(ci) remove python 3.10 from the test matrix; comment out GPU tests for now

## Summary

- Extend `ModelOnDisk` with caching, type hints, default args
- Fail early if there is an error classifying a config

## Related Issues / Discussions

<!--WHEN APPLICABLE: List any related issues or discussions on github or
discord. If this PR closes an issue, please use the "Closes #1234"
format, so that the issue will be automatically closed when the PR
merges.-->

## QA Instructions

<!--WHEN APPLICABLE: Describe how you have tested the changes in this
PR. Provide enough detail that a reviewer can reproduce your tests.-->

## Merge Plan

<!--WHEN APPLICABLE: Large PRs, or PRs that touch sensitive things like
DB schemas, may need some care when merging. For example, a careful
rebase by the change author, timing to not interfere with a pending
release, or a message to contributors on discord after merging.-->

## Checklist

- [ ] _The PR has a short but descriptive title, suitable for a
changelog_
- [ ] _Tests added / updated (if applicable)_
- [ ] _Documentation added / updated (if applicable)_
- [ ] _Updated `What's New` copy (if doing a release after this PR)_

.git-blame-ignore-revs

@@ -1,2 +1,5 @@
 b3dccfaeb636599c02effc377cdd8a87d658256c
 218b6d0546b990fc449c876fb99f44b50c4daa35
+182580ff6970caed400be178c5b888514b75d7f2
+8e9d5c1187b0d36da80571ce4c8ba9b3a37b6c46
+99aac5870e1092b182e6c5f21abcaab6936a4ad1

.gitattributes

@@ -2,4 +2,5 @@
 # Only affects text files and ignores other file types.
 # For more info see: https://www.aleksandrhovhannisyan.com/blog/crlf-vs-lf-normalizing-line-endings-in-git/
 * text=auto
-docker/** text eol=lf
+docker/** text eol=lf
+tests/test_model_probe/stripped_models/** filter=lfs diff=lfs merge=lfs -text

.github/workflows/frontend-checks.yml

@@ -44,7 +44,12 @@ jobs:
       - name: check for changed frontend files
         if: ${{ inputs.always_run != true }}
         id: changed-files
-        uses: tj-actions/changed-files@v42
+        # Pinned to the _hash_ for v45.0.9 to prevent supply-chain attacks.
+        # See:
+        # - CVE-2025-30066
+        # - https://www.stepsecurity.io/blog/harden-runner-detection-tj-actions-changed-files-action-is-compromised
+        # - https://github.com/tj-actions/changed-files/issues/2463
+        uses: tj-actions/changed-files@a284dc1814e3fd07f2e34267fc8f81227ed29fb8
         with:
           files_yaml: |
             frontend:

.github/workflows/frontend-tests.yml

@@ -44,7 +44,12 @@ jobs:
       - name: check for changed frontend files
         if: ${{ inputs.always_run != true }}
         id: changed-files
-        uses: tj-actions/changed-files@v42
+        # Pinned to the _hash_ for v45.0.9 to prevent supply-chain attacks.
+        # See:
+        # - CVE-2025-30066
+        # - https://www.stepsecurity.io/blog/harden-runner-detection-tj-actions-changed-files-action-is-compromised
+        # - https://github.com/tj-actions/changed-files/issues/2463
+        uses: tj-actions/changed-files@a284dc1814e3fd07f2e34267fc8f81227ed29fb8
         with:
           files_yaml: |
             frontend:

.github/workflows/python-checks.yml

@@ -43,7 +43,12 @@ jobs:
       - name: check for changed python files
         if: ${{ inputs.always_run != true }}
         id: changed-files
-        uses: tj-actions/changed-files@v42
+        # Pinned to the _hash_ for v45.0.9 to prevent supply-chain attacks.
+        # See:
+        # - CVE-2025-30066
+        # - https://www.stepsecurity.io/blog/harden-runner-detection-tj-actions-changed-files-action-is-compromised
+        # - https://github.com/tj-actions/changed-files/issues/2463
+        uses: tj-actions/changed-files@a284dc1814e3fd07f2e34267fc8f81227ed29fb8
         with:
           files_yaml: |
             python:
@@ -56,13 +61,13 @@ jobs:
         if: ${{ steps.changed-files.outputs.python_any_changed == 'true' || inputs.always_run == true }}
         uses: actions/setup-python@v5
         with:
-          python-version: '3.10'
+          python-version: '3.12'
           cache: pip
           cache-dependency-path: pyproject.toml
 
       - name: install ruff
         if: ${{ steps.changed-files.outputs.python_any_changed == 'true' || inputs.always_run == true }}
-        run: pip install ruff==0.9.9
+        run: pip install ruff==0.11.2
         shell: bash
 
       - name: ruff check

.github/workflows/python-tests.yml

@@ -39,26 +39,25 @@ jobs:
     strategy:
       matrix:
         python-version:
-          - '3.10'
           - '3.11'
         platform:
-          - linux-cuda-11_7
-          - linux-rocm-5_2
+          # - linux-cuda-12_6
+          # - linux-rocm-6_2
           - linux-cpu
           - macos-default
           - windows-cpu
         include:
-          - platform: linux-cuda-11_7
-            os: ubuntu-22.04
-            github-env: $GITHUB_ENV
-          - platform: linux-rocm-5_2
-            os: ubuntu-22.04
-            extra-index-url: 'https://download.pytorch.org/whl/rocm5.2'
-            github-env: $GITHUB_ENV
+          # - platform: linux-cuda-12_6
+          #   os: ubuntu-24.04
+          #   github-env: $GITHUB_ENV
+          # - platform: linux-rocm-6_2
+          #   os: ubuntu-24.04
+          #   extra-index-url: 'https://download.pytorch.org/whl/rocm6.2'
+          #   github-env: $GITHUB_ENV
           - platform: linux-cpu
-            os: ubuntu-22.04
-            extra-index-url: 'https://download.pytorch.org/whl/cpu'
+            os: ubuntu-24.04
             github-env: $GITHUB_ENV
+            extra-index-url: 'https://download.pytorch.org/whl/cpu'
           - platform: macos-default
             os: macOS-14
             github-env: $GITHUB_ENV
@@ -72,12 +71,18 @@ jobs:
       PIP_USE_PEP517: '1'
     steps:
       - name: checkout
-        uses: actions/checkout@v4
+        # https://github.com/nschloe/action-cached-lfs-checkout
+        uses: nschloe/action-cached-lfs-checkout@f46300cd8952454b9f0a21a3d133d4bd5684cfc2
 
       - name: check for changed python files
         if: ${{ inputs.always_run != true }}
         id: changed-files
-        uses: tj-actions/changed-files@v42
+        # Pinned to the _hash_ for v45.0.9 to prevent supply-chain attacks.
+        # See:
+        # - CVE-2025-30066
+        # - https://www.stepsecurity.io/blog/harden-runner-detection-tj-actions-changed-files-action-is-compromised
+        # - https://github.com/tj-actions/changed-files/issues/2463
+        uses: tj-actions/changed-files@a284dc1814e3fd07f2e34267fc8f81227ed29fb8
         with:
           files_yaml: |
             python:

.github/workflows/typegen-checks.yml

@@ -42,7 +42,12 @@ jobs:
       - name: check for changed files
         if: ${{ inputs.always_run != true }}
         id: changed-files
-        uses: tj-actions/changed-files@v42
+        # Pinned to the _hash_ for v45.0.9 to prevent supply-chain attacks.
+        # See:
+        # - CVE-2025-30066
+        # - https://www.stepsecurity.io/blog/harden-runner-detection-tj-actions-changed-files-action-is-compromised
+        # - https://github.com/tj-actions/changed-files/issues/2463
+        uses: tj-actions/changed-files@a284dc1814e3fd07f2e34267fc8f81227ed29fb8
         with:
           files_yaml: |
             src:

docs/contributing/dev-environment.md

@@ -18,9 +18,19 @@ If you just want to use Invoke, you should use the [launcher][launcher link].
 
 2. [Fork and clone][forking link] the [InvokeAI repo][repo link].
 
-3. Create an directory for user data (images, models, db, etc). This is typically at `~/invokeai`, but if you already have a non-dev install, you may want to create a separate directory for the dev install.
+3. This repository uses Git LFS to manage large files. To ensure all assets are downloaded:
+      - Install git-lfs → [Download here](https://git-lfs.com/)
+      - Enable automatic LFS fetching for this repository:
+        ```shell
+        git config lfs.fetchinclude "*"
+        ```
+        - Fetch files from LFS (only needs to be done once; subsequent `git pull` will fetch changes automatically):
+        ```
+        git lfs pull
+        ```
+4. Create an directory for user data (images, models, db, etc). This is typically at `~/invokeai`, but if you already have a non-dev install, you may want to create a separate directory for the dev install.
 
-4. Follow the [manual install][manual install link] guide, with some modifications to the install command:
+5. Follow the [manual install][manual install link] guide, with some modifications to the install command:
 
       - Use `.` instead of `invokeai` to install from the current directory. You don't need to specify the version.
 
@@ -34,19 +44,19 @@ If you just want to use Invoke, you should use the [launcher][launcher link].
       uv pip install -e ".[dev,test,docs,xformers]" --python 3.11 --python-preference only-managed --index=https://download.pytorch.org/whl/cu124 --reinstall
       ```
 
-5. At this point, you should have Invoke installed, a venv set up and activated, and the server running. But you will see a warning in the terminal that no UI was found. If you go to the URL for the server, you won't get a UI.
+6. At this point, you should have Invoke installed, a venv set up and activated, and the server running. But you will see a warning in the terminal that no UI was found. If you go to the URL for the server, you won't get a UI.
 
       This is because the UI build is not distributed with the source code. You need to build it manually. End the running server instance.
 
       If you only want to edit the docs, you can stop here and skip to the **Documentation** section below.
 
-6. Install the frontend dev toolchain:
+7. Install the frontend dev toolchain:
 
       - [`nodejs`](https://nodejs.org/) (v20+)
 
       - [`pnpm`](https://pnpm.io/8.x/installation) (must be v8 - not v9!)
 
-7. Do a production build of the frontend:
+8. Do a production build of the frontend:
 
       ```sh
       cd <PATH_TO_INVOKEAI_REPO>/invokeai/frontend/web
@@ -54,7 +64,7 @@ If you just want to use Invoke, you should use the [launcher][launcher link].
       pnpm build
       ```
 
-8. Restart the server and navigate to the URL. You should get a UI. After making changes to the python code, restart the server to see those changes.
+9. Restart the server and navigate to the URL. You should get a UI. After making changes to the python code, restart the server to see those changes.
 
 ## Updating the UI
 

invokeai/app/api/routers/app_info.py

@@ -12,6 +12,7 @@ from pydantic import BaseModel, Field
 
 from invokeai.app.api.dependencies import ApiDependencies
 from invokeai.app.invocations.upscale import ESRGAN_MODELS
+from invokeai.app.services.config.config_default import InvokeAIAppConfig, get_config
 from invokeai.app.services.invocation_cache.invocation_cache_common import InvocationCacheStatus
 from invokeai.backend.image_util.infill_methods.patchmatch import PatchMatch
 from invokeai.backend.util.logging import logging
@@ -99,7 +100,7 @@ async def get_app_deps() -> AppDependencyVersions:
 
 
 @app_router.get("/config", operation_id="get_config", status_code=200, response_model=AppConfig)
-async def get_config() -> AppConfig:
+async def get_config_() -> AppConfig:
     infill_methods = ["lama", "tile", "cv2", "color"]  # TODO: add mosaic back
     if PatchMatch.patchmatch_available():
         infill_methods.append("patchmatch")
@@ -121,6 +122,21 @@ async def get_config() -> AppConfig:
     )
 
 
+class InvokeAIAppConfigWithSetFields(BaseModel):
+    """InvokeAI App Config with model fields set"""
+
+    set_fields: set[str] = Field(description="The set fields")
+    config: InvokeAIAppConfig = Field(description="The InvokeAI App Config")
+
+
+@app_router.get(
+    "/runtime_config", operation_id="get_runtime_config", status_code=200, response_model=InvokeAIAppConfigWithSetFields
+)
+async def get_runtime_config() -> InvokeAIAppConfigWithSetFields:
+    config = get_config()
+    return InvokeAIAppConfigWithSetFields(set_fields=config.model_fields_set, config=config)
+
+
 @app_router.get(
     "/logging",
     operation_id="get_log_level",

invokeai/app/api/routers/images.py

@@ -96,6 +96,22 @@ async def upload_image(
         raise HTTPException(status_code=500, detail="Failed to create image")
 
 
+class ImageUploadEntry(BaseModel):
+    image_dto: ImageDTO = Body(description="The image DTO")
+    presigned_url: str = Body(description="The URL to get the presigned URL for the image upload")
+
+
+@images_router.post("/", operation_id="create_image_upload_entry")
+async def create_image_upload_entry(
+    width: int = Body(description="The width of the image"),
+    height: int = Body(description="The height of the image"),
+    board_id: Optional[str] = Body(default=None, description="The board to add this image to, if any"),
+) -> ImageUploadEntry:
+    """Uploads an image from a URL, not implemented"""
+
+    raise HTTPException(status_code=501, detail="Not implemented")
+
+
 @images_router.delete("/i/{image_name}", operation_id="delete_image")
 async def delete_image(
     image_name: str = Path(description="The name of the image to delete"),

invokeai/app/api/routers/model_manager.py

@@ -28,12 +28,10 @@ from invokeai.app.services.model_records import (
     UnknownModelException,
 )
 from invokeai.app.util.suppress_output import SuppressOutput
+from invokeai.backend.model_manager import BaseModelType, ModelFormat, ModelType
 from invokeai.backend.model_manager.config import (
     AnyModelConfig,
-    BaseModelType,
     MainCheckpointConfig,
-    ModelFormat,
-    ModelType,
 )
 from invokeai.backend.model_manager.load.model_cache.cache_stats import CacheStats
 from invokeai.backend.model_manager.metadata.fetch.huggingface import HuggingFaceMetadataFetch

invokeai/app/api/routers/workflows.py

@@ -105,6 +105,7 @@ async def list_workflows(
     categories: Optional[list[WorkflowCategory]] = Query(default=None, description="The categories of workflow to get"),
     tags: Optional[list[str]] = Query(default=None, description="The tags of workflow to get"),
     query: Optional[str] = Query(default=None, description="The text to query by (matches name and description)"),
+    has_been_opened: Optional[bool] = Query(default=None, description="Whether to include/exclude recent workflows"),
 ) -> PaginatedResults[WorkflowRecordListItemWithThumbnailDTO]:
     """Gets a page of workflows"""
     workflows_with_thumbnails: list[WorkflowRecordListItemWithThumbnailDTO] = []
@@ -116,6 +117,7 @@ async def list_workflows(
         query=query,
         categories=categories,
         tags=tags,
+        has_been_opened=has_been_opened,
     )
     for workflow in workflows.items:
         workflows_with_thumbnails.append(
@@ -221,14 +223,29 @@ async def get_workflow_thumbnail(
         raise HTTPException(status_code=404)
 
 
-@workflows_router.get("/counts", operation_id="get_counts")
-async def get_counts(
-    tags: Optional[list[str]] = Query(default=None, description="The tags to include"),
+@workflows_router.get("/counts_by_tag", operation_id="get_counts_by_tag")
+async def get_counts_by_tag(
+    tags: list[str] = Query(description="The tags to get counts for"),
     categories: Optional[list[WorkflowCategory]] = Query(default=None, description="The categories to include"),
-) -> int:
-    """Gets a the count of workflows that include the specified tags and categories"""
+    has_been_opened: Optional[bool] = Query(default=None, description="Whether to include/exclude recent workflows"),
+) -> dict[str, int]:
+    """Counts workflows by tag"""
 
-    return ApiDependencies.invoker.services.workflow_records.get_counts(tags=tags, categories=categories)
+    return ApiDependencies.invoker.services.workflow_records.counts_by_tag(
+        tags=tags, categories=categories, has_been_opened=has_been_opened
+    )
+
+
+@workflows_router.get("/counts_by_category", operation_id="counts_by_category")
+async def counts_by_category(
+    categories: list[WorkflowCategory] = Query(description="The categories to include"),
+    has_been_opened: Optional[bool] = Query(default=None, description="Whether to include/exclude recent workflows"),
+) -> dict[str, int]:
+    """Counts workflows by category"""
+
+    return ApiDependencies.invoker.services.workflow_records.counts_by_category(
+        categories=categories, has_been_opened=has_been_opened
+    )
 
 
 @workflows_router.put(

invokeai/app/invocations/compel.py

@@ -40,10 +40,10 @@ from invokeai.backend.util.devices import TorchDevice
 
 @invocation(
     "compel",
-    title="Prompt",
+    title="Prompt - SD1.5",
     tags=["prompt", "compel"],
     category="conditioning",
-    version="1.2.0",
+    version="1.2.1",
 )
 class CompelInvocation(BaseInvocation):
     """Parse prompt using compel package to conditioning."""
@@ -233,10 +233,10 @@ class SDXLPromptInvocationBase:
 
 @invocation(
     "sdxl_compel_prompt",
-    title="SDXL Prompt",
+    title="Prompt - SDXL",
     tags=["sdxl", "compel", "prompt"],
     category="conditioning",
-    version="1.2.0",
+    version="1.2.1",
 )
 class SDXLCompelPromptInvocation(BaseInvocation, SDXLPromptInvocationBase):
     """Parse prompt using compel package to conditioning."""
@@ -327,10 +327,10 @@ class SDXLCompelPromptInvocation(BaseInvocation, SDXLPromptInvocationBase):
 
 @invocation(
     "sdxl_refiner_compel_prompt",
-    title="SDXL Refiner Prompt",
+    title="Prompt - SDXL Refiner",
     tags=["sdxl", "compel", "prompt"],
     category="conditioning",
-    version="1.1.1",
+    version="1.1.2",
 )
 class SDXLRefinerCompelPromptInvocation(BaseInvocation, SDXLPromptInvocationBase):
     """Parse prompt using compel package to conditioning."""
@@ -376,10 +376,10 @@ class CLIPSkipInvocationOutput(BaseInvocationOutput):
 
 @invocation(
     "clip_skip",
-    title="CLIP Skip",
+    title="Apply CLIP Skip - SD1.5, SDXL",
     tags=["clipskip", "clip", "skip"],
     category="conditioning",
-    version="1.1.0",
+    version="1.1.1",
 )
 class CLIPSkipInvocation(BaseInvocation):
     """Skip layers in clip text_encoder model."""

invokeai/app/invocations/controlnet_image_processors.py

@@ -87,7 +87,7 @@ class ControlOutput(BaseInvocationOutput):
     control: ControlField = OutputField(description=FieldDescriptions.control)
 
 
-@invocation("controlnet", title="ControlNet", tags=["controlnet"], category="controlnet", version="1.1.2")
+@invocation("controlnet", title="ControlNet - SD1.5, SDXL", tags=["controlnet"], category="controlnet", version="1.1.3")
 class ControlNetInvocation(BaseInvocation):
     """Collects ControlNet info to pass to other nodes"""
 

invokeai/app/invocations/create_gradient_mask.py

@@ -19,7 +19,8 @@ from invokeai.app.invocations.image_to_latents import ImageToLatentsInvocation
 from invokeai.app.invocations.model import UNetField, VAEField
 from invokeai.app.services.shared.invocation_context import InvocationContext
 from invokeai.backend.model_manager import LoadedModel
-from invokeai.backend.model_manager.config import MainConfigBase, ModelVariantType
+from invokeai.backend.model_manager.config import MainConfigBase
+from invokeai.backend.model_manager.taxonomy import ModelVariantType
 from invokeai.backend.stable_diffusion.diffusers_pipeline import image_resized_to_grid_as_tensor
 
 

invokeai/app/invocations/denoise_latents.py

@@ -39,8 +39,8 @@ from invokeai.app.invocations.t2i_adapter import T2IAdapterField
 from invokeai.app.services.shared.invocation_context import InvocationContext
 from invokeai.app.util.controlnet_utils import prepare_control_image
 from invokeai.backend.ip_adapter.ip_adapter import IPAdapter
-from invokeai.backend.model_manager import BaseModelType, ModelVariantType
 from invokeai.backend.model_manager.config import AnyModelConfig
+from invokeai.backend.model_manager.taxonomy import BaseModelType, ModelVariantType
 from invokeai.backend.model_patcher import ModelPatcher
 from invokeai.backend.patches.layer_patcher import LayerPatcher
 from invokeai.backend.patches.model_patch_raw import ModelPatchRaw
@@ -127,10 +127,10 @@ def get_scheduler(
 
 @invocation(
     "denoise_latents",
-    title="Denoise Latents",
+    title="Denoise - SD1.5, SDXL",
     tags=["latents", "denoise", "txt2img", "t2i", "t2l", "img2img", "i2i", "l2l"],
     category="latents",
-    version="1.5.3",
+    version="1.5.4",
 )
 class DenoiseLatentsInvocation(BaseInvocation):
     """Denoises noisy latents to decodable images"""

invokeai/app/invocations/fields.py

@@ -59,6 +59,7 @@ class UIType(str, Enum, metaclass=MetaEnum):
     ControlLoRAModel = "ControlLoRAModelField"
     SigLipModel = "SigLipModelField"
     FluxReduxModel = "FluxReduxModelField"
+    LlavaOnevisionModel = "LLaVAModelField"
     # endregion
 
     # region Misc Field Types
@@ -205,6 +206,8 @@ class FieldDescriptions:
     freeu_b2 = "Scaling factor for stage 2 to amplify the contributions of backbone features."
     instantx_control_mode = "The control mode for InstantX ControlNet union models. Ignored for other ControlNet models. The standard mapping is: canny (0), tile (1), depth (2), blur (3), pose (4), gray (5), low quality (6). Negative values will be treated as 'None'."
     flux_redux_conditioning = "FLUX Redux conditioning tensor"
+    vllm_model = "The VLLM model to use"
+    flux_fill_conditioning = "FLUX Fill conditioning tensor"
 
 
 class ImageField(BaseModel):
@@ -274,6 +277,13 @@ class FluxReduxConditioningField(BaseModel):
     )
 
 
+class FluxFillConditioningField(BaseModel):
+    """A FLUX Fill conditioning field."""
+
+    image: ImageField = Field(description="The FLUX Fill reference image.")
+    mask: TensorField = Field(description="The FLUX Fill inpaint mask.")
+
+
 class SD3ConditioningField(BaseModel):
     """A conditioning tensor primitive value"""
 

invokeai/app/invocations/flux_control_lora_loader.py

@@ -1,7 +1,6 @@
 from invokeai.app.invocations.baseinvocation import (
     BaseInvocation,
     BaseInvocationOutput,
-    Classification,
     invocation,
     invocation_output,
 )
@@ -21,11 +20,10 @@ class FluxControlLoRALoaderOutput(BaseInvocationOutput):
 
 @invocation(
     "flux_control_lora_loader",
-    title="Flux Control LoRA",
+    title="Control LoRA - FLUX",
     tags=["lora", "model", "flux"],
     category="model",
-    version="1.1.0",
-    classification=Classification.Prototype,
+    version="1.1.1",
 )
 class FluxControlLoRALoaderInvocation(BaseInvocation):
     """LoRA model and Image to use with FLUX transformer generation."""

