InvokeAI 6.9.0rc3__py3-none-any.whl → 6.10.0rc1__py3-none-any.whl

invokeai/app/api/dependencies.py

@@ -49,6 +49,7 @@ from invokeai.backend.stable_diffusion.diffusion.conditioning_data import (
     FLUXConditioningInfo,
     SD3ConditioningInfo,
     SDXLConditioningInfo,
+    ZImageConditioningInfo,
 )
 from invokeai.backend.util.logging import InvokeAILogger
 from invokeai.version.invokeai_version import __version__
@@ -129,6 +130,7 @@ class ApiDependencies:
                     FLUXConditioningInfo,
                     SD3ConditioningInfo,
                     CogView4ConditioningInfo,
+                    ZImageConditioningInfo,
                 ],
                 ephemeral=True,
             ),

invokeai/app/api/routers/model_manager.py

@@ -28,7 +28,7 @@ from invokeai.app.services.model_records import (
     UnknownModelException,
 )
 from invokeai.app.util.suppress_output import SuppressOutput
-from invokeai.backend.model_manager.configs.factory import AnyModelConfig
+from invokeai.backend.model_manager.configs.factory import AnyModelConfig, ModelConfigFactory
 from invokeai.backend.model_manager.configs.main import (
     Main_Checkpoint_SD1_Config,
     Main_Checkpoint_SD2_Config,
@@ -38,6 +38,7 @@ from invokeai.backend.model_manager.configs.main import (
 from invokeai.backend.model_manager.load.model_cache.cache_stats import CacheStats
 from invokeai.backend.model_manager.metadata.fetch.huggingface import HuggingFaceMetadataFetch
 from invokeai.backend.model_manager.metadata.metadata_base import ModelMetadataWithFiles, UnknownMetadataException
+from invokeai.backend.model_manager.model_on_disk import ModelOnDisk
 from invokeai.backend.model_manager.search import ModelSearch
 from invokeai.backend.model_manager.starter_models import (
     STARTER_BUNDLES,
@@ -191,6 +192,40 @@ async def get_model_record(
         raise HTTPException(status_code=404, detail=str(e))
 
 
+@model_manager_router.post(
+    "/i/{key}/reidentify",
+    operation_id="reidentify_model",
+    responses={
+        200: {
+            "description": "The model configuration was retrieved successfully",
+            "content": {"application/json": {"example": example_model_config}},
+        },
+        400: {"description": "Bad request"},
+        404: {"description": "The model could not be found"},
+    },
+)
+async def reidentify_model(
+    key: Annotated[str, Path(description="Key of the model to reidentify.")],
+) -> AnyModelConfig:
+    """Attempt to reidentify a model by re-probing its weights file."""
+    try:
+        config = ApiDependencies.invoker.services.model_manager.store.get_model(key)
+        models_path = ApiDependencies.invoker.services.configuration.models_path
+        if pathlib.Path(config.path).is_relative_to(models_path):
+            model_path = pathlib.Path(config.path)
+        else:
+            model_path = models_path / config.path
+        mod = ModelOnDisk(model_path)
+        result = ModelConfigFactory.from_model_on_disk(mod)
+        if result.config is None:
+            raise InvalidModelException("Unable to identify model format")
+        result.config.key = config.key  # retain the same key
+        new_config = ApiDependencies.invoker.services.model_manager.store.replace_model(config.key, result.config)
+        return new_config
+    except UnknownModelException as e:
+        raise HTTPException(status_code=404, detail=str(e))
+
+
 class FoundModel(BaseModel):
     path: str = Field(description="Path to the model")
     is_installed: bool = Field(description="Whether or not the model is already installed")
@@ -238,9 +273,10 @@ async def scan_for_models(
             found_model = FoundModel(path=path, is_installed=is_installed)
             scan_results.append(found_model)
     except Exception as e:
+        error_type = type(e).__name__
         raise HTTPException(
             status_code=500,
-            detail=f"An error occurred while searching the directory: {e}",
+            detail=f"An error occurred while searching the directory: {error_type}",
         )
     return scan_results
 
@@ -411,6 +447,59 @@ async def delete_model(
         raise HTTPException(status_code=404, detail=str(e))
 
