chore: prep for v5.9.1

Before FLUX Fill was merged, we didn't do any checks for the model variant. We always returned "normal".

To determine if a model is a FLUX Fill model, we need to check the state dict for a specific key. Initially, this logic was too strict and rejected quantized FLUX models. This issue was resolved, but it turns out there is another failure mode - some fine-tunes use a different key.

This change further reduces the strictness, handling the alternate key and also falling back to "normal" if we don't see either key. This effectively restores the previous probing behaviour for all FLUX models.

Closes #7856
Closes #7859

.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/python-checks.yml

@@ -34,6 +34,9 @@ on:
 
 jobs:
   python-checks:
+    env:
+      # uv requires a venv by default - but for this, we can simply use the system python
+      UV_SYSTEM_PYTHON: 1
     runs-on: ubuntu-latest
     timeout-minutes: 5 # expected run time: <1 min
     steps:
@@ -57,25 +60,19 @@ jobs:
               - '!invokeai/frontend/web/**'
               - 'tests/**'
 
-      - name: setup python
+      - name: setup uv
         if: ${{ steps.changed-files.outputs.python_any_changed == 'true' || inputs.always_run == true }}
-        uses: actions/setup-python@v5
+        uses: astral-sh/setup-uv@v5
         with:
-          python-version: '3.10'
-          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
-        shell: bash
+          version: '0.6.10'
+          enable-cache: true
 
       - name: ruff check
         if: ${{ steps.changed-files.outputs.python_any_changed == 'true' || inputs.always_run == true }}
-        run: ruff check --output-format=github .
+        run: uv tool run ruff@0.11.2 check --output-format=github .
         shell: bash
 
       - name: ruff format
         if: ${{ steps.changed-files.outputs.python_any_changed == 'true' || inputs.always_run == true }}
-        run: ruff format --check .
+        run: uv tool run ruff@0.11.2 format --check .
         shell: bash

.github/workflows/python-tests.yml

@@ -39,24 +39,15 @@ jobs:
     strategy:
       matrix:
         python-version:
-          - '3.10'
           - '3.11'
+          - '3.12'
         platform:
-          - linux-cuda-11_7
-          - linux-rocm-5_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-cpu
-            os: ubuntu-22.04
+            os: ubuntu-24.04
             extra-index-url: 'https://download.pytorch.org/whl/cpu'
             github-env: $GITHUB_ENV
           - platform: macos-default
@@ -70,9 +61,12 @@ jobs:
     timeout-minutes: 15 # expected run time: 2-6 min, depending on platform
     env:
       PIP_USE_PEP517: '1'
+      UV_SYSTEM_PYTHON: 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 }}
@@ -91,20 +85,25 @@ jobs:
               - '!invokeai/frontend/web/**'
               - 'tests/**'
 
+      - name: setup uv
+        if: ${{ steps.changed-files.outputs.python_any_changed == 'true' || inputs.always_run == true }}
+        uses: astral-sh/setup-uv@v5
+        with:
+          version: '0.6.10'
+          enable-cache: true
+          python-version: ${{ matrix.python-version }}
+
       - name: setup python
         if: ${{ steps.changed-files.outputs.python_any_changed == 'true' || inputs.always_run == true }}
         uses: actions/setup-python@v5
         with:
           python-version: ${{ matrix.python-version }}
-          cache: pip
-          cache-dependency-path: pyproject.toml
 
       - name: install dependencies
         if: ${{ steps.changed-files.outputs.python_any_changed == 'true' || inputs.always_run == true }}
         env:
-          PIP_EXTRA_INDEX_URL: ${{ matrix.extra-index-url }}
-        run: >
-          pip3 install --editable=".[test]"
+          UV_INDEX: ${{ matrix.extra-index-url }}
+        run: uv pip install --editable ".[test]"
 
       - name: run pytest
         if: ${{ steps.changed-files.outputs.python_any_changed == 'true' || inputs.always_run == true }}

.github/workflows/typegen-checks.yml

@@ -54,17 +54,25 @@ jobs:
               - 'pyproject.toml'
               - 'invokeai/**'
 
+      - name: setup uv
+        if: ${{ steps.changed-files.outputs.src_any_changed == 'true' || inputs.always_run == true }}
+        uses: astral-sh/setup-uv@v5
+        with:
+          version: '0.6.10'
+          enable-cache: true
+          python-version: '3.11'
+
       - name: setup python
         if: ${{ steps.changed-files.outputs.src_any_changed == 'true' || inputs.always_run == true }}
         uses: actions/setup-python@v5
         with:
-          python-version: '3.10'
-          cache: pip
-          cache-dependency-path: pyproject.toml
+          python-version: '3.11'
 
-      - name: install python dependencies
+      - name: install dependencies
         if: ${{ steps.changed-files.outputs.src_any_changed == 'true' || inputs.always_run == true }}
-        run: pip3 install --use-pep517 --editable="."
+        env:
+          UV_INDEX: ${{ matrix.extra-index-url }}
+        run: uv pip install --editable .
 
       - name: install frontend dependencies
         if: ${{ steps.changed-files.outputs.src_any_changed == 'true' || inputs.always_run == true }}
@@ -77,7 +85,7 @@ jobs:
 
       - name: generate schema
         if: ${{ steps.changed-files.outputs.src_any_changed == 'true' || inputs.always_run == true }}
-        run: make frontend-typegen
+        run: cd invokeai/frontend/web && uv run ../../../scripts/generate_openapi_schema.py | pnpm typegen
         shell: bash
 
       - name: compare files

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/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

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,
 )
@@ -25,7 +24,6 @@ class FluxControlLoRALoaderOutput(BaseInvocationOutput):
     tags=["lora", "model", "flux"],
     category="model",
     version="1.1.1",
-    classification=Classification.Prototype,
 )
 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")
@@ -41,7 +40,6 @@ class FluxModelLoaderOutput(BaseInvocationOutput):
     tags=["model", "flux"],
     category="model",
     version="1.0.6",
-    classification=Classification.Prototype,
 )
 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
@@ -30,7 +30,6 @@ from invokeai.backend.stable_diffusion.diffusion.conditioning_data import Condit
     tags=["prompt", "conditioning", "flux"],
     category="conditioning",
     version="1.1.2",
-    classification=Classification.Prototype,
 )
 class FluxTextEncoderInvocation(BaseInvocation):
     """Encodes and preps a prompt for a flux image."""

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")

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 +1088,131 @@ 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="1.0.1"
+)
+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 the polynomial
+        feather = poly(d_norm)
+
+        # The polynomial fit isn't perfect. Points beyond the fade distance are likely to be slightly less than 1.0,
+        # even though the control points indicate that they should be exactly 1.0. This is due to the nature of the
+        # polynomial fit, which is a best approximation of the control points but not an exact match.
+
+        # When this occurs, the area outside the mask and fade-out will not be 100% transparent. For example, it may
+        # have an alpha value of 1 instead of 0. So we must force pixels at or beyond the fade distance to exactly 1.0.
+
+        # Force pixels at or beyond the fade distance to exactly 1.0
+        feather = numpy.where(d_norm >= 1.0, 1.0, feather)
+
+        # Clip any other values to ensure they're in the valid range [0,1]
+        feather = numpy.clip(feather, 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 +1283,6 @@ 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 +1309,6 @@ 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/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):

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__
 

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):
@@ -126,7 +123,6 @@ class ModelIdentifierOutput(BaseInvocationOutput):
     tags=["model"],
     category="model",
     version="1.0.1",
-    classification=Classification.Prototype,
 )
 class ModelIdentifierInvocation(BaseInvocation):
     """Selects any model, outputting it its identifier. Be careful with this one! The identifier will be accepted as
@@ -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 Model - SD1.5", tags=["model"], category="model", version="1.0.2")
+@invocation("lora_selector", title="Select LoRA", tags=["model"], category="model", version="1.0.3")
 class LoRASelectorInvocation(BaseInvocation):
     """Selects a LoRA model and weight."""
 
@@ -258,7 +254,7 @@ class LoRASelectorInvocation(BaseInvocation):
 
 
 @invocation(
-    "lora_collection_loader", title="LoRA Collection - SD1.5", tags=["model"], category="model", version="1.1.1"
+    "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."""
@@ -322,10 +318,10 @@ class SDXLLoRALoaderOutput(BaseInvocationOutput):
 
 @invocation(
     "sdxl_lora_loader",
-    title="LoRA Model - SDXL",
+    title="Apply LoRA - SDXL",
     tags=["lora", "model"],
     category="model",
-    version="1.0.4",
+    version="1.0.5",
 )
 class SDXLLoRALoaderInvocation(BaseInvocation):
     """Apply selected lora to unet and text_encoder."""
@@ -402,10 +398,10 @@ class SDXLLoRALoaderInvocation(BaseInvocation):
 
 @invocation(
     "sdxl_lora_collection_loader",
-    title="LoRA Collection - SDXL",
+    title="Apply LoRA Collection - SDXL",
     tags=["model"],
     category="model",
-    version="1.1.1",
+    version="1.1.2",
 )
 class SDXLLoRACollectionLoader(BaseInvocation):
     """Applies a collection of SDXL LoRAs to the provided UNet and CLIP models."""

