state_id, node=self
+        )
+
+        context.services.images.save(
+            image_type, image_name, generate_output.image, metadata
+        )
+        return build_image_output(
+            image_type=image_type,
+            image_name=image_name,
+            image=generate_output.image,
         )
 
 
@@ -116,6 +127,19 @@ class ImageToImageInvocation(TextToImageInvocation):
         description="Whether or not the result should be fit to the aspect ratio of the input image",
     )
 
+    def dispatch_progress(
+        self,
+        context: InvocationContext,
+        source_node_id: str,
+        intermediate_state: PipelineIntermediateState,
+    ) -> None:
+        stable_diffusion_step_callback(
+            context=context,
+            intermediate_state=intermediate_state,
+            node=self.dict(),
+            source_node_id=source_node_id,
+        )
+
     def invoke(self, context: InvocationContext) -> ImageOutput:
         image = (
             None
@@ -126,24 +150,28 @@ class ImageToImageInvocation(TextToImageInvocation):
         )
         mask = None
 
-        def step_callback(sample, step=0):
-            self.dispatch_progress(context, sample, step)
-
         # Handle invalid model parameter
-        # TODO: figure out if this can be done via a validator that uses the model_cache
-        # TODO: How to get the default model name now?
-        model = context.services.model_manager.get_model()
-        generator_output = next(
-            Img2Img(model).generate(
+        model = choose_model(context.services.model_manager, self.model)
+
+        # Get the source node id (we are invoking the prepared node)
+        graph_execution_state = context.services.graph_execution_manager.get(
+            context.graph_execution_state_id
+        )
+        source_node_id = graph_execution_state.prepared_source_mapping[self.id]
+
+        outputs = Img2Img(model).generate(
             prompt=self.prompt,
             init_image=image,
             init_mask=mask,
-                step_callback=step_callback,
+            step_callback=partial(self.dispatch_progress, context, source_node_id),
             **self.dict(
                 exclude={"prompt", "image", "mask"}
             ),  # Shorthand for passing all of the parameters above manually
         )
-        )
+
+        # Outputs is an infinite iterator that will return a new InvokeAIGeneratorOutput object
+        # each time it is called. We only need the first one.
+        generator_output = next(outputs)
 
         result_image = generator_output.image
 
@@ -154,11 +182,19 @@ class ImageToImageInvocation(TextToImageInvocation):
         image_name = context.services.images.create_name(
             context.graph_execution_state_id, self.id
         )
-        context.services.images.save(image_type, image_name, result_image)
-        return ImageOutput(
-            image=ImageField(image_type=image_type, image_name=image_name)
+
+        metadata = context.services.metadata.build_metadata(
+            session_id=context.graph_execution_state_id, node=self
         )
 
+        context.services.images.save(image_type, image_name, result_image, metadata)
+        return build_image_output(
+            image_type=image_type,
+            image_name=image_name,
+            image=result_image,
+        )
+
+
 class InpaintInvocation(ImageToImageInvocation):
     """Generates an image using inpaint."""
 
@@ -173,6 +209,19 @@ class InpaintInvocation(ImageToImageInvocation):
         description="The amount by which to replace masked areas with latent noise",
     )
 
+    def dispatch_progress(
+        self,
+        context: InvocationContext,
+        source_node_id: str,
+        intermediate_state: PipelineIntermediateState,
+    ) -> None:
+        stable_diffusion_step_callback(
+            context=context,
+            intermediate_state=intermediate_state,
+            node=self.dict(),
+            source_node_id=source_node_id,
+        )
+
     def invoke(self, context: InvocationContext) -> ImageOutput:
         image = (
             None
@@ -187,24 +236,28 @@ class InpaintInvocation(ImageToImageInvocation):
             else context.services.images.get(self.mask.image_type, self.mask.image_name)
         )
 
-        def step_callback(sample, step=0):
-            self.dispatch_progress(context, sample, step)
-
         # Handle invalid model parameter
-        # TODO: figure out if this can be done via a validator that uses the model_cache
-        # TODO: How to get the default model name now?
-        manager = context.services.model_manager.get_model()
-        generator_output = next(
-            Inpaint(model).generate(
+        model = choose_model(context.services.model_manager, self.model)
+
+        # Get the source node id (we are invoking the prepared node)
+        graph_execution_state = context.services.graph_execution_manager.get(
+            context.graph_execution_state_id
+        )
+        source_node_id = graph_execution_state.prepared_source_mapping[self.id]
+
+        outputs = Inpaint(model).generate(
             prompt=self.prompt,
-                init_image=image,
-                mask_image=mask,
-                step_callback=step_callback,
+            init_img=image,
+            init_mask=mask,
+            step_callback=partial(self.dispatch_progress, context, source_node_id),
             **self.dict(
                 exclude={"prompt", "image", "mask"}
             ),  # Shorthand for passing all of the parameters above manually
         )
-        )
+
+        # Outputs is an infinite iterator that will return a new InvokeAIGeneratorOutput object
+        # each time it is called. We only need the first one.
+        generator_output = next(outputs)
 
         result_image = generator_output.image
 
@@ -215,7 +268,14 @@ class InpaintInvocation(ImageToImageInvocation):
         image_name = context.services.images.create_name(
             context.graph_execution_state_id, self.id
         )
-        context.services.images.save(image_type, image_name, result_image)
-        return ImageOutput(
-            image=ImageField(image_type=image_type, image_name=image_name)
+
+        metadata = context.services.metadata.build_metadata(
+            session_id=context.graph_execution_state_id, node=self
+        )
+
+        context.services.images.save(image_type, image_name, result_image, metadata)
+        return build_image_output(
+            image_type=image_type,
+            image_name=image_name,
+            image=result_image,
         )

invokeai/app/invocations/image.py

@@ -1,43 +1,83 @@
 # Copyright (c) 2022 Kyle Schouviller (https://github.com/kyle0654)
 
-from datetime import datetime, timezone
 from typing import Literal, Optional
 
 import numpy
 from PIL import Image, ImageFilter, ImageOps
 from pydantic import BaseModel, Field
 
-from ..services.image_storage import ImageType
-from ..services.invocation_services import InvocationServices
-from .baseinvocation import BaseInvocation, BaseInvocationOutput, InvocationContext
-
-
-class ImageField(BaseModel):
-    """An image field used for passing image objects between invocations"""
-
-    image_type: str = Field(
-        default=ImageType.RESULT, description="The type of the image"
+from ..models.image import ImageField, ImageType
+from .baseinvocation import (
+    BaseInvocation,
+    BaseInvocationOutput,
+    InvocationContext,
+    InvocationConfig,
 )
-    image_name: Optional[str] = Field(default=None, description="The name of the image")
+
+
+class PILInvocationConfig(BaseModel):
+    """Helper class to provide all PIL invocations with additional config"""
+
+    class Config(InvocationConfig):
+        schema_extra = {
+            "ui": {
+                "tags": ["PIL", "image"],
+            },
+        }
 
 
 class ImageOutput(BaseInvocationOutput):
     """Base class for invocations that output an image"""
+
     # fmt: off
     type: Literal["image"] = "image"
     image:      ImageField = Field(default=None, description="The output image")
+    width:   Optional[int] = Field(default=None, description="The width of the image in pixels")
+    height:  Optional[int] = Field(default=None, description="The height of the image in pixels")
     # fmt: on
 
+    class Config:
+        schema_extra = {
+            "required": ["type", "image", "width", "height", "mode"]
+        }
+
+
+def build_image_output(
+    image_type: ImageType, image_name: str, image: Image.Image
+) -> ImageOutput:
+    """Builds an ImageOutput and its ImageField"""
+    image_field = ImageField(
+        image_name=image_name,
+        image_type=image_type,
+    )
+    return ImageOutput(
+        image=image_field,
+        width=image.width,
+        height=image.height,
+        mode=image.mode,
+    )
+
+
 class MaskOutput(BaseInvocationOutput):
     """Base class for invocations that output a mask"""
+
     # fmt: off
     type: Literal["mask"] = "mask"
     mask:      ImageField = Field(default=None, description="The output mask")
-    #fomt: on
+    # fmt: on
+
+    class Config:
+        schema_extra = {
+            "required": [
+                "type",
+                "mask",
+            ]
+        }
+
 
-# TODO: this isn't really necessary anymore
 class LoadImageInvocation(BaseInvocation):
-    """Load an image from a filename and provide it as output."""
+    """Load an image and provide it as output."""
+
     # fmt: off
     type: Literal["load_image"] = "load_image"
 
@@ -45,10 +85,13 @@ class LoadImageInvocation(BaseInvocation):
     image_type: ImageType = Field(description="The type of the image")
     image_name:       str = Field(description="The name of the image")
     # fmt: on
-
     def invoke(self, context: InvocationContext) -> ImageOutput:
-        return ImageOutput(
-            image=ImageField(image_type=self.image_type, image_name=self.image_name)
+        image = context.services.images.get(self.image_type, self.image_name)
+
+        return build_image_output(
+            image_type=self.image_type,
+            image_name=self.image_name,
+            image=image,
         )
 
 
@@ -69,15 +112,16 @@ class ShowImageInvocation(BaseInvocation):
 
         # TODO: how to handle failure?
 
-        return ImageOutput(
-            image=ImageField(
-                image_type=self.image.image_type, image_name=self.image.image_name
-            )
+        return build_image_output(
+            image_type=self.image.image_type,
+            image_name=self.image.image_name,
+            image=image,
         )
 
 
-class CropImageInvocation(BaseInvocation):
+class CropImageInvocation(BaseInvocation, PILInvocationConfig):
     """Crops an image to a specified box. The box can be outside of the image."""
+
     # fmt: off
     type: Literal["crop"] = "crop"
 
@@ -103,14 +147,22 @@ class CropImageInvocation(BaseInvocation):
         image_name = context.services.images.create_name(
             context.graph_execution_state_id, self.id
         )
-        context.services.images.save(image_type, image_name, image_crop)
-        return ImageOutput(
-            image=ImageField(image_type=image_type, image_name=image_name)
+
+        metadata = context.services.metadata.build_metadata(
+            session_id=context.graph_execution_state_id, node=self
+        )
+        
+        context.services.images.save(image_type, image_name, image_crop, metadata)
+        return build_image_output(
+            image_type=image_type,
+            image_name=image_name,
+            image=image_crop,
         )
 
 
-class PasteImageInvocation(BaseInvocation):
+class PasteImageInvocation(BaseInvocation, PILInvocationConfig):
     """Pastes an image into another image."""
+
     # fmt: off
     type: Literal["paste"] = "paste"
 
@@ -133,7 +185,7 @@ class PasteImageInvocation(BaseInvocation):
             None
             if self.mask is None
             else ImageOps.invert(
-                services.images.get(self.mask.image_type, self.mask.image_name)
+                context.services.images.get(self.mask.image_type, self.mask.image_name)
             )
         )
         # TODO: probably shouldn't invert mask here... should user be required to do it?
@@ -153,14 +205,22 @@ class PasteImageInvocation(BaseInvocation):
         image_name = context.services.images.create_name(
             context.graph_execution_state_id, self.id
         )
-        context.services.images.save(image_type, image_name, new_image)
-        return ImageOutput(
-            image=ImageField(image_type=image_type, image_name=image_name)
+
+        metadata = context.services.metadata.build_metadata(
+            session_id=context.graph_execution_state_id, node=self
+        )
+        
+        context.services.images.save(image_type, image_name, new_image, metadata)
+        return build_image_output(
+            image_type=image_type,
+            image_name=image_name,
+            image=new_image,
         )
 
 
-class MaskFromAlphaInvocation(BaseInvocation):
+class MaskFromAlphaInvocation(BaseInvocation, PILInvocationConfig):
     """Extracts the alpha channel of an image as a mask."""
+
     # fmt: off
     type: Literal["tomask"] = "tomask"
 
@@ -182,11 +242,16 @@ class MaskFromAlphaInvocation(BaseInvocation):
         image_name = context.services.images.create_name(
             context.graph_execution_state_id, self.id
         )
-        context.services.images.save(image_type, image_name, image_mask)
+
+        metadata = context.services.metadata.build_metadata(
+            session_id=context.graph_execution_state_id, node=self
+        )
+
+        context.services.images.save(image_type, image_name, image_mask, metadata)
         return MaskOutput(mask=ImageField(image_type=image_type, image_name=image_name))
 
 
-class BlurInvocation(BaseInvocation):
+class BlurInvocation(BaseInvocation, PILInvocationConfig):
     """Blurs an image"""
 
     # fmt: off
@@ -214,14 +279,20 @@ class BlurInvocation(BaseInvocation):
         image_name = context.services.images.create_name(
             context.graph_execution_state_id, self.id
         )
-        context.services.images.save(image_type, image_name, blur_image)
-        return ImageOutput(
-            image=ImageField(image_type=image_type, image_name=image_name)
+
+        metadata = context.services.metadata.build_metadata(
+            session_id=context.graph_execution_state_id, node=self
+        )
+
+        context.services.images.save(image_type, image_name, blur_image, metadata)
+        return build_image_output(
+            image_type=image_type, image_name=image_name, image=blur_image
         )
 
 
-class LerpInvocation(BaseInvocation):
+class LerpInvocation(BaseInvocation, PILInvocationConfig):
     """Linear interpolation of all pixels of an image"""
+
     # fmt: off
     type: Literal["lerp"] = "lerp"
 
@@ -245,14 +316,20 @@ class LerpInvocation(BaseInvocation):
         image_name = context.services.images.create_name(
             context.graph_execution_state_id, self.id
         )
-        context.services.images.save(image_type, image_name, lerp_image)
-        return ImageOutput(
-            image=ImageField(image_type=image_type, image_name=image_name)
+
+        metadata = context.services.metadata.build_metadata(
+            session_id=context.graph_execution_state_id, node=self
+        )
+
+        context.services.images.save(image_type, image_name, lerp_image, metadata)
+        return build_image_output(
+            image_type=image_type, image_name=image_name, image=lerp_image
         )
 
 
-class InverseLerpInvocation(BaseInvocation):
+class InverseLerpInvocation(BaseInvocation, PILInvocationConfig):
     """Inverse linear interpolation of all pixels of an image"""
+
     # fmt: off
     type: Literal["ilerp"] = "ilerp"
 
@@ -281,7 +358,12 @@ class InverseLerpInvocation(BaseInvocation):
         image_name = context.services.images.create_name(
             context.graph_execution_state_id, self.id
         )
-        context.services.images.save(image_type, image_name, ilerp_image)
-        return ImageOutput(
-            image=ImageField(image_type=image_type, image_name=image_name)
+
+        metadata = context.services.metadata.build_metadata(
+            session_id=context.graph_execution_state_id, node=self
+        )
+
+        context.services.images.save(image_type, image_name, ilerp_image, metadata)
+        return build_image_output(
+            image_type=image_type, image_name=image_name, image=ilerp_image
         )