invokeai/app/invocations/flux_controlnet.py

@@ -3,7 +3,6 @@ from pydantic import BaseModel, Field, field_validator, model_validator
 from invokeai.app.invocations.baseinvocation import (
     BaseInvocation,
     BaseInvocationOutput,
-    Classification,
     invocation,
     invocation_output,
 )
@@ -52,7 +51,6 @@ class FluxControlNetOutput(BaseInvocationOutput):
     tags=["controlnet", "flux"],
     category="controlnet",
     version="1.0.0",
-    classification=Classification.Prototype,
 )
 class FluxControlNetInvocation(BaseInvocation):
     """Collect FLUX ControlNet info to pass to other nodes."""

invokeai/app/invocations/flux_denoise.py

@@ -10,11 +10,12 @@ from PIL import Image
 from torchvision.transforms.functional import resize as tv_resize
 from transformers import CLIPImageProcessor, CLIPVisionModelWithProjection
 
-from invokeai.app.invocations.baseinvocation import BaseInvocation, Classification, invocation
+from invokeai.app.invocations.baseinvocation import BaseInvocation, invocation
 from invokeai.app.invocations.fields import (
     DenoiseMaskField,
     FieldDescriptions,
     FluxConditioningField,
+    FluxFillConditioningField,
     FluxReduxConditioningField,
     ImageField,
     Input,
@@ -48,7 +49,7 @@ from invokeai.backend.flux.sampling_utils import (
     unpack,
 )
 from invokeai.backend.flux.text_conditioning import FluxReduxConditioning, FluxTextConditioning
-from invokeai.backend.model_manager.config import ModelFormat
+from invokeai.backend.model_manager.taxonomy import ModelFormat, ModelVariantType
 from invokeai.backend.patches.layer_patcher import LayerPatcher
 from invokeai.backend.patches.lora_conversions.flux_lora_constants import FLUX_LORA_TRANSFORMER_PREFIX
 from invokeai.backend.patches.model_patch_raw import ModelPatchRaw
@@ -62,8 +63,7 @@ from invokeai.backend.util.devices import TorchDevice
     title="FLUX Denoise",
     tags=["image", "flux"],
     category="image",
-    version="3.2.3",
-    classification=Classification.Prototype,
+    version="3.3.0",
 )
 class FluxDenoiseInvocation(BaseInvocation, WithMetadata, WithBoard):
     """Run denoising process with a FLUX transformer model."""
@@ -109,6 +109,11 @@ class FluxDenoiseInvocation(BaseInvocation, WithMetadata, WithBoard):
         description="FLUX Redux conditioning tensor.",
         input=Input.Connection,
     )
+    fill_conditioning: FluxFillConditioningField | None = InputField(
+        default=None,
+        description="FLUX Fill conditioning.",
+        input=Input.Connection,
+    )
     cfg_scale: float | list[float] = InputField(default=1.0, description=FieldDescriptions.cfg_scale, title="CFG Scale")
     cfg_scale_start_step: int = InputField(
         default=0,
@@ -261,8 +266,19 @@ class FluxDenoiseInvocation(BaseInvocation, WithMetadata, WithBoard):
         if is_schnell and self.control_lora:
             raise ValueError("Control LoRAs cannot be used with FLUX Schnell")
 
-        # Prepare the extra image conditioning tensor if a FLUX structural control image is provided.
-        img_cond = self._prep_structural_control_img_cond(context)
+        # Prepare the extra image conditioning tensor (img_cond) for either FLUX structural control or FLUX Fill.
+        img_cond: torch.Tensor | None = None
+        is_flux_fill = transformer_config.variant == ModelVariantType.Inpaint  # type: ignore
+        if is_flux_fill:
+            img_cond = self._prep_flux_fill_img_cond(
+                context, device=TorchDevice.choose_torch_device(), dtype=inference_dtype
+            )
+        else:
+            if self.fill_conditioning is not None:
+                raise ValueError("fill_conditioning was provided, but the model is not a FLUX Fill model.")
+
+            if self.control_lora is not None:
+                img_cond = self._prep_structural_control_img_cond(context)
 
         inpaint_mask = self._prep_inpaint_mask(context, x)
 
@@ -271,7 +287,6 @@ class FluxDenoiseInvocation(BaseInvocation, WithMetadata, WithBoard):
         # Pack all latent tensors.
         init_latents = pack(init_latents) if init_latents is not None else None
         inpaint_mask = pack(inpaint_mask) if inpaint_mask is not None else None
-        img_cond = pack(img_cond) if img_cond is not None else None
         noise = pack(noise)
         x = pack(x)
 
@@ -664,7 +679,70 @@ class FluxDenoiseInvocation(BaseInvocation, WithMetadata, WithBoard):
         img_cond = einops.rearrange(img_cond, "h w c -> 1 c h w")
 
         vae_info = context.models.load(self.controlnet_vae.vae)
-        return FluxVaeEncodeInvocation.vae_encode(vae_info=vae_info, image_tensor=img_cond)
+        img_cond = FluxVaeEncodeInvocation.vae_encode(vae_info=vae_info, image_tensor=img_cond)
+
+        return pack(img_cond)
+
+    def _prep_flux_fill_img_cond(
+        self, context: InvocationContext, device: torch.device, dtype: torch.dtype
+    ) -> torch.Tensor:
+        """Prepare the FLUX Fill conditioning. This method should be called iff the model is a FLUX Fill model.
+
+        This logic is based on:
+        https://github.com/black-forest-labs/flux/blob/716724eb276d94397be99710a0a54d352664e23b/src/flux/sampling.py#L107-L157
+        """
+        # Validate inputs.
+        if self.fill_conditioning is None:
+            raise ValueError("A FLUX Fill model is being used without fill_conditioning.")
+        # TODO(ryand): We should probable rename controlnet_vae. It's used for more than just ControlNets.
+        if self.controlnet_vae is None:
+            raise ValueError("A FLUX Fill model is being used without controlnet_vae.")
+        if self.control_lora is not None:
+            raise ValueError(
+                "A FLUX Fill model is being used, but a control_lora was provided. Control LoRAs are not compatible with FLUX Fill models."
+            )
+
+        # Log input warnings related to FLUX Fill usage.
+        if self.denoise_mask is not None:
+            context.logger.warning(
+                "Both fill_conditioning and a denoise_mask were provided. You probably meant to use one or the other."
+            )
+        if self.guidance < 25.0:
+            context.logger.warning("A guidance value of ~30.0 is recommended for FLUX Fill models.")
+
+        # Load the conditioning image and resize it to the target image size.
+        cond_img = context.images.get_pil(self.fill_conditioning.image.image_name, mode="RGB")
+        cond_img = cond_img.resize((self.width, self.height), Image.Resampling.BICUBIC)
+        cond_img = np.array(cond_img)
+        cond_img = torch.from_numpy(cond_img).float() / 127.5 - 1.0
+        cond_img = einops.rearrange(cond_img, "h w c -> 1 c h w")
+        cond_img = cond_img.to(device=device, dtype=dtype)
+
+        # Load the mask and resize it to the target image size.
+        mask = context.tensors.load(self.fill_conditioning.mask.tensor_name)
+        # We expect mask to be a bool tensor with shape [1, H, W].
+        assert mask.dtype == torch.bool
+        assert mask.dim() == 3
+        assert mask.shape[0] == 1
+        mask = tv_resize(mask, size=[self.height, self.width], interpolation=tv_transforms.InterpolationMode.NEAREST)
+        mask = mask.to(device=device, dtype=dtype)
+        mask = einops.rearrange(mask, "1 h w -> 1 1 h w")
+
+        # Prepare image conditioning.
+        cond_img = cond_img * (1 - mask)
+        vae_info = context.models.load(self.controlnet_vae.vae)
+        cond_img = FluxVaeEncodeInvocation.vae_encode(vae_info=vae_info, image_tensor=cond_img)
+        cond_img = pack(cond_img)
+
+        # Prepare mask conditioning.
+        mask = mask[:, 0, :, :]
+        # Rearrange mask to a 16-channel representation that matches the shape of the VAE-encoded latent space.
+        mask = einops.rearrange(mask, "b (h ph) (w pw) -> b (ph pw) h w", ph=8, pw=8)
+        mask = pack(mask)
+
+        # Merge image and mask conditioning.
+        img_cond = torch.cat((cond_img, mask), dim=-1)
+        return img_cond
 
     def _normalize_ip_adapter_fields(self) -> list[IPAdapterField]:
         if self.ip_adapter is None:

invokeai/app/invocations/flux_fill.py

@@ -0,0 +1,46 @@
+from invokeai.app.invocations.baseinvocation import (
+    BaseInvocation,
+    BaseInvocationOutput,
+    Classification,
+    invocation,
+    invocation_output,
+)
+from invokeai.app.invocations.fields import (
+    FieldDescriptions,
+    FluxFillConditioningField,
+    InputField,
+    OutputField,
+    TensorField,
+)
+from invokeai.app.invocations.primitives import ImageField
+from invokeai.app.services.shared.invocation_context import InvocationContext
+
+
+@invocation_output("flux_fill_output")
+class FluxFillOutput(BaseInvocationOutput):
+    """The conditioning output of a FLUX Fill invocation."""
+
+    fill_cond: FluxFillConditioningField = OutputField(
+        description=FieldDescriptions.flux_redux_conditioning, title="Conditioning"
+    )
+
+
+@invocation(
+    "flux_fill",
+    title="FLUX Fill Conditioning",
+    tags=["inpaint"],
+    category="inpaint",
+    version="1.0.0",
+    classification=Classification.Beta,
+)
+class FluxFillInvocation(BaseInvocation):
+    """Prepare the FLUX Fill conditioning data."""
+
+    image: ImageField = InputField(description="The FLUX Fill reference image.")
+    mask: TensorField = InputField(
+        description="The bool inpainting mask. Excluded regions should be set to "
+        "False, included regions should be set to True.",
+    )
+
+    def invoke(self, context: InvocationContext) -> FluxFillOutput:
+        return FluxFillOutput(fill_cond=FluxFillConditioningField(image=self.image, mask=self.mask))

invokeai/app/invocations/flux_ip_adapter.py

@@ -4,7 +4,7 @@ from typing import List, Literal, Union
 from pydantic import field_validator, model_validator
 from typing_extensions import Self
 
-from invokeai.app.invocations.baseinvocation import BaseInvocation, Classification, invocation
+from invokeai.app.invocations.baseinvocation import BaseInvocation, invocation
 from invokeai.app.invocations.fields import InputField, UIType
 from invokeai.app.invocations.ip_adapter import (
     CLIP_VISION_MODEL_MAP,
@@ -28,7 +28,6 @@ from invokeai.backend.model_manager.config import (
     tags=["ip_adapter", "control"],
     category="ip_adapter",
     version="1.0.0",
-    classification=Classification.Prototype,
 )
 class FluxIPAdapterInvocation(BaseInvocation):
     """Collects FLUX IP-Adapter info to pass to other nodes."""

invokeai/app/invocations/flux_lora_loader.py

@@ -3,14 +3,13 @@ from typing import Optional
 from invokeai.app.invocations.baseinvocation import (
     BaseInvocation,
     BaseInvocationOutput,
-    Classification,
     invocation,
     invocation_output,
 )
 from invokeai.app.invocations.fields import FieldDescriptions, Input, InputField, OutputField, UIType
 from invokeai.app.invocations.model import CLIPField, LoRAField, ModelIdentifierField, T5EncoderField, TransformerField
 from invokeai.app.services.shared.invocation_context import InvocationContext
-from invokeai.backend.model_manager.config import BaseModelType
+from invokeai.backend.model_manager.taxonomy import BaseModelType
 
 
 @invocation_output("flux_lora_loader_output")
@@ -28,11 +27,10 @@ class FluxLoRALoaderOutput(BaseInvocationOutput):
 
 @invocation(
     "flux_lora_loader",
-    title="FLUX LoRA",
+    title="Apply LoRA - FLUX",
     tags=["lora", "model", "flux"],
     category="model",
-    version="1.2.0",
-    classification=Classification.Prototype,
+    version="1.2.1",
 )
 class FluxLoRALoaderInvocation(BaseInvocation):
     """Apply a LoRA model to a FLUX transformer and/or text encoder."""
@@ -107,11 +105,10 @@ class FluxLoRALoaderInvocation(BaseInvocation):
 
 @invocation(
     "flux_lora_collection_loader",
-    title="FLUX LoRA Collection Loader",
+    title="Apply LoRA Collection - FLUX",
     tags=["lora", "model", "flux"],
     category="model",
-    version="1.3.0",
-    classification=Classification.Prototype,
+    version="1.3.1",
 )
 class FLUXLoRACollectionLoader(BaseInvocation):
     """Applies a collection of LoRAs to a FLUX transformer."""

invokeai/app/invocations/flux_model_loader.py

@@ -3,7 +3,6 @@ from typing import Literal
 from invokeai.app.invocations.baseinvocation import (
     BaseInvocation,
     BaseInvocationOutput,
-    Classification,
     invocation,
     invocation_output,
 )
@@ -17,8 +16,8 @@ from invokeai.app.util.t5_model_identifier import (
 from invokeai.backend.flux.util import max_seq_lengths
 from invokeai.backend.model_manager.config import (
     CheckpointConfigBase,
-    SubModelType,
 )
+from invokeai.backend.model_manager.taxonomy import SubModelType
 
 
 @invocation_output("flux_model_loader_output")
@@ -37,11 +36,10 @@ class FluxModelLoaderOutput(BaseInvocationOutput):
 
 @invocation(
     "flux_model_loader",
-    title="Flux Main Model",
+    title="Main Model - FLUX",
     tags=["model", "flux"],
     category="model",
-    version="1.0.5",
-    classification=Classification.Prototype,
+    version="1.0.6",
 )
 class FluxModelLoaderInvocation(BaseInvocation):
     """Loads a flux base model, outputting its submodels."""

invokeai/app/invocations/flux_redux.py

@@ -23,7 +23,8 @@ from invokeai.app.invocations.primitives import ImageField
 from invokeai.app.services.model_records.model_records_base import ModelRecordChanges
 from invokeai.app.services.shared.invocation_context import InvocationContext
 from invokeai.backend.flux.redux.flux_redux_model import FluxReduxModel
-from invokeai.backend.model_manager.config import AnyModelConfig, BaseModelType, ModelType
+from invokeai.backend.model_manager import BaseModelType, ModelType
+from invokeai.backend.model_manager.config import AnyModelConfig
 from invokeai.backend.model_manager.starter_models import siglip
 from invokeai.backend.sig_lip.sig_lip_pipeline import SigLipPipeline
 from invokeai.backend.util.devices import TorchDevice
@@ -44,7 +45,7 @@ class FluxReduxOutput(BaseInvocationOutput):
     tags=["ip_adapter", "control"],
     category="ip_adapter",
     version="2.0.0",
-    classification=Classification.Prototype,
+    classification=Classification.Beta,
 )
 class FluxReduxInvocation(BaseInvocation):
     """Runs a FLUX Redux model to generate a conditioning tensor."""

invokeai/app/invocations/flux_text_encoder.py

@@ -4,7 +4,7 @@ from typing import Iterator, Literal, Optional, Tuple
 import torch
 from transformers import CLIPTextModel, CLIPTokenizer, T5EncoderModel, T5Tokenizer, T5TokenizerFast
 
-from invokeai.app.invocations.baseinvocation import BaseInvocation, Classification, invocation
+from invokeai.app.invocations.baseinvocation import BaseInvocation, invocation
 from invokeai.app.invocations.fields import (
     FieldDescriptions,
     FluxConditioningField,
@@ -17,7 +17,7 @@ from invokeai.app.invocations.model import CLIPField, T5EncoderField
 from invokeai.app.invocations.primitives import FluxConditioningOutput
 from invokeai.app.services.shared.invocation_context import InvocationContext
 from invokeai.backend.flux.modules.conditioner import HFEncoder
-from invokeai.backend.model_manager.config import ModelFormat
+from invokeai.backend.model_manager import ModelFormat
 from invokeai.backend.patches.layer_patcher import LayerPatcher
 from invokeai.backend.patches.lora_conversions.flux_lora_constants import FLUX_LORA_CLIP_PREFIX, FLUX_LORA_T5_PREFIX
 from invokeai.backend.patches.model_patch_raw import ModelPatchRaw
@@ -26,11 +26,10 @@ from invokeai.backend.stable_diffusion.diffusion.conditioning_data import Condit
 
 @invocation(
     "flux_text_encoder",
-    title="FLUX Text Encoding",
+    title="Prompt - FLUX",
     tags=["prompt", "conditioning", "flux"],
     category="conditioning",
-    version="1.1.1",
-    classification=Classification.Prototype,
+    version="1.1.2",
 )
 class FluxTextEncoderInvocation(BaseInvocation):
     """Encodes and preps a prompt for a flux image."""

invokeai/app/invocations/flux_vae_decode.py

@@ -22,10 +22,10 @@ from invokeai.backend.util.devices import TorchDevice
 
 @invocation(
     "flux_vae_decode",
-    title="FLUX Latents to Image",
+    title="Latents to Image - FLUX",
     tags=["latents", "image", "vae", "l2i", "flux"],
     category="latents",
-    version="1.0.1",
+    version="1.0.2",
 )
 class FluxVaeDecodeInvocation(BaseInvocation, WithMetadata, WithBoard):
     """Generates an image from latents."""

invokeai/app/invocations/flux_vae_encode.py

@@ -19,10 +19,10 @@ from invokeai.backend.util.devices import TorchDevice
 
 @invocation(
     "flux_vae_encode",
-    title="FLUX Image to Latents",
+    title="Image to Latents - FLUX",
     tags=["latents", "image", "vae", "i2l", "flux"],
     category="latents",
-    version="1.0.0",
+    version="1.0.1",
 )
 class FluxVaeEncodeInvocation(BaseInvocation):
     """Encodes an image into latents."""

invokeai/app/invocations/ideal_size.py

@@ -6,7 +6,7 @@ from invokeai.app.invocations.constants import LATENT_SCALE_FACTOR
 from invokeai.app.invocations.fields import FieldDescriptions, InputField, OutputField
 from invokeai.app.invocations.model import UNetField
 from invokeai.app.services.shared.invocation_context import InvocationContext
-from invokeai.backend.model_manager.config import BaseModelType
+from invokeai.backend.model_manager.taxonomy import BaseModelType
 
 
 @invocation_output("ideal_size_output")
@@ -19,9 +19,9 @@ class IdealSizeOutput(BaseInvocationOutput):
 
 @invocation(
     "ideal_size",
-    title="Ideal Size",
+    title="Ideal Size - SD1.5, SDXL",
     tags=["latents", "math", "ideal_size"],
-    version="1.0.4",
+    version="1.0.5",
 )
 class IdealSizeInvocation(BaseInvocation):
     """Calculates the ideal size for generation to avoid duplication"""

invokeai/app/invocations/image.py

@@ -355,7 +355,6 @@ class ImageBlurInvocation(BaseInvocation, WithMetadata, WithBoard):
     tags=["image", "unsharp_mask"],
     category="image",
     version="1.2.2",
-    classification=Classification.Beta,
 )
 class UnsharpMaskInvocation(BaseInvocation, WithMetadata, WithBoard):
     """Applies an unsharp mask filter to an image"""
@@ -1051,7 +1050,7 @@ class MaskFromIDInvocation(BaseInvocation, WithMetadata, WithBoard):
     tags=["image", "mask", "id"],
     category="image",
     version="1.0.0",
-    classification=Classification.Internal,
+    classification=Classification.Deprecated,
 )
 class CanvasV2MaskAndCropInvocation(BaseInvocation, WithMetadata, WithBoard):
     """Handles Canvas V2 image output masking and cropping"""
@@ -1089,6 +1088,118 @@ class CanvasV2MaskAndCropInvocation(BaseInvocation, WithMetadata, WithBoard):
         return ImageOutput.build(image_dto)
 
 
+@invocation(
+    "expand_mask_with_fade", title="Expand Mask with Fade", tags=["image", "mask"], category="image", version="1.0.0"
+)
+class ExpandMaskWithFadeInvocation(BaseInvocation, WithMetadata, WithBoard):
+    """Expands a mask with a fade effect. The mask uses black to indicate areas to keep from the generated image and white for areas to discard.
+    The mask is thresholded to create a binary mask, and then a distance transform is applied to create a fade effect.
+    The fade size is specified in pixels, and the mask is expanded by that amount. The result is a mask with a smooth transition from black to white.
+    If the fade size is 0, the mask is returned as-is.
+    """
+
+    mask: ImageField = InputField(description="The mask to expand")
+    threshold: int = InputField(default=0, ge=0, le=255, description="The threshold for the binary mask (0-255)")
+    fade_size_px: int = InputField(default=32, ge=0, description="The size of the fade in pixels")
+
+    def invoke(self, context: InvocationContext) -> ImageOutput:
+        pil_mask = context.images.get_pil(self.mask.image_name, mode="L")
+
+        if self.fade_size_px == 0:
+            # If the fade size is 0, just return the mask as-is.
+            image_dto = context.images.save(image=pil_mask, image_category=ImageCategory.MASK)
+            return ImageOutput.build(image_dto)
+
+        np_mask = numpy.array(pil_mask)
+
+        # Threshold the mask to create a binary mask - 0 for black, 255 for white
+        # If we don't threshold we can get some weird artifacts
+        np_mask = numpy.where(np_mask > self.threshold, 255, 0).astype(numpy.uint8)
+
+        # Create a mask for the black region (1 where black, 0 otherwise)
+        black_mask = (np_mask == 0).astype(numpy.uint8)
+
+        # Invert the black region
+        bg_mask = 1 - black_mask
+
+        # Create a distance transform of the inverted mask
+        dist = cv2.distanceTransform(bg_mask, cv2.DIST_L2, 5)
+
+        # Normalize distances so that pixels <fade_size_px become a linear gradient (0 to 1)
+        d_norm = numpy.clip(dist / self.fade_size_px, 0, 1)
+
+        # Control points: x values (normalized distance) and corresponding fade pct y values.
+
+        # There are some magic numbers here that are used to create a smooth transition:
+        # - The first point is at 0% of fade size from edge of mask (meaning the edge of the mask), and is 0% fade (black)
+        # - The second point is 1px from the edge of the mask and also has 0% fade, effectively expanding the mask
+        #   by 1px. This fixes an issue where artifacts can occur at the edge of the mask
+        # - The third point is at 20% of the fade size from the edge of the mask and has 20% fade
+        # - The fourth point is at 80% of the fade size from the edge of the mask and has 90% fade
+        # - The last point is at 100% of the fade size from the edge of the mask and has 100% fade (white)
+
+        # x values: 0 = mask edge, 1 = fade_size_px from edge
+        x_control = numpy.array([0.0, 1.0 / self.fade_size_px, 0.2, 0.8, 1.0])
+        # y values: 0 = black, 1 = white
+        y_control = numpy.array([0.0, 0.0, 0.2, 0.9, 1.0])
+
+        # Fit a cubic polynomial that smoothly passes through the control points
+        coeffs = numpy.polyfit(x_control, y_control, 3)
+        poly = numpy.poly1d(coeffs)
+
+        # Evaluate and clip the smooth mapping
+        feather = numpy.clip(poly(d_norm), 0, 1)
+
+        # Build final image.
+        np_result = numpy.where(black_mask == 1, 0, (feather * 255).astype(numpy.uint8))
+
+        # Convert back to PIL, grayscale
+        pil_result = Image.fromarray(np_result.astype(numpy.uint8), mode="L")
+
+        image_dto = context.images.save(image=pil_result, image_category=ImageCategory.MASK)
+
+        return ImageOutput.build(image_dto)
+
+
+@invocation(
+    "apply_mask_to_image",
+    title="Apply Mask to Image",
+    tags=["image", "mask", "blend"],
+    category="image",
+    version="1.0.0",
+)
+class ApplyMaskToImageInvocation(BaseInvocation, WithMetadata, WithBoard):
+    """
+    Extracts a region from a generated image using a mask and blends it seamlessly onto a source image.
+    The mask uses black to indicate areas to keep from the generated image and white for areas to discard.
+    """
+
+    image: ImageField = InputField(description="The image from which to extract the masked region")
+    mask: ImageField = InputField(description="The mask defining the region (black=keep, white=discard)")
+    invert_mask: bool = InputField(
+        default=False,
+        description="Whether to invert the mask before applying it",
+    )
+
+    def invoke(self, context: InvocationContext) -> ImageOutput:
+        # Load images
+        image = context.images.get_pil(self.image.image_name, mode="RGBA")
+        mask = context.images.get_pil(self.mask.image_name, mode="L")
+
+        if self.invert_mask:
+            # Invert the mask if requested
+            mask = ImageOps.invert(mask.copy())
+
+        # Combine the mask as the alpha channel of the image
+        r, g, b, _ = image.split()  # Split the image into RGB and alpha channels
+        result_image = Image.merge("RGBA", (r, g, b, mask))  # Use the mask as the new alpha channel
+
+        # Save the resulting image
+        image_dto = context.images.save(image=result_image)
+
+        return ImageOutput.build(image_dto)
+
+
 @invocation(
     "img_noise",
     title="Add Image Noise",
@@ -1159,7 +1270,6 @@ class ImageNoiseInvocation(BaseInvocation, WithMetadata, WithBoard):
     category="image",
     version="1.0.0",
     tags=["image", "crop"],
-    classification=Classification.Beta,
 )
 class CropImageToBoundingBoxInvocation(BaseInvocation, WithMetadata, WithBoard):
     """Crop an image to the given bounding box. If the bounding box is omitted, the image is cropped to the non-transparent pixels."""
@@ -1186,7 +1296,6 @@ class CropImageToBoundingBoxInvocation(BaseInvocation, WithMetadata, WithBoard):
     category="image",
     version="1.0.0",
     tags=["image", "crop"],
-    classification=Classification.Beta,
 )
 class PasteImageIntoBoundingBoxInvocation(BaseInvocation, WithMetadata, WithBoard):
     """Paste the source image into the target image at the given bounding box.