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
@@ -36,7 +36,6 @@ from invokeai.backend.util.devices import TorchDevice
     tags=["image", "sd3"],
     category="image",
     version="1.1.1",
-    classification=Classification.Prototype,
 )
 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,
@@ -25,7 +25,6 @@ from invokeai.backend.util.devices import TorchDevice
     tags=["image", "latents", "vae", "i2l", "sd3"],
     category="image",
     version="1.0.1",
-    classification=Classification.Prototype,
 )
 class SD3ImageToLatentsInvocation(BaseInvocation, WithMetadata, WithBoard):
     """Generates latents from an image."""

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")
@@ -34,7 +33,6 @@ class Sd3ModelLoaderOutput(BaseInvocationOutput):
     tags=["model", "sd3"],
     category="model",
     version="1.0.1",
-    classification=Classification.Prototype,
 )
 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
@@ -33,7 +33,6 @@ SD3_T5_MAX_SEQ_LEN = 256
     tags=["prompt", "conditioning", "sd3"],
     category="conditioning",
     version="1.0.1",
-    classification=Classification.Prototype,
 )
 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")

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
@@ -56,7 +56,6 @@ def crop_controlnet_data(control_data: ControlNetData, latent_region: TBLR) -> C
     title="Tiled Multi-Diffusion Denoise - SD1.5, SDXL",
     tags=["upscale", "denoise"],
     category="latents",
-    classification=Classification.Beta,
     version="1.0.1",
 )
 class TiledMultiDiffusionDenoiseLatents(BaseInvocation):

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/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/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/flux_state_dict_utils.py

@@ -0,0 +1,23 @@
+from typing import TYPE_CHECKING
+
+if TYPE_CHECKING:
+    from invokeai.backend.model_manager.legacy_probe import CkptType
+
+
+def get_flux_in_channels_from_state_dict(state_dict: "CkptType") -> int | None:
+    """Gets the in channels from the state dict."""
+
+    # "Standard" FLUX models use "img_in.weight", but some community fine tunes use
+    # "model.diffusion_model.img_in.weight". Known models that use the latter key:
+    # - https://civitai.com/models/885098?modelVersionId=990775
+    # - https://civitai.com/models/1018060?modelVersionId=1596255
+    # - https://civitai.com/models/978314/ultrareal-fine-tune?modelVersionId=1413133
+
+    keys = {"img_in.weight", "model.diffusion_model.img_in.weight"}
+
+    for key in keys:
+        val = state_dict.get(key)
+        if val is not None:
+            return val.shape[1]
+
+    return None

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
@@ -14,27 +14,30 @@ from invokeai.backend.flux.controlnet.state_dict_utils import (
     is_state_dict_instantx_controlnet,
     is_state_dict_xlabs_controlnet,
 )
+from invokeai.backend.flux.flux_state_dict_utils import get_flux_in_channels_from_state_dict
 from invokeai.backend.flux.ip_adapter.state_dict_utils import is_state_dict_xlabs_ip_adapter
 from invokeai.backend.flux.redux.flux_redux_state_dict_utils import is_state_dict_likely_flux_redux
 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 +144,7 @@ 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 +420,22 @@ 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 +488,11 @@ 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 +560,34 @@ 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 = get_flux_in_channels_from_state_dict(state_dict)
+
+            if in_channels is None:
+                # If we cannot find the in_channels, we assume that this is a normal variant. Log a warning.
+                logger.warning(
+                    f"{self.model_path} does not have img_in.weight or model.diffusion_model.img_in.weight key. Assuming normal variant."
+                )
+                return ModelVariantType.Normal
+
+            # 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 +800,11 @@ class FluxReduxCheckpointProbe(CheckpointProbeBase):
         return BaseModelType.Flux
 
 
+class LlavaOnevisionCheckpointProbe(CheckpointProbeBase):
+    def get_base_type(self) -> BaseModelType:
+        raise NotImplementedError()
+
+
 ########################################################
 # classes for probing folders
 #######################################################
@@ -1047,6 +1085,11 @@ 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 +1125,7 @@ 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 +1139,6 @@ 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

invokeai/backend/model_manager/load/model_loader_registry.py

@@ -20,13 +20,10 @@ from typing import Callable, Dict, Optional, Tuple, Type, TypeVar
 
 from invokeai.backend.model_manager.config import (
     AnyModelConfig,
-    BaseModelType,
     ModelConfigBase,
-    ModelFormat,
-    ModelType,
-    SubModelType,
 )
 from invokeai.backend.model_manager.load import ModelLoaderBase
+from invokeai.backend.model_manager.taxonomy import BaseModelType, ModelFormat, ModelType, SubModelType
 
 
 class ModelLoaderRegistryBase(ABC):

invokeai/backend/model_manager/load/model_loaders/clip_vision.py

@@ -4,16 +4,12 @@ from typing import Optional
 from transformers import CLIPVisionModelWithProjection
 
 from invokeai.backend.model_manager.config import (
-    AnyModel,
     AnyModelConfig,
-    BaseModelType,
     DiffusersConfigBase,
-    ModelFormat,
-    ModelType,
-    SubModelType,
 )
 from invokeai.backend.model_manager.load.load_default import ModelLoader
 from invokeai.backend.model_manager.load.model_loader_registry import ModelLoaderRegistry
+from invokeai.backend.model_manager.taxonomy import AnyModel, BaseModelType, ModelFormat, ModelType, SubModelType
 
 
 @ModelLoaderRegistry.register(base=BaseModelType.Any, type=ModelType.CLIPVision, format=ModelFormat.Diffusers)

invokeai/backend/model_manager/load/model_loaders/controlnet.py

@@ -5,19 +5,19 @@ from typing import Optional
 
 from diffusers import ControlNetModel
 
-from invokeai.backend.model_manager import (
-    AnyModel,
-    AnyModelConfig,
-)
 from invokeai.backend.model_manager.config import (
-    BaseModelType,
+    AnyModelConfig,
     ControlNetCheckpointConfig,
+)
+from invokeai.backend.model_manager.load.model_loader_registry import ModelLoaderRegistry
+from invokeai.backend.model_manager.load.model_loaders.generic_diffusers import GenericDiffusersLoader
+from invokeai.backend.model_manager.taxonomy import (
+    AnyModel,
+    BaseModelType,
     ModelFormat,
     ModelType,
     SubModelType,
 )
-from invokeai.backend.model_manager.load.model_loader_registry import ModelLoaderRegistry
-from invokeai.backend.model_manager.load.model_loaders.generic_diffusers import GenericDiffusersLoader
 
 
 @ModelLoaderRegistry.register(

invokeai/backend/model_manager/load/model_loaders/flux.py

@@ -27,15 +27,8 @@ from invokeai.backend.flux.model import Flux
 from invokeai.backend.flux.modules.autoencoder import AutoEncoder
 from invokeai.backend.flux.redux.flux_redux_model import FluxReduxModel
 from invokeai.backend.flux.util import ae_params, params
-from invokeai.backend.model_manager import (
-    AnyModel,
-    AnyModelConfig,
-    BaseModelType,
-    ModelFormat,
-    ModelType,
-    SubModelType,
-)
 from invokeai.backend.model_manager.config import (
+    AnyModelConfig,
     CheckpointConfigBase,
     CLIPEmbedDiffusersConfig,
     ControlNetCheckpointConfig,
@@ -51,6 +44,13 @@ from invokeai.backend.model_manager.config import (
 )
 from invokeai.backend.model_manager.load.load_default import ModelLoader
 from invokeai.backend.model_manager.load.model_loader_registry import ModelLoaderRegistry
+from invokeai.backend.model_manager.taxonomy import (
+    AnyModel,
+    BaseModelType,
+    ModelFormat,
+    ModelType,
+    SubModelType,
+)
 from invokeai.backend.model_manager.util.model_util import (
     convert_bundle_to_flux_transformer_checkpoint,
 )

invokeai/backend/model_manager/load/model_loaders/generic_diffusers.py

@@ -8,18 +8,16 @@ from typing import Any, Optional
 from diffusers.configuration_utils import ConfigMixin
 from diffusers.models.modeling_utils import ModelMixin
 