invokeai/app/invocations/latent.py

@@ -0,0 +1,371 @@
+# Copyright (c) 2023 Kyle Schouviller (https://github.com/kyle0654)
+
+import random
+from typing import Literal, Optional
+from pydantic import BaseModel, Field
+import torch
+
+from invokeai.app.invocations.util.choose_model import choose_model
+
+from invokeai.app.util.step_callback import stable_diffusion_step_callback
+
+from ...backend.model_management.model_manager import ModelManager
+from ...backend.util.devices import choose_torch_device, torch_dtype
+from ...backend.stable_diffusion.diffusion.shared_invokeai_diffusion import PostprocessingSettings
+from ...backend.image_util.seamless import configure_model_padding
+from ...backend.prompting.conditioning import get_uc_and_c_and_ec
+from ...backend.stable_diffusion.diffusers_pipeline import ConditioningData, StableDiffusionGeneratorPipeline
+from .baseinvocation import BaseInvocation, BaseInvocationOutput, InvocationContext, InvocationConfig
+import numpy as np
+from ..services.image_storage import ImageType
+from .baseinvocation import BaseInvocation, InvocationContext
+from .image import ImageField, ImageOutput, build_image_output
+from ...backend.stable_diffusion import PipelineIntermediateState
+from diffusers.schedulers import SchedulerMixin as Scheduler
+import diffusers
+from diffusers import DiffusionPipeline
+
+
+class LatentsField(BaseModel):
+    """A latents field used for passing latents between invocations"""
+
+    latents_name: Optional[str] = Field(default=None, description="The name of the latents")
+
+    class Config:
+        schema_extra = {"required": ["latents_name"]}
+
+class LatentsOutput(BaseInvocationOutput):
+    """Base class for invocations that output latents"""
+    #fmt: off
+    type: Literal["latent_output"] = "latent_output"
+    latents: LatentsField            = Field(default=None, description="The output latents")
+    #fmt: on
+
+class NoiseOutput(BaseInvocationOutput):
+    """Invocation noise output"""
+    #fmt: off
+    type: Literal["noise_output"] = "noise_output"
+    noise: LatentsField            = Field(default=None, description="The output noise")
+    #fmt: on
+
+
+# TODO: this seems like a hack
+scheduler_map = dict(
+    ddim=diffusers.DDIMScheduler,
+    dpmpp_2=diffusers.DPMSolverMultistepScheduler,
+    k_dpm_2=diffusers.KDPM2DiscreteScheduler,
+    k_dpm_2_a=diffusers.KDPM2AncestralDiscreteScheduler,
+    k_dpmpp_2=diffusers.DPMSolverMultistepScheduler,
+    k_euler=diffusers.EulerDiscreteScheduler,
+    k_euler_a=diffusers.EulerAncestralDiscreteScheduler,
+    k_heun=diffusers.HeunDiscreteScheduler,
+    k_lms=diffusers.LMSDiscreteScheduler,
+    plms=diffusers.PNDMScheduler,
+)
+
+
+SAMPLER_NAME_VALUES = Literal[
+    tuple(list(scheduler_map.keys()))
+]
+
+
+def get_scheduler(scheduler_name:str, model: StableDiffusionGeneratorPipeline)->Scheduler:
+    scheduler_class = scheduler_map.get(scheduler_name,'ddim')
+    scheduler = scheduler_class.from_config(model.scheduler.config)
+    # hack copied over from generate.py
+    if not hasattr(scheduler, 'uses_inpainting_model'):
+        scheduler.uses_inpainting_model = lambda: False
+    return scheduler
+
+
+def get_noise(width:int, height:int, device:torch.device, seed:int = 0, latent_channels:int=4, use_mps_noise:bool=False, downsampling_factor:int = 8):
+    # limit noise to only the diffusion image channels, not the mask channels
+    input_channels = min(latent_channels, 4)
+    use_device = "cpu" if (use_mps_noise or device.type == "mps") else device
+    generator = torch.Generator(device=use_device).manual_seed(seed)
+    x = torch.randn(
+        [
+            1,
+            input_channels,
+            height // downsampling_factor,
+            width //  downsampling_factor,
+        ],
+        dtype=torch_dtype(device),
+        device=use_device,
+        generator=generator,
+    ).to(device)
+    # if self.perlin > 0.0:
+    #     perlin_noise = self.get_perlin_noise(
+    #         width // self.downsampling_factor, height // self.downsampling_factor
+    #     )
+    #     x = (1 - self.perlin) * x + self.perlin * perlin_noise
+    return x
+
+
+def random_seed():
+    return random.randint(0, np.iinfo(np.uint32).max)
+
+
+class NoiseInvocation(BaseInvocation):
+    """Generates latent noise."""
+
+    type: Literal["noise"] = "noise"
+
+    # Inputs
+    seed:        int = Field(ge=0, le=np.iinfo(np.uint32).max, description="The seed to use", default_factory=random_seed)
+    width:       int = Field(default=512, multiple_of=64, gt=0, description="The width of the resulting noise", )
+    height:      int = Field(default=512, multiple_of=64, gt=0, description="The height of the resulting noise", )
+
+
+    # Schema customisation
+    class Config(InvocationConfig):
+        schema_extra = {
+            "ui": {
+                "tags": ["latents", "noise"],
+            },
+        }
+
+    def invoke(self, context: InvocationContext) -> NoiseOutput:
+        device = torch.device(choose_torch_device())
+        noise = get_noise(self.width, self.height, device, self.seed)
+
+        name = f'{context.graph_execution_state_id}__{self.id}'
+        context.services.latents.set(name, noise)
+        return NoiseOutput(
+            noise=LatentsField(latents_name=name)
+        )
+
+
+# Text to image
+class TextToLatentsInvocation(BaseInvocation):
+    """Generates latents from a prompt."""
+
+    type: Literal["t2l"] = "t2l"
+
+    # Inputs
+    # TODO: consider making prompt optional to enable providing prompt through a link
+    # fmt: off
+    prompt: Optional[str] = Field(description="The prompt to generate an image from")
+    seed:        int = Field(default=-1,ge=-1, le=np.iinfo(np.uint32).max, description="The seed to use (-1 for a random seed)", )
+    noise: Optional[LatentsField] = Field(description="The noise to use")
+    steps:       int = Field(default=10, gt=0, description="The number of steps to use to generate the image")
+    width:       int = Field(default=512, multiple_of=64, gt=0, description="The width of the resulting image", )
+    height:      int = Field(default=512, multiple_of=64, gt=0, description="The height of the resulting image", )
+    cfg_scale: float = Field(default=7.5, gt=0, description="The Classifier-Free Guidance, higher values may result in a result closer to the prompt", )
+    scheduler: SAMPLER_NAME_VALUES = Field(default="k_lms", description="The scheduler to use" )
+    seamless:   bool = Field(default=False, description="Whether or not to generate an image that can tile without seams", )
+    seamless_axes: str = Field(default="", description="The axes to tile the image on, 'x' and/or 'y'")
+    model:       str = Field(default="", description="The model to use (currently ignored)")
+    progress_images: bool = Field(default=False, description="Whether or not to produce progress images during generation",  )
+    # fmt: on
+
+    # Schema customisation
+    class Config(InvocationConfig):
+        schema_extra = {
+            "ui": {
+                "tags": ["latents", "image"],
+                "type_hints": {
+                  "model": "model"
+                }
+            },
+        }
+
+    # TODO: pass this an emitter method or something? or a session for dispatching?
+    def dispatch_progress(
+        self, context: InvocationContext, source_node_id: str, intermediate_state: PipelineIntermediateState
+    ) -> None:
+        stable_diffusion_step_callback(
+            context=context,
+            intermediate_state=intermediate_state,
+            node=self.dict(),
+            source_node_id=source_node_id,
+        )
+
+    def get_model(self, model_manager: ModelManager) -> StableDiffusionGeneratorPipeline:
+        model_info = choose_model(model_manager, self.model)
+        model_name = model_info['model_name']
+        model_hash = model_info['hash']
+        model: StableDiffusionGeneratorPipeline = model_info['model']
+        model.scheduler = get_scheduler(
+            model=model,
+            scheduler_name=self.scheduler
+        )
+
+        if isinstance(model, DiffusionPipeline):
+            for component in [model.unet, model.vae]:
+                configure_model_padding(component,
+                                        self.seamless,
+                                        self.seamless_axes
+                                        )
+        else:
+            configure_model_padding(model,
+                                    self.seamless,
+                                    self.seamless_axes
+                                    )
+
+        return model
+
+
+    def get_conditioning_data(self, model: StableDiffusionGeneratorPipeline) -> ConditioningData:
+        uc, c, extra_conditioning_info = get_uc_and_c_and_ec(self.prompt, model=model)
+        conditioning_data = ConditioningData(
+            uc,
+            c,
+            self.cfg_scale,
+            extra_conditioning_info,
+            postprocessing_settings=PostprocessingSettings(
+                threshold=0.0,#threshold,
+                warmup=0.2,#warmup,
+                h_symmetry_time_pct=None,#h_symmetry_time_pct,
+                v_symmetry_time_pct=None#v_symmetry_time_pct,
+            ),
+        ).add_scheduler_args_if_applicable(model.scheduler, eta=None)#ddim_eta)
+        return conditioning_data
+
+
+    def invoke(self, context: InvocationContext) -> LatentsOutput:
+        noise = context.services.latents.get(self.noise.latents_name)
+
+        # Get the source node id (we are invoking the prepared node)
+        graph_execution_state = context.services.graph_execution_manager.get(context.graph_execution_state_id)
+        source_node_id = graph_execution_state.prepared_source_mapping[self.id]
+
+        def step_callback(state: PipelineIntermediateState):
+            self.dispatch_progress(context, source_node_id, state)
+
+        model = self.get_model(context.services.model_manager)
+        conditioning_data = self.get_conditioning_data(model)
+
+        # TODO: Verify the noise is the right size
+
+        result_latents, result_attention_map_saver = model.latents_from_embeddings(
+            latents=torch.zeros_like(noise, dtype=torch_dtype(model.device)),
+            noise=noise,
+            num_inference_steps=self.steps,
+            conditioning_data=conditioning_data,
+            callback=step_callback
+        )
+
+        # https://discuss.huggingface.co/t/memory-usage-by-later-pipeline-stages/23699
+        torch.cuda.empty_cache()
+
+        name = f'{context.graph_execution_state_id}__{self.id}'
+        context.services.latents.set(name, result_latents)
+        return LatentsOutput(
+            latents=LatentsField(latents_name=name)
+        )
+
+
+class LatentsToLatentsInvocation(TextToLatentsInvocation):
+    """Generates latents using latents as base image."""
+
+    type: Literal["l2l"] = "l2l"
+
+    # Schema customisation
+    class Config(InvocationConfig):
+        schema_extra = {
+            "ui": {
+                "tags": ["latents"],
+                "type_hints": {
+                    "model": "model"
+                }
+            },
+        }
+
+    # Inputs
+    latents: Optional[LatentsField] = Field(description="The latents to use as a base image")
+    strength: float = Field(default=0.5, description="The strength of the latents to use")
+
+    def invoke(self, context: InvocationContext) -> LatentsOutput:
+        noise = context.services.latents.get(self.noise.latents_name)
+        latent = context.services.latents.get(self.latents.latents_name)
+
+        # Get the source node id (we are invoking the prepared node)
+        graph_execution_state = context.services.graph_execution_manager.get(context.graph_execution_state_id)
+        source_node_id = graph_execution_state.prepared_source_mapping[self.id]
+
+        def step_callback(state: PipelineIntermediateState):
+            self.dispatch_progress(context, source_node_id, state)
+
+        model = self.get_model(context.services.model_manager)
+        conditioning_data = self.get_conditioning_data(model)
+
+        # TODO: Verify the noise is the right size
+
+        initial_latents = latent if self.strength < 1.0 else torch.zeros_like(
+            latent, device=model.device, dtype=latent.dtype
+        )
+
+        timesteps, _ = model.get_img2img_timesteps(
+            self.steps,
+            self.strength,
+            device=model.device,
+        )
+
+        result_latents, result_attention_map_saver = model.latents_from_embeddings(
+            latents=initial_latents,
+            timesteps=timesteps,
+            noise=noise,
+            num_inference_steps=self.steps,
+            conditioning_data=conditioning_data,
+            callback=step_callback
+        )
+
+        # https://discuss.huggingface.co/t/memory-usage-by-later-pipeline-stages/23699
+        torch.cuda.empty_cache()
+
+        name = f'{context.graph_execution_state_id}__{self.id}'
+        context.services.latents.set(name, result_latents)
+        return LatentsOutput(
+            latents=LatentsField(latents_name=name)
+        )
+
+
+# Latent to image
+class LatentsToImageInvocation(BaseInvocation):
+    """Generates an image from latents."""
+
+    type: Literal["l2i"] = "l2i"
+
+    # Inputs
+    latents: Optional[LatentsField] = Field(description="The latents to generate an image from")
+    model: str = Field(default="", description="The model to use")
+
+    # Schema customisation
+    class Config(InvocationConfig):
+        schema_extra = {
+            "ui": {
+                "tags": ["latents", "image"],
+                "type_hints": {
+                  "model": "model"
+                }
+            },
+        }
+
+    @torch.no_grad()
+    def invoke(self, context: InvocationContext) -> ImageOutput:
+        latents = context.services.latents.get(self.latents.latents_name)
+
+        # TODO: this only really needs the vae
+        model_info = choose_model(context.services.model_manager, self.model)
+        model: StableDiffusionGeneratorPipeline = model_info['model']
+
+        with torch.inference_mode():
+            np_image = model.decode_latents(latents)
+            image = model.numpy_to_pil(np_image)[0]
+
+            image_type = ImageType.RESULT
+            image_name = context.services.images.create_name(
+                context.graph_execution_state_id, self.id
+            )
+
+            metadata = context.services.metadata.build_metadata(
+                session_id=context.graph_execution_state_id, node=self
+            )
+
+            context.services.images.save(image_type, image_name, image, metadata)
+            return build_image_output(
+                image_type=image_type,
+                image_name=image_name,
+                image=image
+            )

invokeai/app/invocations/math.py

@@ -0,0 +1,75 @@
+# Copyright (c) 2023 Kyle Schouviller (https://github.com/kyle0654)
+
+from typing import Literal
+
+from pydantic import BaseModel, Field
+
+from .baseinvocation import BaseInvocation, BaseInvocationOutput, InvocationContext, InvocationConfig
+
+
+class MathInvocationConfig(BaseModel):
+    """Helper class to provide all math invocations with additional config"""
+
+    # Schema customisation
+    class Config(InvocationConfig):
+        schema_extra = {
+            "ui": {
+                "tags": ["math"],
+            }
+        }
+
+
+class IntOutput(BaseInvocationOutput):
+    """An integer output"""
+    #fmt: off
+    type: Literal["int_output"] = "int_output"
+    a: int = Field(default=None, description="The output integer")
+    #fmt: on
+
+
+class AddInvocation(BaseInvocation, MathInvocationConfig):
+    """Adds two numbers"""
+    #fmt: off
+    type: Literal["add"] = "add"
+    a: int = Field(default=0, description="The first number")
+    b: int = Field(default=0, description="The second number")
+    #fmt: on
+
+    def invoke(self, context: InvocationContext) -> IntOutput:
+        return IntOutput(a=self.a + self.b)
+
+
+class SubtractInvocation(BaseInvocation, MathInvocationConfig):
+    """Subtracts two numbers"""
+    #fmt: off
+    type: Literal["sub"] = "sub"
+    a: int = Field(default=0, description="The first number")
+    b: int = Field(default=0, description="The second number")
+    #fmt: on
+
+    def invoke(self, context: InvocationContext) -> IntOutput:
+        return IntOutput(a=self.a - self.b)
+
+
+class MultiplyInvocation(BaseInvocation, MathInvocationConfig):
+    """Multiplies two numbers"""
+    #fmt: off
+    type: Literal["mul"] = "mul"
+    a: int = Field(default=0, description="The first number")
+    b: int = Field(default=0, description="The second number")
+    #fmt: on
+
+    def invoke(self, context: InvocationContext) -> IntOutput:
+        return IntOutput(a=self.a * self.b)
+
+
+class DivideInvocation(BaseInvocation, MathInvocationConfig):
+    """Divides two numbers"""
+    #fmt: off
+    type: Literal["div"] = "div"
+    a: int = Field(default=0, description="The first number")
+    b: int = Field(default=0, description="The second number")
+    #fmt: on
+
+    def invoke(self, context: InvocationContext) -> IntOutput:
+        return IntOutput(a=int(self.a / self.b))

invokeai/app/invocations/params.py

@@ -0,0 +1,18 @@
+# Copyright (c) 2023 Kyle Schouviller (https://github.com/kyle0654)
+
+from typing import Literal
+from pydantic import Field
+from .baseinvocation import BaseInvocation, BaseInvocationOutput, InvocationContext
+from .math import IntOutput
+
+# Pass-through parameter nodes - used by subgraphs
+
+class ParamIntInvocation(BaseInvocation):
+    """An integer parameter"""
+    #fmt: off
+    type: Literal["param_int"] = "param_int"
+    a: int = Field(default=0, description="The integer value")
+    #fmt: on
+
+    def invoke(self, context: InvocationContext) -> IntOutput:
+        return IntOutput(a=self.a)

invokeai/app/invocations/prompt.py

@@ -12,3 +12,11 @@ class PromptOutput(BaseInvocationOutput):
 
     prompt: str = Field(default=None, description="The output prompt")
     #fmt: on
+
+    class Config:
+        schema_extra = {
+            'required': [
+                'type',
+                'prompt',
+            ]
+        }

invokeai/app/invocations/reconstruct.py

@@ -1,12 +1,11 @@
-from datetime import datetime, timezone
 from typing import Literal, Union
 
 from pydantic import Field
 
-from ..services.image_storage import ImageType
-from ..services.invocation_services import InvocationServices
-from .baseinvocation import BaseInvocation, InvocationContext
-from .image import ImageField, ImageOutput
+from invokeai.app.models.image import ImageField, ImageType
+
+from .baseinvocation import BaseInvocation, InvocationContext, InvocationConfig
+from .image import ImageOutput, build_image_output
 
 class RestoreFaceInvocation(BaseInvocation):
     """Restores faces in an image."""
@@ -18,6 +17,14 @@ class RestoreFaceInvocation(BaseInvocation):
     strength:                float = Field(default=0.75, gt=0, le=1, description="The strength of the restoration"  )
     #fmt: on
     
+    # Schema customisation
+    class Config(InvocationConfig):
+        schema_extra = {
+            "ui": {
+                "tags": ["restoration", "image"],
+            },
+        }
+
     def invoke(self, context: InvocationContext) -> ImageOutput:
         image = context.services.images.get(
             self.image.image_type, self.image.image_name
@@ -36,7 +43,14 @@ class RestoreFaceInvocation(BaseInvocation):
         image_name = context.services.images.create_name(
             context.graph_execution_state_id, self.id
         )
-        context.services.images.save(image_type, image_name, results[0][0])
-        return ImageOutput(
-            image=ImageField(image_type=image_type, image_name=image_name)
+
+        metadata = context.services.metadata.build_metadata(
+            session_id=context.graph_execution_state_id, node=self
+        )
+
+        context.services.images.save(image_type, image_name, results[0][0], metadata)
+        return build_image_output(
+            image_type=image_type,
+            image_name=image_name,
+            image=results[0][0]
         )

invokeai/app/invocations/upscale.py

@@ -1,14 +1,12 @@
 # Copyright (c) 2022 Kyle Schouviller (https://github.com/kyle0654)
 
-from datetime import datetime, timezone
 from typing import Literal, Union
 
 from pydantic import Field
 
-from ..services.image_storage import ImageType
-from ..services.invocation_services import InvocationServices
-from .baseinvocation import BaseInvocation, InvocationContext
-from .image import ImageField, ImageOutput
+from invokeai.app.models.image import ImageField, ImageType
+from .baseinvocation import BaseInvocation, InvocationContext, InvocationConfig
+from .image import ImageOutput, build_image_output
 
 
 class UpscaleInvocation(BaseInvocation):
@@ -22,6 +20,15 @@ class UpscaleInvocation(BaseInvocation):
     level: Literal[2, 4] = Field(default=2, description="The upscale level")
     #fmt: on
 
+
+    # Schema customisation
+    class Config(InvocationConfig):
+        schema_extra = {
+            "ui": {
+                "tags": ["upscaling", "image"],
+            },
+        }
+
     def invoke(self, context: InvocationContext) -> ImageOutput:
         image = context.services.images.get(
             self.image.image_type, self.image.image_name
@@ -40,7 +47,14 @@ class UpscaleInvocation(BaseInvocation):
         image_name = context.services.images.create_name(
             context.graph_execution_state_id, self.id
         )
-        context.services.images.save(image_type, image_name, results[0][0])
-        return ImageOutput(
-            image=ImageField(image_type=image_type, image_name=image_name)
+
+        metadata = context.services.metadata.build_metadata(
+            session_id=context.graph_execution_state_id, node=self
+        )
+
+        context.services.images.save(image_type, image_name, results[0][0], metadata)
+        return build_image_output(
+            image_type=image_type,
+            image_name=image_name,
+            image=results[0][0]
         )

invokeai/app/invocations/util/choose_model.py

@@ -0,0 +1,14 @@
+from invokeai.backend.model_management.model_manager import ModelManager
+
+
+def choose_model(model_manager: ModelManager, model_name: str):
+    """Returns the default model if the `model_name` not a valid model, else returns the selected model."""
+    if model_manager.valid_model(model_name):
+        model = model_manager.get_model(model_name)
+    else:
+        model = model_manager.get_model()
+        print(
+            f"* Warning: '{model_name}' is not a valid model name. Using default model \'{model['model_name']}\' instead."
+        )
+
+    return model

invokeai/app/models/exceptions.py

@@ -0,0 +1,3 @@
+class CanceledException(Exception):
+    """Execution canceled by user."""
+    pass

invokeai/app/models/image.py

@@ -0,0 +1,29 @@
+from enum import Enum
+from typing import Optional
+from pydantic import BaseModel, Field
+
+
+class ImageType(str, Enum):
+    RESULT = "results"
+    INTERMEDIATE = "intermediates"
+    UPLOAD = "uploads"
+
+
+def is_image_type(obj):
+    try:
+        ImageType(obj)
+    except ValueError:
+        return False
+    return True
+
+
+class ImageField(BaseModel):
+    """An image field used for passing image objects between invocations"""
+
+    image_type: ImageType = Field(
+        default=ImageType.RESULT, description="The type of the image"
+    )
+    image_name: Optional[str] = Field(default=None, description="The name of the image")
+
+    class Config:
+        schema_extra = {"required": ["image_type", "image_name"]}

invokeai/app/services/default_graphs.py

@@ -0,0 +1,56 @@
+from ..invocations.latent import LatentsToImageInvocation, NoiseInvocation, TextToLatentsInvocation
+from ..invocations.params import ParamIntInvocation
+from .graph import Edge, EdgeConnection, ExposedNodeInput, ExposedNodeOutput, Graph, LibraryGraph
+from .item_storage import ItemStorageABC
+
+
+default_text_to_image_graph_id = '539b2af5-2b4d-4d8c-8071-e54a3255fc74'
+
+
+def create_text_to_image() -> LibraryGraph:
+    return LibraryGraph(
+        id=default_text_to_image_graph_id,
+        name='t2i',
+        description='Converts text to an image',
+        graph=Graph(
+            nodes={
+                'width': ParamIntInvocation(id='width', a=512),
+                'height': ParamIntInvocation(id='height', a=512),
+                '3': NoiseInvocation(id='3'),
+                '4': TextToLatentsInvocation(id='4'),
+                '5': LatentsToImageInvocation(id='5')
+            },
+            edges=[
+                Edge(source=EdgeConnection(node_id='width', field='a'), destination=EdgeConnection(node_id='3', field='width')),
+                Edge(source=EdgeConnection(node_id='height', field='a'), destination=EdgeConnection(node_id='3', field='height')),
+                Edge(source=EdgeConnection(node_id='width', field='a'), destination=EdgeConnection(node_id='4', field='width')),
+                Edge(source=EdgeConnection(node_id='height', field='a'), destination=EdgeConnection(node_id='4', field='height')),
+                Edge(source=EdgeConnection(node_id='3', field='noise'), destination=EdgeConnection(node_id='4', field='noise')),
+                Edge(source=EdgeConnection(node_id='4', field='latents'), destination=EdgeConnection(node_id='5', field='latents')),
+            ]
+        ),
+        exposed_inputs=[
+            ExposedNodeInput(node_path='4', field='prompt', alias='prompt'),
+            ExposedNodeInput(node_path='width', field='a', alias='width'),
+            ExposedNodeInput(node_path='height', field='a', alias='height')
+        ],
+        exposed_outputs=[
+            ExposedNodeOutput(node_path='5', field='image', alias='image')
+        ])
+
+
+def create_system_graphs(graph_library: ItemStorageABC[LibraryGraph]) -> list[LibraryGraph]:
+    """Creates the default system graphs, or adds new versions if the old ones don't match"""
+
+    graphs: list[LibraryGraph] = list()
+
+    text_to_image = graph_library.get(default_text_to_image_graph_id)
+    
+    # TODO: Check if the graph is the same as the default one, and if not, update it
+    #if text_to_image is None:
+    text_to_image = create_text_to_image()
+    graph_library.set(text_to_image)
+
+    graphs.append(text_to_image)
+
+    return graphs

invokeai/app/services/events.py

@@ -1,10 +1,9 @@
 # Copyright (c) 2022 Kyle Schouviller (https://github.com/kyle0654)
 
-from typing import Any, Dict, TypedDict
+from typing import Any
+from invokeai.app.api.models.images import ProgressImage
+from invokeai.app.util.misc import get_timestamp
 
-ProgressImage = TypedDict(
-    "ProgressImage", {"dataURL": str, "width": int, "height": int}
-)
 
 class EventServiceBase:
     session_event: str = "session_event"
@@ -14,7 +13,8 @@ class EventServiceBase:
     def dispatch(self, event_name: str, payload: Any) -> None:
         pass
 