invokeai/app/invocations/image_to_latents.py

@@ -31,10 +31,10 @@ from invokeai.backend.util.devices import TorchDevice
 
 @invocation(
     "i2l",
-    title="Image to Latents",
+    title="Image to Latents - SD1.5, SDXL",
     tags=["latents", "image", "vae", "i2l"],
     category="latents",
-    version="1.1.0",
+    version="1.1.1",
 )
 class ImageToLatentsInvocation(BaseInvocation):
     """Encodes an image into latents."""

invokeai/app/invocations/ip_adapter.py

@@ -13,10 +13,8 @@ from invokeai.app.services.model_records.model_records_base import ModelRecordCh
 from invokeai.app.services.shared.invocation_context import InvocationContext
 from invokeai.backend.model_manager.config import (
     AnyModelConfig,
-    BaseModelType,
     IPAdapterCheckpointConfig,
     IPAdapterInvokeAIConfig,
-    ModelType,
 )
 from invokeai.backend.model_manager.starter_models import (
     StarterModel,
@@ -24,6 +22,7 @@ from invokeai.backend.model_manager.starter_models import (
     ip_adapter_sd_image_encoder,
     ip_adapter_sdxl_image_encoder,
 )
+from invokeai.backend.model_manager.taxonomy import BaseModelType, ModelType
 
 
 class IPAdapterField(BaseModel):
@@ -69,7 +68,13 @@ CLIP_VISION_MODEL_MAP: dict[Literal["ViT-L", "ViT-H", "ViT-G"], StarterModel] =
 }
 
 
-@invocation("ip_adapter", title="IP-Adapter", tags=["ip_adapter", "control"], category="ip_adapter", version="1.5.0")
+@invocation(
+    "ip_adapter",
+    title="IP-Adapter - SD1.5, SDXL",
+    tags=["ip_adapter", "control"],
+    category="ip_adapter",
+    version="1.5.1",
+)
 class IPAdapterInvocation(BaseInvocation):
     """Collects IP-Adapter info to pass to other nodes."""
 

invokeai/app/invocations/latents_to_image.py

@@ -31,10 +31,10 @@ from invokeai.backend.util.devices import TorchDevice
 
 @invocation(
     "l2i",
-    title="Latents to Image",
+    title="Latents to Image - SD1.5, SDXL",
     tags=["latents", "image", "vae", "l2i"],
     category="latents",
-    version="1.3.1",
+    version="1.3.2",
 )
 class LatentsToImageInvocation(BaseInvocation, WithMetadata, WithBoard):
     """Generates an image from latents."""

invokeai/app/invocations/llava_onevision_vllm.py

@@ -0,0 +1,67 @@
+from typing import Any
+
+import torch
+from PIL.Image import Image
+from pydantic import field_validator
+
+from invokeai.app.invocations.baseinvocation import BaseInvocation, Classification, invocation
+from invokeai.app.invocations.fields import FieldDescriptions, ImageField, InputField, UIComponent, UIType
+from invokeai.app.invocations.model import ModelIdentifierField
+from invokeai.app.invocations.primitives import StringOutput
+from invokeai.app.services.shared.invocation_context import InvocationContext
+from invokeai.backend.llava_onevision_model import LlavaOnevisionModel
+from invokeai.backend.util.devices import TorchDevice
+
+
+@invocation(
+    "llava_onevision_vllm",
+    title="LLaVA OneVision VLLM",
+    tags=["vllm"],
+    category="vllm",
+    version="1.0.0",
+    classification=Classification.Beta,
+)
+class LlavaOnevisionVllmInvocation(BaseInvocation):
+    """Run a LLaVA OneVision VLLM model."""
+
+    images: list[ImageField] | ImageField | None = InputField(default=None, max_length=3, description="Input image.")
+    prompt: str = InputField(
+        default="",
+        description="Input text prompt.",
+        ui_component=UIComponent.Textarea,
+    )
+    vllm_model: ModelIdentifierField = InputField(
+        title="LLaVA Model Type",
+        description=FieldDescriptions.vllm_model,
+        ui_type=UIType.LlavaOnevisionModel,
+    )
+
+    @field_validator("images", mode="before")
+    def listify_images(cls, v: Any) -> list:
+        if v is None:
+            return v
+        if not isinstance(v, list):
+            return [v]
+        return v
+
+    def _get_images(self, context: InvocationContext) -> list[Image]:
+        if self.images is None:
+            return []
+
+        image_fields = self.images if isinstance(self.images, list) else [self.images]
+        return [context.images.get_pil(image_field.image_name, "RGB") for image_field in image_fields]
+
+    @torch.no_grad()
+    def invoke(self, context: InvocationContext) -> StringOutput:
+        images = self._get_images(context)
+
+        with context.models.load(self.vllm_model) as vllm_model:
+            assert isinstance(vllm_model, LlavaOnevisionModel)
+            output = vllm_model.run(
+                prompt=self.prompt,
+                images=images,
+                device=TorchDevice.choose_torch_device(),
+                dtype=TorchDevice.choose_torch_dtype(),
+            )
+
+        return StringOutput(value=output)

invokeai/app/invocations/mask.py

@@ -4,7 +4,6 @@ from PIL import Image
 
 from invokeai.app.invocations.baseinvocation import (
     BaseInvocation,
-    Classification,
     InvocationContext,
     invocation,
 )
@@ -58,7 +57,6 @@ class RectangleMaskInvocation(BaseInvocation, WithMetadata):
     tags=["conditioning"],
     category="conditioning",
     version="1.0.0",
-    classification=Classification.Beta,
 )
 class AlphaMaskToTensorInvocation(BaseInvocation):
     """Convert a mask image to a tensor. Opaque regions are 1 and transparent regions are 0."""
@@ -67,7 +65,7 @@ class AlphaMaskToTensorInvocation(BaseInvocation):
     invert: bool = InputField(default=False, description="Whether to invert the mask.")
 
     def invoke(self, context: InvocationContext) -> MaskOutput:
-        image = context.images.get_pil(self.image.image_name)
+        image = context.images.get_pil(self.image.image_name, mode="RGBA")
         mask = torch.zeros((1, image.height, image.width), dtype=torch.bool)
         if self.invert:
             mask[0] = torch.tensor(np.array(image)[:, :, 3] == 0, dtype=torch.bool)
@@ -87,7 +85,6 @@ class AlphaMaskToTensorInvocation(BaseInvocation):
     tags=["conditioning"],
     category="conditioning",
     version="1.1.0",
-    classification=Classification.Beta,
 )
 class InvertTensorMaskInvocation(BaseInvocation):
     """Inverts a tensor mask."""
@@ -234,7 +231,6 @@ WHITE = ColorField(r=255, g=255, b=255, a=255)
     tags=["mask"],
     category="mask",
     version="1.0.0",
-    classification=Classification.Beta,
 )
 class GetMaskBoundingBoxInvocation(BaseInvocation):
     """Gets the bounding box of the given mask image."""

invokeai/app/invocations/metadata_linked.py

@@ -43,7 +43,7 @@ from invokeai.app.invocations.primitives import BooleanOutput, FloatOutput, Inte
 from invokeai.app.invocations.scheduler import SchedulerOutput
 from invokeai.app.invocations.t2i_adapter import T2IAdapterField, T2IAdapterInvocation
 from invokeai.app.services.shared.invocation_context import InvocationContext
-from invokeai.backend.model_manager.config import ModelType, SubModelType
+from invokeai.backend.model_manager.taxonomy import ModelType, SubModelType
 from invokeai.backend.stable_diffusion.schedulers.schedulers import SCHEDULER_NAME_VALUES
 from invokeai.version import __version__
 
@@ -610,10 +610,10 @@ class LatentsMetaOutput(LatentsOutput, MetadataOutput):
 
 @invocation(
     "denoise_latents_meta",
-    title="Denoise Latents + metadata",
+    title=f"{DenoiseLatentsInvocation.UIConfig.title} + Metadata",
     tags=["latents", "denoise", "txt2img", "t2i", "t2l", "img2img", "i2i", "l2l"],
     category="latents",
-    version="1.1.0",
+    version="1.1.1",
 )
 class DenoiseLatentsMetaInvocation(DenoiseLatentsInvocation, WithMetadata):
     def invoke(self, context: InvocationContext) -> LatentsMetaOutput:
@@ -675,10 +675,10 @@ class DenoiseLatentsMetaInvocation(DenoiseLatentsInvocation, WithMetadata):
 
 @invocation(
     "flux_denoise_meta",
-    title="Flux Denoise + metadata",
+    title=f"{FluxDenoiseInvocation.UIConfig.title} + Metadata",
     tags=["flux", "latents", "denoise", "txt2img", "t2i", "t2l", "img2img", "i2i", "l2l"],
     category="latents",
-    version="1.0.0",
+    version="1.0.1",
 )
 class FluxDenoiseLatentsMetaInvocation(FluxDenoiseInvocation, WithMetadata):
     """Run denoising process with a FLUX transformer model + metadata."""

invokeai/app/invocations/model.py

@@ -6,7 +6,6 @@ from pydantic import BaseModel, Field
 from invokeai.app.invocations.baseinvocation import (
     BaseInvocation,
     BaseInvocationOutput,
-    Classification,
     invocation,
     invocation_output,
 )
@@ -15,10 +14,8 @@ from invokeai.app.services.shared.invocation_context import InvocationContext
 from invokeai.app.shared.models import FreeUConfig
 from invokeai.backend.model_manager.config import (
     AnyModelConfig,
-    BaseModelType,
-    ModelType,
-    SubModelType,
 )
+from invokeai.backend.model_manager.taxonomy import BaseModelType, ModelType, SubModelType
 
 
 class ModelIdentifierField(BaseModel):
@@ -122,11 +119,10 @@ class ModelIdentifierOutput(BaseInvocationOutput):
 
 @invocation(
     "model_identifier",
-    title="Model identifier",
+    title="Any Model",
     tags=["model"],
     category="model",
-    version="1.0.0",
-    classification=Classification.Prototype,
+    version="1.0.1",
 )
 class ModelIdentifierInvocation(BaseInvocation):
     """Selects any model, outputting it its identifier. Be careful with this one! The identifier will be accepted as
@@ -144,10 +140,10 @@ class ModelIdentifierInvocation(BaseInvocation):
 
 @invocation(
     "main_model_loader",
-    title="Main Model",
+    title="Main Model - SD1.5",
     tags=["model"],
     category="model",
-    version="1.0.3",
+    version="1.0.4",
 )
 class MainModelLoaderInvocation(BaseInvocation):
     """Loads a main model, outputting its submodels."""
@@ -181,7 +177,7 @@ class LoRALoaderOutput(BaseInvocationOutput):
     clip: Optional[CLIPField] = OutputField(default=None, description=FieldDescriptions.clip, title="CLIP")
 
 
-@invocation("lora_loader", title="LoRA", tags=["model"], category="model", version="1.0.3")
+@invocation("lora_loader", title="Apply LoRA - SD1.5", tags=["model"], category="model", version="1.0.4")
 class LoRALoaderInvocation(BaseInvocation):
     """Apply selected lora to unet and text_encoder."""
 
@@ -244,7 +240,7 @@ class LoRASelectorOutput(BaseInvocationOutput):
     lora: LoRAField = OutputField(description="LoRA model and weight", title="LoRA")
 
 
-@invocation("lora_selector", title="LoRA Selector", tags=["model"], category="model", version="1.0.1")
+@invocation("lora_selector", title="Select LoRA", tags=["model"], category="model", version="1.0.3")
 class LoRASelectorInvocation(BaseInvocation):
     """Selects a LoRA model and weight."""
 
@@ -257,7 +253,9 @@ class LoRASelectorInvocation(BaseInvocation):
         return LoRASelectorOutput(lora=LoRAField(lora=self.lora, weight=self.weight))
 
 
-@invocation("lora_collection_loader", title="LoRA Collection Loader", tags=["model"], category="model", version="1.1.0")
+@invocation(
+    "lora_collection_loader", title="Apply LoRA Collection - SD1.5", tags=["model"], category="model", version="1.1.2"
+)
 class LoRACollectionLoader(BaseInvocation):
     """Applies a collection of LoRAs to the provided UNet and CLIP models."""
 
@@ -320,10 +318,10 @@ class SDXLLoRALoaderOutput(BaseInvocationOutput):
 
 @invocation(
     "sdxl_lora_loader",
-    title="SDXL LoRA",
+    title="Apply LoRA - SDXL",
     tags=["lora", "model"],
     category="model",
-    version="1.0.3",
+    version="1.0.5",
 )
 class SDXLLoRALoaderInvocation(BaseInvocation):
     """Apply selected lora to unet and text_encoder."""
@@ -400,10 +398,10 @@ class SDXLLoRALoaderInvocation(BaseInvocation):
 
 @invocation(
     "sdxl_lora_collection_loader",
-    title="SDXL LoRA Collection Loader",
+    title="Apply LoRA Collection - SDXL",
     tags=["model"],
     category="model",
-    version="1.1.0",
+    version="1.1.2",
 )
 class SDXLLoRACollectionLoader(BaseInvocation):
     """Applies a collection of SDXL LoRAs to the provided UNet and CLIP models."""
@@ -469,7 +467,9 @@ class SDXLLoRACollectionLoader(BaseInvocation):
         return output
 
 
-@invocation("vae_loader", title="VAE", tags=["vae", "model"], category="model", version="1.0.3")
+@invocation(
+    "vae_loader", title="VAE Model - SD1.5, SDXL, SD3, FLUX", tags=["vae", "model"], category="model", version="1.0.4"
+)
 class VAELoaderInvocation(BaseInvocation):
     """Loads a VAE model, outputting a VaeLoaderOutput"""
 
@@ -496,10 +496,10 @@ class SeamlessModeOutput(BaseInvocationOutput):
 
 @invocation(
     "seamless",
-    title="Seamless",
+    title="Apply Seamless - SD1.5, SDXL",
     tags=["seamless", "model"],
     category="model",
-    version="1.0.1",
+    version="1.0.2",
 )
 class SeamlessModeInvocation(BaseInvocation):
     """Applies the seamless transformation to the Model UNet and VAE."""
@@ -539,7 +539,7 @@ class SeamlessModeInvocation(BaseInvocation):
         return SeamlessModeOutput(unet=unet, vae=vae)
 
 
-@invocation("freeu", title="FreeU", tags=["freeu"], category="unet", version="1.0.1")
+@invocation("freeu", title="Apply FreeU - SD1.5, SDXL", tags=["freeu"], category="unet", version="1.0.2")
 class FreeUInvocation(BaseInvocation):
     """
     Applies FreeU to the UNet. Suggested values (b1/b2/s1/s2):

invokeai/app/invocations/noise.py

@@ -72,10 +72,10 @@ class NoiseOutput(BaseInvocationOutput):
 
 @invocation(
     "noise",
-    title="Noise",
+    title="Create Latent Noise",
     tags=["latents", "noise"],
     category="latents",
-    version="1.0.2",
+    version="1.0.3",
 )
 class NoiseInvocation(BaseInvocation):
     """Generates latent noise."""

invokeai/app/invocations/sd3_denoise.py

@@ -6,7 +6,7 @@ from diffusers.models.transformers.transformer_sd3 import SD3Transformer2DModel
 from torchvision.transforms.functional import resize as tv_resize
 from tqdm import tqdm
 
-from invokeai.app.invocations.baseinvocation import BaseInvocation, Classification, invocation
+from invokeai.app.invocations.baseinvocation import BaseInvocation, invocation
 from invokeai.app.invocations.constants import LATENT_SCALE_FACTOR
 from invokeai.app.invocations.fields import (
     DenoiseMaskField,
@@ -23,7 +23,7 @@ from invokeai.app.invocations.primitives import LatentsOutput
 from invokeai.app.invocations.sd3_text_encoder import SD3_T5_MAX_SEQ_LEN
 from invokeai.app.services.shared.invocation_context import InvocationContext
 from invokeai.backend.flux.sampling_utils import clip_timestep_schedule_fractional
-from invokeai.backend.model_manager.config import BaseModelType
+from invokeai.backend.model_manager import BaseModelType
 from invokeai.backend.sd3.extensions.inpaint_extension import InpaintExtension
 from invokeai.backend.stable_diffusion.diffusers_pipeline import PipelineIntermediateState
 from invokeai.backend.stable_diffusion.diffusion.conditioning_data import SD3ConditioningInfo
@@ -32,11 +32,10 @@ from invokeai.backend.util.devices import TorchDevice
 
 @invocation(
     "sd3_denoise",
-    title="SD3 Denoise",
+    title="Denoise - SD3",
     tags=["image", "sd3"],
     category="image",
-    version="1.1.0",
-    classification=Classification.Prototype,
+    version="1.1.1",
 )
 class SD3DenoiseInvocation(BaseInvocation, WithMetadata, WithBoard):
     """Run denoising process with a SD3 model."""

invokeai/app/invocations/sd3_image_to_latents.py

@@ -2,7 +2,7 @@ import einops
 import torch
 from diffusers.models.autoencoders.autoencoder_kl import AutoencoderKL
 
-from invokeai.app.invocations.baseinvocation import BaseInvocation, Classification, invocation
+from invokeai.app.invocations.baseinvocation import BaseInvocation, invocation
 from invokeai.app.invocations.fields import (
     FieldDescriptions,
     ImageField,
@@ -21,11 +21,10 @@ from invokeai.backend.util.devices import TorchDevice
 
 @invocation(
     "sd3_i2l",
-    title="SD3 Image to Latents",
+    title="Image to Latents - SD3",
     tags=["image", "latents", "vae", "i2l", "sd3"],
     category="image",
-    version="1.0.0",
-    classification=Classification.Prototype,
+    version="1.0.1",
 )
 class SD3ImageToLatentsInvocation(BaseInvocation, WithMetadata, WithBoard):
     """Generates latents from an image."""

invokeai/app/invocations/sd3_latents_to_image.py

@@ -24,10 +24,10 @@ from invokeai.backend.util.devices import TorchDevice
 
 @invocation(
     "sd3_l2i",
-    title="SD3 Latents to Image",
+    title="Latents to Image - SD3",
     tags=["latents", "image", "vae", "l2i", "sd3"],
     category="latents",
-    version="1.3.1",
+    version="1.3.2",
 )
 class SD3LatentsToImageInvocation(BaseInvocation, WithMetadata, WithBoard):
     """Generates an image from latents."""

invokeai/app/invocations/sd3_model_loader.py

@@ -3,7 +3,6 @@ from typing import Optional
 from invokeai.app.invocations.baseinvocation import (
     BaseInvocation,
     BaseInvocationOutput,
-    Classification,
     invocation,
     invocation_output,
 )
@@ -14,7 +13,7 @@ from invokeai.app.util.t5_model_identifier import (
     preprocess_t5_encoder_model_identifier,
     preprocess_t5_tokenizer_model_identifier,
 )
-from invokeai.backend.model_manager.config import SubModelType
+from invokeai.backend.model_manager.taxonomy import SubModelType
 
 
 @invocation_output("sd3_model_loader_output")
@@ -30,11 +29,10 @@ class Sd3ModelLoaderOutput(BaseInvocationOutput):
 
 @invocation(
     "sd3_model_loader",
-    title="SD3 Main Model",
+    title="Main Model - SD3",
     tags=["model", "sd3"],
     category="model",
-    version="1.0.0",
-    classification=Classification.Prototype,
+    version="1.0.1",
 )
 class Sd3ModelLoaderInvocation(BaseInvocation):
     """Loads a SD3 base model, outputting its submodels."""

invokeai/app/invocations/sd3_text_encoder.py

@@ -11,12 +11,12 @@ from transformers import (
     T5TokenizerFast,
 )
 
-from invokeai.app.invocations.baseinvocation import BaseInvocation, Classification, invocation
+from invokeai.app.invocations.baseinvocation import BaseInvocation, invocation
 from invokeai.app.invocations.fields import FieldDescriptions, Input, InputField
 from invokeai.app.invocations.model import CLIPField, T5EncoderField
 from invokeai.app.invocations.primitives import SD3ConditioningOutput
 from invokeai.app.services.shared.invocation_context import InvocationContext
-from invokeai.backend.model_manager.config import ModelFormat
+from invokeai.backend.model_manager.taxonomy import ModelFormat
 from invokeai.backend.patches.layer_patcher import LayerPatcher
 from invokeai.backend.patches.lora_conversions.flux_lora_constants import FLUX_LORA_CLIP_PREFIX
 from invokeai.backend.patches.model_patch_raw import ModelPatchRaw
@@ -29,11 +29,10 @@ SD3_T5_MAX_SEQ_LEN = 256
 
 @invocation(
     "sd3_text_encoder",
-    title="SD3 Text Encoding",
+    title="Prompt - SD3",
     tags=["prompt", "conditioning", "sd3"],
     category="conditioning",
-    version="1.0.0",
-    classification=Classification.Prototype,
+    version="1.0.1",
 )
 class Sd3TextEncoderInvocation(BaseInvocation):
     """Encodes and preps a prompt for a SD3 image."""

invokeai/app/invocations/sdxl.py

@@ -2,7 +2,7 @@ from invokeai.app.invocations.baseinvocation import BaseInvocation, BaseInvocati
 from invokeai.app.invocations.fields import FieldDescriptions, InputField, OutputField, UIType
 from invokeai.app.invocations.model import CLIPField, ModelIdentifierField, UNetField, VAEField
 from invokeai.app.services.shared.invocation_context import InvocationContext
-from invokeai.backend.model_manager import SubModelType
+from invokeai.backend.model_manager.taxonomy import SubModelType
 
 
 @invocation_output("sdxl_model_loader_output")
@@ -24,7 +24,7 @@ class SDXLRefinerModelLoaderOutput(BaseInvocationOutput):
     vae: VAEField = OutputField(description=FieldDescriptions.vae, title="VAE")
 
 
-@invocation("sdxl_model_loader", title="SDXL Main Model", tags=["model", "sdxl"], category="model", version="1.0.3")
+@invocation("sdxl_model_loader", title="Main Model - SDXL", tags=["model", "sdxl"], category="model", version="1.0.4")
 class SDXLModelLoaderInvocation(BaseInvocation):
     """Loads an sdxl base model, outputting its submodels."""
 