-from invokeai.backend.model_manager import (
+from invokeai.backend.model_manager.config import AnyModelConfig, DiffusersConfigBase, InvalidModelConfigException
+from invokeai.backend.model_manager.load.load_default import ModelLoader
+from invokeai.backend.model_manager.load.model_loader_registry import ModelLoaderRegistry
+from invokeai.backend.model_manager.taxonomy import (
     AnyModel,
-    AnyModelConfig,
     BaseModelType,
-    InvalidModelConfigException,
     ModelFormat,
     ModelType,
     SubModelType,
 )
-from invokeai.backend.model_manager.config import DiffusersConfigBase
-from invokeai.backend.model_manager.load.load_default import ModelLoader
-from invokeai.backend.model_manager.load.model_loader_registry import ModelLoaderRegistry
 
 
 @ModelLoaderRegistry.register(base=BaseModelType.Any, type=ModelType.T2IAdapter, format=ModelFormat.Diffusers)

invokeai/backend/model_manager/load/model_loaders/ip_adapter.py

@@ -7,8 +7,9 @@ from typing import Optional
 import torch
 
 from invokeai.backend.ip_adapter.ip_adapter import build_ip_adapter
-from invokeai.backend.model_manager import AnyModel, AnyModelConfig, BaseModelType, ModelFormat, ModelType, SubModelType
+from invokeai.backend.model_manager.config import AnyModelConfig
 from invokeai.backend.model_manager.load import ModelLoader, ModelLoaderRegistry
+from invokeai.backend.model_manager.taxonomy import AnyModel, BaseModelType, ModelFormat, ModelType, SubModelType
 from invokeai.backend.raw_model import RawModel
 
 

invokeai/backend/model_manager/load/model_loaders/llava_onevision.py

@@ -0,0 +1,28 @@
+from pathlib import Path
+from typing import Optional
+
+from invokeai.backend.llava_onevision_model import LlavaOnevisionModel
+from invokeai.backend.model_manager.config import (
+    AnyModelConfig,
+)
+from invokeai.backend.model_manager.load.load_default import ModelLoader
+from invokeai.backend.model_manager.load.model_loader_registry import ModelLoaderRegistry
+from invokeai.backend.model_manager.taxonomy import AnyModel, BaseModelType, ModelFormat, ModelType, SubModelType
+
+
+@ModelLoaderRegistry.register(base=BaseModelType.Any, type=ModelType.LlavaOnevision, format=ModelFormat.Diffusers)
+class LlavaOnevisionModelLoader(ModelLoader):
+    """Class for loading LLaVA Onevision VLLM models."""
+
+    def _load_model(
+        self,
+        config: AnyModelConfig,
+        submodel_type: Optional[SubModelType] = None,
+    ) -> AnyModel:
+        if submodel_type is not None:
+            raise ValueError("Unexpected submodel requested for LLaVA OneVision model.")
+
+        model_path = Path(config.path)
+        model = LlavaOnevisionModel.load_from_path(model_path)
+        model.to(dtype=self._torch_dtype)
+        return model

invokeai/backend/model_manager/load/model_loaders/lora.py

@@ -9,17 +9,17 @@ import torch
 from safetensors.torch import load_file
 
 from invokeai.app.services.config import InvokeAIAppConfig
-from invokeai.backend.model_manager import (
+from invokeai.backend.model_manager.config import AnyModelConfig
+from invokeai.backend.model_manager.load.load_default import ModelLoader
+from invokeai.backend.model_manager.load.model_cache.model_cache import ModelCache
+from invokeai.backend.model_manager.load.model_loader_registry import ModelLoaderRegistry
+from invokeai.backend.model_manager.taxonomy import (
     AnyModel,
-    AnyModelConfig,
     BaseModelType,
     ModelFormat,
     ModelType,
     SubModelType,
 )
-from invokeai.backend.model_manager.load.load_default import ModelLoader
-from invokeai.backend.model_manager.load.model_cache.model_cache import ModelCache
-from invokeai.backend.model_manager.load.model_loader_registry import ModelLoaderRegistry
 from invokeai.backend.patches.lora_conversions.flux_control_lora_utils import (
     is_state_dict_likely_flux_control,
     lora_model_from_flux_control_state_dict,

invokeai/backend/model_manager/load/model_loaders/onnx.py

@@ -5,16 +5,16 @@
 from pathlib import Path
 from typing import Optional
 
-from invokeai.backend.model_manager import (
+from invokeai.backend.model_manager.config import AnyModelConfig
+from invokeai.backend.model_manager.load.model_loader_registry import ModelLoaderRegistry
+from invokeai.backend.model_manager.load.model_loaders.generic_diffusers import GenericDiffusersLoader
+from invokeai.backend.model_manager.taxonomy import (
     AnyModel,
-    AnyModelConfig,
     BaseModelType,
     ModelFormat,
     ModelType,
     SubModelType,
 )
-from invokeai.backend.model_manager.load.model_loader_registry import ModelLoaderRegistry
-from invokeai.backend.model_manager.load.model_loaders.generic_diffusers import GenericDiffusersLoader
 
 
 @ModelLoaderRegistry.register(base=BaseModelType.Any, type=ModelType.ONNX, format=ModelFormat.ONNX)

invokeai/backend/model_manager/load/model_loaders/sig_lip_pipeline.py

@@ -2,15 +2,11 @@ from pathlib import Path
 from typing import Optional
 
 from invokeai.backend.model_manager.config import (
-    AnyModel,
     AnyModelConfig,
-    BaseModelType,
-    ModelFormat,
-    ModelType,
-    SubModelType,
 )
 from invokeai.backend.model_manager.load.load_default import ModelLoader
 from invokeai.backend.model_manager.load.model_loader_registry import ModelLoaderRegistry
+from invokeai.backend.model_manager.taxonomy import AnyModel, BaseModelType, ModelFormat, ModelType, SubModelType
 from invokeai.backend.sig_lip.sig_lip_pipeline import SigLipPipeline
 
 

invokeai/backend/model_manager/load/model_loaders/spandrel_image_to_image.py

@@ -4,15 +4,11 @@ from typing import Optional
 import torch
 
 from invokeai.backend.model_manager.config import (
-    AnyModel,
     AnyModelConfig,
-    BaseModelType,
-    ModelFormat,
-    ModelType,
-    SubModelType,
 )
 from invokeai.backend.model_manager.load.load_default import ModelLoader
 from invokeai.backend.model_manager.load.model_loader_registry import ModelLoaderRegistry
+from invokeai.backend.model_manager.taxonomy import AnyModel, BaseModelType, ModelFormat, ModelType, SubModelType
 from invokeai.backend.spandrel_image_to_image_model import SpandrelImageToImageModel
 
 

invokeai/backend/model_manager/load/model_loaders/stable_diffusion.py

@@ -11,16 +11,8 @@ from diffusers import (
     StableDiffusionXLPipeline,
 )
 
-from invokeai.backend.model_manager import (
-    AnyModel,
-    AnyModelConfig,
-    BaseModelType,
-    ModelFormat,
-    ModelType,
-    ModelVariantType,
-    SubModelType,
-)
 from invokeai.backend.model_manager.config import (
+    AnyModelConfig,
     CheckpointConfigBase,
     DiffusersConfigBase,
     MainCheckpointConfig,
@@ -28,6 +20,14 @@ from invokeai.backend.model_manager.config import (
 from invokeai.backend.model_manager.load.model_cache.model_cache import get_model_cache_key
 from invokeai.backend.model_manager.load.model_loader_registry import ModelLoaderRegistry
 from invokeai.backend.model_manager.load.model_loaders.generic_diffusers import GenericDiffusersLoader
+from invokeai.backend.model_manager.taxonomy import (
+    AnyModel,
+    BaseModelType,
+    ModelFormat,
+    ModelType,
+    ModelVariantType,
+    SubModelType,
+)
 from invokeai.backend.util.silence_warnings import SilenceWarnings
 
 VARIANT_TO_IN_CHANNEL_MAP = {

invokeai/backend/model_manager/load/model_loaders/textual_inversion.py

@@ -4,16 +4,16 @@
 from pathlib import Path
 from typing import Optional
 
-from invokeai.backend.model_manager import (
+from invokeai.backend.model_manager.config import AnyModelConfig
+from invokeai.backend.model_manager.load.load_default import ModelLoader
+from invokeai.backend.model_manager.load.model_loader_registry import ModelLoaderRegistry
+from invokeai.backend.model_manager.taxonomy import (
     AnyModel,
-    AnyModelConfig,
     BaseModelType,
     ModelFormat,
     ModelType,
     SubModelType,
 )
-from invokeai.backend.model_manager.load.load_default import ModelLoader
-from invokeai.backend.model_manager.load.model_loader_registry import ModelLoaderRegistry
 from invokeai.backend.textual_inversion import TextualInversionModelRaw
 