-    def __emit_session_event(self, event_name: str, payload: Dict) -> None:
+    def __emit_session_event(self, event_name: str, payload: dict) -> None:
+        payload["timestamp"] = get_timestamp()
         self.dispatch(
             event_name=EventServiceBase.session_event,
             payload=dict(event=event_name, data=payload),
@@ -25,7 +25,8 @@ class EventServiceBase:
     def emit_generator_progress(
         self,
         graph_execution_state_id: str,
-        invocation_id: str,
+        node: dict,
+        source_node_id: str,
         progress_image: ProgressImage | None,
         step: int,
         total_steps: int,
@@ -35,48 +36,60 @@ class EventServiceBase:
             event_name="generator_progress",
             payload=dict(
                 graph_execution_state_id=graph_execution_state_id,
-                invocation_id=invocation_id,
-                progress_image=progress_image,
+                node=node,
+                source_node_id=source_node_id,
+                progress_image=progress_image.dict() if progress_image is not None else None,
                 step=step,
                 total_steps=total_steps,
             ),
         )
 
     def emit_invocation_complete(
-        self, graph_execution_state_id: str, invocation_id: str, result: Dict
+        self,
+        graph_execution_state_id: str,
+        result: dict,
+        node: dict,
+        source_node_id: str,
     ) -> None:
         """Emitted when an invocation has completed"""
         self.__emit_session_event(
             event_name="invocation_complete",
             payload=dict(
                 graph_execution_state_id=graph_execution_state_id,
-                invocation_id=invocation_id,
+                node=node,
+                source_node_id=source_node_id,
                 result=result,
             ),
         )
 
     def emit_invocation_error(
-        self, graph_execution_state_id: str, invocation_id: str, error: str
+        self,
+        graph_execution_state_id: str,
+        node: dict,
+        source_node_id: str,
+        error: str,
     ) -> None:
         """Emitted when an invocation has completed"""
         self.__emit_session_event(
             event_name="invocation_error",
             payload=dict(
                 graph_execution_state_id=graph_execution_state_id,
-                invocation_id=invocation_id,
+                node=node,
+                source_node_id=source_node_id,
                 error=error,
             ),
         )
 
     def emit_invocation_started(
-        self, graph_execution_state_id: str, invocation_id: str
+        self, graph_execution_state_id: str, node: dict, source_node_id: str
     ) -> None:
         """Emitted when an invocation has started"""
         self.__emit_session_event(
             event_name="invocation_started",
             payload=dict(
                 graph_execution_state_id=graph_execution_state_id,
-                invocation_id=invocation_id,
+                node=node,
+                source_node_id=source_node_id,
             ),
         )
 
@@ -84,5 +97,7 @@ class EventServiceBase:
         """Emitted when a session has completed all invocations"""
         self.__emit_session_event(
             event_name="graph_execution_state_complete",
-            payload=dict(graph_execution_state_id=graph_execution_state_id),
+            payload=dict(
+                graph_execution_state_id=graph_execution_state_id,
+            ),
         )

invokeai/app/services/graph.py

@@ -2,7 +2,6 @@
 
 import copy
 import itertools
-import traceback
 import uuid
 from types import NoneType
 from typing import (
@@ -17,7 +16,7 @@ from typing import (
 )
 
 import networkx as nx
-from pydantic import BaseModel, validator
+from pydantic import BaseModel, root_validator, validator
 from pydantic.fields import Field
 
 from ..invocations import *
@@ -26,7 +25,6 @@ from ..invocations.baseinvocation import (
     BaseInvocationOutput,
     InvocationContext,
 )
-from .invocation_services import InvocationServices
 
 
 class EdgeConnection(BaseModel):
@@ -127,6 +125,13 @@ class NodeAlreadyExecutedError(Exception):
 class GraphInvocationOutput(BaseInvocationOutput):
     type: Literal["graph_output"] = "graph_output"
 
+    class Config:
+        schema_extra = {
+            'required': [
+                'type',
+                'image',
+            ]
+        }
 
 # TODO: Fill this out and move to invocations
 class GraphInvocation(BaseInvocation):
@@ -147,6 +152,13 @@ class IterateInvocationOutput(BaseInvocationOutput):
 
     item: Any = Field(description="The item being iterated over")
 
+    class Config:
+        schema_extra = {
+            'required': [
+                'type',
+                'item',
+            ]
+        }
 
 # TODO: Fill this out and move to invocations
 class IterateInvocation(BaseInvocation):
@@ -169,6 +181,13 @@ class CollectInvocationOutput(BaseInvocationOutput):
 
     collection: list[Any] = Field(description="The collection of input items")
 
+    class Config:
+        schema_extra = {
+            'required': [
+                'type',
+                'collection',
+            ]
+        }
 
 class CollectInvocation(BaseInvocation):
     """Collects values into a collection"""
@@ -194,7 +213,7 @@ InvocationOutputsUnion = Union[BaseInvocationOutput.get_all_subclasses_tuple()]
 
 
 class Graph(BaseModel):
-    id: str = Field(description="The id of this graph", default_factory=uuid.uuid4)
+    id: str = Field(description="The id of this graph", default_factory=lambda: uuid.uuid4().__str__())
     # TODO: use a list (and never use dict in a BaseModel) because pydantic/fastapi hates me
     nodes: dict[str, Annotated[InvocationsUnion, Field(discriminator="type")]] = Field(
         description="The nodes in this graph", default_factory=dict
@@ -262,7 +281,8 @@ class Graph(BaseModel):
         :raises InvalidEdgeError: the provided edge is invalid.
         """
 
-        if self._is_edge_valid(edge) and edge not in self.edges:
+        self._validate_edge(edge)
+        if edge not in self.edges:
             self.edges.append(edge)
         else:
             raise InvalidEdgeError()
@@ -333,7 +353,7 @@ class Graph(BaseModel):
 
         return True
 
-    def _is_edge_valid(self, edge: Edge) -> bool:
+    def _validate_edge(self, edge: Edge):
         """Validates that a new edge doesn't create a cycle in the graph"""
 
         # Validate that the nodes exist (edges may contain node paths, so we can't just check for nodes directly)
@@ -341,54 +361,53 @@ class Graph(BaseModel):
             from_node = self.get_node(edge.source.node_id)
             to_node = self.get_node(edge.destination.node_id)
         except NodeNotFoundError:
-            return False
+            raise InvalidEdgeError("One or both nodes don't exist")
 
         # Validate that an edge to this node+field doesn't already exist
         input_edges = self._get_input_edges(edge.destination.node_id, edge.destination.field)
         if len(input_edges) > 0 and not isinstance(to_node, CollectInvocation):
-            return False
+            raise InvalidEdgeError(f'Edge to node {edge.destination.node_id} field {edge.destination.field} already exists')
 
         # Validate that no cycles would be created
         g = self.nx_graph_flat()
         g.add_edge(edge.source.node_id, edge.destination.node_id)
         if not nx.is_directed_acyclic_graph(g):
-            return False
+            raise InvalidEdgeError(f'Edge creates a cycle in the graph')
 
         # Validate that the field types are compatible
         if not are_connections_compatible(
             from_node, edge.source.field, to_node, edge.destination.field
         ):
-            return False
+            raise InvalidEdgeError(f'Fields are incompatible')
 
         # Validate if iterator output type matches iterator input type (if this edge results in both being set)
         if isinstance(to_node, IterateInvocation) and edge.destination.field == "collection":
             if not self._is_iterator_connection_valid(
                 edge.destination.node_id, new_input=edge.source
             ):
-                return False
+                raise InvalidEdgeError(f'Iterator input type does not match iterator output type')
 
         # Validate if iterator input type matches output type (if this edge results in both being set)
         if isinstance(from_node, IterateInvocation) and edge.source.field == "item":
             if not self._is_iterator_connection_valid(
                 edge.source.node_id, new_output=edge.destination
             ):
-                return False
+                raise InvalidEdgeError(f'Iterator output type does not match iterator input type')
 
         # Validate if collector input type matches output type (if this edge results in both being set)
         if isinstance(to_node, CollectInvocation) and edge.destination.field == "item":
             if not self._is_collector_connection_valid(
                 edge.destination.node_id, new_input=edge.source
             ):
-                return False
+                raise InvalidEdgeError(f'Collector output type does not match collector input type')
 
         # Validate if collector output type matches input type (if this edge results in both being set)
         if isinstance(from_node, CollectInvocation) and edge.source.field == "collection":
             if not self._is_collector_connection_valid(
                 edge.source.node_id, new_output=edge.destination
             ):
-                return False
+                raise InvalidEdgeError(f'Collector input type does not match collector output type')
 
-        return True
 
     def has_node(self, node_path: str) -> bool:
         """Determines whether or not a node exists in the graph."""
@@ -712,7 +731,7 @@ class Graph(BaseModel):
         for sgn in (
             gn for gn in self.nodes.values() if isinstance(gn, GraphInvocation)
         ):
-            sgn.graph.nx_graph_flat(g, self._get_node_path(sgn.id, prefix))
+            g = sgn.graph.nx_graph_flat(g, self._get_node_path(sgn.id, prefix))
 
         # TODO: figure out if iteration nodes need to be expanded
 
@@ -729,9 +748,7 @@ class Graph(BaseModel):
 class GraphExecutionState(BaseModel):
     """Tracks the state of a graph execution"""
 
-    id: str = Field(
-        description="The id of the execution state", default_factory=uuid.uuid4
-    )
+    id: str = Field(description="The id of the execution state", default_factory=lambda: uuid.uuid4().__str__())
 
     # TODO: Store a reference to the graph instead of the actual graph?
     graph: Graph = Field(description="The graph being executed")
@@ -773,9 +790,6 @@ class GraphExecutionState(BaseModel):
         default_factory=dict,
     )
 
-    # Declare all fields as required; necessary for OpenAPI schema generation build.
-    # Technically only fields without a `default_factory` need to be listed here.
-    # See: https://github.com/pydantic/pydantic/discussions/4577
     class Config:
         schema_extra = {
             'required': [
@@ -840,7 +854,8 @@ class GraphExecutionState(BaseModel):
 
     def is_complete(self) -> bool:
         """Returns true if the graph is complete"""
-        return self.has_error() or all((k in self.executed for k in self.graph.nodes))
+        node_ids = set(self.graph.nx_graph_flat().nodes)
+        return self.has_error() or all((k in self.executed for k in node_ids))
 
     def has_error(self) -> bool:
         """Returns true if the graph has any errors"""
@@ -928,11 +943,11 @@ class GraphExecutionState(BaseModel):
 
     def _iterator_graph(self) -> nx.DiGraph:
         """Gets a DiGraph with edges to collectors removed so an ancestor search produces all active iterators for any node"""
-        g = self.graph.nx_graph()
+        g = self.graph.nx_graph_flat()
         collectors = (
             n
             for n in self.graph.nodes
-            if isinstance(self.graph.nodes[n], CollectInvocation)
+            if isinstance(self.graph.get_node(n), CollectInvocation)
         )
         for c in collectors:
             g.remove_edges_from(list(g.in_edges(c)))
@@ -944,7 +959,7 @@ class GraphExecutionState(BaseModel):
         iterators = [
             n
             for n in nx.ancestors(g, node_id)
-            if isinstance(self.graph.nodes[n], IterateInvocation)
+            if isinstance(self.graph.get_node(n), IterateInvocation)
         ]
         return iterators
 
@@ -1048,9 +1063,8 @@ class GraphExecutionState(BaseModel):
                 n
                 for n in prepared_nodes
                 if all(
-                    pit
+                    nx.has_path(execution_graph, pit[0], n)
                     for pit in parent_iterators
-                    if nx.has_path(execution_graph, pit[0], n)
                 )
             ),
             None,
@@ -1081,7 +1095,9 @@ class GraphExecutionState(BaseModel):
 
     # TODO: Add API for modifying underlying graph that checks if the change will be valid given the current execution state
     def _is_edge_valid(self, edge: Edge) -> bool:
-        if not self._is_edge_valid(edge):
+        try:
+            self.graph._validate_edge(edge)
+        except InvalidEdgeError:
             return False
 
         # Invalid if destination has already been prepared or executed
@@ -1127,4 +1143,52 @@ class GraphExecutionState(BaseModel):
         self.graph.delete_edge(edge)
 
 
+class ExposedNodeInput(BaseModel):
+    node_path: str = Field(description="The node path to the node with the input")
+    field: str = Field(description="The field name of the input")
+    alias: str = Field(description="The alias of the input")
+
+
+class ExposedNodeOutput(BaseModel):
+    node_path: str = Field(description="The node path to the node with the output")
+    field: str = Field(description="The field name of the output")
+    alias: str = Field(description="The alias of the output")
+
+class LibraryGraph(BaseModel):
+    id: str = Field(description="The unique identifier for this library graph", default_factory=uuid.uuid4)
+    graph: Graph = Field(description="The graph")
+    name: str = Field(description="The name of the graph")
+    description: str = Field(description="The description of the graph")
+    exposed_inputs: list[ExposedNodeInput] = Field(description="The inputs exposed by this graph", default_factory=list)
+    exposed_outputs: list[ExposedNodeOutput] = Field(description="The outputs exposed by this graph", default_factory=list)
+
+    @validator('exposed_inputs', 'exposed_outputs')
+    def validate_exposed_aliases(cls, v):
+        if len(v) != len(set(i.alias for i in v)):
+            raise ValueError("Duplicate exposed alias")
+        return v
+
+    @root_validator
+    def validate_exposed_nodes(cls, values):
+        graph = values['graph']
+
+        # Validate exposed inputs
+        for exposed_input in values['exposed_inputs']:
+            if not graph.has_node(exposed_input.node_path):
+                raise ValueError(f"Exposed input node {exposed_input.node_path} does not exist")
+            node = graph.get_node(exposed_input.node_path)
+            if get_input_field(node, exposed_input.field) is None:
+                raise ValueError(f"Exposed input field {exposed_input.field} does not exist on node {exposed_input.node_path}")
+
+        # Validate exposed outputs
+        for exposed_output in values['exposed_outputs']:
+            if not graph.has_node(exposed_output.node_path):
+                raise ValueError(f"Exposed output node {exposed_output.node_path} does not exist")
+            node = graph.get_node(exposed_output.node_path)
+            if get_output_field(node, exposed_output.field) is None:
+                raise ValueError(f"Exposed output field {exposed_output.field} does not exist on node {exposed_output.node_path}")
+
+        return values
+
+
 GraphInvocation.update_forward_refs()

invokeai/app/services/image_storage.py

@@ -1,22 +1,24 @@
 # Copyright (c) 2022 Kyle Schouviller (https://github.com/kyle0654)
 
-import datetime
 import os
+from glob import glob
 from abc import ABC, abstractmethod
-from enum import Enum
 from pathlib import Path
 from queue import Queue
-from typing import Dict
+from typing import Dict, List, Tuple
 
 from PIL.Image import Image
-
-from invokeai.backend.image_util import PngWriter
-
-
-class ImageType(str, Enum):
-    RESULT = "results"
-    INTERMEDIATE = "intermediates"
-    UPLOAD = "uploads"
+import PIL.Image as PILImage
+from invokeai.app.api.models.images import ImageResponse, ImageResponseMetadata
+from invokeai.app.models.image import ImageType
+from invokeai.app.services.metadata import (
+    InvokeAIMetadata,
+    MetadataServiceBase,
+    build_invokeai_metadata_pnginfo,
+)
+from invokeai.app.services.item_storage import PaginatedResults
+from invokeai.app.util.misc import get_timestamp
+from invokeai.app.util.thumbnails import get_thumbnail_name, make_thumbnail
 
 
 class ImageStorageBase(ABC):
@@ -24,40 +26,66 @@ class ImageStorageBase(ABC):
 
     @abstractmethod
     def get(self, image_type: ImageType, image_name: str) -> Image:
+        """Retrieves an image as PIL Image."""
+        pass
+
+    @abstractmethod
+    def list(
+        self, image_type: ImageType, page: int = 0, per_page: int = 10
+    ) -> PaginatedResults[ImageResponse]:
+        """Gets a paginated list of images."""
         pass
 
     # TODO: make this a bit more flexible for e.g. cloud storage
     @abstractmethod
-    def get_path(self, image_type: ImageType, image_name: str) -> str:
+    def get_path(
+        self, image_type: ImageType, image_name: str, is_thumbnail: bool = False
+    ) -> str:
+        """Gets the path to an image or its thumbnail."""
+        pass
+
+    # TODO: make this a bit more flexible for e.g. cloud storage
+    @abstractmethod
+    def validate_path(self, path: str) -> bool:
+        """Validates an image path."""
         pass
 
     @abstractmethod
-    def save(self, image_type: ImageType, image_name: str, image: Image) -> None:
+    def save(
+        self,
+        image_type: ImageType,
+        image_name: str,
+        image: Image,
+        metadata: InvokeAIMetadata | None = None,
+    ) -> Tuple[str, str, int]:
+        """Saves an image and a 256x256 WEBP thumbnail. Returns a tuple of the image path, thumbnail path, and created timestamp."""
         pass
 
     @abstractmethod
     def delete(self, image_type: ImageType, image_name: str) -> None:
+        """Deletes an image and its thumbnail (if one exists)."""
         pass
 
     def create_name(self, context_id: str, node_id: str) -> str:
-        return f"{context_id}_{node_id}_{str(int(datetime.datetime.now(datetime.timezone.utc).timestamp()))}.png"
+        """Creates a unique contextual image filename."""
+        return f"{context_id}_{node_id}_{str(get_timestamp())}.png"
 
 
 class DiskImageStorage(ImageStorageBase):
     """Stores images on disk"""
 
     __output_folder: str
-    __pngWriter: PngWriter
     __cache_ids: Queue  # TODO: this is an incredibly naive cache
     __cache: Dict[str, Image]
     __max_cache_size: int
+    __metadata_service: MetadataServiceBase
 
-    def __init__(self, output_folder: str):
+    def __init__(self, output_folder: str, metadata_service: MetadataServiceBase):
         self.__output_folder = output_folder
-        self.__pngWriter = PngWriter(output_folder)
         self.__cache = dict()
         self.__cache_ids = Queue()
         self.__max_cache_size = 10  # TODO: get this from config
+        self.__metadata_service = metadata_service
 
         Path(output_folder).mkdir(parents=True, exist_ok=True)
 
@@ -66,6 +94,61 @@ class DiskImageStorage(ImageStorageBase):
             Path(os.path.join(output_folder, image_type)).mkdir(
                 parents=True, exist_ok=True
             )
+            Path(os.path.join(output_folder, image_type, "thumbnails")).mkdir(
+                parents=True, exist_ok=True
+            )
+
+    def list(
+        self, image_type: ImageType, page: int = 0, per_page: int = 10
+    ) -> PaginatedResults[ImageResponse]:
+        dir_path = os.path.join(self.__output_folder, image_type)
+        image_paths = glob(f"{dir_path}/*.png")
+        count = len(image_paths)
+
+        sorted_image_paths = sorted(
+            glob(f"{dir_path}/*.png"), key=os.path.getctime, reverse=True
+        )
+
+        page_of_image_paths = sorted_image_paths[
+            page * per_page : (page + 1) * per_page
+        ]
+
+        page_of_images: List[ImageResponse] = []
+
+        for path in page_of_image_paths:
+            filename = os.path.basename(path)
+            img = PILImage.open(path)
+
+            invokeai_metadata = self.__metadata_service.get_metadata(img)
+
+            page_of_images.append(
+                ImageResponse(
+                    image_type=image_type.value,
+                    image_name=filename,
+                    # TODO: DiskImageStorage should not be building URLs...?
+                    image_url=f"api/v1/images/{image_type.value}/{filename}",
+                    thumbnail_url=f"api/v1/images/{image_type.value}/thumbnails/{os.path.splitext(filename)[0]}.webp",
+                    # TODO: Creation of this object should happen elsewhere (?), just making it fit here so it works
+                    metadata=ImageResponseMetadata(
+                        created=int(os.path.getctime(path)),
+                        width=img.width,
+                        height=img.height,
+                        invokeai=invokeai_metadata,
+                    ),
+                )
+            )
+
+        page_count_trunc = int(count / per_page)
+        page_count_mod = count % per_page
+        page_count = page_count_trunc if page_count_mod == 0 else page_count_trunc + 1
+
+        return PaginatedResults[ImageResponse](
+            items=page_of_images,
+            page=page,
+            pages=page_count,
+            per_page=per_page,
+            total=count,
+        )
 
     def get(self, image_type: ImageType, image_name: str) -> Image:
         image_path = self.get_path(image_type, image_name)
@@ -73,32 +156,74 @@ class DiskImageStorage(ImageStorageBase):
         if cache_item:
             return cache_item
 
-        image = Image.open(image_path)
+        image = PILImage.open(image_path)
         self.__set_cache(image_path, image)
         return image
 