@@ -58,10 +58,10 @@ class SDXLModelLoaderInvocation(BaseInvocation):
 
 @invocation(
     "sdxl_refiner_model_loader",
-    title="SDXL Refiner Model",
+    title="Refiner Model - SDXL",
     tags=["model", "sdxl", "refiner"],
     category="model",
-    version="1.0.3",
+    version="1.0.4",
 )
 class SDXLRefinerModelLoaderInvocation(BaseInvocation):
     """Loads an sdxl refiner model, outputting its submodels."""

invokeai/app/invocations/t2i_adapter.py

@@ -45,7 +45,11 @@ class T2IAdapterOutput(BaseInvocationOutput):
 
 
 @invocation(
-    "t2i_adapter", title="T2I-Adapter", tags=["t2i_adapter", "control"], category="t2i_adapter", version="1.0.3"
+    "t2i_adapter",
+    title="T2I-Adapter - SD1.5, SDXL",
+    tags=["t2i_adapter", "control"],
+    category="t2i_adapter",
+    version="1.0.4",
 )
 class T2IAdapterInvocation(BaseInvocation):
     """Collects T2I-Adapter info to pass to other nodes."""

invokeai/app/invocations/tiled_multi_diffusion_denoise_latents.py

@@ -7,7 +7,7 @@ from diffusers.models.unets.unet_2d_condition import UNet2DConditionModel
 from diffusers.schedulers.scheduling_utils import SchedulerMixin
 from pydantic import field_validator
 
-from invokeai.app.invocations.baseinvocation import BaseInvocation, Classification, invocation
+from invokeai.app.invocations.baseinvocation import BaseInvocation, invocation
 from invokeai.app.invocations.constants import LATENT_SCALE_FACTOR
 from invokeai.app.invocations.controlnet_image_processors import ControlField
 from invokeai.app.invocations.denoise_latents import DenoiseLatentsInvocation, get_scheduler
@@ -53,11 +53,10 @@ def crop_controlnet_data(control_data: ControlNetData, latent_region: TBLR) -> C
 
 @invocation(
     "tiled_multi_diffusion_denoise_latents",
-    title="Tiled Multi-Diffusion Denoise Latents",
+    title="Tiled Multi-Diffusion Denoise - SD1.5, SDXL",
     tags=["upscale", "denoise"],
     category="latents",
-    classification=Classification.Beta,
-    version="1.0.0",
+    version="1.0.1",
 )
 class TiledMultiDiffusionDenoiseLatents(BaseInvocation):
     """Tiled Multi-Diffusion denoising.

invokeai/app/invocations/tiles.py

@@ -7,7 +7,6 @@ from pydantic import BaseModel
 from invokeai.app.invocations.baseinvocation import (
     BaseInvocation,
     BaseInvocationOutput,
-    Classification,
     invocation,
     invocation_output,
 )
@@ -40,7 +39,6 @@ class CalculateImageTilesOutput(BaseInvocationOutput):
     tags=["tiles"],
     category="tiles",
     version="1.0.1",
-    classification=Classification.Beta,
 )
 class CalculateImageTilesInvocation(BaseInvocation):
     """Calculate the coordinates and overlaps of tiles that cover a target image shape."""
@@ -74,7 +72,6 @@ class CalculateImageTilesInvocation(BaseInvocation):
     tags=["tiles"],
     category="tiles",
     version="1.1.1",
-    classification=Classification.Beta,
 )
 class CalculateImageTilesEvenSplitInvocation(BaseInvocation):
     """Calculate the coordinates and overlaps of tiles that cover a target image shape."""
@@ -117,7 +114,6 @@ class CalculateImageTilesEvenSplitInvocation(BaseInvocation):
     tags=["tiles"],
     category="tiles",
     version="1.0.1",
-    classification=Classification.Beta,
 )
 class CalculateImageTilesMinimumOverlapInvocation(BaseInvocation):
     """Calculate the coordinates and overlaps of tiles that cover a target image shape."""
@@ -168,7 +164,6 @@ class TileToPropertiesOutput(BaseInvocationOutput):
     tags=["tiles"],
     category="tiles",
     version="1.0.1",
-    classification=Classification.Beta,
 )
 class TileToPropertiesInvocation(BaseInvocation):
     """Split a Tile into its individual properties."""
@@ -201,7 +196,6 @@ class PairTileImageOutput(BaseInvocationOutput):
     tags=["tiles"],
     category="tiles",
     version="1.0.1",
-    classification=Classification.Beta,
 )
 class PairTileImageInvocation(BaseInvocation):
     """Pair an image with its tile properties."""
@@ -230,7 +224,6 @@ BLEND_MODES = Literal["Linear", "Seam"]
     tags=["tiles"],
     category="tiles",
     version="1.1.1",
-    classification=Classification.Beta,
 )
 class MergeTilesToImageInvocation(BaseInvocation, WithMetadata, WithBoard):
     """Merge multiple tile images into a single image."""

invokeai/app/run_app.py

@@ -41,16 +41,15 @@ def run_app() -> None:
     )
 
     # Find an open port, and modify the config accordingly.
-    orig_config_port = app_config.port
-    app_config.port = find_open_port(app_config.port)
-    if orig_config_port != app_config.port:
+    first_open_port = find_open_port(app_config.port)
+    if app_config.port != first_open_port:
+        orig_config_port = app_config.port
+        app_config.port = first_open_port
         logger.warning(f"Port {orig_config_port} is already in use. Using port {app_config.port}.")
 
     # Miscellaneous startup tasks.
     apply_monkeypatches()
     register_mime_types()
-    if app_config.dev_reload:
-        enable_dev_reload()
     check_cudnn(logger)
 
     # Initialize the app and event loop.
@@ -61,6 +60,11 @@ def run_app() -> None:
     # core nodes have been imported so that we can catch when a custom node clobbers a core node.
     load_custom_nodes(custom_nodes_path=app_config.custom_nodes_path, logger=logger)
 
+    if app_config.dev_reload:
+        # load_custom_nodes seems to bypass jurrigged's import sniffer, so be sure to call it *after* they're already
+        # imported.
+        enable_dev_reload(custom_nodes_path=app_config.custom_nodes_path)
+
     # Start the server.
     config = uvicorn.Config(
         app=app,

invokeai/app/services/events/events_base.py

@@ -44,7 +44,8 @@ if TYPE_CHECKING:
         SessionQueueItem,
         SessionQueueStatus,
     )
-    from invokeai.backend.model_manager.config import AnyModelConfig, SubModelType
+    from invokeai.backend.model_manager import SubModelType
+    from invokeai.backend.model_manager.config import AnyModelConfig
 
 
 class EventServiceBase:

invokeai/app/services/events/events_common.py

@@ -16,7 +16,8 @@ from invokeai.app.services.session_queue.session_queue_common import (
 )
 from invokeai.app.services.shared.graph import AnyInvocation, AnyInvocationOutput
 from invokeai.app.util.misc import get_timestamp
-from invokeai.backend.model_manager.config import AnyModelConfig, SubModelType
+from invokeai.backend.model_manager import SubModelType
+from invokeai.backend.model_manager.config import AnyModelConfig
 
 if TYPE_CHECKING:
     from invokeai.app.services.download.download_base import DownloadJob

invokeai/app/services/model_install/model_install_common.py

@@ -10,9 +10,9 @@ from typing_extensions import Annotated
 
 from invokeai.app.services.download import DownloadJob, MultiFileDownloadJob
 from invokeai.app.services.model_records import ModelRecordChanges
-from invokeai.backend.model_manager import AnyModelConfig, ModelRepoVariant
-from invokeai.backend.model_manager.config import ModelSourceType
+from invokeai.backend.model_manager.config import AnyModelConfig
 from invokeai.backend.model_manager.metadata import AnyModelRepoMetadata
+from invokeai.backend.model_manager.taxonomy import ModelRepoVariant, ModelSourceType
 
 
 class InstallStatus(str, Enum):

invokeai/app/services/model_install/model_install_default.py

@@ -38,9 +38,9 @@ from invokeai.backend.model_manager.config import (
     AnyModelConfig,
     CheckpointConfigBase,
     InvalidModelConfigException,
-    ModelRepoVariant,
-    ModelSourceType,
+    ModelConfigBase,
 )
+from invokeai.backend.model_manager.legacy_probe import ModelProbe
 from invokeai.backend.model_manager.metadata import (
     AnyModelRepoMetadata,
     HuggingFaceMetadataFetch,
@@ -49,8 +49,8 @@ from invokeai.backend.model_manager.metadata import (
     RemoteModelFile,
 )
 from invokeai.backend.model_manager.metadata.metadata_base import HuggingFaceMetadata
-from invokeai.backend.model_manager.probe import ModelProbe
 from invokeai.backend.model_manager.search import ModelSearch
+from invokeai.backend.model_manager.taxonomy import ModelRepoVariant, ModelSourceType
 from invokeai.backend.util import InvokeAILogger
 from invokeai.backend.util.catch_sigint import catch_sigint
 from invokeai.backend.util.devices import TorchDevice
@@ -182,9 +182,7 @@ class ModelInstallService(ModelInstallServiceBase):
     ) -> str:  # noqa D102
         model_path = Path(model_path)
         config = config or ModelRecordChanges()
-        info: AnyModelConfig = ModelProbe.probe(
-            Path(model_path), config.model_dump(), hash_algo=self._app_config.hashing_algorithm
-        )  # type: ignore
+        info: AnyModelConfig = self._probe(Path(model_path), config)  # type: ignore
 
         if preferred_name := config.name:
             preferred_name = Path(preferred_name).with_suffix(model_path.suffix)
@@ -644,12 +642,22 @@ class ModelInstallService(ModelInstallServiceBase):
         move(old_path, new_path)
         return new_path
 
+    def _probe(self, model_path: Path, config: Optional[ModelRecordChanges] = None):
+        config = config or ModelRecordChanges()
+        hash_algo = self._app_config.hashing_algorithm
+        fields = config.model_dump()
+
+        try:
+            return ModelConfigBase.classify(model_path=model_path, hash_algo=hash_algo, **fields)
+        except InvalidModelConfigException:
+            return ModelProbe.probe(model_path=model_path, fields=fields, hash_algo=hash_algo)  # type: ignore
+
     def _register(
         self, model_path: Path, config: Optional[ModelRecordChanges] = None, info: Optional[AnyModelConfig] = None
     ) -> str:
         config = config or ModelRecordChanges()
 
-        info = info or ModelProbe.probe(model_path, config.model_dump(), hash_algo=self._app_config.hashing_algorithm)  # type: ignore
+        info = info or self._probe(model_path, config)
 
         model_path = model_path.resolve()
 

invokeai/app/services/model_load/model_load_base.py

@@ -5,9 +5,10 @@ from abc import ABC, abstractmethod
 from pathlib import Path
 from typing import Callable, Optional
 
-from invokeai.backend.model_manager import AnyModel, AnyModelConfig, SubModelType
+from invokeai.backend.model_manager.config import AnyModelConfig
 from invokeai.backend.model_manager.load import LoadedModel, LoadedModelWithoutConfig
 from invokeai.backend.model_manager.load.model_cache.model_cache import ModelCache
+from invokeai.backend.model_manager.taxonomy import AnyModel, SubModelType
 
 
 class ModelLoadServiceBase(ABC):

invokeai/app/services/model_load/model_load_default.py

@@ -11,7 +11,7 @@ from torch import load as torch_load
 from invokeai.app.services.config import InvokeAIAppConfig
 from invokeai.app.services.invoker import Invoker
 from invokeai.app.services.model_load.model_load_base import ModelLoadServiceBase
-from invokeai.backend.model_manager import AnyModel, AnyModelConfig, SubModelType
+from invokeai.backend.model_manager.config import AnyModelConfig
 from invokeai.backend.model_manager.load import (
     LoadedModel,
     LoadedModelWithoutConfig,
@@ -20,6 +20,7 @@ from invokeai.backend.model_manager.load import (
 )
 from invokeai.backend.model_manager.load.model_cache.model_cache import ModelCache
 from invokeai.backend.model_manager.load.model_loaders.generic_diffusers import GenericDiffusersLoader
+from invokeai.backend.model_manager.taxonomy import AnyModel, SubModelType
 from invokeai.backend.util.devices import TorchDevice
 from invokeai.backend.util.logging import InvokeAILogger
 
@@ -85,8 +86,11 @@ class ModelLoadService(ModelLoadServiceBase):
 
         def torch_load_file(checkpoint: Path) -> AnyModel:
             scan_result = scan_file_path(checkpoint)
-            if scan_result.infected_files != 0 or scan_result.scan_err:
-                raise Exception("The model at {checkpoint} is potentially infected by malware. Aborting load.")
+            if scan_result.infected_files != 0:
+                raise Exception(f"The model at {checkpoint} is potentially infected by malware. Aborting load.")
+            if scan_result.scan_err:
+                raise Exception(f"Error scanning model at {checkpoint} for malware. Aborting load.")
+
             result = torch_load(checkpoint, map_location="cpu")
             return result
 

invokeai/app/services/model_manager/__init__.py

@@ -1,16 +1,12 @@
 """Initialization file for model manager service."""
 
 from invokeai.app.services.model_manager.model_manager_default import ModelManagerService, ModelManagerServiceBase
-from invokeai.backend.model_manager import AnyModel, AnyModelConfig, BaseModelType, ModelType, SubModelType
+from invokeai.backend.model_manager import AnyModelConfig
 from invokeai.backend.model_manager.load import LoadedModel
 
 __all__ = [
     "ModelManagerServiceBase",
     "ModelManagerService",
-    "AnyModel",
     "AnyModelConfig",
-    "BaseModelType",
-    "ModelType",
-    "SubModelType",
     "LoadedModel",
 ]

invokeai/app/services/model_records/model_records_base.py

@@ -14,10 +14,12 @@ from invokeai.app.services.shared.pagination import PaginatedResults
 from invokeai.app.util.model_exclude_null import BaseModelExcludeNull
 from invokeai.backend.model_manager.config import (
     AnyModelConfig,
-    BaseModelType,
-    ClipVariantType,
     ControlAdapterDefaultSettings,
     MainModelDefaultSettings,
+)
+from invokeai.backend.model_manager.taxonomy import (
+    BaseModelType,
+    ClipVariantType,
     ModelFormat,
     ModelSourceType,
     ModelType,

invokeai/app/services/model_records/model_records_sql.py

@@ -60,11 +60,9 @@ from invokeai.app.services.shared.pagination import PaginatedResults
 from invokeai.app.services.shared.sqlite.sqlite_database import SqliteDatabase
 from invokeai.backend.model_manager.config import (
     AnyModelConfig,
-    BaseModelType,
     ModelConfigFactory,
-    ModelFormat,
-    ModelType,
 )
+from invokeai.backend.model_manager.taxonomy import BaseModelType, ModelFormat, ModelType
 
 
 class ModelRecordServiceSQL(ModelRecordServiceBase):

invokeai/app/services/session_queue/session_queue_common.py

@@ -570,7 +570,10 @@ ValueToInsertTuple: TypeAlias = tuple[
     str | None,  # destination (optional)
     int | None,  # retried_from_item_id (optional, this is always None for new items)
 ]
-"""A type alias for the tuple of values to insert into the session queue table."""
+"""A type alias for the tuple of values to insert into the session queue table.
+
+**If you change this, be sure to update the `enqueue_batch` and `retry_items_by_id` methods in the session queue service!**
+"""
 
 
 def prepare_values_to_insert(

invokeai/app/services/session_queue/session_queue_sqlite.py

@@ -27,6 +27,7 @@ from invokeai.app.services.session_queue.session_queue_common import (
     SessionQueueItemDTO,
     SessionQueueItemNotFoundError,
     SessionQueueStatus,
+    ValueToInsertTuple,
     calc_session_count,
     prepare_values_to_insert,
 )
@@ -689,7 +690,7 @@ class SqliteSessionQueue(SessionQueueBase):
         """Retries the given queue items"""
         try:
             cursor = self._conn.cursor()
-            values_to_insert: list[tuple] = []
+            values_to_insert: list[ValueToInsertTuple] = []
             retried_item_ids: list[int] = []
 
             for item_id in item_ids:
@@ -715,16 +716,16 @@ class SqliteSessionQueue(SessionQueueBase):
                     else queue_item.item_id
                 )
 
-                value_to_insert = (
+                value_to_insert: ValueToInsertTuple = (
                     queue_item.queue_id,
-                    queue_item.batch_id,
-                    queue_item.destination,
-                    field_values_json,
-                    queue_item.origin,
-                    queue_item.priority,
-                    workflow_json,
                     cloned_session_json,
                     cloned_session.id,
+                    queue_item.batch_id,
+                    field_values_json,
+                    queue_item.priority,
+                    workflow_json,
+                    queue_item.origin,
+                    queue_item.destination,
                     retried_from_item_id,
                 )
                 values_to_insert.append(value_to_insert)

invokeai/app/services/shared/invocation_context.py

@@ -20,14 +20,10 @@ from invokeai.app.services.session_processor.session_processor_common import Pro
 from invokeai.app.services.shared.sqlite.sqlite_common import SQLiteDirection
 from invokeai.app.util.step_callback import flux_step_callback, stable_diffusion_step_callback
 from invokeai.backend.model_manager.config import (
-    AnyModel,
     AnyModelConfig,
-    BaseModelType,
-    ModelFormat,
-    ModelType,
-    SubModelType,
 )
 from invokeai.backend.model_manager.load.load_base import LoadedModel, LoadedModelWithoutConfig
+from invokeai.backend.model_manager.taxonomy import AnyModel, BaseModelType, ModelFormat, ModelType, SubModelType
 from invokeai.backend.stable_diffusion.diffusers_pipeline import PipelineIntermediateState
 from invokeai.backend.stable_diffusion.diffusion.conditioning_data import ConditioningFieldData
 

invokeai/app/services/shared/sqlite/sqlite_util.py

@@ -20,6 +20,7 @@ from invokeai.app.services.shared.sqlite_migrator.migrations.migration_14 import
 from invokeai.app.services.shared.sqlite_migrator.migrations.migration_15 import build_migration_15
 from invokeai.app.services.shared.sqlite_migrator.migrations.migration_16 import build_migration_16
 from invokeai.app.services.shared.sqlite_migrator.migrations.migration_17 import build_migration_17
+from invokeai.app.services.shared.sqlite_migrator.migrations.migration_18 import build_migration_18
 from invokeai.app.services.shared.sqlite_migrator.sqlite_migrator_impl import SqliteMigrator
 
 
@@ -57,6 +58,7 @@ def init_db(config: InvokeAIAppConfig, logger: Logger, image_files: ImageFileSto
     migrator.register_migration(build_migration_15())
     migrator.register_migration(build_migration_16())
     migrator.register_migration(build_migration_17())
+    migrator.register_migration(build_migration_18())
     migrator.run_migrations()
 
     return db

invokeai/app/services/shared/sqlite_migrator/migrations/migration_18.py

@@ -0,0 +1,47 @@
+import sqlite3
+
+from invokeai.app.services.shared.sqlite_migrator.sqlite_migrator_common import Migration
+
+
+class Migration18Callback:
+    def __call__(self, cursor: sqlite3.Cursor) -> None:
+        self._make_workflow_opened_at_nullable(cursor)
+
+    def _make_workflow_opened_at_nullable(self, cursor: sqlite3.Cursor) -> None:
+        """
+        Make the `opened_at` column nullable in the `workflow_library` table. This is accomplished by:
+            - Dropping the existing `idx_workflow_library_opened_at` index (must be done before dropping the column)
+            - Dropping the existing `opened_at` column
+            - Adding a new nullable column `opened_at` (no data migration needed, all values will be NULL)
+            - Adding a new `idx_workflow_library_opened_at` index on the `opened_at` column
+        """
+        # For index renaming in SQLite, we need to drop and recreate
+        cursor.execute("DROP INDEX IF EXISTS idx_workflow_library_opened_at;")
+        # Rename existing column to deprecated
+        cursor.execute("ALTER TABLE workflow_library DROP COLUMN opened_at;")
+        # Add new nullable column - all values will be NULL - no migration of data needed
+        cursor.execute("ALTER TABLE workflow_library ADD COLUMN opened_at DATETIME;")
+        # Create new index on the new column
+        cursor.execute(
+            "CREATE INDEX idx_workflow_library_opened_at ON workflow_library(opened_at);",
+        )
+
+
+def build_migration_18() -> Migration:
+    """
+    Build the migration from database version 17 to 18.
+
+    This migration does the following:
+        - Make the `opened_at` column nullable in the `workflow_library` table. This is accomplished by:
+            - Dropping the existing `idx_workflow_library_opened_at` index (must be done before dropping the column)
+            - Dropping the existing `opened_at` column
+            - Adding a new nullable column `opened_at` (no data migration needed, all values will be NULL)
+            - Adding a new `idx_workflow_library_opened_at` index on the `opened_at` column
+    """
+    migration_18 = Migration(
+        from_version=17,
+        to_version=18,
+        callback=Migration18Callback(),
+    )
+
+    return migration_18

invokeai/app/services/workflow_records/default_workflows/ESRGAN Upscaling with Canny ControlNet.json

@@ -1,11 +1,11 @@
 {
   "id": "default_686bb1d0-d086-4c70-9fa3-2f600b922023",
-  "name": "ESRGAN Upscaling with Canny ControlNet",
+  "name": "Upscaler - SD1.5, ESRGAN",
   "author": "InvokeAI",
-  "description": "Sample workflow for using Upscaling with ControlNet with SD1.5",
+  "description": "Sample workflow for using ESRGAN to upscale with ControlNet with SD1.5",
   "version": "2.1.0",
   "contact": "invoke@invoke.ai",
-  "tags": "upscaling, controlnet, default",
+  "tags": "sd1.5, upscaling, control",
   "notes": "",
   "exposedFields": [
     {
@@ -185,14 +185,7 @@
           },
           "control_model": {
             "name": "control_model",
-            "label": "Control Model (select Canny)",
-            "value": {
-              "key": "a7b9c76f-4bc5-42aa-b918-c1c458a5bb24",
-              "hash": "blake3:260c7f8e10aefea9868cfc68d89970e91033bd37132b14b903e70ee05ebf530e",
-              "name": "sd-controlnet-canny",
-              "base": "sd-1",
-              "type": "controlnet"
-            }
+            "label": "Control Model (select Canny)"
           },
           "control_weight": {
             "name": "control_weight",
@@ -295,14 +288,7 @@
         "inputs": {
           "model": {
             "name": "model",
-            "label": "",
-            "value": {
-              "key": "5cd43ca0-dd0a-418d-9f7e-35b2b9d5e106",
-              "hash": "blake3:6987f323017f597213cc3264250edf57056d21a40a0a85d83a1a33a7d44dc41a",
-              "name": "Deliberate_v5",
-              "base": "sd-1",
-              "type": "main"
-            }
+            "label": ""
           }
         },
         "isOpen": true,
@@ -849,4 +835,4 @@
       "targetHandle": "image_resolution"
     }
   ]
-}
+}

invokeai/app/services/workflow_records/default_workflows/FLUX Image to Image.json