 

invokeai/backend/model_manager/load/model_loaders/vae.py

@@ -5,15 +5,16 @@ from typing import Optional
 
 from diffusers import AutoencoderKL
 
-from invokeai.backend.model_manager import (
-    AnyModelConfig,
+from invokeai.backend.model_manager.config import AnyModelConfig, VAECheckpointConfig
+from invokeai.backend.model_manager.load.model_loader_registry import ModelLoaderRegistry
+from invokeai.backend.model_manager.load.model_loaders.generic_diffusers import GenericDiffusersLoader
+from invokeai.backend.model_manager.taxonomy import (
+    AnyModel,
     BaseModelType,
     ModelFormat,
     ModelType,
+    SubModelType,
 )
-from invokeai.backend.model_manager.config import AnyModel, SubModelType, VAECheckpointConfig
-from invokeai.backend.model_manager.load.model_loader_registry import ModelLoaderRegistry
-from invokeai.backend.model_manager.load.model_loaders.generic_diffusers import GenericDiffusersLoader
 
 
 @ModelLoaderRegistry.register(base=BaseModelType.Any, type=ModelType.VAE, format=ModelFormat.Diffusers)

invokeai/backend/model_manager/load/model_util.py

@@ -15,7 +15,8 @@ from invokeai.backend.image_util.depth_anything.depth_anything_pipeline import D
 from invokeai.backend.image_util.grounding_dino.grounding_dino_pipeline import GroundingDinoPipeline
 from invokeai.backend.image_util.segment_anything.segment_anything_pipeline import SegmentAnythingPipeline
 from invokeai.backend.ip_adapter.ip_adapter import IPAdapter
-from invokeai.backend.model_manager.config import AnyModel
+from invokeai.backend.llava_onevision_model import LlavaOnevisionModel
+from invokeai.backend.model_manager.taxonomy import AnyModel
 from invokeai.backend.onnx.onnx_runtime import IAIOnnxRuntimeModel
 from invokeai.backend.patches.model_patch_raw import ModelPatchRaw
 from invokeai.backend.sig_lip.sig_lip_pipeline import SigLipPipeline
@@ -50,6 +51,7 @@ def calc_model_size_by_data(logger: logging.Logger, model: AnyModel) -> int:
             SegmentAnythingPipeline,
             DepthAnythingPipeline,
             SigLipPipeline,
+            LlavaOnevisionModel,
         ),
     ):
         return model.calc_size()

invokeai/backend/model_manager/metadata/fetch/fetch_base.py

@@ -17,12 +17,12 @@ from typing import Optional
 from pydantic.networks import AnyHttpUrl
 from requests.sessions import Session
 
-from invokeai.backend.model_manager import ModelRepoVariant
 from invokeai.backend.model_manager.metadata.metadata_base import (
     AnyModelRepoMetadata,
     AnyModelRepoMetadataValidator,
     BaseMetadata,
 )
+from invokeai.backend.model_manager.taxonomy import ModelRepoVariant
 
 
 class ModelMetadataFetchBase(ABC):

invokeai/backend/model_manager/metadata/fetch/huggingface.py

@@ -24,7 +24,6 @@ from huggingface_hub.errors import RepositoryNotFoundError, RevisionNotFoundErro
 from pydantic.networks import AnyHttpUrl
 from requests.sessions import Session
 
-from invokeai.backend.model_manager.config import ModelRepoVariant
 from invokeai.backend.model_manager.metadata.fetch.fetch_base import ModelMetadataFetchBase
 from invokeai.backend.model_manager.metadata.metadata_base import (
     AnyModelRepoMetadata,
@@ -32,6 +31,7 @@ from invokeai.backend.model_manager.metadata.metadata_base import (
     RemoteModelFile,
     UnknownMetadataException,
 )
+from invokeai.backend.model_manager.taxonomy import ModelRepoVariant
 
 HF_MODEL_RE = r"https?://huggingface.co/([\w\-.]+/[\w\-.]+)"
 

invokeai/backend/model_manager/metadata/metadata_base.py

@@ -23,7 +23,7 @@ from pydantic.networks import AnyHttpUrl
 from requests.sessions import Session
 from typing_extensions import Annotated
 
-from invokeai.backend.model_manager import ModelRepoVariant
+from invokeai.backend.model_manager.taxonomy import ModelRepoVariant
 from invokeai.backend.model_manager.util.select_hf_files import filter_files
 
 

invokeai/backend/model_manager/starter_models.py

@@ -2,7 +2,7 @@ from typing import Optional
 
 from pydantic import BaseModel
 
-from invokeai.backend.model_manager.config import BaseModelType, ModelFormat, ModelType
+from invokeai.backend.model_manager.taxonomy import BaseModelType, ModelFormat, ModelType
 
 
 class StarterModelWithoutDependencies(BaseModel):
@@ -614,6 +614,26 @@ flux_redux = StarterModel(
 )
 # endregion
 
+# region LlavaOnevisionModel
+llava_onevision = StarterModel(
+    name="LLaVA Onevision Qwen2 0.5B",
+    base=BaseModelType.Any,
+    source="llava-hf/llava-onevision-qwen2-0.5b-ov-hf",
+    description="LLaVA Onevision VLLM model",
+    type=ModelType.LlavaOnevision,
+)
+# endregion
+
+# region FLUX Fill
+flux_fill = StarterModel(
+    name="FLUX Fill",
+    base=BaseModelType.Flux,
+    source="black-forest-labs/FLUX.1-Fill-dev::flux1-fill-dev.safetensors",
+    description="FLUX Fill model (for inpainting).",
+    type=ModelType.Main,
+)
+# endregion
+
 # List of starter models, displayed on the frontend.
 # The order/sort of this list is not changed by the frontend - set it how you want it here.
 STARTER_MODELS: list[StarterModel] = [
@@ -683,6 +703,8 @@ STARTER_MODELS: list[StarterModel] = [
     clip_l_encoder,
     siglip,
     flux_redux,
+    llava_onevision,
+    flux_fill,
 ]
 
 sd1_bundle: list[StarterModel] = [
@@ -731,6 +753,7 @@ flux_bundle: list[StarterModel] = [
     flux_canny_control_lora,
     flux_depth_control_lora,
     flux_redux,
+    flux_fill,
 ]
 
 STARTER_BUNDLES: dict[str, list[StarterModel]] = {

invokeai/backend/model_manager/taxonomy.py

@@ -0,0 +1,129 @@
+from enum import Enum
+from typing import Dict, TypeAlias, Union
+
+import diffusers
+import onnxruntime as ort
+import torch
+from diffusers import ModelMixin
+
+from invokeai.backend.raw_model import RawModel
+
+# 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
+]
+
+
+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"
+    LlavaOnevision = "llava_onevision"
+
+
+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"
+
+
+AnyVariant: TypeAlias = Union[ModelVariantType, ClipVariantType, None]

invokeai/backend/model_manager/util/model_util.py

@@ -1,14 +1,14 @@
-"""Utilities for parsing model files, used mostly by probe.py"""
+"""Utilities for parsing model files, used mostly by legacy_probe.py"""
 
 import json
 from pathlib import Path
 from typing import Dict, Optional, Union
 
+import picklescan.scanner as pscan
 import safetensors
 import torch
-from picklescan.scanner import scan_file_path
 
-from invokeai.backend.model_manager.config import ClipVariantType
+from invokeai.backend.model_manager.taxonomy import ClipVariantType
 from invokeai.backend.quantization.gguf.loaders import gguf_sd_loader
 
 
@@ -57,9 +57,12 @@ def read_checkpoint_meta(path: Union[str, Path], scan: bool = True) -> Dict[str,
         checkpoint = gguf_sd_loader(Path(path), compute_dtype=torch.float32)
     else:
         if scan:
-            scan_result = scan_file_path(path)
-            if scan_result.infected_files != 0 or scan_result.scan_err:
-                raise Exception(f'The model file "{path}" is potentially infected by malware. Aborting import.')
+            scan_result = pscan.scan_file_path(path)
+            if scan_result.infected_files != 0:
+                raise Exception(f"The model at {path} is potentially infected by malware. Aborting import.")
+            if scan_result.scan_err:
+                raise Exception(f"Error scanning model at {path} for malware. Aborting import.")
+
         checkpoint = torch.load(path, map_location=torch.device("meta"))
     return checkpoint
 

invokeai/backend/model_manager/util/select_hf_files.py

@@ -17,7 +17,7 @@ from dataclasses import dataclass
 from pathlib import Path
 from typing import Dict, List, Optional, Set
 