     # TODO: make this a bit more flexible for e.g. cloud storage
-    def get_path(self, image_type: ImageType, image_name: str) -> str:
-        path = os.path.join(self.__output_folder, image_type, image_name)
+    def get_path(
+        self, image_type: ImageType, image_name: str, is_thumbnail: bool = False
+    ) -> str:
+        # strip out any relative path shenanigans
+        basename = os.path.basename(image_name)
+
+        if is_thumbnail:
+            path = os.path.join(
+                self.__output_folder, image_type, "thumbnails", basename
+            )
+        else:
+            path = os.path.join(self.__output_folder, image_type, basename)
+
         return path
 
-    def save(self, image_type: ImageType, image_name: str, image: Image) -> None:
-        image_subpath = os.path.join(image_type, image_name)
-        self.__pngWriter.save_image_and_prompt_to_png(
-            image, "", image_subpath, None
-        )  # TODO: just pass full path to png writer
+    def validate_path(self, path: str) -> bool:
+        try:
+            os.stat(path)
+            return True
+        except Exception:
+            return False
 
+    def save(
+        self,
+        image_type: ImageType,
+        image_name: str,
+        image: Image,
+        metadata: InvokeAIMetadata | None = None,
+    ) -> Tuple[str, str, int]:
         image_path = self.get_path(image_type, image_name)
+
+        # TODO: Reading the image and then saving it strips the metadata...
+        if metadata:
+            pnginfo = build_invokeai_metadata_pnginfo(metadata=metadata)
+            image.save(image_path, "PNG", pnginfo=pnginfo)
+        else:
+            image.save(image_path) # this saved image has an empty info
+
+        thumbnail_name = get_thumbnail_name(image_name)
+        thumbnail_path = self.get_path(image_type, thumbnail_name, is_thumbnail=True)
+        thumbnail_image = make_thumbnail(image)
+        thumbnail_image.save(thumbnail_path)
+
         self.__set_cache(image_path, image)
+        self.__set_cache(thumbnail_path, thumbnail_image)
+
+        return (image_path, thumbnail_path, int(os.path.getctime(image_path)))
 
     def delete(self, image_type: ImageType, image_name: str) -> None:
         image_path = self.get_path(image_type, image_name)
+        thumbnail_path = self.get_path(image_type, image_name, True)
         if os.path.exists(image_path):
             os.remove(image_path)
 
         if image_path in self.__cache:
             del self.__cache[image_path]
 
+        if os.path.exists(thumbnail_path):
+            os.remove(thumbnail_path)
+
+        if thumbnail_path in self.__cache:
+            del self.__cache[thumbnail_path]
+
     def __get_cache(self, image_name: str) -> Image:
         return None if image_name not in self.__cache else self.__cache[image_name]
 

invokeai/app/services/invocation_queue.py

@@ -1,27 +1,17 @@
 # Copyright (c) 2022 Kyle Schouviller (https://github.com/kyle0654)
 
+import time
 from abc import ABC, abstractmethod
 from queue import Queue
 
+from pydantic import BaseModel, Field
 
-# TODO: make this serializable
-class InvocationQueueItem:
-    # session_id: str
-    graph_execution_state_id: str
-    invocation_id: str
-    invoke_all: bool
 
-    def __init__(
-        self,
-        # session_id: str,
-        graph_execution_state_id: str,
-        invocation_id: str,
-        invoke_all: bool = False,
-    ):
-        # self.session_id = session_id
-        self.graph_execution_state_id = graph_execution_state_id
-        self.invocation_id = invocation_id
-        self.invoke_all = invoke_all
+class InvocationQueueItem(BaseModel):
+    graph_execution_state_id: str = Field(description="The ID of the graph execution state")
+    invocation_id: str = Field(description="The ID of the node being invoked")
+    invoke_all: bool = Field(default=False)
+    timestamp: float = Field(default_factory=time.time)
 
 
 class InvocationQueueABC(ABC):
@@ -35,15 +25,44 @@ class InvocationQueueABC(ABC):
     def put(self, item: InvocationQueueItem | None) -> None:
         pass
 
+    @abstractmethod
+    def cancel(self, graph_execution_state_id: str) -> None:
+        pass
+
+    @abstractmethod
+    def is_canceled(self, graph_execution_state_id: str) -> bool:
+        pass
+
 
 class MemoryInvocationQueue(InvocationQueueABC):
     __queue: Queue
+    __cancellations: dict[str, float]
 
     def __init__(self):
         self.__queue = Queue()
+        self.__cancellations = dict()
 
     def get(self) -> InvocationQueueItem:
-        return self.__queue.get()
+        item = self.__queue.get()
+
+        while isinstance(item, InvocationQueueItem) \
+            and item.graph_execution_state_id in self.__cancellations \
+            and self.__cancellations[item.graph_execution_state_id] > item.timestamp:
+            item = self.__queue.get()
+
+        # Clear old items
+        for graph_execution_state_id in list(self.__cancellations.keys()):
+            if self.__cancellations[graph_execution_state_id] < item.timestamp:
+                del self.__cancellations[graph_execution_state_id]
+
+        return item
 
     def put(self, item: InvocationQueueItem | None) -> None:
         self.__queue.put(item)
+
+    def cancel(self, graph_execution_state_id: str) -> None:
+        if graph_execution_state_id not in self.__cancellations:
+            self.__cancellations[graph_execution_state_id] = time.time()
+
+    def is_canceled(self, graph_execution_state_id: str) -> bool:
+        return graph_execution_state_id in self.__cancellations

invokeai/app/services/invocation_services.py

@@ -1,7 +1,9 @@
 # Copyright (c) 2022 Kyle Schouviller (https://github.com/kyle0654)
+from invokeai.app.services.metadata import MetadataServiceBase
 from invokeai.backend import ModelManager
 
 from .events import EventServiceBase
+from .latent_storage import LatentsStorageBase
 from .image_storage import ImageStorageBase
 from .restoration_services import RestorationServices
 from .invocation_queue import InvocationQueueABC
@@ -11,12 +13,15 @@ class InvocationServices:
     """Services that can be used by invocations"""
 
     events: EventServiceBase
+    latents: LatentsStorageBase
     images: ImageStorageBase
+    metadata: MetadataServiceBase
     queue: InvocationQueueABC
     model_manager: ModelManager
     restoration: RestorationServices
 
     # NOTE: we must forward-declare any types that include invocations, since invocations can use services
+    graph_library: ItemStorageABC["LibraryGraph"]
     graph_execution_manager: ItemStorageABC["GraphExecutionState"]
     processor: "InvocationProcessorABC"
 
@@ -24,16 +29,22 @@ class InvocationServices:
             self,
             model_manager: ModelManager,
             events: EventServiceBase,
+            latents: LatentsStorageBase,
             images: ImageStorageBase,
+            metadata: MetadataServiceBase,
             queue: InvocationQueueABC,
+            graph_library: ItemStorageABC["LibraryGraph"],
             graph_execution_manager: ItemStorageABC["GraphExecutionState"],
             processor: "InvocationProcessorABC",
             restoration: RestorationServices,
     ):
         self.model_manager = model_manager
         self.events = events
+        self.latents = latents
         self.images = images
+        self.metadata = metadata
         self.queue = queue
+        self.graph_library = graph_library
         self.graph_execution_manager = graph_execution_manager
         self.processor = processor
         self.restoration = restoration

invokeai/app/services/invoker.py

@@ -33,7 +33,6 @@ class Invoker:
         self.services.graph_execution_manager.set(graph_execution_state)
 
         # Queue the invocation
-        print(f"queueing item {invocation.id}")
         self.services.queue.put(
             InvocationQueueItem(
                 # session_id    = session.id,
@@ -51,6 +50,10 @@ class Invoker:
         self.services.graph_execution_manager.set(new_state)
         return new_state
     
+    def cancel(self, graph_execution_state_id: str) -> None:
+        """Cancels the given execution state"""
+        self.services.queue.cancel(graph_execution_state_id)
+
     def __start_service(self, service) -> None:
         # Call start() method on any services that have it
         start_op = getattr(service, "start", None)

invokeai/app/services/latent_storage.py

@@ -0,0 +1,93 @@
+# Copyright (c) 2023 Kyle Schouviller (https://github.com/kyle0654)
+
+import os
+from abc import ABC, abstractmethod
+from pathlib import Path
+from queue import Queue
+from typing import Dict
+
+import torch
+
+class LatentsStorageBase(ABC):
+    """Responsible for storing and retrieving latents."""
+
+    @abstractmethod
+    def get(self, name: str) -> torch.Tensor:
+        pass
+
+    @abstractmethod
+    def set(self, name: str, data: torch.Tensor) -> None:
+        pass
+
+    @abstractmethod
+    def delete(self, name: str) -> None:
+        pass
+
+
+class ForwardCacheLatentsStorage(LatentsStorageBase):
+    """Caches the latest N latents in memory, writing-thorugh to and reading from underlying storage"""
+    
+    __cache: Dict[str, torch.Tensor]
+    __cache_ids: Queue
+    __max_cache_size: int
+    __underlying_storage: LatentsStorageBase
+
+    def __init__(self, underlying_storage: LatentsStorageBase, max_cache_size: int = 20):
+        self.__underlying_storage = underlying_storage
+        self.__cache = dict()
+        self.__cache_ids = Queue()
+        self.__max_cache_size = max_cache_size
+
+    def get(self, name: str) -> torch.Tensor:
+        cache_item = self.__get_cache(name)
+        if cache_item is not None:
+            return cache_item
+
+        latent = self.__underlying_storage.get(name)
+        self.__set_cache(name, latent)
+        return latent
+
+    def set(self, name: str, data: torch.Tensor) -> None:
+        self.__underlying_storage.set(name, data)
+        self.__set_cache(name, data)
+
+    def delete(self, name: str) -> None:
+        self.__underlying_storage.delete(name)
+        if name in self.__cache:
+            del self.__cache[name]
+
+    def __get_cache(self, name: str) -> torch.Tensor|None:
+        return None if name not in self.__cache else self.__cache[name]
+
+    def __set_cache(self, name: str, data: torch.Tensor):
+        if not name in self.__cache:
+            self.__cache[name] = data
+            self.__cache_ids.put(name)
+            if self.__cache_ids.qsize() > self.__max_cache_size:
+                self.__cache.pop(self.__cache_ids.get())
+
+
+class DiskLatentsStorage(LatentsStorageBase):
+    """Stores latents in a folder on disk without caching"""
+
+    __output_folder: str
+
+    def __init__(self, output_folder: str):
+        self.__output_folder = output_folder
+        Path(output_folder).mkdir(parents=True, exist_ok=True)
+
+    def get(self, name: str) -> torch.Tensor:
+        latent_path = self.get_path(name)
+        return torch.load(latent_path)
+
+    def set(self, name: str, data: torch.Tensor) -> None:
+        latent_path = self.get_path(name)
+        torch.save(data, latent_path)
+
+    def delete(self, name: str) -> None:
+        latent_path = self.get_path(name)
+        os.remove(latent_path)
+
+    def get_path(self, name: str) -> str:
+        return os.path.join(self.__output_folder, name)
+    

invokeai/app/services/metadata.py

@@ -0,0 +1,96 @@
+import json
+from abc import ABC, abstractmethod
+from typing import Any, Dict, Optional, TypedDict
+from PIL import Image, PngImagePlugin
+from pydantic import BaseModel
+
+from invokeai.app.models.image import ImageType, is_image_type
+
+
+class MetadataImageField(TypedDict):
+    """Pydantic-less ImageField, used for metadata parsing."""
+
+    image_type: ImageType
+    image_name: str
+
+
+class MetadataLatentsField(TypedDict):
+    """Pydantic-less LatentsField, used for metadata parsing."""
+
+    latents_name: str
+
+
+# TODO: This is a placeholder for `InvocationsUnion` pending resolution of circular imports
+NodeMetadata = Dict[
+    str, str | int | float | bool | MetadataImageField | MetadataLatentsField
+]
+
+
+class InvokeAIMetadata(TypedDict, total=False):
+    """InvokeAI-specific metadata format."""
+
+    session_id: Optional[str]
+    node: Optional[NodeMetadata]
+
+
+def build_invokeai_metadata_pnginfo(
+    metadata: InvokeAIMetadata | None,
+) -> PngImagePlugin.PngInfo:
+    """Builds a PngInfo object with key `"invokeai"` and value `metadata`"""
+    pnginfo = PngImagePlugin.PngInfo()
+
+    if metadata is not None:
+        pnginfo.add_text("invokeai", json.dumps(metadata))
+
+    return pnginfo
+
+
+class MetadataServiceBase(ABC):
+    @abstractmethod
+    def get_metadata(self, image: Image.Image) -> InvokeAIMetadata | None:
+        """Gets the InvokeAI metadata from a PIL Image, skipping invalid values"""
+        pass
+
+    @abstractmethod
+    def build_metadata(
+        self, session_id: str, node: BaseModel
+    ) -> InvokeAIMetadata | None:
+        """Builds an InvokeAIMetadata object"""
+        pass
+
+
+class PngMetadataService(MetadataServiceBase):
+    """Handles loading and building metadata for images."""
+
+    # TODO: Use `InvocationsUnion` to **validate** metadata as representing a fully-functioning node
+    def _load_metadata(self, image: Image.Image) -> dict | None:
+        """Loads a specific info entry from a PIL Image."""
+
+        try:
+            info = image.info.get("invokeai")
+
+            if type(info) is not str:
+                return None
+
+            loaded_metadata = json.loads(info)
+
+            if type(loaded_metadata) is not dict:
+                return None
+
+            if len(loaded_metadata.items()) == 0:
+                return None
+
+            return loaded_metadata
+        except:
+            return None
+
+    def get_metadata(self, image: Image.Image) -> dict | None:
+        """Retrieves an image's metadata as a dict"""
+        loaded_metadata = self._load_metadata(image)
+
+        return loaded_metadata
+
+    def build_metadata(self, session_id: str, node: BaseModel) -> InvokeAIMetadata:
+        metadata = InvokeAIMetadata(session_id=session_id, node=node.dict())
+
+        return metadata

invokeai/app/services/processor.py

@@ -4,7 +4,7 @@ from threading import Event, Thread
 from ..invocations.baseinvocation import InvocationContext
 from .invocation_queue import InvocationQueueItem
 from .invoker import InvocationProcessorABC, Invoker
-
+from ..models.exceptions import CanceledException
 
 class DefaultInvocationProcessor(InvocationProcessorABC):
     __invoker_thread: Thread
@@ -43,10 +43,14 @@ class DefaultInvocationProcessor(InvocationProcessorABC):
                     queue_item.invocation_id
                 )
 
+                # get the source node id to provide to clients (the prepared node id is not as useful)
+                source_node_id = graph_execution_state.prepared_source_mapping[invocation.id]
+
                 # Send starting event
                 self.__invoker.services.events.emit_invocation_started(
                     graph_execution_state_id=graph_execution_state.id,
-                    invocation_id=invocation.id,
+                    node=invocation.dict(),
+                    source_node_id=source_node_id
                 )
 
                 # Invoke
@@ -58,6 +62,12 @@ class DefaultInvocationProcessor(InvocationProcessorABC):
                         )
                     )
 
+                    # Check queue to see if this is canceled, and skip if so
+                    if self.__invoker.services.queue.is_canceled(
+                        graph_execution_state.id
+                    ):
+                        continue
+
                     # Save outputs and history
                     graph_execution_state.complete(invocation.id, outputs)
 
@@ -69,13 +79,17 @@ class DefaultInvocationProcessor(InvocationProcessorABC):
                     # Send complete event
                     self.__invoker.services.events.emit_invocation_complete(
                         graph_execution_state_id=graph_execution_state.id,
-                        invocation_id=invocation.id,
+                        node=invocation.dict(),
+                        source_node_id=source_node_id,
                         result=outputs.dict(),
                     )
 
                 except KeyboardInterrupt:
                     pass
 
+                except CanceledException:
+                    pass
+
                 except Exception as e:
                     error = traceback.format_exc()
 
@@ -90,12 +104,19 @@ class DefaultInvocationProcessor(InvocationProcessorABC):
                     # Send error event
                     self.__invoker.services.events.emit_invocation_error(
                         graph_execution_state_id=graph_execution_state.id,
-                        invocation_id=invocation.id,
+                        node=invocation.dict(),
+                        source_node_id=source_node_id,
                         error=error,
                     )
 
                     pass
                 
+                # Check queue to see if this is canceled, and skip if so
+                if self.__invoker.services.queue.is_canceled(
+                    graph_execution_state.id
+                ):
+                    continue
+
                 # Queue any further commands if invoking all
                 is_complete = graph_execution_state.is_complete()
                 if queue_item.invoke_all and not is_complete:

invokeai/app/services/sqlite.py

@@ -59,6 +59,7 @@ class SqliteItemStorage(ItemStorageABC, Generic[T]):
                 f"""INSERT OR REPLACE INTO {self._table_name} (item) VALUES (?);""",
                 (item.json(),),
             )
+            self._conn.commit()
         finally:
             self._lock.release()
         self._on_changed(item)
@@ -84,6 +85,7 @@ class SqliteItemStorage(ItemStorageABC, Generic[T]):
             self._cursor.execute(
                 f"""DELETE FROM {self._table_name} WHERE id = ?;""", (str(id),)
             )
+            self._conn.commit()
         finally:
             self._lock.release()
         self._on_deleted(id)

invokeai/app/util/misc.py

@@ -0,0 +1,5 @@
+import datetime
+
+
+def get_timestamp():
+    return int(datetime.datetime.now(datetime.timezone.utc).timestamp())

invokeai/app/util/step_callback.py

@@ -0,0 +1,55 @@
+from invokeai.app.api.models.images import ProgressImage
+from invokeai.app.models.exceptions import CanceledException
+from ..invocations.baseinvocation import InvocationContext
+from ...backend.util.util import image_to_dataURL
+from ...backend.generator.base import Generator
+from ...backend.stable_diffusion import PipelineIntermediateState
+
+
+def stable_diffusion_step_callback(
+    context: InvocationContext,
+    intermediate_state: PipelineIntermediateState,
+    node: dict,
+    source_node_id: str,
+):
+    if context.services.queue.is_canceled(context.graph_execution_state_id):
+        raise CanceledException
+
+    # Some schedulers report not only the noisy latents at the current timestep,
+    # but also their estimate so far of what the de-noised latents will be. Use
+    # that estimate if it is available.
+    if intermediate_state.predicted_original is not None:
+        sample = intermediate_state.predicted_original
+    else:
+        sample = intermediate_state.latents
+
+    # TODO: This does not seem to be needed any more?
+    # # txt2img provides a Tensor in the step_callback
+    # # img2img provides a PipelineIntermediateState
+    # if isinstance(sample, PipelineIntermediateState):
+    #     # this was an img2img
+    #     print('img2img')
+    #     latents = sample.latents
+    #     step = sample.step
+    # else:
+    #     print('txt2img')
+    #     latents = sample
+    #     step = intermediate_state.step
+
+    # TODO: only output a preview image when requested
+    image = Generator.sample_to_lowres_estimated_image(sample)
+
+    (width, height) = image.size
+    width *= 8
+    height *= 8
+
+    dataURL = image_to_dataURL(image, image_format="JPEG")
+
+    context.services.events.emit_generator_progress(
+        graph_execution_state_id=context.graph_execution_state_id,
+        node=node,
+        source_node_id=source_node_id,
+        progress_image=ProgressImage(width=width, height=height, dataURL=dataURL),
+        step=intermediate_state.step,
+        total_steps=node["steps"],
+    )

invokeai/app/util/thumbnails.py

@@ -0,0 +1,15 @@
+import os
+from PIL import Image
+
+
+def get_thumbnail_name(image_name: str) -> str:
+    """Formats given an image name, returns the appropriate thumbnail image name"""
+    thumbnail_name = os.path.splitext(image_name)[0] + ".webp"
+    return thumbnail_name
+
+
+def make_thumbnail(image: Image.Image, size: int = 256) -> Image.Image:
+    """Makes a thumbnail from a PIL Image"""
+    thumbnail = image.copy()
+    thumbnail.thumbnail(size=(size, size))
+    return thumbnail

invokeai/backend/args.py

@@ -561,7 +561,7 @@ class Args(object):
             "--autoimport",
             default=None,
             type=str,
-            help="Check the indicated directory for .ckpt/.safetensors weights files at startup and import directly",
+            help="(DEPRECATED - NONFUNCTIONAL). Check the indicated directory for .ckpt/.safetensors weights files at startup and import directly",
         )
         model_group.add_argument(
             "--autoconvert",

invokeai/backend/config/model_install_backend.py

@@ -67,7 +67,6 @@ def install_requested_models(
     scan_directory: Path = None,
     external_models: List[str] = None,
     scan_at_startup: bool = False,
-    convert_to_diffusers: bool = False,
     precision: str = "float16",
     purge_deleted: bool = False,
     config_file_path: Path = None,
@@ -113,7 +112,6 @@ def install_requested_models(
             try:
                 model_manager.heuristic_import(
                     path_url_or_repo,
-                    convert=convert_to_diffusers,
                     commit_to_conf=config_file_path,
                 )
             except KeyboardInterrupt:
@@ -122,7 +120,7 @@ def install_requested_models(
                 pass
 
     if scan_at_startup and scan_directory.is_dir():
-        argument = "--autoconvert" if convert_to_diffusers else "--autoimport"
+        argument = "--autoconvert"
         initfile = Path(Globals.root, Globals.initfile)
         replacement = Path(Globals.root, f"{Globals.initfile}.new")
         directory = str(scan_directory).replace("\\", "/")

invokeai/backend/generator/base.py

@@ -21,7 +21,7 @@ from PIL import Image, ImageChops, ImageFilter
 from accelerate.utils import set_seed
 from diffusers import DiffusionPipeline
 from tqdm import trange
-from typing import List, Iterator, Type
+from typing import Callable, List, Iterator, Optional, Type
 from dataclasses import dataclass, field
 from diffusers.schedulers import SchedulerMixin as Scheduler
 