@@ -1,11 +1,11 @@
 {
   "id": "default_cbf0e034-7b54-4b2c-b670-3b1e2e4b4a88",
-  "name": "FLUX Image to Image",
+  "name": "Image to Image - FLUX",
   "author": "InvokeAI",
   "description": "A simple image-to-image workflow using a FLUX dev model. ",
   "version": "1.1.0",
   "contact": "",
-  "tags": "image2image, flux, image-to-image, image to image",
+  "tags": "flux, image to image",
   "notes": "Prerequisite model downloads: T5 Encoder, CLIP-L Encoder, and FLUX VAE. Quantized and un-quantized versions can be found in the starter models tab within your Model Manager. We recommend using FLUX dev models for image-to-image workflows. The image-to-image performance with FLUX schnell models is poor.",
   "exposedFields": [
     {
@@ -201,36 +201,15 @@
           },
           "t5_encoder_model": {
             "name": "t5_encoder_model",
-            "label": "",
-            "value": {
-              "key": "d18d5575-96b6-4da3-b3d8-eb58308d6705",
-              "hash": "random:f2f9ed74acdfb4bf6fec200e780f6c25f8dd8764a35e65d425d606912fdf573a",
-              "name": "t5_bnb_int8_quantized_encoder",
-              "base": "any",
-              "type": "t5_encoder"
-            }
+            "label": ""
           },
           "clip_embed_model": {
             "name": "clip_embed_model",
-            "label": "",
-            "value": {
-              "key": "5a19d7e5-8d98-43cd-8a81-87515e4b3b4e",
-              "hash": "random:4bd08514c08fb6ff04088db9aeb45def3c488e8b5fd09a35f2cc4f2dc346f99f",
-              "name": "clip-vit-large-patch14",
-              "base": "any",
-              "type": "clip_embed"
-            }
+            "label": ""
           },
           "vae_model": {
             "name": "vae_model",
-            "label": "",
-            "value": {
-              "key": "9172beab-5c1d-43f0-b2f0-6e0b956710d9",
-              "hash": "random:c54dde288e5fa2e6137f1c92e9d611f598049e6f16e360207b6d96c9f5a67ba0",
-              "name": "FLUX.1-schnell_ae",
-              "base": "flux",
-              "type": "vae"
-            }
+            "label": ""
           }
         }
       },

invokeai/app/services/workflow_records/default_workflows/Face Detailer with IP-Adapter & Canny (See Note in Details).json

@@ -1,11 +1,11 @@
 {
   "id": "default_dec5a2e9-f59c-40d9-8869-a056751d79b8",
-  "name": "Face Detailer with IP-Adapter & Canny (See Note in Details)",
+  "name": "Face Detailer - SD1.5",
   "author": "kosmoskatten",
   "description": "A workflow to add detail to and improve faces. This workflow is most effective when used with a model that creates realistic outputs. ",
   "version": "2.1.0",
   "contact": "invoke@invoke.ai",
-  "tags": "face detailer, IP-Adapter, Canny",
+  "tags": "sd1.5, reference image, control",
   "notes": "Set this image as the blur mask: https://i.imgur.com/Gxi61zP.png",
   "exposedFields": [
     {
@@ -136,14 +136,7 @@
           },
           "control_model": {
             "name": "control_model",
-            "label": "Control Model (select canny)",
-            "value": {
-              "key": "5bdaacf7-a7a3-4fb8-b394-cc0ffbb8941d",
-              "hash": "blake3:260c7f8e10aefea9868cfc68d89970e91033bd37132b14b903e70ee05ebf530e",
-              "name": "sd-controlnet-canny",
-              "base": "sd-1",
-              "type": "controlnet"
-            }
+            "label": "Control Model (select canny)"
           },
           "control_weight": {
             "name": "control_weight",
@@ -197,14 +190,7 @@
           },
           "ip_adapter_model": {
             "name": "ip_adapter_model",
-            "label": "IP-Adapter Model (select IP Adapter Face)",
-            "value": {
-              "key": "1cc210bb-4d0a-4312-b36c-b5d46c43768e",
-              "hash": "blake3:3d669dffa7471b357b4df088b99ffb6bf4d4383d5e0ef1de5ec1c89728a3d5a5",
-              "name": "ip_adapter_sd15",
-              "base": "sd-1",
-              "type": "ip_adapter"
-            }
+            "label": "IP-Adapter Model (select IP Adapter Face)"
           },
           "clip_vision_model": {
             "name": "clip_vision_model",

invokeai/app/services/workflow_records/default_workflows/Flux Text to Image.json

@@ -1,11 +1,11 @@
 {
   "id": "default_444fe292-896b-44fd-bfc6-c0b5d220fffc",
-  "name": "FLUX Text to Image",
+  "name": "Text to Image - FLUX",
   "author": "InvokeAI",
   "description": "A simple text-to-image workflow using FLUX dev or schnell models.",
   "version": "1.1.0",
   "contact": "",
-  "tags": "text2image, flux, text to image",
+  "tags": "flux, text to image",
   "notes": "Prerequisite model downloads: T5 Encoder, CLIP-L Encoder, and FLUX VAE. Quantized and un-quantized versions can be found in the starter models tab within your Model Manager. We recommend 4 steps for FLUX schnell models and 30 steps for FLUX dev models.",
   "exposedFields": [
     {
@@ -169,36 +169,15 @@
           },
           "t5_encoder_model": {
             "name": "t5_encoder_model",
-            "label": "",
-            "value": {
-              "key": "d18d5575-96b6-4da3-b3d8-eb58308d6705",
-              "hash": "random:f2f9ed74acdfb4bf6fec200e780f6c25f8dd8764a35e65d425d606912fdf573a",
-              "name": "t5_bnb_int8_quantized_encoder",
-              "base": "any",
-              "type": "t5_encoder"
-            }
+            "label": ""
           },
           "clip_embed_model": {
             "name": "clip_embed_model",
-            "label": "",
-            "value": {
-              "key": "5a19d7e5-8d98-43cd-8a81-87515e4b3b4e",
-              "hash": "random:4bd08514c08fb6ff04088db9aeb45def3c488e8b5fd09a35f2cc4f2dc346f99f",
-              "name": "clip-vit-large-patch14",
-              "base": "any",
-              "type": "clip_embed"
-            }
+            "label": ""
           },
           "vae_model": {
             "name": "vae_model",
-            "label": "",
-            "value": {
-              "key": "9172beab-5c1d-43f0-b2f0-6e0b956710d9",
-              "hash": "random:c54dde288e5fa2e6137f1c92e9d611f598049e6f16e360207b6d96c9f5a67ba0",
-              "name": "FLUX.1-schnell_ae",
-              "base": "flux",
-              "type": "vae"
-            }
+            "label": ""
           }
         }
       },

invokeai/app/services/workflow_records/default_workflows/Multi ControlNet (Canny & Depth).json

@@ -1,11 +1,11 @@
 {
   "id": "default_2d05e719-a6b9-4e64-9310-b875d3b2f9d2",
-  "name": "Multi ControlNet (Canny & Depth)",
+  "name": "Text to Image -  SD1.5, Control",
   "author": "InvokeAI",
   "description": "A sample workflow using canny & depth ControlNets to guide the generation process. ",
   "version": "2.1.0",
   "contact": "invoke@invoke.ai",
-  "tags": "ControlNet, canny, depth",
+  "tags": "sd1.5, control, text to image",
   "notes": "",
   "exposedFields": [
     {
@@ -217,14 +217,7 @@
           },
           "control_model": {
             "name": "control_model",
-            "label": "Control Model (select canny)",
-            "value": {
-              "key": "5bdaacf7-a7a3-4fb8-b394-cc0ffbb8941d",
-              "hash": "blake3:260c7f8e10aefea9868cfc68d89970e91033bd37132b14b903e70ee05ebf530e",
-              "name": "sd-controlnet-canny",
-              "base": "sd-1",
-              "type": "controlnet"
-            }
+            "label": "Control Model (select canny)"
           },
           "control_weight": {
             "name": "control_weight",
@@ -371,14 +364,7 @@
           },
           "control_model": {
             "name": "control_model",
-            "label": "Control Model (select depth)",
-            "value": {
-              "key": "87e8855c-671f-4c9e-bbbb-8ed47ccb4aac",
-              "hash": "blake3:2550bf22a53942dfa28ab2fed9d10d80851112531f44d977168992edf9d0534c",
-              "name": "control_v11f1p_sd15_depth",
-              "base": "sd-1",
-              "type": "controlnet"
-            }
+            "label": "Control Model (select depth)"
           },
           "control_weight": {
             "name": "control_weight",

invokeai/app/services/workflow_records/default_workflows/MultiDiffusion SD1.5.json

@@ -1,11 +1,11 @@
 {
   "id": "default_f96e794f-eb3e-4d01-a960-9b4e43402bcf",
-  "name": "MultiDiffusion SD1.5",
+  "name": "Upscaler - SD1.5, MultiDiffusion",
   "author": "Invoke",
   "description": "A workflow to upscale an input image with tiled upscaling, using SD1.5 based models.",
   "version": "1.0.0",
   "contact": "invoke@invoke.ai",
-  "tags": "tiled, upscaling, sdxl",
+  "tags": "sd1.5, upscaling",
   "notes": "",
   "exposedFields": [
     {
@@ -135,14 +135,7 @@
         "inputs": {
           "model": {
             "name": "model",
-            "label": "",
-            "value": {
-              "key": "e7b402e5-62e5-4acb-8c39-bee6bdb758ab",
-              "hash": "c8659e796168d076368256b57edbc1b48d6dafc1712f1bb37cc57c7c06889a6b",
-              "name": "526mix",
-              "base": "sd-1",
-              "type": "main"
-            }
+            "label": ""
           }
         }
       },
@@ -384,21 +377,11 @@
           },
           "image": {
             "name": "image",
-            "label": "Image to Upscale",
-            "value": {
-              "image_name": "ee7009f7-a35d-488b-a2a6-21237ef5ae05.png"
-            }
+            "label": "Image to Upscale"
           },
           "image_to_image_model": {
             "name": "image_to_image_model",
-            "label": "",
-            "value": {
-              "key": "38bb1a29-8ede-42ba-b77f-64b3478896eb",
-              "hash": "blake3:e52fdbee46a484ebe9b3b20ea0aac0a35a453ab6d0d353da00acfd35ce7a91ed",
-              "name": "4xNomosWebPhoto_esrgan",
-              "base": "sdxl",
-              "type": "spandrel_image_to_image"
-            }
+            "label": ""
           },
           "tile_size": {
             "name": "tile_size",
@@ -437,14 +420,7 @@
         "inputs": {
           "model": {
             "name": "model",
-            "label": "ControlNet Model - Choose a Tile ControlNet",
-            "value": {
-              "key": "20645e4d-ef97-4c5a-9243-b834a3483925",
-              "hash": "f0812e13758f91baf4e54b7dbb707b70642937d3b2098cd2b94cc36d3eba308e",
-              "name": "tile",
-              "base": "sd-1",
-              "type": "controlnet"
-            }
+            "label": "ControlNet Model - Choose a Tile ControlNet"
           }
         }
       },

invokeai/app/services/workflow_records/default_workflows/MultiDiffusion SDXL.json

@@ -1,11 +1,11 @@
 {
   "id": "default_35658541-6d41-4a20-8ec5-4bf2561faed0",
-  "name": "MultiDiffusion SDXL",
+  "name": "Upscaler - SDXL, MultiDiffusion",
   "author": "Invoke",
   "description": "A workflow to upscale an input image with tiled upscaling, using SDXL based models.",
   "version": "1.1.0",
   "contact": "invoke@invoke.ai",
-  "tags": "tiled, upscaling, sdxl",
+  "tags": "sdxl, upscaling",
   "notes": "",
   "exposedFields": [
     {
@@ -341,14 +341,7 @@
         "inputs": {
           "model": {
             "name": "model",
-            "label": "ControlNet Model - Choose a Tile ControlNet",
-            "value": {
-              "key": "74f4651f-0ace-4b7b-b616-e98360257797",
-              "hash": "blake3:167a5b84583aaed3e5c8d660b45830e82e1c602743c689d3c27773c6c8b85b4a",
-              "name": "controlnet-tile-sdxl-1.0",
-              "base": "sdxl",
-              "type": "controlnet"
-            }
+            "label": "ControlNet Model - Choose a Tile ControlNet"
           }
         }
       },
@@ -801,14 +794,7 @@
         "inputs": {
           "vae_model": {
             "name": "vae_model",
-            "label": "",
-            "value": {
-              "key": "ff926845-090e-4d46-b81e-30289ee47474",
-              "hash": "9705ab1c31fa96b308734214fb7571a958621c7a9247eed82b7d277145f8d9fa",
-              "name": "VAEFix",
-              "base": "sdxl",
-              "type": "vae"
-            }
+            "label": ""
           }
         }
       },
@@ -832,14 +818,7 @@
         "inputs": {
           "model": {
             "name": "model",
-            "label": "SDXL Model",
-            "value": {
-              "key": "ab191f73-68d2-492c-8aec-b438a8cf0f45",
-              "hash": "blake3:2d50e940627e3bf555f015280ec0976d5c1fa100f7bc94e95ffbfc770e98b6fe",
-              "name": "CustomXLv7",
-              "base": "sdxl",
-              "type": "main"
-            }
+            "label": "SDXL Model"
           }
         }
       },

invokeai/app/services/workflow_records/default_workflows/Prompt from File.json

@@ -1,11 +1,11 @@
 {
   "id": "default_d7a1c60f-ca2f-4f90-9e33-75a826ca6d8f",
-  "name": "Prompt from File",
+  "name": "Text to Image - SD1.5, Prompt from File",
   "author": "InvokeAI",
   "description": "Sample workflow using Prompt from File node",
   "version": "2.1.0",
   "contact": "invoke@invoke.ai",
-  "tags": "text2image, prompt from file, default, text to image",
+  "tags": "sd1.5, text to image",
   "notes": "",
   "exposedFields": [
     {
@@ -513,4 +513,4 @@
       "targetHandle": "vae"
     }
   ]
-}
+}

invokeai/app/services/workflow_records/default_workflows/README.md

@@ -10,6 +10,7 @@ _default workflows_ on app startup.
   An exception will be raised during sync if this is not set correctly.
 - Default workflows appear on the "Default Workflows" tab of the Workflow
   Library.
+- Default workflows should not reference any resources that are user-created or installed. That includes images and models. For example, if a default workflow references Juggernaut as an SDXL model, when a user loads the workflow, even if they have a version of Juggernaut installed, it will have a different UUID. They may see a warning. So, it's best to ship default workflows without any references to these types of resources.
 
 After adding or updating default workflows, you **must** start the app up and
 load them to ensure:

invokeai/app/services/workflow_records/default_workflows/SD3.5 Text to Image.json

@@ -1,11 +1,11 @@
 {
   "id": "default_dbe46d95-22aa-43fb-9c16-94400d0ce2fd",
-  "name": "SD3.5 Text to Image",
+  "name": "Text to Image - SD3.5",
   "author": "InvokeAI",
   "description": "Sample text to image workflow for Stable Diffusion 3.5",
   "version": "1.0.0",
   "contact": "invoke@invoke.ai",
-  "tags": "text2image, SD3.5, text to image",
+  "tags": "SD3.5, text to image",
   "notes": "",
   "exposedFields": [
     {
@@ -38,14 +38,7 @@
         "inputs": {
           "model": {
             "name": "model",
-            "label": "",
-            "value": {
-              "key": "f7b20be9-92a8-4cfb-bca4-6c3b5535c10b",
-              "hash": "placeholder",
-              "name": "stable-diffusion-3.5-medium",
-              "base": "sd-3",
-              "type": "main"
-            }
+            "label": ""
           },
           "t5_encoder_model": {
             "name": "t5_encoder_model",

invokeai/app/services/workflow_records/default_workflows/Text to Image - SD1.5.json

@@ -5,7 +5,7 @@
   "description": "Sample text to image workflow for Stable Diffusion 1.5/2",
   "version": "2.1.0",
   "contact": "invoke@invoke.ai",
-  "tags": "text2image, SD1.5, SD2, text to image",
+  "tags": "SD1.5, text to image",
   "notes": "",
   "exposedFields": [
     {
@@ -417,4 +417,4 @@
       "targetHandle": "vae"
     }
   ]
-}
+}

invokeai/app/services/workflow_records/default_workflows/Text to Image - SDXL.json

@@ -5,7 +5,7 @@
   "description": "Sample text to image workflow for SDXL",
   "version": "2.1.0",
   "contact": "invoke@invoke.ai",
-  "tags": "text2image, SDXL, text to image",
+  "tags": "SDXL, text to image",
   "notes": "",
   "exposedFields": [
     {
@@ -46,14 +46,7 @@
         "inputs": {
           "vae_model": {
             "name": "vae_model",
-            "label": "VAE (use the FP16 model)",
-            "value": {
-              "key": "f20f9e5c-1bce-4c46-a84d-34ebfa7df069",
-              "hash": "blake3:9705ab1c31fa96b308734214fb7571a958621c7a9247eed82b7d277145f8d9fa",
-              "name": "sdxl-vae-fp16-fix",
-              "base": "sdxl",
-              "type": "vae"
-            }
+            "label": "VAE (use the FP16 model)"
           }
         },
         "isOpen": true,
@@ -203,14 +196,7 @@
         "inputs": {
           "model": {
             "name": "model",
-            "label": "",
-            "value": {
-              "key": "4a63b226-e8ff-4da4-854e-0b9f04b562ba",
-              "hash": "blake3:d279309ea6e5ee6e8fd52504275865cc280dac71cbf528c5b07c98b888bddaba",
-              "name": "dreamshaper-xl-v2-turbo",
-              "base": "sdxl",
-              "type": "main"
-            }
+            "label": ""
           }
         },
         "isOpen": true,
@@ -715,4 +701,4 @@
       "targetHandle": "style"
     }
   ]
-}
+}

invokeai/app/services/workflow_records/default_workflows/Text to Image with LoRA.json

@@ -1,11 +1,11 @@
 {
   "id": "default_e71d153c-2089-43c7-bd2c-f61f37d4c1c1",
-  "name": "Text to Image with LoRA",
+  "name": "Text to Image - SD1.5, LoRA",
   "author": "InvokeAI",
   "description": "Simple text to image workflow with a LoRA",
   "version": "2.1.0",
   "contact": "invoke@invoke.ai",
-  "tags": "text to image, lora, text to image",
+  "tags": "sd1.5, text to image, lora",
   "notes": "",
   "exposedFields": [
     {

invokeai/app/services/workflow_records/default_workflows/Tiled Upscaling (Beta).json

@@ -1,11 +1,11 @@
 {
   "id": "default_43b0d7f7-6a12-4dcf-a5a4-50c940cbee29",
-  "name": "Tiled Upscaling (Beta)",
+  "name": "Upscaler - SD1.5, Tiled",
   "author": "Invoke",
   "description": "A workflow to upscale an input image with tiled upscaling. ",
   "version": "2.1.0",
   "contact": "invoke@invoke.ai",
-  "tags": "tiled, upscaling, sd1.5",
+  "tags": "sd1.5, upscaling",
   "notes": "",
   "exposedFields": [
     {
@@ -86,14 +86,7 @@
           },
           "ip_adapter_model": {
             "name": "ip_adapter_model",
-            "label": "IP-Adapter Model (select ip_adapter_sd15)",
-            "value": {
-              "key": "1cc210bb-4d0a-4312-b36c-b5d46c43768e",
-              "hash": "blake3:3d669dffa7471b357b4df088b99ffb6bf4d4383d5e0ef1de5ec1c89728a3d5a5",
-              "name": "ip_adapter_sd15",
-              "base": "sd-1",
-              "type": "ip_adapter"
-            }
+            "label": "IP-Adapter Model (select ip_adapter_sd15)"
           },
           "clip_vision_model": {
             "name": "clip_vision_model",
@@ -201,14 +194,7 @@
           },
           "control_model": {
             "name": "control_model",
-            "label": "Control Model (select contro_v11f1e_sd15_tile)",
-            "value": {
-              "key": "773843c8-db1f-4502-8f65-59782efa7960",
-              "hash": "blake3:f0812e13758f91baf4e54b7dbb707b70642937d3b2098cd2b94cc36d3eba308e",
-              "name": "control_v11f1e_sd15_tile",
-              "base": "sd-1",
-              "type": "controlnet"
-            }
+            "label": "Control Model (select control_v11f1e_sd15_tile)"
           },
           "control_weight": {
             "name": "control_weight",
@@ -1816,4 +1802,4 @@
       "targetHandle": "unet"
     }
   ]
-}
+}

invokeai/app/services/workflow_records/workflow_records_base.py

@@ -46,17 +46,28 @@ class WorkflowRecordsStorageBase(ABC):
         per_page: Optional[int],
         query: Optional[str],
         tags: Optional[list[str]],
+        has_been_opened: Optional[bool],
     ) -> PaginatedResults[WorkflowRecordListItemDTO]:
         """Gets many workflows."""
         pass
 
     @abstractmethod
-    def get_counts(
+    def counts_by_category(
         self,
-        tags: Optional[list[str]],
-        categories: Optional[list[WorkflowCategory]],
-    ) -> int:
-        """Gets the count of workflows for the given tags and categories."""
+        categories: list[WorkflowCategory],
+        has_been_opened: Optional[bool] = None,
+    ) -> dict[str, int]:
+        """Gets a dictionary of counts for each of the provided categories."""
+        pass
+
+    @abstractmethod
+    def counts_by_tag(
+        self,
+        tags: list[str],
+        categories: Optional[list[WorkflowCategory]] = None,
+        has_been_opened: Optional[bool] = None,
+    ) -> dict[str, int]:
+        """Gets a dictionary of counts for each of the provided tags."""
         pass
 
     @abstractmethod

invokeai/app/services/workflow_records/workflow_records_common.py

@@ -1,6 +1,6 @@
 import datetime
 from enum import Enum
-from typing import Any, Union
+from typing import Any, Optional, Union
 
 import semver
 from pydantic import BaseModel, ConfigDict, Field, JsonValue, TypeAdapter, field_validator
@@ -98,7 +98,9 @@ class WorkflowRecordDTOBase(BaseModel):
     name: str = Field(description="The name of the workflow.")
     created_at: Union[datetime.datetime, str] = Field(description="The created timestamp of the workflow.")
     updated_at: Union[datetime.datetime, str] = Field(description="The updated timestamp of the workflow.")
-    opened_at: Union[datetime.datetime, str] = Field(description="The opened timestamp of the workflow.")
+    opened_at: Optional[Union[datetime.datetime, str]] = Field(
+        default=None, description="The opened timestamp of the workflow."
+    )
 
 
 class WorkflowRecordDTO(WorkflowRecordDTOBase):

invokeai/app/services/workflow_records/workflow_records_sqlite.py

@@ -118,6 +118,7 @@ class SqliteWorkflowRecordsStorage(WorkflowRecordsStorageBase):
         per_page: Optional[int] = None,
         query: Optional[str] = None,
         tags: Optional[list[str]] = None,
+        has_been_opened: Optional[bool] = None,
     ) -> PaginatedResults[WorkflowRecordListItemDTO]:
         # sanitize!
         assert order_by in WorkflowRecordOrderBy
@@ -175,6 +176,11 @@ class SqliteWorkflowRecordsStorage(WorkflowRecordsStorageBase):
             conditions.append(tags_condition)
             params.extend(tags_params)
 