 
+class BulkDeleteModelsRequest(BaseModel):
+    """Request body for bulk model deletion."""
+
+    keys: List[str] = Field(description="List of model keys to delete")
+
+
+class BulkDeleteModelsResponse(BaseModel):
+    """Response body for bulk model deletion."""
+
+    deleted: List[str] = Field(description="List of successfully deleted model keys")
+    failed: List[dict] = Field(description="List of failed deletions with error messages")
+
+
+@model_manager_router.post(
+    "/i/bulk_delete",
+    operation_id="bulk_delete_models",
+    responses={
+        200: {"description": "Models deleted (possibly with some failures)"},
+    },
+    status_code=200,
+)
+async def bulk_delete_models(
+    request: BulkDeleteModelsRequest = Body(description="List of model keys to delete"),
+) -> BulkDeleteModelsResponse:
+    """
+    Delete multiple model records from database.
+
+    The configuration records will be removed. The corresponding weights files will be
+    deleted as well if they reside within the InvokeAI "models" directory.
+    Returns a list of successfully deleted keys and failed deletions with error messages.
+    """
+    logger = ApiDependencies.invoker.services.logger
+    installer = ApiDependencies.invoker.services.model_manager.install
+
+    deleted = []
+    failed = []
+
+    for key in request.keys:
+        try:
+            installer.delete(key)
+            deleted.append(key)
+            logger.info(f"Deleted model: {key}")
+        except UnknownModelException as e:
+            logger.error(f"Failed to delete model {key}: {str(e)}")
+            failed.append({"key": key, "error": str(e)})
+        except Exception as e:
+            logger.error(f"Failed to delete model {key}: {str(e)}")
+            failed.append({"key": key, "error": str(e)})
+
+    logger.info(f"Bulk delete completed: {len(deleted)} deleted, {len(failed)} failed")
+    return BulkDeleteModelsResponse(deleted=deleted, failed=failed)
+
+
 @model_manager_router.delete(
     "/i/{key}/image",
     operation_id="delete_model_image",

invokeai/app/api/routers/workflows.py

@@ -223,6 +223,15 @@ async def get_workflow_thumbnail(
         raise HTTPException(status_code=404)
 
 
+@workflows_router.get("/tags", operation_id="get_all_tags")
+async def get_all_tags(
+    categories: Optional[list[WorkflowCategory]] = Query(default=None, description="The categories to include"),
+) -> list[str]:
+    """Gets all unique tags from workflows"""
+
+    return ApiDependencies.invoker.services.workflow_records.get_all_tags(categories=categories)
+
+
 @workflows_router.get("/counts_by_tag", operation_id="get_counts_by_tag")
 async def get_counts_by_tag(
     tags: list[str] = Query(description="The tags to get counts for"),

invokeai/app/invocations/fields.py

@@ -154,6 +154,7 @@ class FieldDescriptions:
     clip = "CLIP (tokenizer, text encoder, LoRAs) and skipped layer count"
     t5_encoder = "T5 tokenizer and text encoder"
     glm_encoder = "GLM (THUDM) tokenizer and text encoder"
+    qwen3_encoder = "Qwen3 tokenizer and text encoder"
     clip_embed_model = "CLIP Embed loader"
     clip_g_model = "CLIP-G Embed loader"
     unet = "UNet (scheduler, LoRAs)"
@@ -169,6 +170,7 @@ class FieldDescriptions:
     flux_model = "Flux model (Transformer) to load"
     sd3_model = "SD3 model (MMDiTX) to load"
     cogview4_model = "CogView4 model (Transformer) to load"
+    z_image_model = "Z-Image model (Transformer) to load"
     sdxl_main_model = "SDXL Main model (UNet, VAE, CLIP1, CLIP2) to load"
     sdxl_refiner_model = "SDXL Refiner Main Modde (UNet, VAE, CLIP2) to load"
     onnx_main_model = "ONNX Main model (UNet, VAE, CLIP) to load"
@@ -241,6 +243,12 @@ class BoardField(BaseModel):
     board_id: str = Field(description="The id of the board")
 
 
+class StylePresetField(BaseModel):
+    """A style preset primitive field"""
+
+    style_preset_id: str = Field(description="The id of the style preset")
+
+
 class DenoiseMaskField(BaseModel):
     """An inpaint mask field"""
 
@@ -321,6 +329,17 @@ class CogView4ConditioningField(BaseModel):
     conditioning_name: str = Field(description="The name of conditioning tensor")
 
 
+class ZImageConditioningField(BaseModel):
+    """A Z-Image conditioning tensor primitive value"""
+
+    conditioning_name: str = Field(description="The name of conditioning tensor")
+    mask: Optional[TensorField] = Field(
+        default=None,
+        description="The mask associated with this conditioning tensor for regional prompting. "
+        "Excluded regions should be set to False, included regions should be set to True.",
+    )
+
+
 class ConditioningField(BaseModel):
     """A conditioning tensor primitive value"""
 

invokeai/app/invocations/image_to_latents.py

@@ -1,5 +1,6 @@
 from contextlib import nullcontext
 from functools import singledispatchmethod
+from typing import Literal
 
 import einops
 import torch
@@ -20,7 +21,7 @@ from invokeai.app.invocations.fields import (
     Input,
     InputField,
 )
-from invokeai.app.invocations.model import VAEField
+from invokeai.app.invocations.model import BaseModelType, VAEField
 from invokeai.app.invocations.primitives import LatentsOutput
 from invokeai.app.services.shared.invocation_context import InvocationContext
 from invokeai.backend.model_manager.load.load_base import LoadedModel
@@ -29,13 +30,21 @@ from invokeai.backend.stable_diffusion.vae_tiling import patch_vae_tiling_params
 from invokeai.backend.util.devices import TorchDevice
 from invokeai.backend.util.vae_working_memory import estimate_vae_working_memory_sd15_sdxl
 