@@ -35,23 +35,23 @@ downsampling = 8
 
 @dataclass
 class InvokeAIGeneratorBasicParams:
-    seed: int=None
+    seed: Optional[int]=None
     width: int=512
     height: int=512
-    cfg_scale: int=7.5
+    cfg_scale: float=7.5
     steps: int=20
     ddim_eta: float=0.0
-    scheduler: int='ddim'
+    scheduler: str='ddim'
     precision: str='float16'
     perlin: float=0.0
-    threshold: int=0.0
+    threshold: float=0.0
     seamless: bool=False
     seamless_axes: List[str]=field(default_factory=lambda: ['x', 'y'])
-    h_symmetry_time_pct: float=None
-    v_symmetry_time_pct: float=None
+    h_symmetry_time_pct: Optional[float]=None
+    v_symmetry_time_pct: Optional[float]=None
     variation_amount: float = 0.0
     with_variations: list=field(default_factory=list)
-    safety_checker: SafetyChecker=None
+    safety_checker: Optional[SafetyChecker]=None
 
 @dataclass
 class InvokeAIGeneratorOutput:
@@ -61,10 +61,10 @@ class InvokeAIGeneratorOutput:
     and the model hash, as well as all the generate() parameters that went into
     generating the image (in .params, also available as attributes)
     '''
-    image: Image
+    image: Image.Image
     seed: int
     model_hash: str
-    attention_maps_images: List[Image]
+    attention_maps_images: List[Image.Image]
     params: Namespace
 
 # we are interposing a wrapper around the original Generator classes so that
@@ -92,8 +92,8 @@ class InvokeAIGenerator(metaclass=ABCMeta):
 
     def generate(self,
                  prompt: str='',
-                 callback: callable=None,
-                 step_callback: callable=None,
+                 callback: Optional[Callable]=None,
+                 step_callback: Optional[Callable]=None,
                  iterations: int=1,
                  **keyword_args,
                  )->Iterator[InvokeAIGeneratorOutput]:
@@ -206,10 +206,10 @@ class Txt2Img(InvokeAIGenerator):
 # ------------------------------------
 class Img2Img(InvokeAIGenerator):
     def generate(self,
-               init_image: Image | torch.FloatTensor,
+               init_image: Image.Image | torch.FloatTensor,
                strength: float=0.75,
                **keyword_args
-               )->List[InvokeAIGeneratorOutput]:
+               )->Iterator[InvokeAIGeneratorOutput]:
         return super().generate(init_image=init_image,
                                 strength=strength,
                                 **keyword_args
@@ -223,7 +223,7 @@ class Img2Img(InvokeAIGenerator):
 # Takes all the arguments of Img2Img and adds the mask image and the seam/infill stuff
 class Inpaint(Img2Img):
     def generate(self,
-                 mask_image: Image | torch.FloatTensor,
+                 mask_image: Image.Image | torch.FloatTensor,
                  # Seam settings - when 0, doesn't fill seam
                  seam_size: int = 0,
                  seam_blur: int = 0,
@@ -236,7 +236,7 @@ class Inpaint(Img2Img):
                  inpaint_height=None,
                  inpaint_fill: tuple(int) = (0x7F, 0x7F, 0x7F, 0xFF),
                  **keyword_args
-                 )->List[InvokeAIGeneratorOutput]:
+                 )->Iterator[InvokeAIGeneratorOutput]:
         return super().generate(
             mask_image=mask_image,
             seam_size=seam_size,
@@ -263,7 +263,7 @@ class Embiggen(Txt2Img):
             embiggen: list=None,
             embiggen_tiles: list = None,
             strength: float=0.75,
-            **kwargs)->List[InvokeAIGeneratorOutput]:
+            **kwargs)->Iterator[InvokeAIGeneratorOutput]:
         return super().generate(embiggen=embiggen,
                                 embiggen_tiles=embiggen_tiles,
                                 strength=strength,

invokeai/backend/model_management/__init__.py

@@ -7,3 +7,4 @@ from .convert_ckpt_to_diffusers import (
 )
 from .model_manager import ModelManager
 
+

invokeai/backend/model_management/convert_ckpt_to_diffusers.py

@@ -1264,10 +1264,10 @@ def load_pipeline_from_original_stable_diffusion_ckpt(
                 cache_dir=cache_dir,
             )
             pipe = pipeline_class(
-                vae=vae,
-                text_encoder=text_model,
+                vae=vae.to(precision),
+                text_encoder=text_model.to(precision),
                 tokenizer=tokenizer,
-                unet=unet,
+                unet=unet.to(precision),
                 scheduler=scheduler,
                 safety_checker=None,
                 feature_extractor=None,

invokeai/backend/model_management/model_manager.py

@@ -1,4 +1,4 @@
-"""
+"""enum
 Manage a cache of Stable Diffusion model files for fast switching.
 They are moved between GPU and CPU as necessary. If CPU memory falls
 below a preset minimum, the least recently used model will be
@@ -15,17 +15,21 @@ import sys
 import textwrap
 import time
 import warnings
-from enum import Enum
+from enum import Enum, auto
 from pathlib import Path
 from shutil import move, rmtree
-from typing import Any, Optional, Union
+from typing import Any, Optional, Union, Callable
 
 import safetensors
 import safetensors.torch
 import torch
 import transformers
-from diffusers import AutoencoderKL
-from diffusers import logging as dlogging
+from diffusers import (
+    AutoencoderKL,
+    UNet2DConditionModel,
+    SchedulerMixin,
+    logging as dlogging,
+)    
 from huggingface_hub import scan_cache_dir
 from omegaconf import OmegaConf
 from omegaconf.dictconfig import DictConfig
@@ -33,23 +37,44 @@ from picklescan.scanner import scan_file_path
 
 from invokeai.backend.globals import Globals, global_cache_dir
 
-from ..stable_diffusion import StableDiffusionGeneratorPipeline
+from transformers import (
+    CLIPTextModel,
+    CLIPTokenizer,
+    CLIPFeatureExtractor,
+)
+from diffusers.pipelines.stable_diffusion.safety_checker import (
+    StableDiffusionSafetyChecker,
+    )
+from ..stable_diffusion import (
+    StableDiffusionGeneratorPipeline,
+)
 from ..util import CUDA_DEVICE, ask_user, download_with_resume
 
+
 class SDLegacyType(Enum):
-    V1 = 1
-    V1_INPAINT = 2
-    V2   = 3
-    V2_e = 4
-    V2_v = 5
-    UNKNOWN = 99
+    V1 = auto()
+    V1_INPAINT = auto()
+    V2 = auto()
+    V2_e = auto()
+    V2_v = auto()
+    UNKNOWN = auto()
+
+class SDModelComponent(Enum):
+    vae="vae"
+    text_encoder="text_encoder"
+    tokenizer="tokenizer"
+    unet="unet"
+    scheduler="scheduler"
+    safety_checker="safety_checker"
+    feature_extractor="feature_extractor"
     
 DEFAULT_MAX_MODELS = 2
 
 class ModelManager(object):
-    '''
+    """
     Model manager handles loading, caching, importing, deleting, converting, and editing models.
-    '''
+    """
+
     def __init__(
         self,
         config: OmegaConf | Path,
@@ -88,13 +113,23 @@ class ModelManager(object):
         return model_name in self.config
 
     def get_model(self, model_name: str = None) -> dict:
-        """
-        Given a model named identified in models.yaml, return
-        the model object. If in RAM will load into GPU VRAM.
-        If on disk, will load from there.
+        """Given a model named identified in models.yaml, return a dict
+        containing the model object and some of its key features. If
+        in RAM will load into GPU VRAM.  If on disk, will load from
+        there.
+        The dict has the following keys:
+        'model': The StableDiffusionGeneratorPipeline object
+        'model_name': The name of the model in models.yaml
+        'width': The width of images trained by this model
+        'height': The height of images trained by this model
+        'hash': A unique hash of this model's files on disk.
         """
         if not model_name:
-            return self.get_model(self.current_model) if self.current_model else self.get_model(self.default_model())
+            return (
+                self.get_model(self.current_model)
+                if self.current_model
+                else self.get_model(self.default_model())
+            )
 
         if not self.valid_model(model_name):
             print(
@@ -135,6 +170,81 @@ class ModelManager(object):
             "hash": hash,
         }
 
+    def get_model_vae(self, model_name: str=None)->AutoencoderKL:
+        """Given a model name identified in models.yaml, load the model into
+        GPU if necessary and return its assigned VAE as an
+        AutoencoderKL object. If no model name is provided, return the
+        vae from the model currently in the GPU.
+        """
+        return self._get_sub_model(model_name, SDModelComponent.vae)
+                                   
+    def get_model_tokenizer(self, model_name: str=None)->CLIPTokenizer:
+        """Given a model name identified in models.yaml, load the model into
+        GPU if necessary and return its assigned CLIPTokenizer. If no
+        model name is provided, return the tokenizer from the model
+        currently in the GPU.
+        """
+        return self._get_sub_model(model_name, SDModelComponent.tokenizer)
+                                   
+    def get_model_unet(self, model_name: str=None)->UNet2DConditionModel:
+        """Given a model name identified in models.yaml, load the model into
+        GPU if necessary and return its assigned UNet2DConditionModel. If no model
+        name is provided, return the UNet from the model
+        currently in the GPU.  
+        """
+        return self._get_sub_model(model_name, SDModelComponent.unet)
+
+    def get_model_text_encoder(self, model_name: str=None)->CLIPTextModel:
+        """Given a model name identified in models.yaml, load the model into
+        GPU if necessary and return its assigned CLIPTextModel. If no
+        model name is provided, return the text encoder from the model
+        currently in the GPU.
+        """
+        return self._get_sub_model(model_name, SDModelComponent.text_encoder)
+
+    def get_model_feature_extractor(self, model_name: str=None)->CLIPFeatureExtractor:
+        """Given a model name identified in models.yaml, load the model into
+        GPU if necessary and return its assigned CLIPFeatureExtractor. If no
+        model name is provided, return the text encoder from the model
+        currently in the GPU.
+        """
+        return self._get_sub_model(model_name, SDModelComponent.feature_extractor)
+
+    def get_model_scheduler(self, model_name: str=None)->SchedulerMixin:
+        """Given a model name identified in models.yaml, load the model into
+        GPU if necessary and return its assigned scheduler. If no
+        model name is provided, return the text encoder from the model
+        currently in the GPU.
+        """
+        return self._get_sub_model(model_name, SDModelComponent.scheduler)
+    
+    def _get_sub_model(
+            self,
+            model_name: str=None,
+            model_part: SDModelComponent=SDModelComponent.vae,
+    ) -> Union[
+        AutoencoderKL,
+        CLIPTokenizer,
+        CLIPFeatureExtractor,
+        UNet2DConditionModel,
+        CLIPTextModel,
+        StableDiffusionSafetyChecker,
+    ]:
+        """Given a model name identified in models.yaml, and the part of the
+        model you wish to retrieve, return that part. Parts are in an Enum
+        class named SDModelComponent, and consist of:
+        SDModelComponent.vae
+        SDModelComponent.text_encoder
+        SDModelComponent.tokenizer
+        SDModelComponent.unet
+        SDModelComponent.scheduler
+        SDModelComponent.safety_checker
+        SDModelComponent.feature_extractor
+        """
+        model_dict = self.get_model(model_name)
+        model = model_dict["model"]
+        return getattr(model, model_part.value)
+
     def default_model(self) -> str | None:
         """
         Returns the name of the default model, or None
@@ -454,14 +564,18 @@ class ModelManager(object):
         from . import load_pipeline_from_original_stable_diffusion_ckpt
 
         try:
-            if self.list_models()[self.current_model]['status'] == 'active':
+            if self.list_models()[self.current_model]["status"] == "active":
                 self.offload_model(self.current_model)
         except Exception as e:
             pass
 
         vae_path = None
         if vae:
-            vae_path = vae if os.path.isabs(vae) else os.path.normpath(os.path.join(Globals.root, vae))
+            vae_path = (
+                vae
+                if os.path.isabs(vae)
+                else os.path.normpath(os.path.join(Globals.root, vae))
+            )
         if self._has_cuda():
             torch.cuda.empty_cache()
         pipeline = load_pipeline_from_original_stable_diffusion_ckpt(
@@ -571,9 +685,7 @@ class ModelManager(object):
         models.yaml file.
         """
         model_name = model_name or Path(repo_or_path).stem
-        model_description = (
-            description or f"Imported diffusers model {model_name}"
-        )
+        model_description = description or f"Imported diffusers model {model_name}"
         new_config = dict(
             description=model_description,
             vae=vae,
@@ -602,7 +714,7 @@ class ModelManager(object):
         SDLegacyType.V2_v   (V2 using 'v_prediction' prediction type)
         SDLegacyType.UNKNOWN
         """
-        global_step = checkpoint.get('global_step')
+        global_step = checkpoint.get("global_step")
         state_dict = checkpoint.get("state_dict") or checkpoint
 
         try:
@@ -630,14 +742,13 @@ class ModelManager(object):
     def heuristic_import(
         self,
         path_url_or_repo: str,
-            convert: bool = True,
         model_name: str = None,
         description: str = None,
         model_config_file: Path = None,
         commit_to_conf: Path = None,
+        config_file_callback: Callable[[Path], Path] = None,
     ) -> str:
-        """
-        Accept a string which could be:
+        """Accept a string which could be:
            - a HF diffusers repo_id
            - a URL pointing to a legacy .ckpt or .safetensors file
            - a local path pointing to a legacy .ckpt or .safetensors file
@@ -651,16 +762,20 @@ class ModelManager(object):
         The model_name and/or description can be provided. If not, they will
         be generated automatically.
 
-        If convert is true, legacy models will be converted to diffusers
-        before importing.
-
         If commit_to_conf is provided, the newly loaded model will be written
         to the `models.yaml` file at the indicated path. Otherwise, the changes
         will only remain in memory.
 
-        The (potentially derived) name of the model is returned on success, or None
-        on failure. When multiple models are added from a directory, only the last
-        imported one is returned.
+        The routine will do its best to figure out the config file
+        needed to convert legacy checkpoint file, but if it can't it
+        will call the config_file_callback routine, if provided. The
+        callback accepts a single argument, the Path to the checkpoint
+        file, and returns a Path to the config file to use.
+
+        The (potentially derived) name of the model is returned on
+        success, or None on failure. When multiple models are added
+        from a directory, only the last imported one is returned.
+
         """
         model_path: Path = None
         thing = path_url_or_repo  # to save typing
@@ -707,7 +822,7 @@ class ModelManager(object):
                     Path(thing).rglob("*.safetensors")
                 ):
                     if model_name := self.heuristic_import(
-                        str(m), convert, commit_to_conf=commit_to_conf
+                        str(m), commit_to_conf=commit_to_conf
                     ):
                         print(f" >> {model_name} successfully imported")
                 return model_name
@@ -735,7 +850,7 @@ class ModelManager(object):
 
         # another round of heuristics to guess the correct config file.
         checkpoint = None
-        if model_path.suffix.endswith((".ckpt",".pt")):
+        if model_path.suffix in [".ckpt", ".pt"]:
             self.scan_model(model_path, model_path)
             checkpoint = torch.load(model_path)
         else:
@@ -743,6 +858,12 @@ class ModelManager(object):
 
         # additional probing needed if no config file provided
         if model_config_file is None:
+            # look for a like-named .yaml file in same directory
+            if model_path.with_suffix(".yaml").exists():
+                model_config_file = model_path.with_suffix(".yaml")
+                print(f"   | Using config file {model_config_file.name}")
+
+            else:
                 model_type = self.probe_model_type(checkpoint)
                 if model_type == SDLegacyType.V1:
                     print("   | SD-v1 model detected")
@@ -752,24 +873,19 @@ class ModelManager(object):
                 elif model_type == SDLegacyType.V1_INPAINT:
                     print("   | SD-v1 inpainting model detected")
                     model_config_file = Path(
-                    Globals.root, "configs/stable-diffusion/v1-inpainting-inference.yaml"
+                        Globals.root,
+                        "configs/stable-diffusion/v1-inpainting-inference.yaml",
                     )
                 elif model_type == SDLegacyType.V2_v:
-                print(
-                    "   | SD-v2-v model detected; model will be converted to diffusers format"
-                )
+                    print("   | SD-v2-v model detected")
                     model_config_file = Path(
                         Globals.root, "configs/stable-diffusion/v2-inference-v.yaml"
                     )
-                convert = True
                 elif model_type == SDLegacyType.V2_e:
-                print(
-                    "   | SD-v2-e model detected; model will be converted to diffusers format"
-                )
+                    print("   | SD-v2-e model detected")
                     model_config_file = Path(
                         Globals.root, "configs/stable-diffusion/v2-inference.yaml"
                     )
-                convert = True
                 elif model_type == SDLegacyType.V2:
                     print(
                         f"** {thing} is a V2 checkpoint file, but its parameterization cannot be determined. Please provide configuration file path."
@@ -781,13 +897,29 @@ class ModelManager(object):
                     )
                     return
 
+        if not model_config_file and config_file_callback:
+            model_config_file = config_file_callback(model_path)
+
+        # despite our best efforts, we could not find a model config file, so give up
+        if not model_config_file:
+            return
+
+        # look for a custom vae, a like-named file ending with .vae in the same directory
+        vae_path = None
+        for suffix in ["pt", "ckpt", "safetensors"]:
+            if (model_path.with_suffix(f".vae.{suffix}")).exists():
+                vae_path = model_path.with_suffix(f".vae.{suffix}")
+                print(f"   | Using VAE file {vae_path.name}")
+        vae = None if vae_path else dict(repo_id="stabilityai/sd-vae-ft-mse")
+
         diffuser_path = Path(
             Globals.root, "models", Globals.converted_ckpts_dir, model_path.stem
         )
         model_name = self.convert_and_import(
             model_path,
             diffusers_path=diffuser_path,
-            vae=dict(repo_id="stabilityai/sd-vae-ft-mse"),
+            vae=vae,
+            vae_path=str(vae_path),
             model_name=model_name,
             model_description=description,
             original_config_file=model_config_file,
@@ -829,8 +961,8 @@ class ModelManager(object):
             return
 
         model_name = model_name or diffusers_path.name
-        model_description = model_description or f"Optimized version of {model_name}"
-        print(f">> Optimizing {model_name} (30-60s)")
+        model_description = model_description or f"Converted version of {model_name}"
+        print(f"   | Converting {model_name} to diffusers (30-60s)")
         try:
             # By passing the specified VAE to the conversion function, the autoencoder
             # will be built into the model rather than tacked on afterward via the config file
@@ -848,7 +980,7 @@ class ModelManager(object):
                 scan_needed=scan_needed,
             )
             print(
-                f"   | Success. Optimized model is now located at {str(diffusers_path)}"
+                f"   | Success. Converted model is now located at {str(diffusers_path)}"
             )
             print(f"   | Writing new config file entry for {model_name}")
             new_config = dict(
@@ -953,15 +1085,15 @@ class ModelManager(object):
         legacy_locations = [
             Path(
                 models_dir,
-                "CompVis/stable-diffusion-safety-checker/models--CompVis--stable-diffusion-safety-checker"
+                "CompVis/stable-diffusion-safety-checker/models--CompVis--stable-diffusion-safety-checker",
             ),
             Path(models_dir, "bert-base-uncased/models--bert-base-uncased"),
             Path(
                 models_dir,
-                "openai/clip-vit-large-patch14/models--openai--clip-vit-large-patch14"
+                "openai/clip-vit-large-patch14/models--openai--clip-vit-large-patch14",
             ),
         ]
-        legacy_locations.extend(list(global_cache_dir("diffusers").glob('*')))
+        legacy_locations.extend(list(global_cache_dir("diffusers").glob("*")))
 
         legacy_layout = False
         for model in legacy_locations:
@@ -980,7 +1112,7 @@ class ModelManager(object):
 >> make adjustments, please press ctrl-C now to abort and relaunch InvokeAI when you are ready.
 >> Otherwise press <enter> to continue."""
         )
-        input('continue> ')
+        input("continue> ")
 