+        if has_been_opened:
+            conditions.append("opened_at IS NOT NULL")
+        elif has_been_opened is False:
+            conditions.append("opened_at IS NULL")
+
         # Ignore whitespace in the query
         stripped_query = query.strip() if query else None
         if stripped_query:
@@ -230,54 +236,105 @@ class SqliteWorkflowRecordsStorage(WorkflowRecordsStorageBase):
             total=total,
         )
 
-    def get_counts(
+    def counts_by_tag(
         self,
-        tags: Optional[list[str]],
-        categories: Optional[list[WorkflowCategory]],
-    ) -> int:
+        tags: list[str],
+        categories: Optional[list[WorkflowCategory]] = None,
+        has_been_opened: Optional[bool] = None,
+    ) -> dict[str, int]:
+        if not tags:
+            return {}
+
         cursor = self._conn.cursor()
+        result: dict[str, int] = {}
+        # Base conditions for categories and selected tags
+        base_conditions: list[str] = []
+        base_params: list[str | int] = []
 
-        # Start with an empty list of conditions and params
-        conditions: list[str] = []
-        params: list[str | int] = []
-
-        if tags:
-            # Construct a list of conditions for each tag
-            tags_conditions = ["tags LIKE ?" for _ in tags]
-            tags_conditions_joined = " OR ".join(tags_conditions)
-            tags_condition = f"({tags_conditions_joined})"
-
-            # And the params for the tags, case-insensitive
-            tags_params = [f"%{t.strip()}%" for t in tags]
-
-            conditions.append(tags_condition)
-            params.extend(tags_params)
-
+        # Add category conditions
         if categories:
-            # Ensure all categories are valid (is this necessary?)
             assert all(c in WorkflowCategory for c in categories)
-
-            # Construct a placeholder string for the number of categories
             placeholders = ", ".join("?" for _ in categories)
+            base_conditions.append(f"category IN ({placeholders})")
+            base_params.extend([category.value for category in categories])
 
-            # Construct the condition string & params
-            conditions.append(f"category IN ({placeholders})")
-            params.extend([category.value for category in categories])
+        if has_been_opened:
+            base_conditions.append("opened_at IS NOT NULL")
+        elif has_been_opened is False:
+            base_conditions.append("opened_at IS NULL")
 
-        stmt = """--sql
-            SELECT COUNT(*)
-            FROM workflow_library
-            """
+        # For each tag to count, run a separate query
+        for tag in tags:
+            # Start with the base conditions
+            conditions = base_conditions.copy()
+            params = base_params.copy()
 
-        if conditions:
-            # If there are conditions, add a WHERE clause and then join the conditions
-            stmt += " WHERE "
+            # Add this specific tag condition
+            conditions.append("tags LIKE ?")
+            params.append(f"%{tag.strip()}%")
 
-            all_conditions = " AND ".join(conditions)
-            stmt += all_conditions
+            # Construct the full query
+            stmt = """--sql
+                SELECT COUNT(*)
+                FROM workflow_library
+                """
 
-        cursor.execute(stmt, tuple(params))
-        return cursor.fetchone()[0]
+            if conditions:
+                stmt += " WHERE " + " AND ".join(conditions)
+
+            cursor.execute(stmt, params)
+            count = cursor.fetchone()[0]
+            result[tag] = count
+
+        return result
+
+    def counts_by_category(
+        self,
+        categories: list[WorkflowCategory],
+        has_been_opened: Optional[bool] = None,
+    ) -> dict[str, int]:
+        cursor = self._conn.cursor()
+        result: dict[str, int] = {}
+        # Base conditions for categories
+        base_conditions: list[str] = []
+        base_params: list[str | int] = []
+
+        # Add category conditions
+        if categories:
+            assert all(c in WorkflowCategory for c in categories)
+            placeholders = ", ".join("?" for _ in categories)
+            base_conditions.append(f"category IN ({placeholders})")
+            base_params.extend([category.value for category in categories])
+
+        if has_been_opened:
+            base_conditions.append("opened_at IS NOT NULL")
+        elif has_been_opened is False:
+            base_conditions.append("opened_at IS NULL")
+
+        # For each category to count, run a separate query
+        for category in categories:
+            # Start with the base conditions
+            conditions = base_conditions.copy()
+            params = base_params.copy()
+
+            # Add this specific category condition
+            conditions.append("category = ?")
+            params.append(category.value)
+
+            # Construct the full query
+            stmt = """--sql
+                SELECT COUNT(*)
+                FROM workflow_library
+                """
+
+            if conditions:
+                stmt += " WHERE " + " AND ".join(conditions)
+
+            cursor.execute(stmt, params)
+            count = cursor.fetchone()[0]
+            result[category.value] = count
+
+        return result
 
     def update_opened_at(self, workflow_id: str) -> None:
         try:

invokeai/app/services/workflow_thumbnails/workflow_thumbnails_base.py

@@ -8,12 +8,12 @@ class WorkflowThumbnailServiceBase(ABC):
     """Base class for workflow thumbnail services"""
 
     @abstractmethod
-    def get_path(self, workflow_id: str) -> Path:
+    def get_path(self, workflow_id: str, with_hash: bool = True) -> Path:
         """Gets the path to a workflow thumbnail"""
         pass
 
     @abstractmethod
-    def get_url(self, workflow_id: str) -> str | None:
+    def get_url(self, workflow_id: str, with_hash: bool = True) -> str | None:
         """Gets the URL of a workflow thumbnail"""
         pass
 

invokeai/app/services/workflow_thumbnails/workflow_thumbnails_disk.py

@@ -41,7 +41,7 @@ class WorkflowThumbnailFileStorageDisk(WorkflowThumbnailServiceBase):
         except Exception as e:
             raise WorkflowThumbnailFileSaveException from e
 
-    def get_path(self, workflow_id: str) -> Path:
+    def get_path(self, workflow_id: str, with_hash: bool = True) -> Path:
         workflow = self._invoker.services.workflow_records.get(workflow_id).workflow
         if workflow.meta.category is WorkflowCategory.Default:
             default_thumbnails_dir = Path(__file__).parent / Path("default_workflow_thumbnails")
@@ -51,7 +51,7 @@ class WorkflowThumbnailFileStorageDisk(WorkflowThumbnailServiceBase):
 
         return path
 
-    def get_url(self, workflow_id: str) -> str | None:
+    def get_url(self, workflow_id: str, with_hash: bool = True) -> str | None:
         path = self.get_path(workflow_id)
         if not self._validate_path(path):
             return
@@ -59,7 +59,8 @@ class WorkflowThumbnailFileStorageDisk(WorkflowThumbnailServiceBase):
         url = self._invoker.services.urls.get_workflow_thumbnail_url(workflow_id)
 
         # The image URL never changes, so we must add random query string to it to prevent caching
-        url += f"?{uuid_string()}"
+        if with_hash:
+            url += f"?{uuid_string()}"
 
         return url
 

invokeai/app/util/startup_utils.py

@@ -1,6 +1,7 @@
 import logging
 import mimetypes
 import socket
+from pathlib import Path
 
 import torch
 
@@ -33,7 +34,16 @@ def check_cudnn(logger: logging.Logger) -> None:
             )
 
 
-def enable_dev_reload() -> None:
+def invokeai_source_dir() -> Path:
+    # `invokeai.__file__` doesn't always work for editable installs
+    this_module_path = Path(__file__).resolve()
+    # https://youtrack.jetbrains.com/issue/PY-38382/Unresolved-reference-spec-but-this-is-standard-builtin
+    # noinspection PyUnresolvedReferences
+    depth = len(__spec__.parent.split("."))
+    return this_module_path.parents[depth - 1]
+
+
+def enable_dev_reload(custom_nodes_path=None) -> None:
     """Enable hot reloading on python file changes during development."""
     from invokeai.backend.util.logging import InvokeAILogger
 
@@ -44,7 +54,10 @@ def enable_dev_reload() -> None:
             'Can\'t start `--dev_reload` because jurigged is not found; `pip install -e ".[dev]"` to include development dependencies.'
         ) from e
     else:
-        jurigged.watch(logger=InvokeAILogger.get_logger(name="jurigged").info)
+        paths = [str(invokeai_source_dir() / "*.py")]
+        if custom_nodes_path:
+            paths.append(str(custom_nodes_path / "*.py"))
+        jurigged.watch(pattern=paths, logger=InvokeAILogger.get_logger(name="jurigged").info)
 
 
 def apply_monkeypatches() -> None:

invokeai/app/util/step_callback.py

@@ -5,7 +5,7 @@ import torch
 from PIL import Image
 
 from invokeai.app.services.session_processor.session_processor_common import CanceledException
-from invokeai.backend.model_manager.config import BaseModelType
+from invokeai.backend.model_manager.taxonomy import BaseModelType
 from invokeai.backend.stable_diffusion.diffusers_pipeline import PipelineIntermediateState
 
 # fast latents preview matrix for sdxl

invokeai/app/util/t5_model_identifier.py

@@ -1,5 +1,5 @@
 from invokeai.app.invocations.model import ModelIdentifierField
-from invokeai.backend.model_manager.config import BaseModelType, SubModelType
+from invokeai.backend.model_manager.taxonomy import BaseModelType, SubModelType
 
 
 def preprocess_t5_encoder_model_identifier(model_identifier: ModelIdentifierField) -> ModelIdentifierField:

invokeai/app/util/ti_utils.py

@@ -4,7 +4,7 @@ from typing import List, Tuple
 import invokeai.backend.util.logging as logger
 from invokeai.app.services.model_records import UnknownModelException
 from invokeai.app.services.shared.invocation_context import InvocationContext
-from invokeai.backend.model_manager.config import BaseModelType, ModelType
+from invokeai.backend.model_manager.taxonomy import BaseModelType, ModelType
 from invokeai.backend.textual_inversion import TextualInversionModelRaw
 
 

invokeai/backend/flux/util.py

@@ -20,6 +20,7 @@ class ModelSpec:
 
 max_seq_lengths: Dict[str, Literal[256, 512]] = {
     "flux-dev": 512,
+    "flux-dev-fill": 512,
     "flux-schnell": 256,
 }
 
@@ -68,4 +69,19 @@ params = {
         qkv_bias=True,
         guidance_embed=False,
     ),
+    "flux-dev-fill": FluxParams(
+        in_channels=384,
+        out_channels=64,
+        vec_in_dim=768,
+        context_in_dim=4096,
+        hidden_size=3072,
+        mlp_ratio=4.0,
+        num_heads=24,
+        depth=19,
+        depth_single_blocks=38,
+        axes_dim=[16, 56, 56],
+        theta=10_000,
+        qkv_bias=True,
+        guidance_embed=True,
+    ),
 }

invokeai/backend/image_util/infill_methods/lama.py

@@ -6,8 +6,8 @@ import torch
 from PIL import Image
 
 import invokeai.backend.util.logging as logger
-from invokeai.backend.model_manager.config import AnyModel
 from invokeai.backend.model_manager.load.model_cache.utils import get_effective_device
+from invokeai.backend.model_manager.taxonomy import AnyModel
 
 
 def norm_img(np_img):

invokeai/backend/image_util/normal_bae/nets/submodules/efficientnet_repo/geffnet/conv2d_layers.py

@@ -16,7 +16,7 @@ import torch
 import torch.nn as nn
 import torch.nn.functional as F
 
-from .config import *
+from .config import is_exportable, is_scriptable
 
 
 # From PyTorch internals

invokeai/backend/image_util/normal_bae/nets/submodules/efficientnet_repo/geffnet/efficientnet_builder.py

@@ -5,8 +5,8 @@ Copyright 2020 Ross Wightman
 import re
 from copy import deepcopy
 
-from .conv2d_layers import *
-from geffnet.activations import *
+from .conv2d_layers import CondConv2d, get_condconv_initializer, math, partial, select_conv2d
+from geffnet.activations import F, get_act_layer, nn, sigmoid, torch
 
 __all__ = ['get_bn_args_tf', 'resolve_bn_args', 'resolve_se_args', 'resolve_act_layer', 'make_divisible',
            'round_channels', 'drop_connect', 'SqueezeExcite', 'ConvBnAct', 'DepthwiseSeparableConv',

invokeai/backend/image_util/normal_bae/nets/submodules/efficientnet_repo/geffnet/gen_efficientnet.py

@@ -32,7 +32,9 @@ import torch.nn.functional as F
 from .config import layer_config_kwargs, is_scriptable
 from .conv2d_layers import select_conv2d
 from .helpers import load_pretrained
-from .efficientnet_builder import *
+from .efficientnet_builder import (BN_EPS_TF_DEFAULT, EfficientNetBuilder, decode_arch_def,
+                                   initialize_weight_default, initialize_weight_goog,
+                                   resolve_act_layer, resolve_bn_args, round_channels)
 
 __all__ = ['GenEfficientNet', 'mnasnet_050', 'mnasnet_075', 'mnasnet_100', 'mnasnet_b1', 'mnasnet_140',
            'semnasnet_050', 'semnasnet_075', 'semnasnet_100', 'mnasnet_a1', 'semnasnet_140', 'mnasnet_small',

invokeai/backend/image_util/normal_bae/nets/submodules/efficientnet_repo/geffnet/mobilenetv3.py

@@ -13,7 +13,9 @@ from .activations import get_act_fn, get_act_layer, HardSwish
 from .config import layer_config_kwargs
 from .conv2d_layers import select_conv2d
 from .helpers import load_pretrained
-from .efficientnet_builder import *
+from .efficientnet_builder import (BN_EPS_TF_DEFAULT, EfficientNetBuilder, decode_arch_def,
+                                   initialize_weight_default, initialize_weight_goog,
+                                   resolve_act_layer, resolve_bn_args, round_channels)
 
 __all__ = ['mobilenetv3_rw', 'mobilenetv3_large_075', 'mobilenetv3_large_100', 'mobilenetv3_large_minimal_100',
            'mobilenetv3_small_075', 'mobilenetv3_small_100', 'mobilenetv3_small_minimal_100',

invokeai/backend/image_util/realesrgan/realesrgan.py

@@ -10,7 +10,7 @@ from cv2.typing import MatLike
 from tqdm import tqdm
 
 from invokeai.backend.image_util.basicsr.rrdbnet_arch import RRDBNet
-from invokeai.backend.model_manager.config import AnyModel
+from invokeai.backend.model_manager.taxonomy import AnyModel
 from invokeai.backend.util.devices import TorchDevice
 
 """

invokeai/backend/llava_onevision_model.py

@@ -0,0 +1,56 @@
+from pathlib import Path
+from typing import Optional
+
+import torch
+from PIL.Image import Image
+from transformers import AutoProcessor, LlavaOnevisionForConditionalGeneration, LlavaOnevisionProcessor
+
+from invokeai.backend.raw_model import RawModel
+
+
+class LlavaOnevisionModel(RawModel):
+    def __init__(self, vllm_model: LlavaOnevisionForConditionalGeneration, processor: LlavaOnevisionProcessor):
+        self._vllm_model = vllm_model
+        self._processor = processor
+
+    @classmethod
+    def load_from_path(cls, path: str | Path):
+        vllm_model = LlavaOnevisionForConditionalGeneration.from_pretrained(path, local_files_only=True)
+        assert isinstance(vllm_model, LlavaOnevisionForConditionalGeneration)
+        processor = AutoProcessor.from_pretrained(path, local_files_only=True)
+        assert isinstance(processor, LlavaOnevisionProcessor)
+        return cls(vllm_model, processor)
+
+    def run(self, prompt: str, images: list[Image], device: torch.device, dtype: torch.dtype) -> str:
+        # TODO(ryand): Tune the max number of images that are useful for the model.
+        if len(images) > 3:
+            raise ValueError(
+                f"{len(images)} images were provided as input to the LLaVA OneVision model. "
+                "Pass <=3 images for good performance."
+            )
+
+        # Define a chat history and use `apply_chat_template` to get correctly formatted prompt.
+        # "content" is a list of dicts with types "text" or "image".
+        content = [{"type": "text", "text": prompt}]
+        # Add the correct number of images.
+        for _ in images:
+            content.append({"type": "image"})
+
+        conversation = [{"role": "user", "content": content}]
+        prompt = self._processor.apply_chat_template(conversation, add_generation_prompt=True)
+        inputs = self._processor(images=images or None, text=prompt, return_tensors="pt").to(device=device, dtype=dtype)
+        output = self._vllm_model.generate(**inputs, max_new_tokens=400, do_sample=False)
+        output_str: str = self._processor.decode(output[0][2:], skip_special_tokens=True)
+        # The output_str will include the prompt, so we extract the response.
+        response = output_str.split("assistant\n", 1)[1].strip()
+        return response
+
+    def to(self, device: Optional[torch.device] = None, dtype: Optional[torch.dtype] = None) -> None:
+        self._vllm_model.to(device=device, dtype=dtype)
+
+    def calc_size(self) -> int:
+        """Get size of the model in memory in bytes."""
+        # HACK(ryand): Fix this issue with circular imports.
+        from invokeai.backend.model_manager.load.model_util import calc_module_size
+
+        return calc_module_size(self._vllm_model)

invokeai/backend/model_manager/__init__.py

@@ -1,33 +1,43 @@
 """Re-export frequently-used symbols from the Model Manager backend."""
 
 from invokeai.backend.model_manager.config import (
-    AnyModel,
     AnyModelConfig,
-    BaseModelType,
     InvalidModelConfigException,
+    ModelConfigBase,
     ModelConfigFactory,
+)
+from invokeai.backend.model_manager.legacy_probe import ModelProbe
+from invokeai.backend.model_manager.load import LoadedModel
+from invokeai.backend.model_manager.search import ModelSearch
+from invokeai.backend.model_manager.taxonomy import (
+    AnyModel,
+    AnyVariant,
+    BaseModelType,
+    ClipVariantType,
     ModelFormat,
     ModelRepoVariant,
+    ModelSourceType,
     ModelType,
     ModelVariantType,
     SchedulerPredictionType,
     SubModelType,
 )
-from invokeai.backend.model_manager.load import LoadedModel
-from invokeai.backend.model_manager.probe import ModelProbe
-from invokeai.backend.model_manager.search import ModelSearch
 
 __all__ = [
-    "AnyModel",
     "AnyModelConfig",
-    "BaseModelType",
-    "ModelRepoVariant",
     "InvalidModelConfigException",
     "LoadedModel",
     "ModelConfigFactory",
-    "ModelFormat",
     "ModelProbe",
     "ModelSearch",
+    "ModelConfigBase",
+    "AnyModel",
+    "AnyVariant",
+    "BaseModelType",
+    "ClipVariantType",
+    "ModelFormat",
+    "ModelRepoVariant",
+    "ModelSourceType",
     "ModelType",
     "ModelVariantType",
     "SchedulerPredictionType",

invokeai/backend/model_manager/config.py

@@ -20,147 +20,56 @@ Validation errors will raise an InvalidModelConfigException error.
 
 """
 
+# pyright: reportIncompatibleVariableOverride=false
+import json
+import logging
 import time
+from abc import ABC, abstractmethod
 from enum import Enum
-from typing import Literal, Optional, Type, TypeAlias, Union
+from inspect import isabstract
+from pathlib import Path
+from typing import ClassVar, Literal, Optional, TypeAlias, Union
 
-import diffusers
-import onnxruntime as ort
+import safetensors.torch
 import torch
-from diffusers.models.modeling_utils import ModelMixin
+from picklescan.scanner import scan_file_path
 from pydantic import BaseModel, ConfigDict, Discriminator, Field, Tag, TypeAdapter
 from typing_extensions import Annotated, Any, Dict
 
 from invokeai.app.util.misc import uuid_string
 from invokeai.backend.model_hash.hash_validator import validate_hash
-from invokeai.backend.raw_model import RawModel
+from invokeai.backend.model_hash.model_hash import HASHING_ALGORITHMS, ModelHash
+from invokeai.backend.model_manager.taxonomy import (
+    AnyVariant,
+    BaseModelType,
+    ClipVariantType,
+    ModelFormat,
+    ModelRepoVariant,
+    ModelSourceType,
+    ModelType,
+    ModelVariantType,
+    SchedulerPredictionType,
+    SubModelType,
+)
+from invokeai.backend.quantization.gguf.loaders import gguf_sd_loader
 from invokeai.backend.stable_diffusion.schedulers.schedulers import SCHEDULER_NAME_VALUES
+from invokeai.backend.util.silence_warnings import SilenceWarnings
 
-# ModelMixin is the base class for all diffusers and transformers models
-# RawModel is the InvokeAI wrapper class for ip_adapters, loras, textual_inversion and onnx runtime
-AnyModel = Union[
-    ModelMixin, RawModel, torch.nn.Module, Dict[str, torch.Tensor], diffusers.DiffusionPipeline, ort.InferenceSession
-]
+logger = logging.getLogger(__name__)
 
 
 class InvalidModelConfigException(Exception):
-    """Exception for when config parser doesn't recognized this combination of model type and format."""
+    """Exception for when config parser doesn't recognize this combination of model type and format."""
 