-from invokeai.backend.model_manager.config import ModelRepoVariant
+from invokeai.backend.model_manager.taxonomy import ModelRepoVariant
 
 
 def filter_files(

invokeai/backend/stable_diffusion/extensions/t2i_adapter.py

@@ -8,7 +8,7 @@ from diffusers import T2IAdapter
 from PIL.Image import Image
 
 from invokeai.app.util.controlnet_utils import prepare_control_image
-from invokeai.backend.model_manager import BaseModelType
+from invokeai.backend.model_manager.taxonomy import BaseModelType
 from invokeai.backend.stable_diffusion.diffusion.conditioning_data import ConditioningMode
 from invokeai.backend.stable_diffusion.extension_callback_type import ExtensionCallbackType
 from invokeai.backend.stable_diffusion.extensions.base import ExtensionBase, callback

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

@@ -114,7 +114,9 @@
         "layout": "Layout",
         "board": "Ordner",
         "combinatorial": "Kombinatorisch",
-        "saveChanges": "Änderungen speichern"
+        "saveChanges": "Änderungen speichern",
+        "error_withCount_one": "{{count}} Fehler",
+        "error_withCount_other": "{{count}} Fehler"
     },
     "gallery": {
         "galleryImageSize": "Bildgröße",
@@ -764,10 +766,10 @@
         "layerCopiedToClipboard": "Ebene in die Zwischenablage kopiert",
         "sentToCanvas": "An Leinwand gesendet",
         "problemDeletingWorkflow": "Problem beim Löschen des Arbeitsablaufs",
-        "uploadFailedInvalidUploadDesc_withCount_one": "Es darf maximal 1 PNG- oder JPEG-Bild sein.",
-        "uploadFailedInvalidUploadDesc_withCount_other": "Es dürfen maximal {{count}} PNG- oder JPEG-Bilder sein.",
+        "uploadFailedInvalidUploadDesc_withCount_one": "Darf maximal 1 PNG-, JPEG- oder WEBP-Bild sein.",
+        "uploadFailedInvalidUploadDesc_withCount_other": "Dürfen maximal {{count}} PNG-, JPEG- oder WEBP-Bild sein.",
         "problemRetrievingWorkflow": "Problem beim Abrufen des Arbeitsablaufs",
-        "uploadFailedInvalidUploadDesc": "Müssen PNG- oder JPEG-Bilder sein.",
+        "uploadFailedInvalidUploadDesc": "Müssen PNG-, JPEG- oder WEBP-Bilder sein.",
         "pasteSuccess": "Eingefügt in {{destination}}",
         "pasteFailed": "Einfügen fehlgeschlagen",
         "unableToCopy": "Kopieren nicht möglich",
@@ -1259,7 +1261,6 @@
         "nodePack": "Knoten-Pack",
         "loadWorkflow": "Lade Workflow",
         "snapToGrid": "Am Gitternetz einrasten",
-        "unknownOutput": "Unbekannte Ausgabe: {{name}}",
         "updateNode": "Knoten updaten",
         "edge": "Rand / Kante",
         "sourceNodeDoesNotExist": "Ungültiger Rand: Quell- / Ausgabe-Knoten {{node}} existiert nicht",
@@ -1325,7 +1326,9 @@
         "description": "Beschreibung",
         "loadWorkflowDesc": "Arbeitsablauf laden?",
         "loadWorkflowDesc2": "Ihr aktueller Arbeitsablauf enthält nicht gespeicherte Änderungen.",
-        "loadingTemplates": "Lade {{name}}"
+        "loadingTemplates": "Lade {{name}}",
+        "missingSourceOrTargetHandle": "Fehlender Quell- oder Zielgriff",
+        "missingSourceOrTargetNode": "Fehlender Quell- oder Zielknoten"
     },
     "hrf": {
         "enableHrf": "Korrektur für hohe Auflösungen",

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

@@ -194,7 +194,10 @@
         "combinatorial": "Combinatorial",
         "layout": "Layout",
         "row": "Row",
-        "column": "Column"
+        "column": "Column",
+        "value": "Value",
+        "label": "Label",
+        "systemInformation": "System Information"
     },
     "hrf": {
         "hrf": "High Resolution Fix",
@@ -846,6 +849,7 @@
         "starterModels": "Starter Models",
         "starterModelsInModelManager": "Starter Models can be found in Model Manager",
         "controlLora": "Control LoRA",
+        "llavaOnevision": "LLaVA OneVision",
         "syncModels": "Sync Models",
         "textualInversions": "Textual Inversions",
         "triggerPhrases": "Trigger Phrases",
@@ -1013,7 +1017,10 @@
         "unknownNodeType": "Unknown node type",
         "unknownTemplate": "Unknown Template",
         "unknownInput": "Unknown input: {{name}}",
-        "unknownOutput": "Unknown output: {{name}}",
+        "missingField_withName": "Missing field \"{{name}}\"",
+        "unexpectedField_withName": "Unexpected field \"{{name}}\"",
+        "unknownField_withName": "Unknown field \"{{name}}\"",
+        "unknownFieldEditWorkflowToFix_withName": "Workflow contains an unknown field \"{{name}}\".\nEdit the workflow to fix the issue.",
         "updateNode": "Update Node",
         "updateApp": "Update App",
         "loadingTemplates": "Loading {{name}}",
@@ -1298,7 +1305,8 @@
         "problemDeletingWorkflow": "Problem Deleting Workflow",
         "unableToCopy": "Unable to Copy",
         "unableToCopyDesc": "Your browser does not support clipboard access. Firefox users may be able to fix this by following ",
-        "unableToCopyDesc_theseSteps": "these steps"
+        "unableToCopyDesc_theseSteps": "these steps",
+        "fluxFillIncompatibleWithT2IAndI2I": "FLUX Fill is not compatible with Text to Image or Image to Image. Use other FLUX models for these tasks."
     },
     "popovers": {
         "clipSkip": {
@@ -1700,6 +1708,7 @@
         "shared": "Shared",
         "browseWorkflows": "Browse Workflows",
         "deselectAll": "Deselect All",
+        "recommended": "Recommended For You",
         "opened": "Opened",
         "openWorkflow": "Open Workflow",
         "updated": "Updated",
@@ -1738,6 +1747,7 @@
         "openLibrary": "Open Library",
         "workflowThumbnail": "Workflow Thumbnail",
         "saveChanges": "Save Changes",
+        "emptyStringPlaceholder": "<empty string>",
         "builder": {
             "deleteAllElements": "Delete All Form Elements",
             "resetAllNodeFields": "Reset All Node Fields",
@@ -1762,6 +1772,9 @@
             "singleLine": "Single Line",
             "multiLine": "Multi Line",
             "slider": "Slider",
+            "dropdown": "Dropdown",
+            "addOption": "Add Option",
+            "resetOptions": "Reset Options",
             "both": "Both",
             "emptyRootPlaceholderViewMode": "Click Edit to start building a form for this workflow.",
             "emptyRootPlaceholderEditMode": "Drag a form element or node field here to get started.",
@@ -1948,7 +1961,8 @@
             "rgNegativePromptNotSupported": "Negative Prompt not supported for selected base model",
             "rgReferenceImagesNotSupported": "regional Reference Images not supported for selected base model",
             "rgAutoNegativeNotSupported": "Auto-Negative not supported for selected base model",
-            "rgNoRegion": "no region drawn"
+            "rgNoRegion": "no region drawn",
+            "fluxFillIncompatibleWithControlLoRA": "Control LoRA is not compatible with FLUX Fill"
         },
         "errors": {
             "unableToFindImage": "Unable to find image",
@@ -2330,8 +2344,9 @@
     "whatsNew": {
         "whatsNewInInvoke": "What's New in Invoke",
         "items": [
-            "Workflows: New and improved Workflow Library.",
-            "FLUX: Support for FLUX Redux in Workflows and Canvas."
+            "Workflows: Support for custom string drop-downs in Workflow Builder.",
+            "FLUX: Support for FLUX Fill in Workflows and Canvas.",
+            "LLaVA OneVision VLLM: Beta support in Workflows."
         ],
         "readReleaseNotes": "Read Release Notes",
         "watchRecentReleaseVideos": "Watch Recent Release Videos",