         # transformer files get moved into the hub directory
         if cls._is_huggingface_hub_directory_present():
@@ -1072,7 +1204,7 @@ class ModelManager(object):
         return self.device.type == "cuda"
 
     def _diffuser_sha256(
-        self, name_or_path: Union[str, Path], chunksize=4096
+        self, name_or_path: Union[str, Path], chunksize=16777216
     ) -> Union[str, bytes]:
         path = None
         if isinstance(name_or_path, Path):

invokeai/backend/stable_diffusion/concepts_lib.py

@@ -57,7 +57,7 @@ class HuggingFaceConceptsLibrary(object):
                 self.concept_list.extend(list(local_concepts_to_add))
                 return self.concept_list
             return self.concept_list
-        else:
+        elif Globals.internet_available is True:
             try:
                 models = self.hf_api.list_models(
                     filter=ModelFilter(model_name="sd-concepts-library/")
@@ -73,6 +73,8 @@ class HuggingFaceConceptsLibrary(object):
                     " ** You may load .bin and .pt file(s) manually using the --embedding_directory argument."
                 )
             return self.concept_list
+        else: 
+            return self.concept_list
 
     def get_concept_model_path(self, concept_name: str) -> str:
         """

invokeai/backend/stable_diffusion/textual_inversion_manager.py

@@ -1,16 +1,26 @@
-import os
 import traceback
 from dataclasses import dataclass
 from pathlib import Path
-from typing import Optional, Union
+from typing import Optional, Union, List
 
+import safetensors.torch
 import torch
+
 from compel.embeddings_provider import BaseTextualInversionManager
 from picklescan.scanner import scan_file_path
 from transformers import CLIPTextModel, CLIPTokenizer
 
 from .concepts_lib import HuggingFaceConceptsLibrary
 
+@dataclass
+class EmbeddingInfo:
+    name: str
+    embedding: torch.Tensor
+    num_vectors_per_token: int
+    token_dim: int 
+    trained_steps: int = None
+    trained_model_name: str = None
+    trained_model_checksum: str = None
 
 @dataclass
 class TextualInversion:
@@ -72,37 +82,17 @@ class TextualInversionManager(BaseTextualInversionManager):
         if str(ckpt_path).endswith(".DS_Store"):
             return
 
-        try:
-            scan_result = scan_file_path(str(ckpt_path))
-            if scan_result.infected_files == 1:
+        embedding_list = self._parse_embedding(str(ckpt_path))
+        for embedding_info in embedding_list:
+            if (self.text_encoder.get_input_embeddings().weight.data[0].shape[0] != embedding_info.token_dim):
                 print(
-                    f"\n### Security Issues Found in Model: {scan_result.issues_count}"
+                    f"   ** Notice: {ckpt_path.parents[0].name}/{ckpt_path.name} was trained on a model with an incompatible token dimension: {self.text_encoder.get_input_embeddings().weight.data[0].shape[0]} vs {embedding_info.token_dim}."
                 )
-                print("### For your safety, InvokeAI will not load this embed.")
-                return
-        except Exception:
-            print(
-                f"### {ckpt_path.parents[0].name}/{ckpt_path.name} is damaged or corrupt."
-            )
-            return
-
-        embedding_info = self._parse_embedding(str(ckpt_path))
-
-        if embedding_info is None:
-            # We've already put out an error message about the bad embedding in _parse_embedding, so just return.
-            return
-        elif (
-            self.text_encoder.get_input_embeddings().weight.data[0].shape[0]
-            != embedding_info["token_dim"]
-        ):
-            print(
-                f"** Notice: {ckpt_path.parents[0].name}/{ckpt_path.name} was trained on a model with an incompatible token dimension: {self.text_encoder.get_input_embeddings().weight.data[0].shape[0]} vs {embedding_info['token_dim']}."
-            )
-            return
+                continue
 
             # Resolve the situation in which an earlier embedding has claimed the same
             # trigger string. We replace the trigger with '<source_file>', as we used to.
-        trigger_str = embedding_info["name"]
+            trigger_str = embedding_info.name
             sourcefile = (
                 f"{ckpt_path.parent.name}/{ckpt_path.name}"
                 if ckpt_path.name == "learned_embeds.bin"
@@ -123,7 +113,7 @@ class TextualInversionManager(BaseTextualInversionManager):
             try:
                 self._add_textual_inversion(
                     trigger_str,
-                embedding_info["embedding"],
+                    embedding_info.embedding,
                     defer_injecting_tokens=defer_injecting_tokens,
                 )
                 # remember which source file claims this trigger
@@ -309,111 +299,130 @@ class TextualInversionManager(BaseTextualInversionManager):
 
         return token_id
 
-    def _parse_embedding(self, embedding_file: str):
-        file_type = embedding_file.split(".")[-1]
-        if file_type == "pt":
-            return self._parse_embedding_pt(embedding_file)
-        elif file_type == "bin":
-            return self._parse_embedding_bin(embedding_file)
-        else:
-            print(f"** Notice: unrecognized embedding file format: {embedding_file}")
-            return None
 
-    def _parse_embedding_pt(self, embedding_file):
-        embedding_ckpt = torch.load(embedding_file, map_location="cpu")
-        embedding_info = {}
-
-        # Check if valid embedding file
-        if "string_to_token" and "string_to_param" in embedding_ckpt:
-            # Catch variants that do not have the expected keys or values.
+    def _parse_embedding(self, embedding_file: str)->List[EmbeddingInfo]:
+        suffix = Path(embedding_file).suffix
         try:
-                embedding_info["name"] = embedding_ckpt["name"] or os.path.basename(
-                    os.path.splitext(embedding_file)[0]
+            if suffix in [".pt",".ckpt",".bin"]:
+                scan_result = scan_file_path(embedding_file)
+                if scan_result.infected_files > 0:
+                    print(
+                        f"   ** Security Issues Found in Model: {scan_result.issues_count}"
                     )
+                    print("   ** For your safety, InvokeAI will not load this embed.")
+                    return list()
+                ckpt = torch.load(embedding_file,map_location="cpu")
+            else:
+                ckpt = safetensors.torch.load_file(embedding_file)
+        except Exception as e:
+            print(f"   ** Notice: unrecognized embedding file format: {embedding_file}: {e}")
+            return list()
         
-                # Check num of embeddings and warn user only the first will be used
-                embedding_info["num_of_embeddings"] = len(
-                    embedding_ckpt["string_to_token"]
-                )
-                if embedding_info["num_of_embeddings"] > 1:
-                    print(">> More than 1 embedding found. Will use the first one")
+        # try to figure out what kind of embedding file it is and parse accordingly
+        keys = list(ckpt.keys())
+        if all(x in keys for x in ['string_to_token','string_to_param','name','step']):
+            return self._parse_embedding_v1(ckpt, embedding_file)     # example rem_rezero.pt
         
-                embedding = list(embedding_ckpt["string_to_param"].values())[0]
-            except (AttributeError, KeyError):
-                return self._handle_broken_pt_variants(embedding_ckpt, embedding_file)
+        elif all(x in keys for x in ['string_to_token','string_to_param']):
+            return self._parse_embedding_v2(ckpt, embedding_file)     # example midj-strong.pt
         
-            embedding_info["embedding"] = embedding
-            embedding_info["num_vectors_per_token"] = embedding.size()[0]
-            embedding_info["token_dim"] = embedding.size()[1]
-
-            try:
-                embedding_info["trained_steps"] = embedding_ckpt["step"]
-                embedding_info["trained_model_name"] = embedding_ckpt[
-                    "sd_checkpoint_name"
-                ]
-                embedding_info["trained_model_checksum"] = embedding_ckpt[
-                    "sd_checkpoint"
-                ]
-            except AttributeError:
-                print(">> No Training Details Found. Passing ...")
-
-        # .pt files found at https://cyberes.github.io/stable-diffusion-textual-inversion-models/
-        # They are actually .bin files
-        elif len(embedding_ckpt.keys()) == 1:
-            embedding_info = self._parse_embedding_bin(embedding_file)
+        elif 'emb_params' in keys:
+            return self._parse_embedding_v3(ckpt, embedding_file)     # example easynegative.safetensors
         
         else:
-            print(">> Invalid embedding format")
-            embedding_info = None
+            return self._parse_embedding_v4(ckpt, embedding_file)     # usually a '.bin' file
 
-        return embedding_info
+    def _parse_embedding_v1(self, embedding_ckpt: dict, file_path: str)->List[EmbeddingInfo]:
+        basename = Path(file_path).stem
+        print(f'   | Loading v1 embedding file: {basename}')
 
-    def _parse_embedding_bin(self, embedding_file):
-        embedding_ckpt = torch.load(embedding_file, map_location="cpu")
-        embedding_info = {}
-
-        if list(embedding_ckpt.keys()) == 0:
-            print(">> Invalid concepts file")
-            embedding_info = None
+        embeddings = list()
+        token_counter = -1
+        for token,embedding in embedding_ckpt["string_to_param"].items():
+            if token_counter < 0:
+                trigger = embedding_ckpt["name"]
+            elif token_counter == 0:
+                trigger = f'<basename>'
             else:
-            for token in list(embedding_ckpt.keys()):
-                embedding_info["name"] = (
-                    token
-                    or f"<{os.path.basename(os.path.splitext(embedding_file)[0])}>"
+                trigger = f'<{basename}-{int(token_counter:=token_counter)}>'
+            token_counter += 1
+            embedding_info = EmbeddingInfo(
+                name = trigger,
+                embedding = embedding,
+                num_vectors_per_token = embedding.size()[0],
+                token_dim = embedding.size()[1],
+                trained_steps = embedding_ckpt["step"],
+                trained_model_name = embedding_ckpt["sd_checkpoint_name"],
+                trained_model_checksum = embedding_ckpt["sd_checkpoint"]
             )
-                embedding_info["embedding"] = embedding_ckpt[token]
-                embedding_info[
-                    "num_vectors_per_token"
-                ] = 1  # All Concepts seem to default to 1
-                embedding_info["token_dim"] = embedding_info["embedding"].size()[0]
+            embeddings.append(embedding_info)
+        return embeddings
 
-        return embedding_info
+    def _parse_embedding_v2 (
+        self, embedding_ckpt: dict, file_path: str
+    ) -> List[EmbeddingInfo]:
+        """
+        This handles embedding .pt file variant #2.
+        """
+        basename = Path(file_path).stem
+        print(f'   | Loading v2 embedding file: {basename}')
+        embeddings = list()
         
-    def _handle_broken_pt_variants(
-        self, embedding_ckpt: dict, embedding_file: str
-    ) -> dict:
-        """
-        This handles the broken .pt file variants. We only know of one at present.
-        """
-        embedding_info = {}
         if isinstance(
             list(embedding_ckpt["string_to_token"].values())[0], torch.Tensor
         ):
-            for token in list(embedding_ckpt["string_to_token"].keys()):
-                embedding_info["name"] = (
-                    token
-                    if token != "*"
-                    else f"<{os.path.basename(os.path.splitext(embedding_file)[0])}>"
+            token_counter = 0
+            for token,embedding in embedding_ckpt["string_to_param"].items():
+                trigger = token if token != '*' \
+                    else f'<{basename}>' if token_counter == 0 \
+                         else f'<{basename}-{int(token_counter:=token_counter+1)}>'
+                embedding_info = EmbeddingInfo(
+                    name = trigger,
+                    embedding = embedding,
+                    num_vectors_per_token = embedding.size()[0],
+                    token_dim = embedding.size()[1],
                 )
-                embedding_info["embedding"] = embedding_ckpt[
-                    "string_to_param"
-                ].state_dict()[token]
-                embedding_info["num_vectors_per_token"] = embedding_info[
-                    "embedding"
-                ].shape[0]
-                embedding_info["token_dim"] = embedding_info["embedding"].size()[1]
+                embeddings.append(embedding_info)
         else:
-            print(">> Invalid embedding format")
-            embedding_info = None
+            print(f"   ** {basename}: Unrecognized embedding format")
 
-        return embedding_info
+        return embeddings
+
+    def _parse_embedding_v3(self, embedding_ckpt: dict, file_path: str)->List[EmbeddingInfo]:
+        """
+        Parse 'version 3' of the .pt textual inversion embedding files.
+        """
+        basename = Path(file_path).stem
+        print(f'   | Loading v3 embedding file: {basename}')
+        embedding = embedding_ckpt['emb_params']
+        embedding_info = EmbeddingInfo(
+            name = f'<{basename}>',
+            embedding = embedding,
+            num_vectors_per_token = embedding.size()[0],
+            token_dim = embedding.size()[1],
+        )
+        return [embedding_info]
+    
+    def _parse_embedding_v4(self, embedding_ckpt: dict, filepath: str)->List[EmbeddingInfo]:
+        """
+        Parse 'version 4' of the textual inversion embedding files. This one
+        is usually associated with .bin files trained by HuggingFace diffusers.
+        """
+        basename = Path(filepath).stem
+        short_path = Path(filepath).parents[0].name+'/'+Path(filepath).name
+        
+        print(f'   | Loading v4 embedding file: {short_path}')
+        
+        embeddings = list()
+        if list(embedding_ckpt.keys()) == 0:
+            print(f"   ** Invalid embeddings file: {short_path}")
+        else:
+            for token,embedding in embedding_ckpt.items():
+                embedding_info = EmbeddingInfo(
+                    name = token or f"<{basename}>",
+                    embedding = embedding,
+                    num_vectors_per_token = 1,  # All Concepts seem to default to 1
+                    token_dim = embedding.size()[0],
+                )
+                embeddings.append(embedding_info)
+        return embeddings

invokeai/backend/web/invoke_ai_web_server.py

@@ -1022,7 +1022,7 @@ class InvokeAIWebServer:
                     "RGB"
                 )
 
-            def image_progress(sample, step):
+            def image_progress(intermediate_state: PipelineIntermediateState):
                 if self.canceled.is_set():
                     raise CanceledException
 
@@ -1030,6 +1030,14 @@ class InvokeAIWebServer:
                 nonlocal generation_parameters
                 nonlocal progress
 
+                step = intermediate_state.step
+                if intermediate_state.predicted_original is not None:
+                    # Some schedulers report not only the noisy latents at the current timestep,
+                    # but also their estimate so far of what the de-noised latents will be.
+                    sample = intermediate_state.predicted_original
+                else:
+                    sample = intermediate_state.latents
+
                 generation_messages = {
                     "txt2img": "common.statusGeneratingTextToImage",
                     "img2img": "common.statusGeneratingImageToImage",
@@ -1302,16 +1310,9 @@ class InvokeAIWebServer:
 
                 progress.set_current_iteration(progress.current_iteration + 1)
 
-            def diffusers_step_callback_adapter(*cb_args, **kwargs):
-                if isinstance(cb_args[0], PipelineIntermediateState):
-                    progress_state: PipelineIntermediateState = cb_args[0]
-                    return image_progress(progress_state.latents, progress_state.step)
-                else:
-                    return image_progress(*cb_args, **kwargs)
-
             self.generate.prompt2image(
                 **generation_parameters,
-                step_callback=diffusers_step_callback_adapter,
+                step_callback=image_progress,
                 image_callback=image_done,
             )
 

invokeai/frontend/CLI/CLI.py

@@ -158,14 +158,9 @@ def main():
         report_model_error(opt, e)
 
     # try to autoconvert new models
-    if path := opt.autoimport:
-        gen.model_manager.heuristic_import(
-            str(path), convert=False, commit_to_conf=opt.conf
-        )
-
     if path := opt.autoconvert:
         gen.model_manager.heuristic_import(
-            str(path), convert=True, commit_to_conf=opt.conf
+            str(path), commit_to_conf=opt.conf
         )
 
     # web server loops forever
@@ -581,6 +576,7 @@ def do_command(command: str, gen, opt: Args, completer) -> tuple:
 
     elif command.startswith("!replay"):
         file_path = command.replace("!replay", "", 1).strip()
+        file_path = os.path.join(opt.outdir, file_path)
         if infile is None and os.path.isfile(file_path):
             infile = open(file_path, "r", encoding="utf-8")
         completer.add_history(command)
@@ -626,7 +622,7 @@ def set_default_output_dir(opt: Args, completer: Completer):
     completer.set_default_dir(opt.outdir)
 
 
-def import_model(model_path: str, gen, opt, completer, convert=False):
+def import_model(model_path: str, gen, opt, completer):
     """
     model_path can be (1) a URL to a .ckpt file; (2) a local .ckpt file path;
     (3) a huggingface repository id; or (4) a local directory containing a
@@ -657,7 +653,6 @@ def import_model(model_path: str, gen, opt, completer, convert=False):
         model_path,
         model_name=model_name,
         description=model_desc,
-        convert=convert,
     )
 
     if not imported_name:
@@ -666,7 +661,6 @@ def import_model(model_path: str, gen, opt, completer, convert=False):
                 model_path,
                 model_name=model_name,
                 description=model_desc,
-                convert=convert,
                 model_config_file=config_file,
             )
     if not imported_name:
@@ -757,7 +751,6 @@ def _get_model_name_and_desc(
     )
     return model_name, model_description
 
-
 def convert_model(model_name_or_path: Union[Path, str], gen, opt, completer):
     model_name_or_path = model_name_or_path.replace("\\", "/")  # windows
     manager = gen.model_manager
@@ -788,7 +781,7 @@ def convert_model(model_name_or_path: Union[Path, str], gen, opt, completer):
         )
     else:
         try:
-            import_model(model_name_or_path, gen, opt, completer, convert=True)
+            import_model(model_name_or_path, gen, opt, completer)
         except KeyboardInterrupt:
             return
 

invokeai/frontend/install/model_install.py

@@ -199,17 +199,6 @@ class addModelsForm(npyscreen.FormMultiPage):
             relx=4,
             scroll_exit=True,
         )
-        self.nextrely += 1
-        self.convert_models = self.add_widget_intelligent(
-            npyscreen.TitleSelectOne,
-            name="== CONVERT IMPORTED MODELS INTO DIFFUSERS==",
-            values=["Keep original format", "Convert to diffusers"],
-            value=0,
-            begin_entry_at=4,
-            max_height=4,
-            hidden=True,  # will appear when imported models box is edited
-            scroll_exit=True,
-        )
         self.cancel = self.add_widget_intelligent(
             npyscreen.ButtonPress,
             name="CANCEL",
@@ -244,8 +233,6 @@ class addModelsForm(npyscreen.FormMultiPage):
             self.show_directory_fields.addVisibleWhenSelected(i)
 
         self.show_directory_fields.when_value_edited = self._clear_scan_directory
-        self.import_model_paths.when_value_edited = self._show_hide_convert
-        self.autoload_directory.when_value_edited = self._show_hide_convert
 
     def resize(self):
         super().resize()
@@ -256,13 +243,6 @@ class addModelsForm(npyscreen.FormMultiPage):
         if not self.show_directory_fields.value:
             self.autoload_directory.value = ""
 
-    def _show_hide_convert(self):
-        model_paths = self.import_model_paths.value or ""
-        autoload_directory = self.autoload_directory.value or ""
-        self.convert_models.hidden = (
-            len(model_paths) == 0 and len(autoload_directory) == 0
-        )
-
     def _get_starter_model_labels(self) -> List[str]:
         window_width, window_height = get_terminal_size()
         label_width = 25
@@ -322,7 +302,6 @@ class addModelsForm(npyscreen.FormMultiPage):
         .scan_directory: Path to a directory of models to scan and import
         .autoscan_on_startup:  True if invokeai should scan and import at startup time
         .import_model_paths:   list of URLs, repo_ids and file paths to import
-        .convert_to_diffusers: if True, convert legacy checkpoints into diffusers
         """
         # we're using a global here rather than storing the result in the parentapp
         # due to some bug in npyscreen that is causing attributes to be lost
@@ -359,7 +338,6 @@ class addModelsForm(npyscreen.FormMultiPage):
 