-
-class BaseModelType(str, Enum):
-    """Base model type."""
-
-    Any = "any"
-    StableDiffusion1 = "sd-1"
-    StableDiffusion2 = "sd-2"
-    StableDiffusion3 = "sd-3"
-    StableDiffusionXL = "sdxl"
-    StableDiffusionXLRefiner = "sdxl-refiner"
-    Flux = "flux"
-    # Kandinsky2_1 = "kandinsky-2.1"
-
-
-class ModelType(str, Enum):
-    """Model type."""
-
-    ONNX = "onnx"
-    Main = "main"
-    VAE = "vae"
-    LoRA = "lora"
-    ControlLoRa = "control_lora"
-    ControlNet = "controlnet"  # used by model_probe
-    TextualInversion = "embedding"
-    IPAdapter = "ip_adapter"
-    CLIPVision = "clip_vision"
-    CLIPEmbed = "clip_embed"
-    T2IAdapter = "t2i_adapter"
-    T5Encoder = "t5_encoder"
-    SpandrelImageToImage = "spandrel_image_to_image"
-    SigLIP = "siglip"
-    FluxRedux = "flux_redux"
-
-
-class SubModelType(str, Enum):
-    """Submodel type."""
-
-    UNet = "unet"
-    Transformer = "transformer"
-    TextEncoder = "text_encoder"
-    TextEncoder2 = "text_encoder_2"
-    TextEncoder3 = "text_encoder_3"
-    Tokenizer = "tokenizer"
-    Tokenizer2 = "tokenizer_2"
-    Tokenizer3 = "tokenizer_3"
-    VAE = "vae"
-    VAEDecoder = "vae_decoder"
-    VAEEncoder = "vae_encoder"
-    Scheduler = "scheduler"
-    SafetyChecker = "safety_checker"
-
-
-class ClipVariantType(str, Enum):
-    """Variant type."""
-
-    L = "large"
-    G = "gigantic"
-
-
-class ModelVariantType(str, Enum):
-    """Variant type."""
-
-    Normal = "normal"
-    Inpaint = "inpaint"
-    Depth = "depth"
-
-
-class ModelFormat(str, Enum):
-    """Storage format of model."""
-
-    Diffusers = "diffusers"
-    Checkpoint = "checkpoint"
-    LyCORIS = "lycoris"
-    ONNX = "onnx"
-    Olive = "olive"
-    EmbeddingFile = "embedding_file"
-    EmbeddingFolder = "embedding_folder"
-    InvokeAI = "invokeai"
-    T5Encoder = "t5_encoder"
-    BnbQuantizedLlmInt8b = "bnb_quantized_int8b"
-    BnbQuantizednf4b = "bnb_quantized_nf4b"
-    GGUFQuantized = "gguf_quantized"
-
-
-class SchedulerPredictionType(str, Enum):
-    """Scheduler prediction type."""
-
-    Epsilon = "epsilon"
-    VPrediction = "v_prediction"
-    Sample = "sample"
-
-
-class ModelRepoVariant(str, Enum):
-    """Various hugging face variants on the diffusers format."""
-
-    Default = ""  # model files without "fp16" or other qualifier
-    FP16 = "fp16"
-    FP32 = "fp32"
-    ONNX = "onnx"
-    OpenVINO = "openvino"
-    Flax = "flax"
-
-
-class ModelSourceType(str, Enum):
-    """Model source type."""
-
-    Path = "path"
-    Url = "url"
-    HFRepoID = "hf_repo_id"
+    pass
 
 
 DEFAULTS_PRECISION = Literal["fp16", "fp32"]
 
 
-AnyVariant: TypeAlias = Union[ModelVariantType, ClipVariantType, None]
+class FSLayout(Enum):
+    FILE = "file"
+    DIRECTORY = "directory"
 
 
 class SubmodelDefinition(BaseModel):
@@ -190,12 +99,117 @@ class MainModelDefaultSettings(BaseModel):
 class ControlAdapterDefaultSettings(BaseModel):
     # This could be narrowed to controlnet processor nodes, but they change. Leaving this a string is safer.
     preprocessor: str | None
-
     model_config = ConfigDict(extra="forbid")
 
 
-class ModelConfigBase(BaseModel):
-    """Base class for model configuration information."""
+class ModelOnDisk:
+    """A utility class representing a model stored on disk."""
+
+    def __init__(self, path: Path, hash_algo: HASHING_ALGORITHMS = "blake3_single"):
+        self.path = path
+        # TODO: Revisit checkpoint vs diffusers terminology
+        self.layout = FSLayout.DIRECTORY if path.is_dir() else FSLayout.FILE
+        if self.path.suffix in {".safetensors", ".bin", ".pt", ".ckpt"}:
+            self.name = path.stem
+        else:
+            self.name = path.name
+        self.hash_algo = hash_algo
+        self._state_dict_cache = {}
+
+    def hash(self) -> str:
+        return ModelHash(algorithm=self.hash_algo).hash(self.path)
+
+    def size(self) -> int:
+        if self.layout == FSLayout.FILE:
+            return self.path.stat().st_size
+        return sum(file.stat().st_size for file in self.path.rglob("*"))
+
+    def component_paths(self) -> set[Path]:
+        if self.layout == FSLayout.FILE:
+            return {self.path}
+        extensions = {".safetensors", ".pt", ".pth", ".ckpt", ".bin", ".gguf"}
+        return {f for f in self.path.rglob("*") if f.suffix in extensions}
+
+    def repo_variant(self) -> Optional[ModelRepoVariant]:
+        if self.layout == FSLayout.FILE:
+            return None
+
+        weight_files = list(self.path.glob("**/*.safetensors"))
+        weight_files.extend(list(self.path.glob("**/*.bin")))
+        for x in weight_files:
+            if ".fp16" in x.suffixes:
+                return ModelRepoVariant.FP16
+            if "openvino_model" in x.name:
+                return ModelRepoVariant.OpenVINO
+            if "flax_model" in x.name:
+                return ModelRepoVariant.Flax
+            if x.suffix == ".onnx":
+                return ModelRepoVariant.ONNX
+        return ModelRepoVariant.Default
+
+    def load_state_dict(self, path: Optional[Path] = None) -> Dict[str | int, Any]:
+        if path in self._state_dict_cache:
+            return self._state_dict_cache[path]
+
+        if not path:
+            components = list(self.component_paths())
+            match components:
+                case []:
+                    raise ValueError("No weight files found for this model")
+                case [p]:
+                    path = p
+                case ps if len(ps) >= 2:
+                    raise ValueError(
+                        f"Multiple weight files found for this model: {ps}. "
+                        f"Please specify the intended file using the 'path' argument"
+                    )
+
+        with SilenceWarnings():
+            if path.suffix.endswith((".ckpt", ".pt", ".pth", ".bin")):
+                scan_result = scan_file_path(path)
+                if scan_result.infected_files != 0 or scan_result.scan_err:
+                    raise RuntimeError(f"The model {path.stem} is potentially infected by malware. Aborting import.")
+                checkpoint = torch.load(path, map_location="cpu")
+                assert isinstance(checkpoint, dict)
+            elif path.suffix.endswith(".gguf"):
+                checkpoint = gguf_sd_loader(path, compute_dtype=torch.float32)
+            elif path.suffix.endswith(".safetensors"):
+                checkpoint = safetensors.torch.load_file(path)
+            else:
+                raise ValueError(f"Unrecognized model extension: {path.suffix}")
+
+        state_dict = checkpoint.get("state_dict", checkpoint)
+        self._state_dict_cache[path] = state_dict
+        return state_dict
+
+
+class MatchSpeed(int, Enum):
+    """Represents the estimated runtime speed of a config's 'matches' method."""
+
+    FAST = 0
+    MED = 1
+    SLOW = 2
+
+
+class ModelConfigBase(ABC, BaseModel):
+    """
+    Abstract Base class for model configurations.
+
+    To create a new config type, inherit from this class and implement its interface:
+    - (mandatory) override methods 'matches' and 'parse'
+    - (mandatory) define fields 'type' and 'format' as class attributes
+
+    - (optional) override method 'get_tag'
+    - (optional) override field _MATCH_SPEED
+
+    See MinimalConfigExample in test_model_probe.py for an example implementation.
+    """
+
+    @staticmethod
+    def json_schema_extra(schema: dict[str, Any]) -> None:
+        schema["required"].extend(["key", "type", "format"])
+
+    model_config = ConfigDict(validate_assignment=True, json_schema_extra=json_schema_extra)
 
     key: str = Field(description="A unique key for this model.", default_factory=uuid_string)
     hash: str = Field(description="The hash of the model file(s).")
@@ -203,27 +217,133 @@ class ModelConfigBase(BaseModel):
         description="Path to the model on the filesystem. Relative paths are relative to the Invoke root directory."
     )
     name: str = Field(description="Name of the model.")
+    type: ModelType = Field(description="Model type")
+    format: ModelFormat = Field(description="Model format")
     base: BaseModelType = Field(description="The base model.")
-    description: Optional[str] = Field(description="Model description", default=None)
     source: str = Field(description="The original source of the model (path, URL or repo_id).")
     source_type: ModelSourceType = Field(description="The type of source")
+
+    description: Optional[str] = Field(description="Model description", default=None)
     source_api_response: Optional[str] = Field(
         description="The original API response from the source, as stringified JSON.", default=None
     )
     cover_image: Optional[str] = Field(description="Url for image to preview model", default=None)
-
-    @staticmethod
-    def json_schema_extra(schema: dict[str, Any], model_class: Type[BaseModel]) -> None:
-        schema["required"].extend(["key", "type", "format"])
-
-    model_config = ConfigDict(validate_assignment=True, json_schema_extra=json_schema_extra)
     submodels: Optional[Dict[SubModelType, SubmodelDefinition]] = Field(
         description="Loadable submodels in this model", default=None
     )
 
+    _USING_LEGACY_PROBE: ClassVar[set] = set()
+    _USING_CLASSIFY_API: ClassVar[set] = set()
+    _MATCH_SPEED: ClassVar[MatchSpeed] = MatchSpeed.MED
 
-class CheckpointConfigBase(ModelConfigBase):
-    """Model config for checkpoint-style models."""
+    def __init_subclass__(cls, **kwargs):
+        super().__init_subclass__(**kwargs)
+        if issubclass(cls, LegacyProbeMixin):
+            ModelConfigBase._USING_LEGACY_PROBE.add(cls)
+        else:
+            ModelConfigBase._USING_CLASSIFY_API.add(cls)
+
+    @staticmethod
+    def all_config_classes():
+        subclasses = ModelConfigBase._USING_LEGACY_PROBE | ModelConfigBase._USING_CLASSIFY_API
+        concrete = {cls for cls in subclasses if not isabstract(cls)}
+        return concrete
+
+    @staticmethod
+    def classify(model_path: Path, hash_algo: HASHING_ALGORITHMS = "blake3_single", **overrides):
+        """
+        Returns the best matching ModelConfig instance from a model's file/folder path.
+        Raises InvalidModelConfigException if no valid configuration is found.
+        Created to deprecate ModelProbe.probe
+        """
+        candidates = ModelConfigBase._USING_CLASSIFY_API
+        sorted_by_match_speed = sorted(candidates, key=lambda cls: cls._MATCH_SPEED)
+        mod = ModelOnDisk(model_path, hash_algo)
+
+        for config_cls in sorted_by_match_speed:
+            try:
+                if not config_cls.matches(mod):
+                    continue
+            except Exception as e:
+                logger.warning(f"Unexpected exception while matching {mod.name} to '{config_cls.__name__}': {e}")
+                continue
+            else:
+                return config_cls.from_model_on_disk(mod, **overrides)
+
+        raise InvalidModelConfigException("No valid config found")
+
+    @classmethod
+    def get_tag(cls) -> Tag:
+        type = cls.model_fields["type"].default.value
+        format = cls.model_fields["format"].default.value
+        return Tag(f"{type}.{format}")
+
+    @classmethod
+    @abstractmethod
+    def parse(cls, mod: ModelOnDisk) -> dict[str, Any]:
+        """Returns a dictionary with the fields needed to construct the model.
+        Raises InvalidModelConfigException if the model is invalid.
+        """
+        pass
+
+    @classmethod
+    @abstractmethod
+    def matches(cls, mod: ModelOnDisk) -> bool:
+        """Performs a quick check to determine if the config matches the model.
+        This doesn't need to be a perfect test - the aim is to eliminate unlikely matches quickly before parsing."""
+        pass
+
+    @staticmethod
+    def cast_overrides(overrides: dict[str, Any]):
+        """Casts user overrides from str to Enum"""
+        if "type" in overrides:
+            overrides["type"] = ModelType(overrides["type"])
+
+        if "format" in overrides:
+            overrides["format"] = ModelFormat(overrides["format"])
+
+        if "base" in overrides:
+            overrides["base"] = BaseModelType(overrides["base"])
+
+        if "source_type" in overrides:
+            overrides["source_type"] = ModelSourceType(overrides["source_type"])
+
+    @classmethod
+    def from_model_on_disk(cls, mod: ModelOnDisk, **overrides):
+        """Creates an instance of this config or raises InvalidModelConfigException."""
+        fields = cls.parse(mod)
+        cls.cast_overrides(overrides)
+        fields.update(overrides)
+
+        type = fields.get("type") or cls.model_fields["type"].default
+        base = fields.get("base") or cls.model_fields["base"].default
+
+        fields["path"] = mod.path.as_posix()
+        fields["source"] = fields.get("source") or fields["path"]
+        fields["source_type"] = fields.get("source_type") or ModelSourceType.Path
+        fields["name"] = name = fields.get("name") or mod.name
+        fields["hash"] = fields.get("hash") or mod.hash()
+        fields["key"] = fields.get("key") or uuid_string()
+        fields["description"] = fields.get("description") or f"{base.value} {type.value} model {name}"
+        fields["repo_variant"] = fields.get("repo_variant") or mod.repo_variant()
+
+        return cls(**fields)
+
+
+class LegacyProbeMixin:
+    """Mixin for classes using the legacy probe for model classification."""
+
+    @classmethod
+    def matches(cls, *args, **kwargs):
+        raise NotImplementedError(f"Method 'matches' not implemented for {cls.__name__}")
+
+    @classmethod
+    def parse(cls, *args, **kwargs):
+        raise NotImplementedError(f"Method 'parse' not implemented for {cls.__name__}")
+
+
+class CheckpointConfigBase(ABC, BaseModel):
+    """Base class for checkpoint-style models."""
 
     format: Literal[ModelFormat.Checkpoint, ModelFormat.BnbQuantizednf4b, ModelFormat.GGUFQuantized] = Field(
         description="Format of the provided checkpoint model", default=ModelFormat.Checkpoint
@@ -234,153 +354,109 @@ class CheckpointConfigBase(ModelConfigBase):
     )
 
 
-class DiffusersConfigBase(ModelConfigBase):
-    """Model config for diffusers-style models."""
+class DiffusersConfigBase(ABC, BaseModel):
+    """Base class for diffusers-style models."""
 
     format: Literal[ModelFormat.Diffusers] = ModelFormat.Diffusers
     repo_variant: Optional[ModelRepoVariant] = ModelRepoVariant.Default
 
 
-class LoRAConfigBase(ModelConfigBase):
+class LoRAConfigBase(ABC, BaseModel):
+    """Base class for LoRA models."""
+
     type: Literal[ModelType.LoRA] = ModelType.LoRA
     trigger_phrases: Optional[set[str]] = Field(description="Set of trigger phrases for this model", default=None)
 
 
-class T5EncoderConfigBase(ModelConfigBase):
+class T5EncoderConfigBase(ABC, BaseModel):
+    """Base class for diffusers-style models."""
+
     type: Literal[ModelType.T5Encoder] = ModelType.T5Encoder
 
 
-class T5EncoderConfig(T5EncoderConfigBase):
+class T5EncoderConfig(T5EncoderConfigBase, LegacyProbeMixin, ModelConfigBase):
     format: Literal[ModelFormat.T5Encoder] = ModelFormat.T5Encoder
 
-    @staticmethod
-    def get_tag() -> Tag:
-        return Tag(f"{ModelType.T5Encoder.value}.{ModelFormat.T5Encoder.value}")
 
-
-class T5EncoderBnbQuantizedLlmInt8bConfig(T5EncoderConfigBase):
+class T5EncoderBnbQuantizedLlmInt8bConfig(T5EncoderConfigBase, LegacyProbeMixin, ModelConfigBase):
     format: Literal[ModelFormat.BnbQuantizedLlmInt8b] = ModelFormat.BnbQuantizedLlmInt8b
 
-    @staticmethod
-    def get_tag() -> Tag:
-        return Tag(f"{ModelType.T5Encoder.value}.{ModelFormat.BnbQuantizedLlmInt8b.value}")
 
-
-class LoRALyCORISConfig(LoRAConfigBase):
+class LoRALyCORISConfig(LoRAConfigBase, LegacyProbeMixin, ModelConfigBase):
     """Model config for LoRA/Lycoris models."""
 
     format: Literal[ModelFormat.LyCORIS] = ModelFormat.LyCORIS
 
-    @staticmethod
-    def get_tag() -> Tag:
-        return Tag(f"{ModelType.LoRA.value}.{ModelFormat.LyCORIS.value}")
 
-
-class ControlAdapterConfigBase(BaseModel):
+class ControlAdapterConfigBase(ABC, BaseModel):
     default_settings: Optional[ControlAdapterDefaultSettings] = Field(
         description="Default settings for this model", default=None
     )
 
 
-class ControlLoRALyCORISConfig(ModelConfigBase, ControlAdapterConfigBase):
+class ControlLoRALyCORISConfig(ControlAdapterConfigBase, LegacyProbeMixin, ModelConfigBase):
     """Model config for Control LoRA models."""
 
     type: Literal[ModelType.ControlLoRa] = ModelType.ControlLoRa
     trigger_phrases: Optional[set[str]] = Field(description="Set of trigger phrases for this model", default=None)
     format: Literal[ModelFormat.LyCORIS] = ModelFormat.LyCORIS
 
-    @staticmethod
-    def get_tag() -> Tag:
-        return Tag(f"{ModelType.ControlLoRa.value}.{ModelFormat.LyCORIS.value}")
 
-
-class ControlLoRADiffusersConfig(ModelConfigBase, ControlAdapterConfigBase):
+class ControlLoRADiffusersConfig(ControlAdapterConfigBase, LegacyProbeMixin, ModelConfigBase):
     """Model config for Control LoRA models."""
 
     type: Literal[ModelType.ControlLoRa] = ModelType.ControlLoRa
     trigger_phrases: Optional[set[str]] = Field(description="Set of trigger phrases for this model", default=None)
     format: Literal[ModelFormat.Diffusers] = ModelFormat.Diffusers
 
-    @staticmethod
-    def get_tag() -> Tag:
-        return Tag(f"{ModelType.ControlLoRa.value}.{ModelFormat.Diffusers.value}")
 
-
-class LoRADiffusersConfig(LoRAConfigBase):
+class LoRADiffusersConfig(LoRAConfigBase, LegacyProbeMixin, ModelConfigBase):
     """Model config for LoRA/Diffusers models."""
 
     format: Literal[ModelFormat.Diffusers] = ModelFormat.Diffusers
 
-    @staticmethod
-    def get_tag() -> Tag:
-        return Tag(f"{ModelType.LoRA.value}.{ModelFormat.Diffusers.value}")
 
-
-class VAECheckpointConfig(CheckpointConfigBase):
+class VAECheckpointConfig(CheckpointConfigBase, LegacyProbeMixin, ModelConfigBase):
     """Model config for standalone VAE models."""
 
     type: Literal[ModelType.VAE] = ModelType.VAE
 
-    @staticmethod
-    def get_tag() -> Tag:
-        return Tag(f"{ModelType.VAE.value}.{ModelFormat.Checkpoint.value}")
 
-
-class VAEDiffusersConfig(ModelConfigBase):
+class VAEDiffusersConfig(LegacyProbeMixin, ModelConfigBase):
     """Model config for standalone VAE models (diffusers version)."""
 
     type: Literal[ModelType.VAE] = ModelType.VAE
     format: Literal[ModelFormat.Diffusers] = ModelFormat.Diffusers
 
-    @staticmethod
-    def get_tag() -> Tag:
-        return Tag(f"{ModelType.VAE.value}.{ModelFormat.Diffusers.value}")
 
-
-class ControlNetDiffusersConfig(DiffusersConfigBase, ControlAdapterConfigBase):
+class ControlNetDiffusersConfig(DiffusersConfigBase, ControlAdapterConfigBase, LegacyProbeMixin, ModelConfigBase):
     """Model config for ControlNet models (diffusers version)."""
 
     type: Literal[ModelType.ControlNet] = ModelType.ControlNet
     format: Literal[ModelFormat.Diffusers] = ModelFormat.Diffusers
 
-    @staticmethod
-    def get_tag() -> Tag:
-        return Tag(f"{ModelType.ControlNet.value}.{ModelFormat.Diffusers.value}")
 
-
-class ControlNetCheckpointConfig(CheckpointConfigBase, ControlAdapterConfigBase):
+class ControlNetCheckpointConfig(CheckpointConfigBase, ControlAdapterConfigBase, LegacyProbeMixin, ModelConfigBase):
     """Model config for ControlNet models (diffusers version)."""
 
     type: Literal[ModelType.ControlNet] = ModelType.ControlNet
 
-    @staticmethod
-    def get_tag() -> Tag:
-        return Tag(f"{ModelType.ControlNet.value}.{ModelFormat.Checkpoint.value}")
 
-
-class TextualInversionFileConfig(ModelConfigBase):
+class TextualInversionFileConfig(LegacyProbeMixin, ModelConfigBase):
     """Model config for textual inversion embeddings."""
 
     type: Literal[ModelType.TextualInversion] = ModelType.TextualInversion
     format: Literal[ModelFormat.EmbeddingFile] = ModelFormat.EmbeddingFile
 
-    @staticmethod
-    def get_tag() -> Tag:
-        return Tag(f"{ModelType.TextualInversion.value}.{ModelFormat.EmbeddingFile.value}")
 
-
-class TextualInversionFolderConfig(ModelConfigBase):
+class TextualInversionFolderConfig(LegacyProbeMixin, ModelConfigBase):
     """Model config for textual inversion embeddings."""
 
     type: Literal[ModelType.TextualInversion] = ModelType.TextualInversion
     format: Literal[ModelFormat.EmbeddingFolder] = ModelFormat.EmbeddingFolder
 
-    @staticmethod
-    def get_tag() -> Tag:
-        return Tag(f"{ModelType.TextualInversion.value}.{ModelFormat.EmbeddingFolder.value}")
 
-
-class MainConfigBase(ModelConfigBase):
+class MainConfigBase(ABC, BaseModel):
     type: Literal[ModelType.Main] = ModelType.Main
     trigger_phrases: Optional[set[str]] = Field(description="Set of trigger phrases for this model", default=None)
     default_settings: Optional[MainModelDefaultSettings] = Field(
@@ -389,167 +465,146 @@ class MainConfigBase(ModelConfigBase):
     variant: AnyVariant = ModelVariantType.Normal
 
 
-class MainCheckpointConfig(CheckpointConfigBase, MainConfigBase):
+class MainCheckpointConfig(CheckpointConfigBase, MainConfigBase, LegacyProbeMixin, ModelConfigBase):
     """Model config for main checkpoint models."""
 
     prediction_type: SchedulerPredictionType = SchedulerPredictionType.Epsilon
     upcast_attention: bool = False
 
-    @staticmethod
-    def get_tag() -> Tag:
-        return Tag(f"{ModelType.Main.value}.{ModelFormat.Checkpoint.value}")
 
-
-class MainBnbQuantized4bCheckpointConfig(CheckpointConfigBase, MainConfigBase):
+class MainBnbQuantized4bCheckpointConfig(CheckpointConfigBase, MainConfigBase, LegacyProbeMixin, ModelConfigBase):
     """Model config for main checkpoint models."""
 
+    format: Literal[ModelFormat.BnbQuantizednf4b] = ModelFormat.BnbQuantizednf4b
     prediction_type: SchedulerPredictionType = SchedulerPredictionType.Epsilon
     upcast_attention: bool = False
 
-    def __init__(self, *args, **kwargs):
-        super().__init__(*args, **kwargs)
-        self.format = ModelFormat.BnbQuantizednf4b
 
-    @staticmethod
-    def get_tag() -> Tag:
-        return Tag(f"{ModelType.Main.value}.{ModelFormat.BnbQuantizednf4b.value}")
-
-
-class MainGGUFCheckpointConfig(CheckpointConfigBase, MainConfigBase):
+class MainGGUFCheckpointConfig(CheckpointConfigBase, MainConfigBase, LegacyProbeMixin, ModelConfigBase):
     """Model config for main checkpoint models."""
 
+    format: Literal[ModelFormat.GGUFQuantized] = ModelFormat.GGUFQuantized
     prediction_type: SchedulerPredictionType = SchedulerPredictionType.Epsilon
     upcast_attention: bool = False
 
-    def __init__(self, *args, **kwargs):
-        super().__init__(*args, **kwargs)
-        self.format = ModelFormat.GGUFQuantized
 
-    @staticmethod
-    def get_tag() -> Tag:
-        return Tag(f"{ModelType.Main.value}.{ModelFormat.GGUFQuantized.value}")
-
-
-class MainDiffusersConfig(DiffusersConfigBase, MainConfigBase):
+class MainDiffusersConfig(DiffusersConfigBase, MainConfigBase, LegacyProbeMixin, ModelConfigBase):
     """Model config for main diffusers models."""
 