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

@@ -1653,7 +1653,6 @@
         "collectionFieldType": "{{name}} (Collection)",
         "newWorkflow": "Nouveau Workflow",
         "reorderLinearView": "Réorganiser la vue linéaire",
-        "unknownOutput": "Sortie inconnue : {{name}}",
         "outputFieldTypeParseError": "Impossible d'analyser le type du champ de sortie {{node}}.{{field}} ({{message}})",
         "unsupportedMismatchedUnion": "type CollectionOrScalar non concordant avec les types de base {{firstType}} et {{secondType}}",
         "unableToParseFieldType": "impossible d'analyser le type de champ",

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

@@ -113,7 +113,9 @@
         "saveChanges": "Salva modifiche",
         "error_withCount_one": "{{count}} errore",
         "error_withCount_many": "{{count}} errori",
-        "error_withCount_other": "{{count}} errori"
+        "error_withCount_other": "{{count}} errori",
+        "value": "Valore",
+        "label": "Etichetta"
     },
     "gallery": {
         "galleryImageSize": "Dimensione dell'immagine",
@@ -779,7 +781,8 @@
         "enableModelDescriptions": "Abilita le descrizioni dei modelli nei menu a discesa",
         "modelDescriptionsDisabled": "Descrizioni dei modelli nei menu a discesa disabilitate",
         "modelDescriptionsDisabledDesc": "Le descrizioni dei modelli nei menu a discesa sono state disabilitate. Abilitale nelle Impostazioni.",
-        "showDetailedInvocationProgress": "Mostra dettagli avanzamento"
+        "showDetailedInvocationProgress": "Mostra dettagli avanzamento",
+        "enableHighlightFocusedRegions": "Evidenzia le regioni interessate"
     },
     "toast": {
         "uploadFailed": "Caricamento fallito",
@@ -847,7 +850,8 @@
         "pasteSuccess": "Incollato su {{destination}}",
         "unableToCopy": "Impossibile copiare",
         "unableToCopyDesc": "Il tuo browser non supporta l'accesso agli appunti. Gli utenti di Firefox potrebbero risolvere il problema seguendo ",
-        "unableToCopyDesc_theseSteps": "questi passaggi"
+        "unableToCopyDesc_theseSteps": "questi passaggi",
+        "fluxFillIncompatibleWithT2IAndI2I": "FLUX Fill non è compatibile con Testo a Immagine o Immagine a Immagine. Per queste attività, utilizzare altri modelli FLUX."
     },
     "accessibility": {
         "invokeProgressBar": "Barra di avanzamento generazione",
@@ -968,7 +972,6 @@
         "unableToGetWorkflowVersion": "Impossibile ottenere la versione dello schema del flusso di lavoro",
         "nodePack": "Pacchetto di nodi",
         "unableToExtractSchemaNameFromRef": "Impossibile estrarre il nome dello schema dal riferimento",
-        "unknownOutput": "Output sconosciuto: {{name}}",
         "unknownNodeType": "Tipo di nodo sconosciuto",
         "targetNodeDoesNotExist": "Connessione non valida: il nodo di destinazione/input {{node}} non esiste",
         "unknownFieldType": "$t(nodes.unknownField) tipo: {{type}}",
@@ -1038,7 +1041,11 @@
         "generatorImages_many": "{{count}} immagini",
         "generatorImages_other": "{{count}} immagini",
         "generatorImagesFromBoard": "Immagini dalla Bacheca",
-        "missingSourceOrTargetNode": "Nodo sorgente o di destinazione mancante"
+        "missingSourceOrTargetNode": "Nodo sorgente o di destinazione mancante",
+        "unknownField_withName": "Campo \"{{name}}\" sconosciuto",
+        "missingField_withName": "Campo \"{{name}}\" mancante",
+        "unknownFieldEditWorkflowToFix_withName": "Il flusso di lavoro contiene un campo \"{{name}}\" sconosciuto .\nModifica il flusso di lavoro per risolvere il problema.",
+        "unexpectedField_withName": "Campo \"{{name}}\" inaspettato"
     },
     "boards": {
         "autoAddBoard": "Aggiungi automaticamente bacheca",
@@ -1776,7 +1783,12 @@
             "text": "Testo",
             "numberInput": "Ingresso numerico",
             "containerRowLayout": "Contenitore (disposizione riga)",
-            "containerColumnLayout": "Contenitore (disposizione colonna)"
+            "containerColumnLayout": "Contenitore (disposizione colonna)",
+            "minimum": "Minimo",
+            "maximum": "Massimo",
+            "dropdown": "Elenco a discesa",
+            "addOption": "Aggiungi opzione",
+            "resetOptions": "Reimposta opzioni"
         },
         "loadMore": "Carica altro",
         "searchPlaceholder": "Cerca per nome, descrizione o etichetta",
@@ -1791,7 +1803,9 @@
         "private": "Privato",
         "deselectAll": "Deseleziona tutto",
         "noRecentWorkflows": "Nessun flusso di lavoro recente",
-        "view": "Visualizza"
+        "view": "Visualizza",
+        "recommended": "Consigliato per te",
+        "emptyStringPlaceholder": "<stringa vuota>"
     },
     "accordions": {
         "compositing": {
@@ -2235,7 +2249,8 @@
             "rgNegativePromptNotSupported": "Prompt negativo non supportato per il modello base selezionato",
             "ipAdapterIncompatibleBaseModel": "modello base dell'immagine di riferimento incompatibile",
             "ipAdapterNoImageSelected": "nessuna immagine di riferimento selezionata",
-            "rgAutoNegativeNotSupported": "Auto-Negativo non supportato per il modello base selezionato"
+            "rgAutoNegativeNotSupported": "Auto-Negativo non supportato per il modello base selezionato",
+            "fluxFillIncompatibleWithControlLoRA": "Il controllo LoRA non è compatibile con FLUX Fill"
         },
         "pasteTo": "Incolla su",
         "pasteToBboxDesc": "Nuovo livello (nel riquadro di delimitazione)",
@@ -2350,8 +2365,8 @@
         "watchRecentReleaseVideos": "Guarda i video su questa versione",
         "watchUiUpdatesOverview": "Guarda le novità dell'interfaccia",
         "items": [
-            "Gestione della memoria: nuova impostazione per gli utenti con GPU Nvidia per ridurre l'utilizzo della VRAM.",
-            "Prestazioni: continui miglioramenti alle prestazioni e alla reattività complessive dell'applicazione."
+            "Flussi di lavoro: nuova e migliorata libreria dei flussi di lavoro.",
+            "FLUX: supporto per FLUX Redux e FLUX Fill in Flussi di lavoro e Tela."
         ]
     },
     "system": {

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

@@ -425,7 +425,6 @@
         "newWorkflow": "Nieuwe werkstroom",
         "unknownErrorValidatingWorkflow": "Onbekende fout bij valideren werkstroom",
         "unsupportedAnyOfLength": "te veel union-leden ({{count}})",
-        "unknownOutput": "Onbekende uitvoer: {{name}}",
         "viewMode": "Gebruik in lineaire weergave",
         "unableToExtractSchemaNameFromRef": "fout bij het extraheren van de schemanaam via de ref",
         "unsupportedMismatchedUnion": "niet-overeenkomende soort CollectionOrScalar met basissoorten {{firstType}} en {{secondType}}",

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

@@ -879,7 +879,6 @@
         "unableToExtractSchemaNameFromRef": "невозможно извлечь имя схемы из ссылки",
         "executionStateError": "Ошибка",
         "prototypeDesc": "Этот вызов является прототипом. Он может претерпевать изменения при обновлении приложения и может быть удален в любой момент.",
-        "unknownOutput": "Неизвестный вывод: {{name}}",
         "executionStateCompleted": "Выполнено",
         "node": "Узел",
         "workflowAuthor": "Автор",