         # URLs and the like
         selections.import_model_paths = self.import_model_paths.value.split()
-        selections.convert_to_diffusers = self.convert_models.value[0] == 1
 
 
 class AddModelApplication(npyscreen.NPSAppManaged):
@@ -372,7 +350,6 @@ class AddModelApplication(npyscreen.NPSAppManaged):
             scan_directory=None,
             autoscan_on_startup=None,
             import_model_paths=None,
-            convert_to_diffusers=None,
         )
 
     def onStart(self):
@@ -393,7 +370,6 @@ def process_and_execute(opt: Namespace, selections: Namespace):
     directory_to_scan = selections.scan_directory
     scan_at_startup = selections.autoscan_on_startup
     potential_models_to_install = selections.import_model_paths
-    convert_to_diffusers = selections.convert_to_diffusers
 
     install_requested_models(
         install_initial_models=models_to_install,
@@ -401,7 +377,6 @@ def process_and_execute(opt: Namespace, selections: Namespace):
         scan_directory=Path(directory_to_scan) if directory_to_scan else None,
         external_models=potential_models_to_install,
         scan_at_startup=scan_at_startup,
-        convert_to_diffusers=convert_to_diffusers,
         precision="float32"
         if opt.full_precision
         else choose_precision(torch.device(choose_torch_device())),

invokeai/frontend/web/.eslintignore

@@ -6,3 +6,5 @@ stats.html
 index.html
 .yarn/
 *.scss
+src/services/api/
+src/services/fixtures/*

invokeai/frontend/web/.prettierignore

@@ -3,4 +3,8 @@ dist/
 node_modules/
 patches/
 stats.html
+index.html
 .yarn/
+*.scss
+src/services/api/
+src/services/fixtures/*

invokeai/frontend/web/dist/assets/ThemeLocaleProvider-7791bd91.js

@@ -1,4 +1,4 @@
-import{j as y,cN as Ie,r as _,cO as bt,q as Lr,cP as o,cQ as b,cR as v,cS as S,cT as Vr,cU as ut,cV as vt,cM as ft,cW as mt,n as gt,cX as ht,E as pt}from"./index-f7f41e1f.js";import{d as yt,i as St,T as xt,j as $t,h as kt}from"./storeHooks-eaf47ae3.js";var Or=`
+import{j as y,cO as Ie,r as _,cP as bt,q as Lr,cQ as o,cR as b,cS as v,cT as S,cU as Vr,cV as ut,cW as vt,cN as ft,cX as mt,n as gt,cY as ht,E as pt}from"./index-e53e8108.js";import{d as yt,i as St,T as xt,j as $t,h as kt}from"./storeHooks-5cde7d31.js";var Or=`
   :root {
     --chakra-vh: 100vh;
   }

invokeai/frontend/web/dist/index.html

@@ -12,7 +12,7 @@
         margin: 0;
       }
     </style>
-    <script type="module" crossorigin src="./assets/index-f7f41e1f.js"></script>
+    <script type="module" crossorigin src="./assets/index-e53e8108.js"></script>
     <link rel="stylesheet" href="./assets/index-5483945c.css">
   </head>
 

invokeai/frontend/web/dist/locales/ar.json

@@ -8,7 +8,6 @@
         "darkTheme": "داكن",
         "lightTheme": "فاتح",
         "greenTheme": "أخضر",
-        "text2img": "نص إلى صورة",
         "img2img": "صورة إلى صورة",
         "unifiedCanvas": "لوحة موحدة",
         "nodes": "عقد",

invokeai/frontend/web/dist/locales/de.json

@@ -7,7 +7,6 @@
         "darkTheme": "Dunkel",
         "lightTheme": "Hell",
         "greenTheme": "Grün",
-        "text2img": "Text zu Bild",
         "img2img": "Bild zu Bild",
         "nodes": "Knoten",
         "langGerman": "Deutsch",

invokeai/frontend/web/dist/locales/en.json

@@ -505,7 +505,9 @@
         "info": "Info",
         "deleteImage": "Delete Image",
         "initialImage": "Initial Image",
-        "showOptionsPanel": "Show Options Panel"
+        "showOptionsPanel": "Show Options Panel",
+        "hidePreview": "Hide Preview",
+        "showPreview": "Show Preview"
     },
     "settings": {
         "models": "Models",

invokeai/frontend/web/dist/locales/es.json

@@ -8,7 +8,6 @@
         "darkTheme": "Oscuro",
         "lightTheme": "Claro",
         "greenTheme": "Verde",
-        "text2img": "Texto a Imagen",
         "img2img": "Imagen a Imagen",
         "unifiedCanvas": "Lienzo Unificado",
         "nodes": "Nodos",
@@ -70,7 +69,11 @@
         "langHebrew": "Hebreo",
         "pinOptionsPanel": "Pin del panel de opciones",
         "loading": "Cargando",
-        "loadingInvokeAI": "Cargando invocar a la IA"
+        "loadingInvokeAI": "Cargando invocar a la IA",
+        "postprocessing": "Tratamiento posterior",
+        "txt2img": "De texto a imagen",
+        "accept": "Aceptar",
+        "cancel": "Cancelar"
     },
     "gallery": {
         "generations": "Generaciones",
@@ -404,7 +407,8 @@
         "none": "ninguno",
         "pickModelType": "Elige el tipo de modelo",
         "v2_768": "v2 (768px)",
-        "addDifference": "Añadir una diferencia"
+        "addDifference": "Añadir una diferencia",
+        "scanForModels": "Buscar modelos"
     },
     "parameters": {
         "images": "Imágenes",
@@ -574,7 +578,7 @@
         "autoSaveToGallery": "Guardar automáticamente en galería",
         "saveBoxRegionOnly": "Guardar solo región dentro de la caja",
         "limitStrokesToBox": "Limitar trazos a la caja",
-        "showCanvasDebugInfo": "Mostrar información de depuración de lienzo",
+        "showCanvasDebugInfo": "Mostrar la información adicional del lienzo",
         "clearCanvasHistory": "Limpiar historial de lienzo",
         "clearHistory": "Limpiar historial",
         "clearCanvasHistoryMessage": "Limpiar el historial de lienzo también restablece completamente el lienzo unificado. Esto incluye todo el historial de deshacer/rehacer, las imágenes en el área de preparación y la capa base del lienzo.",

invokeai/frontend/web/dist/locales/fr.json

@@ -8,7 +8,6 @@
         "darkTheme": "Sombre",
         "lightTheme": "Clair",
         "greenTheme": "Vert",
-        "text2img": "Texte en image",
         "img2img": "Image en image",
         "unifiedCanvas": "Canvas unifié",
         "nodes": "Nœuds",
@@ -47,7 +46,19 @@
         "statusLoadingModel": "Chargement du modèle",
         "statusModelChanged": "Modèle changé",
         "discordLabel": "Discord",
-        "githubLabel": "Github"
+        "githubLabel": "Github",
+        "accept": "Accepter",
+        "statusMergingModels": "Mélange des modèles",
+        "loadingInvokeAI": "Chargement de Invoke AI",
+        "cancel": "Annuler",
+        "langEnglish": "Anglais",
+        "statusConvertingModel": "Conversion du modèle",
+        "statusModelConverted": "Modèle converti",
+        "loading": "Chargement",
+        "pinOptionsPanel": "Épingler la page d'options",
+        "statusMergedModels": "Modèles mélangés",
+        "txt2img": "Texte vers image",
+        "postprocessing": "Post-Traitement"
     },
     "gallery": {
         "generations": "Générations",
@@ -518,5 +529,15 @@
         "betaDarkenOutside": "Assombrir à l'extérieur",
         "betaLimitToBox": "Limiter à la boîte",
         "betaPreserveMasked": "Conserver masqué"
+    },
+    "accessibility": {
+        "uploadImage": "Charger une image",
+        "reset": "Réinitialiser",
+        "nextImage": "Image suivante",
+        "previousImage": "Image précédente",
+        "useThisParameter": "Utiliser ce paramètre",
+        "zoomIn": "Zoom avant",
+        "zoomOut": "Zoom arrière",
+        "showOptionsPanel": "Montrer la page d'options"
     }
 }

invokeai/frontend/web/dist/locales/he.json

@@ -125,7 +125,6 @@
         "langSimplifiedChinese": "סינית",
         "langUkranian": "אוקראינית",
         "langSpanish": "ספרדית",
-        "text2img": "טקסט לתמונה",
         "img2img": "תמונה לתמונה",
         "unifiedCanvas": "קנבס מאוחד",
         "nodes": "צמתים",

invokeai/frontend/web/dist/locales/it.json

@@ -8,7 +8,6 @@
         "darkTheme": "Scuro",
         "lightTheme": "Chiaro",
         "greenTheme": "Verde",
-        "text2img": "Testo a Immagine",
         "img2img": "Immagine a Immagine",
         "unifiedCanvas": "Tela unificata",
         "nodes": "Nodi",
@@ -70,7 +69,11 @@
         "loading": "Caricamento in corso",
         "oceanTheme": "Oceano",
         "langHebrew": "Ebraico",
-        "loadingInvokeAI": "Caricamento Invoke AI"
+        "loadingInvokeAI": "Caricamento Invoke AI",
+        "postprocessing": "Post Elaborazione",
+        "txt2img": "Testo a Immagine",
+        "accept": "Accetta",
+        "cancel": "Annulla"
     },
     "gallery": {
         "generations": "Generazioni",
@@ -404,7 +407,8 @@
         "v2_768": "v2 (768px)",
         "none": "niente",
         "addDifference": "Aggiungi differenza",
-        "pickModelType": "Scegli il tipo di modello"
+        "pickModelType": "Scegli il tipo di modello",
+        "scanForModels": "Cerca modelli"
     },
     "parameters": {
         "images": "Immagini",
@@ -574,7 +578,7 @@
         "autoSaveToGallery": "Salvataggio automatico nella Galleria",
         "saveBoxRegionOnly": "Salva solo l'area di selezione",
         "limitStrokesToBox": "Limita i tratti all'area di selezione",
-        "showCanvasDebugInfo": "Mostra informazioni di debug della Tela",
+        "showCanvasDebugInfo": "Mostra ulteriori informazioni sulla Tela",
         "clearCanvasHistory": "Cancella cronologia Tela",
         "clearHistory": "Cancella la cronologia",
         "clearCanvasHistoryMessage": "La cancellazione della cronologia della tela lascia intatta la tela corrente, ma cancella in modo irreversibile la cronologia degli annullamenti e dei ripristini.",
@@ -612,7 +616,7 @@
         "copyMetadataJson": "Copia i metadati JSON",
         "exitViewer": "Esci dal visualizzatore",
         "zoomIn": "Zoom avanti",
-        "zoomOut": "Zoom Indietro",
+        "zoomOut": "Zoom indietro",
         "rotateCounterClockwise": "Ruotare in senso antiorario",
         "rotateClockwise": "Ruotare in senso orario",
         "flipHorizontally": "Capovolgi orizzontalmente",

invokeai/frontend/web/dist/locales/ko.json

@@ -11,7 +11,6 @@
         "langArabic": "العربية",
         "langEnglish": "English",
         "langDutch": "Nederlands",
-        "text2img": "텍스트->이미지",
         "unifiedCanvas": "통합 캔버스",
         "langFrench": "Français",
         "langGerman": "Deutsch",

invokeai/frontend/web/dist/locales/nl.json

@@ -8,7 +8,6 @@
         "darkTheme": "Donker",
         "lightTheme": "Licht",
         "greenTheme": "Groen",
-        "text2img": "Tekst naar afbeelding",
         "img2img": "Afbeelding naar afbeelding",
         "unifiedCanvas": "Centraal canvas",
         "nodes": "Knooppunten",

invokeai/frontend/web/dist/locales/pl.json

@@ -8,7 +8,6 @@
         "darkTheme": "Ciemny",
         "lightTheme": "Jasny",
         "greenTheme": "Zielony",
-        "text2img": "Tekst na obraz",
         "img2img": "Obraz na obraz",
         "unifiedCanvas": "Tryb uniwersalny",
         "nodes": "Węzły",

invokeai/frontend/web/dist/locales/pt.json

@@ -20,7 +20,6 @@
         "langSpanish": "Espanhol",
         "langRussian": "Русский",
         "langUkranian": "Украї́нська",
-        "text2img": "Texto para Imagem",
         "img2img": "Imagem para Imagem",
         "unifiedCanvas": "Tela Unificada",
         "nodes": "Nós",

invokeai/frontend/web/dist/locales/pt_BR.json

@@ -8,7 +8,6 @@
         "darkTheme": "Noite",
         "lightTheme": "Dia",
         "greenTheme": "Verde",
-        "text2img": "Texto Para Imagem",
         "img2img": "Imagem Para Imagem",
         "unifiedCanvas": "Tela Unificada",
         "nodes": "Nódulos",

invokeai/frontend/web/dist/locales/ru.json

@@ -8,7 +8,6 @@
         "darkTheme": "Темная",
         "lightTheme": "Светлая",
         "greenTheme": "Зеленая",
-        "text2img": "Изображение из текста (text2img)",
         "img2img": "Изображение в изображение (img2img)",
         "unifiedCanvas": "Универсальный холст",
         "nodes": "Ноды",

invokeai/frontend/web/dist/locales/uk.json

@@ -8,7 +8,6 @@
         "darkTheme": "Темна",
         "lightTheme": "Світла",
         "greenTheme": "Зелена",
-        "text2img": "Зображення із тексту (text2img)",
         "img2img": "Зображення із зображення (img2img)",
         "unifiedCanvas": "Універсальне полотно",
         "nodes": "Вузли",

invokeai/frontend/web/dist/locales/zh_CN.json

@@ -8,7 +8,6 @@
         "darkTheme": "暗色",
         "lightTheme": "亮色",
         "greenTheme": "绿色",
-        "text2img": "文字到图像",
         "img2img": "图像到图像",
         "unifiedCanvas": "统一画布",
         "nodes": "节点",

invokeai/frontend/web/dist/locales/zh_Hant.json

@@ -33,7 +33,6 @@
         "langBrPortuguese": "巴西葡萄牙語",
         "langRussian": "俄語",
         "langSpanish": "西班牙語",
-        "text2img": "文字到圖像",
         "unifiedCanvas": "統一畫布"
     }
 }

invokeai/frontend/web/docs/API_CLIENT.md

@@ -0,0 +1,87 @@
+# Generated axios API client
+
+- [Generated axios API client](#generated-axios-api-client)
+  - [Generation](#generation)
+    - [Generate the API client from the nodes web server](#generate-the-api-client-from-the-nodes-web-server)
+    - [Generate the API client from JSON](#generate-the-api-client-from-json)
+      - [Getting the JSON from the nodes web server](#getting-the-json-from-the-nodes-web-server)
+      - [Getting the JSON with a python script](#getting-the-json-with-a-python-script)
+      - [Generate the API client](#generate-the-api-client)
+  - [The generated client](#the-generated-client)
+  - [API client customisation](#api-client-customisation)
+
+This API client is generated by an [openapi code generator](https://github.com/ferdikoomen/openapi-typescript-codegen).
+
+All files in `invokeai/frontend/web/src/services/api/` are made by the generator.
+
+## Generation
+
+The axios client may be generated by from the OpenAPI schema from the nodes web server, or from JSON.
+
+### Generate the API client from the nodes web server
+
+We need to start the nodes web server, which serves the OpenAPI schema to the generator.
+
+1. Start the nodes web server.
+
+```bash
+# from the repo root
+python scripts/invoke-new.py --web
+```
+
+2. Generate the API client.
+
+```bash
+# from invokeai/frontend/web/
+yarn api:web
+```
+
+### Generate the API client from JSON
+
+The JSON can be acquired from the nodes web server, or with a python script.
+
+#### Getting the JSON from the nodes web server
+
+Start the nodes web server as described above, then download the file.
+
+```bash
+# from invokeai/frontend/web/
+curl http://localhost:9090/openapi.json -o openapi.json
+```
+
+#### Getting the JSON with a python script
+
+Run this python script from the repo root, so it can access the nodes server modules.
+
+The script will output `openapi.json` in the repo root. Then we need to move it to `invokeai/frontend/web/`.
+
+```bash
+# from the repo root
+python invokeai/app/util/generate_openapi_json.py
+mv invokeai/app/util/openapi.json invokeai/frontend/web/services/fixtures/
+```
+
+#### Generate the API client
+
+Now we can generate the API client from the JSON.
+
+```bash
+# from invokeai/frontend/web/
+yarn api:file
+```
+
+## The generated client
+
+The client will be written to `invokeai/frontend/web/services/api/`:
+
+- `axios` client
+- TS types
+- An easily parseable schema, which we can use to generate UI
+
+## API client customisation
+
+The generator has a default `request.ts` file that implements a base `axios` client. The generated client uses this base client.
+
+One shortcoming of this is base client is it does not provide response headers unless the response body is empty. To fix this, we provide our own lightly-patched `request.ts`.
+
+To access the headers, call `getHeaders(response)` on any response from the generated api client. This function is exported from `invokeai/frontend/web/src/services/util/getHeaders.ts`.

invokeai/frontend/web/docs/EVENTS.md

@@ -0,0 +1,21 @@
+# Events
+
+Events via `socket.io`
+
+## `actions.ts`
+
+Redux actions for all socket events. Payloads all include a timestamp, and optionally some other data.
+
+Any reducer (or middleware) can respond to the actions.
+
+## `middleware.ts`
+
+Redux middleware for events.
+
+Handles dispatching the event actions. Only put logic here if it can't really go anywhere else.
+
+For example, on connect we want to load images to the gallery if it's not populated. This requires dispatching a thunk, so we need to directly dispatch this in the middleware.
+
+## `types.ts`
+
+Hand-written types for the socket events. Cannot generate these from the server, but fortunately they are few and simple.

invokeai/frontend/web/docs/NODE_EDITOR.md

@@ -0,0 +1,17 @@
+# Node Editor Design
+
+WIP
+
+nodes
+
+everything in `src/features/nodes/`
+
+have a look at `state.nodes.invocation`
+
+- on socket connect, if no schema saved, fetch `localhost:9090/openapi.json`, save JSON to `state.nodes.schema`
+- on fulfilled schema fetch, `parseSchema()` the schema. this outputs a `Record<string, Invocation>` which is saved to `state.nodes.invocations` - `Invocation` is like a template for the node
+- when you add a node, the the `Invocation` template is passed to `InvocationComponent.tsx` to build the UI component for that node
+- inputs/outputs have field types - and each field type gets an `FieldComponent` which includes a dispatcher to write state changes to redux `nodesSlice`
+- `reactflow` sends changes to nodes/edges to redux
+- to invoke, `buildNodesGraph()` state, then send this
+- changed onClick Invoke button actions to build the schema, then when schema builds it dispatches the actual network request to create the session - see `session.ts`

invokeai/frontend/web/docs/PACKAGE_SCRIPTS.md

@@ -0,0 +1,29 @@
+# Package Scripts
+
+WIP walkthrough of `package.json` scripts.
+
+## `theme` & `theme:watch`
+
+These run the Chakra CLI to generate types for the theme, or watch for code change and re-generate the types.
+
+The CLI essentially monkeypatches Chakra's files in `node_modules`.
+
+## `postinstall`
+
+The `postinstall` script patches a few packages and runs the Chakra CLI to generate types for the theme.
+
+### Patch `@chakra-ui/cli`
+
+See: <https://github.com/chakra-ui/chakra-ui/issues/7394>
+
+### Patch `redux-persist`
+
+We want to persist the canvas state to `localStorage` but many canvas operations change data very quickly, so we need to debounce the writes to `localStorage`.
+
+`redux-persist` is unfortunately unmaintained. The repo's current code is nonfunctional, but the last release's code depends on a package that was removed from `npm` for being malware, so we cannot just fork it.
+
+So, we have to patch it directly. Perhaps a better way would be to write a debounced storage adapter, but I couldn't figure out how to do that.
+
+### Patch `redux-deep-persist`
+
+This package makes blacklisting and whitelisting persist configs very simple, but we have to patch it to match `redux-persist` for the types to work.

invokeai/frontend/web/docs/README.md

@@ -1,10 +1,16 @@
 # InvokeAI Web UI
 
+- [InvokeAI Web UI](#invokeai-web-ui)
+  - [Stack](#stack)
+  - [Contributing](#contributing)
+    - [Dev Environment](#dev-environment)
+    - [Production builds](#production-builds)
+
 The UI is a fairly straightforward Typescript React app. The only really fancy stuff is the Unified Canvas.
 
 Code in `invokeai/frontend/web/` if you want to have a look.
 
-## Details
+## Stack
 
 State management is Redux via [Redux Toolkit](https://github.com/reduxjs/redux-toolkit). Communication with server is a mix of HTTP and [socket.io](https://github.com/socketio/socket.io-client) (with a custom redux middleware to help).
 