-    @staticmethod
-    def get_tag() -> Tag:
-        return Tag(f"{ModelType.Main.value}.{ModelFormat.Diffusers.value}")
+    pass
 
 
-class IPAdapterBaseConfig(ModelConfigBase):
+class IPAdapterConfigBase(ABC, BaseModel):
     type: Literal[ModelType.IPAdapter] = ModelType.IPAdapter
 
 
-class IPAdapterInvokeAIConfig(IPAdapterBaseConfig):
+class IPAdapterInvokeAIConfig(IPAdapterConfigBase, LegacyProbeMixin, ModelConfigBase):
     """Model config for IP Adapter diffusers format models."""
 
     # TODO(ryand): Should we deprecate this field? From what I can tell, it hasn't been probed correctly for a long
     # time. Need to go through the history to make sure I'm understanding this fully.
     image_encoder_model_id: str
-    format: Literal[ModelFormat.InvokeAI]
-
-    @staticmethod
-    def get_tag() -> Tag:
-        return Tag(f"{ModelType.IPAdapter.value}.{ModelFormat.InvokeAI.value}")
+    format: Literal[ModelFormat.InvokeAI] = ModelFormat.InvokeAI
 
 
-class IPAdapterCheckpointConfig(IPAdapterBaseConfig):
+class IPAdapterCheckpointConfig(IPAdapterConfigBase, LegacyProbeMixin, ModelConfigBase):
     """Model config for IP Adapter checkpoint format models."""
 
-    format: Literal[ModelFormat.Checkpoint]
-
-    @staticmethod
-    def get_tag() -> Tag:
-        return Tag(f"{ModelType.IPAdapter.value}.{ModelFormat.Checkpoint.value}")
+    format: Literal[ModelFormat.Checkpoint] = ModelFormat.Checkpoint
 
 
 class CLIPEmbedDiffusersConfig(DiffusersConfigBase):
     """Model config for Clip Embeddings."""
 
+    variant: ClipVariantType = Field(description="Clip variant for this model")
     type: Literal[ModelType.CLIPEmbed] = ModelType.CLIPEmbed
     format: Literal[ModelFormat.Diffusers] = ModelFormat.Diffusers
-    variant: ClipVariantType = ClipVariantType.L
-
-    @staticmethod
-    def get_tag() -> Tag:
-        return Tag(f"{ModelType.CLIPEmbed.value}.{ModelFormat.Diffusers.value}")
 
 
-class CLIPGEmbedDiffusersConfig(CLIPEmbedDiffusersConfig):
+class CLIPGEmbedDiffusersConfig(CLIPEmbedDiffusersConfig, LegacyProbeMixin, ModelConfigBase):
     """Model config for CLIP-G Embeddings."""
 
-    variant: ClipVariantType = ClipVariantType.G
+    variant: Literal[ClipVariantType.G] = ClipVariantType.G
 
-    @staticmethod
-    def get_tag() -> Tag:
-        return Tag(f"{ModelType.CLIPEmbed.value}.{ModelFormat.Diffusers.value}.{ClipVariantType.G}")
+    @classmethod
+    def get_tag(cls) -> Tag:
+        return Tag(f"{ModelType.CLIPEmbed.value}.{ModelFormat.Diffusers.value}.{ClipVariantType.G.value}")
 
 
-class CLIPLEmbedDiffusersConfig(CLIPEmbedDiffusersConfig):
+class CLIPLEmbedDiffusersConfig(CLIPEmbedDiffusersConfig, LegacyProbeMixin, ModelConfigBase):
     """Model config for CLIP-L Embeddings."""
 
-    variant: ClipVariantType = ClipVariantType.L
+    variant: Literal[ClipVariantType.L] = ClipVariantType.L
 
-    @staticmethod
-    def get_tag() -> Tag:
-        return Tag(f"{ModelType.CLIPEmbed.value}.{ModelFormat.Diffusers.value}.{ClipVariantType.L}")
+    @classmethod
+    def get_tag(cls) -> Tag:
+        return Tag(f"{ModelType.CLIPEmbed.value}.{ModelFormat.Diffusers.value}.{ClipVariantType.L.value}")
 
 
-class CLIPVisionDiffusersConfig(DiffusersConfigBase):
+class CLIPVisionDiffusersConfig(DiffusersConfigBase, LegacyProbeMixin, ModelConfigBase):
     """Model config for CLIPVision."""
 
     type: Literal[ModelType.CLIPVision] = ModelType.CLIPVision
     format: Literal[ModelFormat.Diffusers] = ModelFormat.Diffusers
 
-    @staticmethod
-    def get_tag() -> Tag:
-        return Tag(f"{ModelType.CLIPVision.value}.{ModelFormat.Diffusers.value}")
 
-
-class T2IAdapterConfig(DiffusersConfigBase, ControlAdapterConfigBase):
+class T2IAdapterConfig(DiffusersConfigBase, ControlAdapterConfigBase, LegacyProbeMixin, ModelConfigBase):
     """Model config for T2I."""
 
     type: Literal[ModelType.T2IAdapter] = ModelType.T2IAdapter
     format: Literal[ModelFormat.Diffusers] = ModelFormat.Diffusers
 
-    @staticmethod
-    def get_tag() -> Tag:
-        return Tag(f"{ModelType.T2IAdapter.value}.{ModelFormat.Diffusers.value}")
 
-
-class SpandrelImageToImageConfig(ModelConfigBase):
+class SpandrelImageToImageConfig(LegacyProbeMixin, ModelConfigBase):
     """Model config for Spandrel Image to Image models."""
 
+    _MATCH_SPEED: ClassVar[MatchSpeed] = MatchSpeed.SLOW  # requires loading the model from disk
+
     type: Literal[ModelType.SpandrelImageToImage] = ModelType.SpandrelImageToImage
     format: Literal[ModelFormat.Checkpoint] = ModelFormat.Checkpoint
 
-    @staticmethod
-    def get_tag() -> Tag:
-        return Tag(f"{ModelType.SpandrelImageToImage.value}.{ModelFormat.Checkpoint.value}")
 
-
-class SigLIPConfig(DiffusersConfigBase):
+class SigLIPConfig(DiffusersConfigBase, LegacyProbeMixin, ModelConfigBase):
     """Model config for SigLIP."""
 
     type: Literal[ModelType.SigLIP] = ModelType.SigLIP
     format: Literal[ModelFormat.Diffusers] = ModelFormat.Diffusers
 
-    @staticmethod
-    def get_tag() -> Tag:
-        return Tag(f"{ModelType.SigLIP.value}.{ModelFormat.Diffusers.value}")
 
-
-class FluxReduxConfig(ModelConfigBase):
+class FluxReduxConfig(LegacyProbeMixin, ModelConfigBase):
     """Model config for FLUX Tools Redux model."""
 
     type: Literal[ModelType.FluxRedux] = ModelType.FluxRedux
     format: Literal[ModelFormat.Checkpoint] = ModelFormat.Checkpoint
 
-    @staticmethod
-    def get_tag() -> Tag:
-        return Tag(f"{ModelType.FluxRedux.value}.{ModelFormat.Checkpoint.value}")
+
+class LlavaOnevisionConfig(DiffusersConfigBase, ModelConfigBase):
+    """Model config for Llava Onevision models."""
+
+    type: Literal[ModelType.LlavaOnevision] = ModelType.LlavaOnevision
+    format: Literal[ModelFormat.Diffusers] = ModelFormat.Diffusers
+
+    @classmethod
+    def matches(cls, mod: ModelOnDisk) -> bool:
+        if mod.layout == FSLayout.FILE:
+            return False
+
+        config_path = mod.path / "config.json"
+        try:
+            with open(config_path, "r") as file:
+                config = json.load(file)
+        except FileNotFoundError:
+            return False
+
+        architectures = config.get("architectures")
+        return architectures and architectures[0] == "LlavaOnevisionForConditionalGeneration"
+
+    @classmethod
+    def parse(cls, mod: ModelOnDisk) -> dict[str, Any]:
+        return {
+            "base": BaseModelType.Any,
+            "variant": ModelVariantType.Normal,
+        }
 
 
 def get_model_discriminator_value(v: Any) -> str:
@@ -557,22 +612,40 @@ def get_model_discriminator_value(v: Any) -> str:
     Computes the discriminator value for a model config.
     https://docs.pydantic.dev/latest/concepts/unions/#discriminated-unions-with-callable-discriminator
     """
-    format_ = None
-    type_ = None
+    format_ = type_ = variant_ = None
+
     if isinstance(v, dict):
         format_ = v.get("format")
         if isinstance(format_, Enum):
             format_ = format_.value
+
         type_ = v.get("type")
         if isinstance(type_, Enum):
             type_ = type_.value
+
+        variant_ = v.get("variant")
+        if isinstance(variant_, Enum):
+            variant_ = variant_.value
     else:
         format_ = v.format.value
         type_ = v.type.value
-    v = f"{type_}.{format_}"
-    return v
+        variant_ = getattr(v, "variant", None)
+        if variant_:
+            variant_ = variant_.value
+
+    # Ideally, each config would be uniquely identified with a combination of fields
+    # i.e. (type, format, variant) without any special cases. Alas...
+
+    # Previously, CLIPEmbed did not have any variants, meaning older database entries lack a variant field.
+    # To maintain compatibility, we default to ClipVariantType.L in this case.
+    if type_ == ModelType.CLIPEmbed.value and format_ == ModelFormat.Diffusers.value:
+        variant_ = variant_ or ClipVariantType.L.value
+        return f"{type_}.{format_}.{variant_}"
+    return f"{type_}.{format_}"
 
 
+# The types are listed explicitly because IDEs/LSPs can't identify the correct types
+# when AnyModelConfig is constructed dynamically using ModelConfigBase.all_config_classes
 AnyModelConfig = Annotated[
     Union[
         Annotated[MainDiffusersConfig, MainDiffusersConfig.get_tag()],
@@ -596,11 +669,11 @@ AnyModelConfig = Annotated[
         Annotated[T2IAdapterConfig, T2IAdapterConfig.get_tag()],
         Annotated[SpandrelImageToImageConfig, SpandrelImageToImageConfig.get_tag()],
         Annotated[CLIPVisionDiffusersConfig, CLIPVisionDiffusersConfig.get_tag()],
-        Annotated[CLIPEmbedDiffusersConfig, CLIPEmbedDiffusersConfig.get_tag()],
         Annotated[CLIPLEmbedDiffusersConfig, CLIPLEmbedDiffusersConfig.get_tag()],
         Annotated[CLIPGEmbedDiffusersConfig, CLIPGEmbedDiffusersConfig.get_tag()],
         Annotated[SigLIPConfig, SigLIPConfig.get_tag()],
         Annotated[FluxReduxConfig, FluxReduxConfig.get_tag()],
+        Annotated[LlavaOnevisionConfig, LlavaOnevisionConfig.get_tag()],
     ],
     Discriminator(get_model_discriminator_value),
 ]
@@ -609,39 +682,12 @@ AnyModelConfigValidator = TypeAdapter(AnyModelConfig)
 AnyDefaultSettings: TypeAlias = Union[MainModelDefaultSettings, ControlAdapterDefaultSettings]
 
 
-class ModelConfigFactory(object):
-    """Class for parsing config dicts into StableDiffusion Config obects."""
-
-    @classmethod
-    def make_config(
-        cls,
-        model_data: Union[Dict[str, Any], AnyModelConfig],
-        key: Optional[str] = None,
-        dest_class: Optional[Type[ModelConfigBase]] = None,
-        timestamp: Optional[float] = None,
-    ) -> AnyModelConfig:
-        """
-        Return the appropriate config object from raw dict values.
-
-        :param model_data: A raw dict corresponding the obect fields to be
-        parsed into a ModelConfigBase obect (or descendent), or a ModelConfigBase
-        object, which will be passed through unchanged.
-        :param dest_class: The config class to be returned. If not provided, will
-        be selected automatically.
-        """
-        model: Optional[ModelConfigBase] = None
-        if isinstance(model_data, ModelConfigBase):
-            model = model_data
-        elif dest_class:
-            model = dest_class.model_validate(model_data)
-        else:
-            # mypy doesn't typecheck TypeAdapters well?
-            model = AnyModelConfigValidator.validate_python(model_data)  # type: ignore
-        assert model is not None
-        if key:
-            model.key = key
-        if isinstance(model, CheckpointConfigBase) and timestamp is not None:
+class ModelConfigFactory:
+    @staticmethod
+    def make_config(model_data: Dict[str, Any], timestamp: Optional[float] = None) -> AnyModelConfig:
+        """Return the appropriate config object from raw dict values."""
+        model = AnyModelConfigValidator.validate_python(model_data)  # type: ignore
+        if isinstance(model, CheckpointConfigBase) and timestamp:
             model.converted_at = timestamp
-        if model:
-            validate_hash(model.hash)
+        validate_hash(model.hash)
         return model  # type: ignore

invokeai/backend/model_manager/legacy_probe.py

@@ -3,10 +3,10 @@ import re
 from pathlib import Path
 from typing import Any, Callable, Dict, Literal, Optional, Union
 
+import picklescan.scanner as pscan
 import safetensors.torch
 import spandrel
 import torch
-from picklescan.scanner import scan_file_path
 
 import invokeai.backend.util.logging as logger
 from invokeai.app.util.misc import uuid_string
@@ -19,22 +19,24 @@ from invokeai.backend.flux.redux.flux_redux_state_dict_utils import is_state_dic
 from invokeai.backend.model_hash.model_hash import HASHING_ALGORITHMS, ModelHash
 from invokeai.backend.model_manager.config import (
     AnyModelConfig,
-    AnyVariant,
-    BaseModelType,
     ControlAdapterDefaultSettings,
     InvalidModelConfigException,
     MainModelDefaultSettings,
     ModelConfigFactory,
+    SubmodelDefinition,
+)
+from invokeai.backend.model_manager.load.model_loaders.generic_diffusers import ConfigLoader
+from invokeai.backend.model_manager.taxonomy import (
+    AnyVariant,
+    BaseModelType,
     ModelFormat,
     ModelRepoVariant,
     ModelSourceType,
     ModelType,
     ModelVariantType,
     SchedulerPredictionType,
-    SubmodelDefinition,
     SubModelType,
 )
-from invokeai.backend.model_manager.load.model_loaders.generic_diffusers import ConfigLoader
 from invokeai.backend.model_manager.util.model_util import (
     get_clip_variant_type,
     lora_token_vector_length,
@@ -141,6 +143,7 @@ class ModelProbe(object):
         "SD3Transformer2DModel": ModelType.Main,
         "CLIPTextModelWithProjection": ModelType.CLIPEmbed,
         "SiglipModel": ModelType.SigLIP,
+        "LlavaOnevisionForConditionalGeneration": ModelType.LlavaOnevision,
     }
 
     TYPE2VARIANT: Dict[ModelType, Callable[[str], Optional[AnyVariant]]] = {ModelType.CLIPEmbed: get_clip_variant_type}
@@ -416,20 +419,22 @@ class ModelProbe(object):
                 # TODO: Decide between dev/schnell
                 checkpoint = ModelProbe._scan_and_load_checkpoint(model_path)
                 state_dict = checkpoint.get("state_dict") or checkpoint
+
+                # HACK: For FLUX, config_file is used as a key into invokeai.backend.flux.util.params during model
+                # loading. When FLUX support was first added, it was decided that this was the easiest way to support
+                # the various FLUX formats rather than adding new model types/formats. Be careful when modifying this in
+                # the future.
                 if (
                     "guidance_in.out_layer.weight" in state_dict
                     or "model.diffusion_model.guidance_in.out_layer.weight" in state_dict
                 ):
-                    # For flux, this is a key in invokeai.backend.flux.util.params
-                    #   Due to model type and format being the descriminator for model configs this
-                    #   is used rather than attempting to support flux with separate model types and format
-                    #   If changed in the future, please fix me
-                    config_file = "flux-dev"
+                    if variant_type == ModelVariantType.Normal:
+                        config_file = "flux-dev"
+                    elif variant_type == ModelVariantType.Inpaint:
+                        config_file = "flux-dev-fill"
+                    else:
+                        raise ValueError(f"Unexpected FLUX variant type: {variant_type}")
                 else:
-                    # For flux, this is a key in invokeai.backend.flux.util.params
-                    #   Due to model type and format being the discriminator for model configs this
-                    #   is used rather than attempting to support flux with separate model types and format
-                    #   If changed in the future, please fix me
                     config_file = "flux-schnell"
             else:
                 config_file = LEGACY_CONFIGS[base_type][variant_type]
@@ -482,9 +487,11 @@ class ModelProbe(object):
         and option to exit if an infected file is identified.
         """
         # scan model
-        scan_result = scan_file_path(checkpoint)
-        if scan_result.infected_files != 0 or scan_result.scan_err:
-            raise Exception("The model {model_name} is potentially infected by malware. Aborting import.")
+        scan_result = pscan.scan_file_path(checkpoint)
+        if scan_result.infected_files != 0:
+            raise Exception(f"The model {model_name} is potentially infected by malware. Aborting import.")
+        if scan_result.scan_err:
+            raise Exception(f"Error scanning model {model_name} for malware. Aborting import.")
 
 
 # Probing utilities
@@ -552,9 +559,27 @@ class CheckpointProbeBase(ProbeBase):
     def get_variant_type(self) -> ModelVariantType:
         model_type = ModelProbe.get_model_type_from_checkpoint(self.model_path, self.checkpoint)
         base_type = self.get_base_type()
-        if model_type != ModelType.Main or base_type == BaseModelType.Flux:
+        if model_type != ModelType.Main:
             return ModelVariantType.Normal
         state_dict = self.checkpoint.get("state_dict") or self.checkpoint
+
+        if base_type == BaseModelType.Flux:
+            in_channels = state_dict["img_in.weight"].shape[1]
+
+            # FLUX Model variant types are distinguished by input channels:
+            # - Unquantized Dev and Schnell have in_channels=64
+            # - BNB-NF4 Dev and Schnell have in_channels=1
+            # - FLUX Fill has in_channels=384
+            # - Unsure of quantized FLUX Fill models
+            # - Unsure of GGUF-quantized models
+            if in_channels == 384:
+                # This is a FLUX Fill model. FLUX Fill needs special handling throughout the application. The variant
+                # type is used to determine whether to use the fill model or the base model.
+                return ModelVariantType.Inpaint
+            else:
+                # Fall back on "normal" variant type for all other FLUX models.
+                return ModelVariantType.Normal
+
         in_channels = state_dict["model.diffusion_model.input_blocks.0.0.weight"].shape[1]
         if in_channels == 9:
             return ModelVariantType.Inpaint
@@ -767,6 +792,11 @@ class FluxReduxCheckpointProbe(CheckpointProbeBase):
         return BaseModelType.Flux
 
 
+class LlavaOnevisionCheckpointProbe(CheckpointProbeBase):
+    def get_base_type(self) -> BaseModelType:
+        raise NotImplementedError()
+
+
 ########################################################
 # classes for probing folders
 #######################################################
@@ -1047,6 +1077,11 @@ class FluxReduxFolderProbe(FolderProbeBase):
         raise NotImplementedError()
 
 
+class LlaveOnevisionFolderProbe(FolderProbeBase):
+    def get_base_type(self) -> BaseModelType:
+        return BaseModelType.Any
+
+
 class T2IAdapterFolderProbe(FolderProbeBase):
     def get_base_type(self) -> BaseModelType:
         config_file = self.model_path / "config.json"
@@ -1082,6 +1117,7 @@ ModelProbe.register_probe("diffusers", ModelType.T2IAdapter, T2IAdapterFolderPro
 ModelProbe.register_probe("diffusers", ModelType.SpandrelImageToImage, SpandrelImageToImageFolderProbe)
 ModelProbe.register_probe("diffusers", ModelType.SigLIP, SigLIPFolderProbe)
 ModelProbe.register_probe("diffusers", ModelType.FluxRedux, FluxReduxFolderProbe)
+ModelProbe.register_probe("diffusers", ModelType.LlavaOnevision, LlaveOnevisionFolderProbe)
 
 ModelProbe.register_probe("checkpoint", ModelType.Main, PipelineCheckpointProbe)
 ModelProbe.register_probe("checkpoint", ModelType.VAE, VaeCheckpointProbe)
@@ -1095,5 +1131,6 @@ ModelProbe.register_probe("checkpoint", ModelType.T2IAdapter, T2IAdapterCheckpoi
 ModelProbe.register_probe("checkpoint", ModelType.SpandrelImageToImage, SpandrelImageToImageCheckpointProbe)
 ModelProbe.register_probe("checkpoint", ModelType.SigLIP, SigLIPCheckpointProbe)
 ModelProbe.register_probe("checkpoint", ModelType.FluxRedux, FluxReduxCheckpointProbe)
+ModelProbe.register_probe("checkpoint", ModelType.LlavaOnevision, LlavaOnevisionCheckpointProbe)
 
 ModelProbe.register_probe("onnx", ModelType.ONNX, ONNXFolderProbe)

invokeai/backend/model_manager/load/load_base.py

@@ -13,12 +13,11 @@ import torch
 
 from invokeai.app.services.config import InvokeAIAppConfig
 from invokeai.backend.model_manager.config import (
-    AnyModel,
     AnyModelConfig,
-    SubModelType,
 )
 from invokeai.backend.model_manager.load.model_cache.cache_record import CacheRecord
 from invokeai.backend.model_manager.load.model_cache.model_cache import ModelCache
+from invokeai.backend.model_manager.taxonomy import AnyModel, SubModelType
 
 
 class LoadedModelWithoutConfig:

invokeai/backend/model_manager/load/load_default.py

@@ -6,18 +6,16 @@ from pathlib import Path
 from typing import Optional
 
 from invokeai.app.services.config import InvokeAIAppConfig
-from invokeai.backend.model_manager import (
-    AnyModel,
-    AnyModelConfig,
-    InvalidModelConfigException,
-    SubModelType,
-)
-from invokeai.backend.model_manager.config import DiffusersConfigBase
+from invokeai.backend.model_manager.config import AnyModelConfig, DiffusersConfigBase, InvalidModelConfigException
 from invokeai.backend.model_manager.load.load_base import LoadedModel, ModelLoaderBase
 from invokeai.backend.model_manager.load.model_cache.cache_record import CacheRecord
 from invokeai.backend.model_manager.load.model_cache.model_cache import ModelCache, get_model_cache_key
 from invokeai.backend.model_manager.load.model_util import calc_model_size_by_fs
 from invokeai.backend.model_manager.load.optimizations import skip_torch_weight_init
+from invokeai.backend.model_manager.taxonomy import (
+    AnyModel,
+    SubModelType,
+)
 from invokeai.backend.util.devices import TorchDevice
 
 

invokeai/backend/model_manager/load/model_cache/model_cache.py

@@ -9,7 +9,6 @@ from typing import Any, Callable, Dict, List, Optional
 import psutil
 import torch
 
-from invokeai.backend.model_manager import AnyModel, SubModelType
 from invokeai.backend.model_manager.load.memory_snapshot import MemorySnapshot
 from invokeai.backend.model_manager.load.model_cache.cache_record import CacheRecord
 from invokeai.backend.model_manager.load.model_cache.cache_stats import CacheStats
@@ -23,6 +22,7 @@ from invokeai.backend.model_manager.load.model_cache.torch_module_autocast.torch
     apply_custom_layers_to_model,
 )
 from invokeai.backend.model_manager.load.model_util import calc_model_size_by_data
+from invokeai.backend.model_manager.taxonomy import AnyModel, SubModelType
 from invokeai.backend.util.devices import TorchDevice
 from invokeai.backend.util.logging import InvokeAILogger
 from invokeai.backend.util.prefix_logger_adapter import PrefixedLoggerAdapter