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

@@ -236,7 +236,8 @@
         "layout": "Bố Cục",
         "row": "Hàng",
         "board": "Bảng",
-        "saveChanges": "Lưu Thay Đổi"
+        "saveChanges": "Lưu Thay Đổi",
+        "error_withCount_other": "{{count}} lỗi"
     },
     "prompt": {
         "addPromptTrigger": "Thêm Prompt Trigger",
@@ -769,7 +770,8 @@
         "urlForbiddenErrorMessage": "Bạn có thể cần yêu cầu quyền truy cập từ trang web đang cung cấp model.",
         "urlUnauthorizedErrorMessage": "Bạn có thể cần thiếp lập một token API để dùng được model này.",
         "fluxRedux": "FLUX Redux",
-        "sigLip": "SigLIP"
+        "sigLip": "SigLIP",
+        "llavaOnevision": "LLaVA OneVision"
     },
     "metadata": {
         "guidance": "Hướng Dẫn",
@@ -892,7 +894,6 @@
         "targetNodeFieldDoesNotExist": "Kết nối không phù hợp: đích đến/đầu vào của vùng {{node}}.{{field}} không tồn tại",
         "missingTemplate": "Node không hợp lệ: node {{node}} thuộc loại {{type}} bị thiếu mẫu trình bày (chưa tải?)",
         "unsupportedMismatchedUnion": "Dạng số lượng dữ liệu không khớp với {{firstType}} và {{secondType}}",
-        "unknownOutput": "Đầu Ra Không Rõ: {{name}}",
         "betaDesc": "Trình kích hoạt này vẫn trong giai đoạn beta. Cho đến khi ổn định, nó có thể phá hỏng thay đổi trong khi cập nhật ứng dụng. Chúng tôi dự định hỗ trợ trình kích hoạt này về lâu dài.",
         "cannotConnectInputToInput": "Không thế kết nối đầu vào với đầu vào",
         "showEdgeLabelsHelp": "Hiển thị tên trên kết nối, chỉ ra những node được kết nối",
@@ -1019,7 +1020,11 @@
         "downloadWorkflowError": "Lỗi tải xuống workflow",
         "generatorImagesFromBoard": "Ảnh Từ Bảng",
         "generatorImagesCategory": "Phân Loại",
-        "generatorImages_other": "{{count}} ảnh"
+        "generatorImages_other": "{{count}} ảnh",
+        "unknownField_withName": "Vùng Dữ Liệu Không Rõ \"{{name}}\"",
+        "unexpectedField_withName": "Sai Vùng Dữ Liệu \"{{name}}\"",
+        "unknownFieldEditWorkflowToFix_withName": "Workflow chứa vùng dữ liệu không rõ \"{{name}}\".\nHãy biên tập workflow để sửa lỗi.",
+        "missingField_withName": "Thiếu Vùng Dữ Liệu \"{{name}}\""
     },
     "popovers": {
         "paramCFGRescaleMultiplier": {
@@ -1618,7 +1623,8 @@
         "displayInProgress": "Hiển Thị Hình Ảnh Đang Xử Lý",
         "intermediatesClearedFailed": "Có Vấn Đề Khi Dọn Sạch Sản Phẩm Trung Gian",
         "enableInvisibleWatermark": "Bật Chế Độ Ẩn Watermark",
-        "showDetailedInvocationProgress": "Hiện Dữ Liệu Xử Lý"
+        "showDetailedInvocationProgress": "Hiện Dữ Liệu Xử Lý",
+        "enableHighlightFocusedRegions": "Nhấn Mạnh Khu Vực Chỉ Định"
     },
     "sdxl": {
         "loading": "Đang Tải...",
@@ -2048,7 +2054,8 @@
             "rgNegativePromptNotSupported": "Lệnh Tiêu Cực không được hỗ trợ cho model cơ sở được chọn",
             "rgReferenceImagesNotSupported": "Ảnh Mẫu Khu Vực không được hỗ trợ cho model cơ sở được chọn",
             "rgAutoNegativeNotSupported": "Tự Động Đảo Chiều không được hỗ trợ cho model cơ sở được chọn",
-            "rgNoRegion": "không có khu vực được vẽ"
+            "rgNoRegion": "không có khu vực được vẽ",
+            "fluxFillIncompatibleWithControlLoRA": "LoRA Điều Khiển Được không tương tích với FLUX Fill"
         },
         "pasteTo": "Dán Vào",
         "pasteToAssets": "Tài Nguyên",
@@ -2199,7 +2206,8 @@
         "unableToCopyDesc_theseSteps": "các bước sau",
         "unableToCopyDesc": "Trình duyệt của bạn không hỗ trợ tính năng clipboard. Người dùng Firefox có thể khắc phục theo ",
         "pasteSuccess": "Dán Vào {{destination}}",
-        "pasteFailed": "Dán Thất Bại"
+        "pasteFailed": "Dán Thất Bại",
+        "fluxFillIncompatibleWithT2IAndI2I": "FLUX Fill không tương tích với Từ Ngữ Sang Hình Ảnh và Hình Ảnh Sang Hình Ảnh. Dùng model FLUX khác cho các tính năng này."
     },
     "ui": {
         "tabs": {
@@ -2288,7 +2296,11 @@
             "container": "Hộp Chứa",
             "heading": "Đầu Dòng",
             "text": "Văn Bản",
-            "divider": "Gạch Chia"
+            "divider": "Gạch Chia",
+            "minimum": "Tối Thiểu",
+            "maximum": "Tối Đa",
+            "containerRowLayout": "Hộp Chứa (bố cục hàng)",
+            "containerColumnLayout": "Hộp Chứa (bố cục cột)"
         },
         "yourWorkflows": "Workflow Của Bạn",
         "browseWorkflows": "Khám Phá Workflow",
@@ -2300,7 +2312,11 @@
         "filterByTags": "Lọc Theo Nhãn",
         "recentlyOpened": "Mở Gần Đây",
         "private": "Cá Nhân",
-        "loadMore": "Tải Thêm"
+        "loadMore": "Tải Thêm",
+        "view": "Xem",
+        "deselectAll": "Huỷ Chọn Tất Cả",
+        "noRecentWorkflows": "Không Có Workflows Gần Đây",
+        "recommended": "Có Thể Bạn Sẽ Cần"
     },
     "upscaling": {
         "missingUpscaleInitialImage": "Thiếu ảnh dùng để upscale",
@@ -2327,7 +2343,8 @@
         "gettingStartedSeries": "Cần thêm hướng dẫn? Xem thử <LinkComponent>Bắt Đầu Làm Quen</LinkComponent> để biết thêm mẹo khai thác toàn bộ tiềm năng của Invoke Studio.",
         "toGetStarted": "Để bắt đầu, hãy nhập lệnh vào hộp và nhấp chuột vào <StrongComponent>Kích Hoạt</StrongComponent> để tạo ra bức ảnh đầu tiên. Chọn một mẫu trình bày cho lệnh để cải thiện kết quả. Bạn có thể chọn để lưu ảnh trực tiếp vào <StrongComponent>Thư Viện Ảnh</StrongComponent> hoặc chỉnh sửa chúng ở <StrongComponent>Canvas</StrongComponent>.",
         "noModelsInstalled": "Dường như bạn chưa tải model nào cả! Bạn có thể <DownloadStarterModelsButton>tải xuống các model khởi đầu</DownloadStarterModelsButton> hoặc <ImportModelsButton>nhập vào thêm model</ImportModelsButton>.",
-        "lowVRAMMode": "Cho hiệu suất tốt nhất, hãy làm theo <LinkComponent>hướng dẫn VRAM Thấp</LinkComponent> của chúng tôi."
+        "lowVRAMMode": "Cho hiệu suất tốt nhất, hãy làm theo <LinkComponent>hướng dẫn VRAM Thấp</LinkComponent> của chúng tôi.",
+        "toGetStartedWorkflow": "Để bắt đầu, hãy điền vào khu vực bên trái và bấm <StrongComponent>Kích Hoạt</StrongComponent> nhằm tạo sinh ảnh. Muốn khám phá thêm workflow? Nhấp vào <StrongComponent>icon thư mục</StrongComponent> nằm cạnh tiêu đề workflow để xem một dãy các mẫu trình bày khác."
     },
     "whatsNew": {
         "whatsNewInInvoke": "Có Gì Mới Ở Invoke",
@@ -2335,8 +2352,8 @@
         "watchRecentReleaseVideos": "Xem Video Phát Hành Mới Nhất",
         "watchUiUpdatesOverview": "Xem Tổng Quan Về Những Cập Nhật Cho Giao Diện Người Dùng",
         "items": [
-            "Trình Quản Lý Bộ Nhớ: Thiết lập mới cho người dùng với GPU Nvidia để giảm lượng VRAM sử dụng.",
-            "Hiệu suất: Các cải thiện tiếp theo nhằm gói gọn hiệu suất và khả năng phản hồi của ứng dụng."
+            "Workflow: Thư Viện Workflow mới và đã được cải tiến.",
+            "FLUX: Hỗ trợ FLUX Redux & FLUX Fill trong Workflow và Canvas."
         ]
     },
     "upsell": {