@@ -32,7 +38,7 @@ Start everything in dev mode:
 
 1. Start the dev server: `yarn dev`
 2. Start the InvokeAI UI per usual: `invokeai --web`
-3. Point your browser to the dev server address e.g. `http://localhost:5173/`
+3. Point your browser to the dev server address e.g. <http://localhost:5173/>
 
 ### Production builds
 

invokeai/frontend/web/index.d.ts

@@ -1,6 +1,7 @@
 import React, { PropsWithChildren } from 'react';
 import { IAIPopoverProps } from '../web/src/common/components/IAIPopover';
 import { IAIIconButtonProps } from '../web/src/common/components/IAIIconButton';
+import { InvokeTabName } from 'features/ui/store/tabMap';
 
 export {};
 
@@ -64,9 +65,25 @@ declare module '@invoke-ai/invoke-ai-ui' {
   declare class SettingsModal extends React.Component<SettingsModalProps> {
     public constructor(props: SettingsModalProps);
   }
+
+  declare class StatusIndicator extends React.Component<StatusIndicatorProps> {
+    public constructor(props: StatusIndicatorProps);
   }
 
-declare function Invoke(props: PropsWithChildren): JSX.Element;
+  declare class ModelSelect extends React.Component<ModelSelectProps> {
+    public constructor(props: ModelSelectProps);
+  }
+}
+
+interface InvokeProps extends PropsWithChildren {
+  apiUrl?: string;
+  disabledPanels?: string[];
+  disabledTabs?: InvokeTabName[];
+  token?: string;
+  shouldTransformUrls?: boolean;
+}
+
+declare function Invoke(props: InvokeProps): JSX.Element;
 
 export {
   ThemeChanger,
@@ -74,5 +91,7 @@ export {
   IAIPopover,
   IAIIconButton,
   SettingsModal,
+  StatusIndicator,
+  ModelSelect,
 };
 export = Invoke;

invokeai/frontend/web/package.json

@@ -5,7 +5,10 @@
   "scripts": {
     "prepare": "cd ../../../ && husky install invokeai/frontend/web/.husky",
     "dev": "concurrently \"vite dev\" \"yarn run theme:watch\"",
+    "dev:nodes": "concurrently \"vite dev --mode nodes\" \"yarn run theme:watch\"",
     "build": "yarn run lint && vite build",
+    "api:web": "openapi -i http://localhost:9090/openapi.json -o src/services/api --client axios --useOptions --useUnionTypes --exportSchemas true --indent 2 --request src/services/fixtures/request.ts",
+    "api:file": "openapi -i src/services/fixtures/openapi.json -o src/services/api --client axios --useOptions --useUnionTypes --exportSchemas true --indent 2 --request src/services/fixtures/request.ts",
     "preview": "vite preview",
     "lint:madge": "madge --circular src/main.tsx",
     "lint:eslint": "eslint --max-warnings=0 .",
@@ -41,9 +44,11 @@
     "@chakra-ui/react": "^2.5.1",
     "@chakra-ui/styled-system": "^2.6.1",
     "@chakra-ui/theme-tools": "^2.0.16",
+    "@dagrejs/graphlib": "^2.1.12",
     "@emotion/react": "^11.10.6",
     "@emotion/styled": "^11.10.6",
-    "@reduxjs/toolkit": "^1.9.2",
+    "@fontsource/inter": "^4.5.15",
+    "@reduxjs/toolkit": "^1.9.3",
     "chakra-ui-contextmenu": "^1.0.5",
     "dateformat": "^5.0.3",
     "formik": "^2.2.9",
@@ -67,15 +72,17 @@
     "react-redux": "^8.0.5",
     "react-transition-group": "^4.4.5",
     "react-zoom-pan-pinch": "^2.6.1",
+    "reactflow": "^11.7.0",
     "redux-deep-persist": "^1.0.7",
+    "redux-dynamic-middlewares": "^2.2.0",
     "redux-persist": "^6.0.0",
     "socket.io-client": "^4.6.0",
     "use-image": "^1.1.0",
     "uuid": "^9.0.0"
   },
   "devDependencies": {
-    "@fontsource/inter": "^4.5.15",
     "@types/dateformat": "^5.0.0",
+    "@types/lodash": "^4.14.194",
     "@types/react": "^18.0.28",
     "@types/react-dom": "^18.0.11",
     "@types/react-transition-group": "^4.4.5",
@@ -83,6 +90,7 @@
     "@typescript-eslint/eslint-plugin": "^5.52.0",
     "@typescript-eslint/parser": "^5.52.0",
     "@vitejs/plugin-react-swc": "^3.2.0",
+    "axios": "^1.3.4",
     "babel-plugin-transform-imports": "^2.0.0",
     "concurrently": "^7.6.0",
     "eslint": "^8.34.0",
@@ -90,13 +98,17 @@
     "eslint-plugin-prettier": "^4.2.1",
     "eslint-plugin-react": "^7.32.2",
     "eslint-plugin-react-hooks": "^4.6.0",
+    "form-data": "^4.0.0",
     "husky": "^8.0.3",
     "lint-staged": "^13.1.2",
     "madge": "^6.0.0",
+    "openapi-types": "^12.1.0",
+    "openapi-typescript-codegen": "^0.23.0",
     "postinstall-postinstall": "^2.1.0",
     "prettier": "^2.8.4",
     "rollup-plugin-visualizer": "^5.9.0",
     "terser": "^5.16.4",
+    "typescript": "4.9.5",
     "vite": "^4.1.2",
     "vite-plugin-eslint": "^1.8.1",
     "vite-tsconfig-paths": "^4.0.5",

invokeai/frontend/web/public/locales/en.json

@@ -52,6 +52,7 @@
         "txt2img": "Text To Image",
         "img2img": "Image To Image",
         "unifiedCanvas": "Unified Canvas",
+        "linear": "Linear",
         "nodes": "Nodes",
         "postprocessing": "Post Processing",
         "nodesDesc": "A node based system for the generation of images is under development currently. Stay tuned for updates about this amazing feature.",
@@ -505,7 +506,9 @@
         "info": "Info",
         "deleteImage": "Delete Image",
         "initialImage": "Initial Image",
-        "showOptionsPanel": "Show Options Panel"
+        "showOptionsPanel": "Show Options Panel",
+        "hidePreview": "Hide Preview",
+        "showPreview": "Show Preview"
     },
     "settings": {
         "models": "Models",
@@ -522,6 +525,10 @@
         "resetComplete": "Web UI has been reset. Refresh the page to reload."
     },
     "toast": {
+        "serverError": "Server Error",
+        "disconnected": "Disconnected from Server",
+        "connected": "Connected to Server",
+        "canceled": "Processing Canceled",
         "tempFoldersEmptied": "Temp Folder Emptied",
         "uploadFailed": "Upload failed",
         "uploadFailedMultipleImagesDesc": "Multiple images pasted, may only upload one image at a time",

invokeai/frontend/web/src/app/App.tsx

@@ -13,16 +13,42 @@ import { Box, Flex, Grid, Portal, useColorMode } from '@chakra-ui/react';
 import { APP_HEIGHT, APP_WIDTH } from 'theme/util/constants';
 import ImageGalleryPanel from 'features/gallery/components/ImageGalleryPanel';
 import Lightbox from 'features/lightbox/components/Lightbox';
-import { useAppSelector } from './storeHooks';
+import { useAppDispatch, useAppSelector } from './storeHooks';
 import { PropsWithChildren, useEffect } from 'react';
+import { setDisabledPanels, setDisabledTabs } from 'features/ui/store/uiSlice';
+import { InvokeTabName } from 'features/ui/store/tabMap';
+import { shouldTransformUrlsChanged } from 'features/system/store/systemSlice';
 
 keepGUIAlive();
 
-const App = (props: PropsWithChildren) => {
+interface Props extends PropsWithChildren {
+  options: {
+    disabledPanels: string[];
+    disabledTabs: InvokeTabName[];
+    shouldTransformUrls?: boolean;
+  };
+}
+
+const App = (props: Props) => {
   useToastWatcher();
 
   const currentTheme = useAppSelector((state) => state.ui.currentTheme);
   const { setColorMode } = useColorMode();
+  const dispatch = useAppDispatch();
+
+  useEffect(() => {
+    dispatch(setDisabledPanels(props.options.disabledPanels));
+  }, [dispatch, props.options.disabledPanels]);
+
+  useEffect(() => {
+    dispatch(setDisabledTabs(props.options.disabledTabs));
+  }, [dispatch, props.options.disabledTabs]);
+
+  useEffect(() => {
+    dispatch(
+      shouldTransformUrlsChanged(Boolean(props.options.shouldTransformUrls))
+    );
+  }, [dispatch, props.options.shouldTransformUrls]);
 
   useEffect(() => {
     setColorMode(['light'].includes(currentTheme) ? 'light' : 'dark');

invokeai/frontend/web/src/app/invokeai.d.ts

@@ -14,6 +14,8 @@
 
 import { InvokeTabName } from 'features/ui/store/tabMap';
 import { IRect } from 'konva/lib/types';
+import { ImageMetadata, ImageType } from 'services/api';
+import { AnyInvocation } from 'services/events/types';
 
 /**
  * TODO:
@@ -113,7 +115,7 @@ export declare type Metadata = SystemGenerationMetadata & {
 };
 
 // An Image has a UUID, url, modified timestamp, width, height and maybe metadata
-export declare type Image = {
+export declare type _Image = {
   uuid: string;
   url: string;
   thumbnail: string;
@@ -124,11 +126,23 @@ export declare type Image = {
   category: GalleryCategory;
   isBase64?: boolean;
   dreamPrompt?: 'string';
+  name?: string;
+};
+
+/**
+ * ResultImage
+ */
+export declare type Image = {
+  name: string;
+  type: ImageType;
+  url: string;
+  thumbnail: string;
+  metadata: ImageMetadata;
 };
 
 // GalleryImages is an array of Image.
 export declare type GalleryImages = {
-  images: Array<Image>;
+  images: Array<_Image>;
 };
 
 /**
@@ -275,7 +289,7 @@ export declare type SystemStatusResponse = SystemStatus;
 
 export declare type SystemConfigResponse = SystemConfig;
 
-export declare type ImageResultResponse = Omit<Image, 'uuid'> & {
+export declare type ImageResultResponse = Omit<_Image, 'uuid'> & {
   boundingBox?: IRect;
   generationMode: InvokeTabName;
 };
@@ -296,7 +310,7 @@ export declare type ErrorResponse = {
 };
 
 export declare type GalleryImagesResponse = {
-  images: Array<Omit<Image, 'uuid'>>;
+  images: Array<Omit<_Image, 'uuid'>>;
   areMoreImagesAvailable: boolean;
   category: GalleryCategory;
 };

invokeai/frontend/web/src/app/selectors/readinessSelector.ts

@@ -20,6 +20,7 @@ export const readinessSelector = createSelector(
       seedWeights,
       initialImage,
       seed,
+      isImageToImageEnabled,
     } = generation;
 
     const { isProcessing, isConnected } = system;
@@ -33,7 +34,7 @@ export const readinessSelector = createSelector(
       reasonsWhyNotReady.push('Missing prompt');
     }
 
-    if (activeTabName === 'img2img' && !initialImage) {
+    if (isImageToImageEnabled && !initialImage) {
       isReady = false;
       reasonsWhyNotReady.push('No initial image selected');
     }

invokeai/frontend/web/src/app/socketio/actions.ts

@@ -13,9 +13,13 @@ import { InvokeTabName } from 'features/ui/store/tabMap';
 export const generateImage = createAction<InvokeTabName>(
   'socketio/generateImage'
 );
-export const runESRGAN = createAction<InvokeAI.Image>('socketio/runESRGAN');
-export const runFacetool = createAction<InvokeAI.Image>('socketio/runFacetool');
-export const deleteImage = createAction<InvokeAI.Image>('socketio/deleteImage');
+export const runESRGAN = createAction<InvokeAI._Image>('socketio/runESRGAN');
+export const runFacetool = createAction<InvokeAI._Image>(
+  'socketio/runFacetool'
+);
+export const deleteImage = createAction<InvokeAI._Image>(
+  'socketio/deleteImage'
+);
 export const requestImages = createAction<GalleryCategory>(
   'socketio/requestImages'
 );

invokeai/frontend/web/src/app/socketio/emitters.ts

@@ -91,7 +91,7 @@ const makeSocketIOEmitters = (
         })
       );
     },
-    emitRunESRGAN: (imageToProcess: InvokeAI.Image) => {
+    emitRunESRGAN: (imageToProcess: InvokeAI._Image) => {
       dispatch(setIsProcessing(true));
 
       const {
@@ -119,7 +119,7 @@ const makeSocketIOEmitters = (
         })
       );
     },
-    emitRunFacetool: (imageToProcess: InvokeAI.Image) => {
+    emitRunFacetool: (imageToProcess: InvokeAI._Image) => {
       dispatch(setIsProcessing(true));
 
       const {
@@ -150,7 +150,7 @@ const makeSocketIOEmitters = (
         })
       );
     },
-    emitDeleteImage: (imageToDelete: InvokeAI.Image) => {
+    emitDeleteImage: (imageToDelete: InvokeAI._Image) => {
       const { url, uuid, category, thumbnail } = imageToDelete;
       dispatch(removeImage(imageToDelete));
       socketio.emit('deleteImage', url, thumbnail, uuid, category);

invokeai/frontend/web/src/app/socketio/listeners.ts

@@ -34,8 +34,9 @@ import type { RootState } from 'app/store';
 import { addImageToStagingArea } from 'features/canvas/store/canvasSlice';
 import {
   clearInitialImage,
+  initialImageSelected,
   setInfillMethod,
-  setInitialImage,
+  // setInitialImage,
   setMaskPath,
 } from 'features/parameters/store/generationSlice';
 import { tabMap } from 'features/ui/store/tabMap';
@@ -142,15 +143,17 @@ const makeSocketIOListeners = (
           }
         }
 
-        if (shouldLoopback) {
-          const activeTabName = tabMap[activeTab];
-          switch (activeTabName) {
-            case 'img2img': {
-              dispatch(setInitialImage(newImage));
-              break;
-            }
-          }
-        }
+        // TODO: fix
+        // if (shouldLoopback) {
+        //   const activeTabName = tabMap[activeTab];
+        //   switch (activeTabName) {
+        //     case 'img2img': {
+        //       dispatch(initialImageSelected(newImage.uuid));
+        //       // dispatch(setInitialImage(newImage));
+        //       break;
+        //     }
+        //   }
+        // }
 
         dispatch(clearIntermediateImage());
 
@@ -262,7 +265,7 @@ const makeSocketIOListeners = (
        */
 
       // Generate a UUID for each image
-      const preparedImages = images.map((image): InvokeAI.Image => {
+      const preparedImages = images.map((image): InvokeAI._Image => {
         return {
           uuid: uuidv4(),
           ...image,
@@ -334,7 +337,7 @@ const makeSocketIOListeners = (
 
       if (
         initialImage === url ||
-        (initialImage as InvokeAI.Image)?.url === url
+        (initialImage as InvokeAI._Image)?.url === url
       ) {
         dispatch(clearInitialImage());
       }

invokeai/frontend/web/src/app/socketio/middleware.ts

@@ -29,6 +29,8 @@ export const socketioMiddleware = () => {
     path: `${window.location.pathname}socket.io`,
   });
 
+  socketio.disconnect();
+
   let areListenersSet = false;
 
   const middleware: Middleware = (store) => (next) => (action) => {

invokeai/frontend/web/src/app/store.ts

@@ -2,18 +2,32 @@ import { combineReducers, configureStore } from '@reduxjs/toolkit';
 
 import { persistReducer } from 'redux-persist';
 import storage from 'redux-persist/lib/storage'; // defaults to localStorage for web
-
+import dynamicMiddlewares from 'redux-dynamic-middlewares';
 import { getPersistConfig } from 'redux-deep-persist';
 
 import canvasReducer from 'features/canvas/store/canvasSlice';
 import galleryReducer from 'features/gallery/store/gallerySlice';
+import resultsReducer from 'features/gallery/store/resultsSlice';
+import uploadsReducer from 'features/gallery/store/uploadsSlice';
 import lightboxReducer from 'features/lightbox/store/lightboxSlice';
 import generationReducer from 'features/parameters/store/generationSlice';
 import postprocessingReducer from 'features/parameters/store/postprocessingSlice';
 import systemReducer from 'features/system/store/systemSlice';
 import uiReducer from 'features/ui/store/uiSlice';
+import modelsReducer from 'features/system/store/modelSlice';
+import nodesReducer from 'features/nodes/store/nodesSlice';
 
 import { socketioMiddleware } from './socketio/middleware';
+import { socketMiddleware } from 'services/events/middleware';
+import { canvasBlacklist } from 'features/canvas/store/canvasPersistBlacklist';
+import { galleryBlacklist } from 'features/gallery/store/galleryPersistBlacklist';
+import { generationBlacklist } from 'features/parameters/store/generationPersistBlacklist';
+import { lightboxBlacklist } from 'features/lightbox/store/lightboxPersistBlacklist';
+import { modelsBlacklist } from 'features/system/store/modelsPersistBlacklist';
+import { nodesBlacklist } from 'features/nodes/store/nodesPersistBlacklist';
+import { postprocessingBlacklist } from 'features/parameters/store/postprocessingPersistBlacklist';
+import { systemBlacklist } from 'features/system/store/systemPersistsBlacklist';
+import { uiBlacklist } from 'features/ui/store/uiPersistBlacklist';
 
 /**
  * redux-persist provides an easy and reliable way to persist state across reloads.
@@ -29,49 +43,18 @@ import { socketioMiddleware } from './socketio/middleware';
  * The necesssary nested persistors with blacklists are configured below.
  */
 
-const canvasBlacklist = [
-  'cursorPosition',
-  'isCanvasInitialized',
-  'doesCanvasNeedScaling',
-].map((blacklistItem) => `canvas.${blacklistItem}`);
-
-const systemBlacklist = [
-  'currentIteration',
-  'currentStatus',
-  'currentStep',
-  'isCancelable',
-  'isConnected',
-  'isESRGANAvailable',
-  'isGFPGANAvailable',
-  'isProcessing',
-  'socketId',
-  'totalIterations',
-  'totalSteps',
-  'openModel',
-  'cancelOptions.cancelAfter',
-].map((blacklistItem) => `system.${blacklistItem}`);
-
-const galleryBlacklist = [
-  'categories',
-  'currentCategory',
-  'currentImage',
-  'currentImageUuid',
-  'shouldAutoSwitchToNewImages',
-  'intermediateImage',
-].map((blacklistItem) => `gallery.${blacklistItem}`);
-
-const lightboxBlacklist = ['isLightboxOpen'].map(
-  (blacklistItem) => `lightbox.${blacklistItem}`
-);
-
 const rootReducer = combineReducers({
-  generation: generationReducer,
-  postprocessing: postprocessingReducer,
-  gallery: galleryReducer,
-  system: systemReducer,
   canvas: canvasReducer,
-  ui: uiReducer,
+  gallery: galleryReducer,
+  generation: generationReducer,
   lightbox: lightboxReducer,
+  models: modelsReducer,
+  nodes: nodesReducer,
+  postprocessing: postprocessingReducer,
+  results: resultsReducer,
+  system: systemReducer,
+  ui: uiReducer,
+  uploads: uploadsReducer,
 });
 
 const rootPersistConfig = getPersistConfig({
@@ -80,23 +63,40 @@ const rootPersistConfig = getPersistConfig({
   rootReducer,
   blacklist: [
     ...canvasBlacklist,
-    ...systemBlacklist,
     ...galleryBlacklist,
+    ...generationBlacklist,
     ...lightboxBlacklist,
+    ...modelsBlacklist,
+    ...nodesBlacklist,
+    ...postprocessingBlacklist,
+    // ...resultsBlacklist,
+    'results',
+    ...systemBlacklist,
+    ...uiBlacklist,
+    // ...uploadsBlacklist,
+    'uploads',
   ],
   debounce: 300,
 });
 
 const persistedReducer = persistReducer(rootPersistConfig, rootReducer);
 
-// Continue with store setup
+// TODO: rip the old middleware out when nodes is complete
+export function buildMiddleware() {
+  if (import.meta.env.MODE === 'nodes' || import.meta.env.MODE === 'package') {
+    return socketMiddleware();
+  } else {
+    return socketioMiddleware();
+  }
+}
+
 export const store = configureStore({
   reducer: persistedReducer,
   middleware: (getDefaultMiddleware) =>
     getDefaultMiddleware({
       immutableCheck: false,
       serializableCheck: false,
-    }).concat(socketioMiddleware()),
+    }).concat(dynamicMiddlewares),
   devTools: {
     // Uncommenting these very rapidly called actions makes the redux dev tools output much more readable
     actionsDenylist: [