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

@@ -908,7 +908,6 @@
         "unableToGetWorkflowVersion": "无法获取工作流架构版本",
         "nodePack": "节点包",
         "unableToExtractSchemaNameFromRef": "无法从参考中提取架构名",
-        "unknownOutput": "未知输出：{{name}}",
         "unknownErrorValidatingWorkflow": "验证工作流时出现未知错误",
         "collectionFieldType": "{{name}}(合集)",
         "unknownNodeType": "未知节点类型",

invokeai/frontend/web/src/app/hooks/useStudioInitAction.ts

@@ -12,6 +12,12 @@ import { sentImageToCanvas } from 'features/gallery/store/actions';
 import { parseAndRecallAllMetadata } from 'features/metadata/util/handlers';
 import { $hasTemplates } from 'features/nodes/store/nodesSlice';
 import { $isWorkflowLibraryModalOpen } from 'features/nodes/store/workflowLibraryModal';
+import {
+  $workflowLibraryTagOptions,
+  workflowLibraryTagsReset,
+  workflowLibraryTagToggled,
+  workflowLibraryViewChanged,
+} from 'features/nodes/store/workflowLibrarySlice';
 import { $isStylePresetsMenuOpen, activeStylePresetIdChanged } from 'features/stylePresets/store/stylePresetSlice';
 import { toast } from 'features/toast/toast';
 import { activeTabCanvasRightPanelChanged, setActiveTab } from 'features/ui/store/uiSlice';
@@ -30,9 +36,17 @@ type SendToCanvasAction = _StudioInitAction<'sendToCanvas', { imageName: string
 type UseAllParametersAction = _StudioInitAction<'useAllParameters', { imageName: string }>;
 type StudioDestinationAction = _StudioInitAction<
   'goToDestination',
-  { destination: 'generation' | 'canvas' | 'workflows' | 'upscaling' | 'viewAllWorkflows' | 'viewAllStylePresets' }
+  {
+    destination:
+      | 'generation'
+      | 'canvas'
+      | 'workflows'
+      | 'upscaling'
+      | 'viewAllWorkflows'
+      | 'viewAllWorkflowsRecommended'
+      | 'viewAllStylePresets';
+  }
 >;
-
 // Use global state to show loader until we are ready to render the studio.
 export const $didStudioInit = atom(false);
 
@@ -58,6 +72,7 @@ export const useStudioInitAction = (action?: StudioInitAction) => {
   const didParseOpenAPISchema = useStore($hasTemplates);
   const store = useAppStore();
   const loadWorkflowWithDialog = useLoadWorkflowWithDialog();
+  const workflowLibraryTagOptions = useStore($workflowLibraryTagOptions);
 
   const handleSendToCanvas = useCallback(
     async (imageName: string) => {
@@ -173,6 +188,18 @@ export const useStudioInitAction = (action?: StudioInitAction) => {
           store.dispatch(setActiveTab('workflows'));
           $isWorkflowLibraryModalOpen.set(true);
           break;
+        case 'viewAllWorkflowsRecommended':
+          // Go to the workflows tab and open the workflow library modal with the recommended workflows view
+          store.dispatch(setActiveTab('workflows'));
+          $isWorkflowLibraryModalOpen.set(true);
+          store.dispatch(workflowLibraryViewChanged('defaults'));
+          store.dispatch(workflowLibraryTagsReset());
+          for (const tag of workflowLibraryTagOptions) {
+            if (tag.recommended) {
+              store.dispatch(workflowLibraryTagToggled(tag.label));
+            }
+          }
+          break;
         case 'viewAllStylePresets':
           // Go to the canvas tab and open the style presets menu
           store.dispatch(setActiveTab('canvas'));
@@ -180,7 +207,7 @@ export const useStudioInitAction = (action?: StudioInitAction) => {
           break;
       }
     },
-    [store]
+    [store, workflowLibraryTagOptions]
   );
 
   const handleStudioInitAction = useCallback(

invokeai/frontend/web/src/app/store/middleware/listenerMiddleware/listeners/imageUploaded.ts

@@ -1,6 +1,8 @@
+import { isAnyOf } from '@reduxjs/toolkit';
 import { logger } from 'app/logging/logger';
 import type { AppStartListening } from 'app/store/middleware/listenerMiddleware';
 import type { RootState } from 'app/store/store';
+import { imageUploadedClientSide } from 'features/gallery/store/actions';
 import { selectListBoardsQueryArgs } from 'features/gallery/store/gallerySelectors';
 import { boardIdSelected, galleryViewChanged } from 'features/gallery/store/gallerySlice';
 import { toast } from 'features/toast/toast';
@@ -8,7 +10,8 @@ import { t } from 'i18next';
 import { omit } from 'lodash-es';
 import { boardsApi } from 'services/api/endpoints/boards';
 import { imagesApi } from 'services/api/endpoints/images';
-
+import type { ImageDTO } from 'services/api/types';
+import { getCategories, getListImagesUrl } from 'services/api/util';
 const log = logger('gallery');
 
 /**
@@ -34,19 +37,56 @@ let lastUploadedToastTimeout: number | null = null;
 
 export const addImageUploadedFulfilledListener = (startAppListening: AppStartListening) => {
   startAppListening({
-    matcher: imagesApi.endpoints.uploadImage.matchFulfilled,
+    matcher: isAnyOf(imagesApi.endpoints.uploadImage.matchFulfilled, imageUploadedClientSide),
     effect: (action, { dispatch, getState }) => {
-      const imageDTO = action.payload;
+      let imageDTO: ImageDTO;
+      let silent;
+      let isFirstUploadOfBatch = true;
+
+      if (imageUploadedClientSide.match(action)) {
+        imageDTO = action.payload.imageDTO;
+        silent = action.payload.silent;
+        isFirstUploadOfBatch = action.payload.isFirstUploadOfBatch;
+      } else if (imagesApi.endpoints.uploadImage.matchFulfilled(action)) {
+        imageDTO = action.payload;
+        silent = action.meta.arg.originalArgs.silent;
+        isFirstUploadOfBatch = action.meta.arg.originalArgs.isFirstUploadOfBatch ?? true;
+      } else {
+        return;
+      }
+
+      if (silent || imageDTO.is_intermediate) {
+        // If the image is silent or intermediate, we don't want to show a toast
+        return;
+      }
+
+      if (imageUploadedClientSide.match(action)) {
+        const categories = getCategories(imageDTO);
+        const boardId = imageDTO.board_id ?? 'none';
+        dispatch(
+          imagesApi.util.invalidateTags([
+            {
+              type: 'ImageList',
+              id: getListImagesUrl({
+                board_id: boardId,
+                categories,
+              }),
+            },
+            {
+              type: 'Board',
+              id: boardId,
+            },
+            {
+              type: 'BoardImagesTotal',
+              id: boardId,
+            },
+          ])
+        );
+      }
       const state = getState();
 
       log.debug({ imageDTO }, 'Image uploaded');
 
-      if (action.meta.arg.originalArgs.silent || imageDTO.is_intermediate) {
-        // When a "silent" upload is requested, or the image is intermediate, we can skip all post-upload actions,
-        // like toasts and switching the gallery view
-        return;
-      }
-
       const boardId = imageDTO.board_id ?? 'none';
 
       const DEFAULT_UPLOADED_TOAST = {
@@ -80,7 +120,7 @@ export const addImageUploadedFulfilledListener = (startAppListening: AppStartLis
        *
        * Default to true to not require _all_ image upload handlers to set this value
        */
-      const isFirstUploadOfBatch = action.meta.arg.originalArgs.isFirstUploadOfBatch ?? true;
+
       if (isFirstUploadOfBatch) {
         dispatch(boardIdSelected({ boardId }));
         dispatch(galleryViewChanged('assets'));

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

@@ -73,6 +73,7 @@ export type AppConfig = {
   maxUpscaleDimension?: number;
   allowPrivateBoards: boolean;
   allowPrivateStylePresets: boolean;
+  allowClientSideUpload: boolean;
   disabledTabs: TabName[];
   disabledFeatures: AppFeature[];
   disabledSDFeatures: SDFeature[];
@@ -81,7 +82,6 @@ export type AppConfig = {
   metadataFetchDebounce?: number;
   workflowFetchDebounce?: number;
   isLocal?: boolean;
-  maxImageUploadCount?: number;
   sd: {
     defaultModel?: string;
     disabledControlNetModels: string[];

invokeai/frontend/web/src/common/components/OverlayScrollbars/ScrollableContent.tsx

@@ -19,7 +19,7 @@ const styles: CSSProperties = { position: 'absolute', top: 0, left: 0, right: 0,
 
 const ScrollableContent = ({ children, maxHeight, overflowX = 'hidden', overflowY = 'scroll' }: Props) => {
   const overlayscrollbarsOptions = useMemo(
-    () => getOverlayScrollbarsParams(overflowX, overflowY).options,
+    () => getOverlayScrollbarsParams({ overflowX, overflowY }).options,
     [overflowX, overflowY]
   );
   const [os, osRef] = useState<OverlayScrollbarsComponentRef | null>(null);