+"""
+SDXL VAE color compensation values determined experimentally to reduce color drift.
+If more reliable values are found in the future (e.g. individual color channels), they can be updated.
+SD1.5, TAESD, TAESDXL VAEs distort in less predictable ways, so no compensation is offered at this time.
+"""
+COMPENSATION_OPTIONS = Literal["None", "SDXL"]
+COLOR_COMPENSATION_MAP = {"None": [1, 0], "SDXL": [1.015, -0.002]}
+
 
 @invocation(
     "i2l",
     title="Image to Latents - SD1.5, SDXL",
     tags=["latents", "image", "vae", "i2l"],
     category="latents",
-    version="1.1.1",
+    version="1.2.0",
 )
 class ImageToLatentsInvocation(BaseInvocation):
     """Encodes an image into latents."""
@@ -52,6 +61,10 @@ class ImageToLatentsInvocation(BaseInvocation):
     # offer a way to directly set None values.
     tile_size: int = InputField(default=0, multiple_of=8, description=FieldDescriptions.vae_tile_size)
     fp32: bool = InputField(default=False, description=FieldDescriptions.fp32)
+    color_compensation: COMPENSATION_OPTIONS = InputField(
+        default="None",
+        description="Apply VAE scaling compensation when encoding images (reduces color drift).",
+    )
 
     @classmethod
     def vae_encode(
@@ -62,7 +75,7 @@ class ImageToLatentsInvocation(BaseInvocation):
         image_tensor: torch.Tensor,
         tile_size: int = 0,
     ) -> torch.Tensor:
-        assert isinstance(vae_info.model, (AutoencoderKL, AutoencoderTiny))
+        assert isinstance(vae_info.model, (AutoencoderKL, AutoencoderTiny)), "VAE must be of type SD-1.5 or SDXL"
         estimated_working_memory = estimate_vae_working_memory_sd15_sdxl(
             operation="encode",
             image_tensor=image_tensor,
@@ -71,7 +84,7 @@ class ImageToLatentsInvocation(BaseInvocation):
             fp32=upcast,
         )
         with vae_info.model_on_device(working_mem_bytes=estimated_working_memory) as (_, vae):
-            assert isinstance(vae, (AutoencoderKL, AutoencoderTiny))
+            assert isinstance(vae, (AutoencoderKL, AutoencoderTiny)), "VAE must be of type SD-1.5 or SDXL"
             orig_dtype = vae.dtype
             if upcast:
                 vae.to(dtype=torch.float32)
@@ -127,9 +140,14 @@ class ImageToLatentsInvocation(BaseInvocation):
         image = context.images.get_pil(self.image.image_name)
 
         vae_info = context.models.load(self.vae.vae)
-        assert isinstance(vae_info.model, (AutoencoderKL, AutoencoderTiny))
+        assert isinstance(vae_info.model, (AutoencoderKL, AutoencoderTiny)), "VAE must be of type SD-1.5 or SDXL"
 
         image_tensor = image_resized_to_grid_as_tensor(image.convert("RGB"))
+
+        if self.color_compensation != "None" and vae_info.config.base == BaseModelType.StableDiffusionXL:
+            scale, bias = COLOR_COMPENSATION_MAP[self.color_compensation]
+            image_tensor = image_tensor * scale + bias
+
         if image_tensor.dim() == 3:
             image_tensor = einops.rearrange(image_tensor, "c h w -> 1 c h w")
 

invokeai/app/invocations/latents_to_image.py

@@ -2,12 +2,6 @@ from contextlib import nullcontext
 
 import torch
 from diffusers.image_processor import VaeImageProcessor
-from diffusers.models.attention_processor import (
-    AttnProcessor2_0,
-    LoRAAttnProcessor2_0,
-    LoRAXFormersAttnProcessor,
-    XFormersAttnProcessor,
-)
 from diffusers.models.autoencoders.autoencoder_kl import AutoencoderKL
 from diffusers.models.autoencoders.autoencoder_tiny import AutoencoderTiny
 
@@ -77,26 +71,9 @@ class LatentsToImageInvocation(BaseInvocation, WithMetadata, WithBoard):
             assert isinstance(vae, (AutoencoderKL, AutoencoderTiny))
             latents = latents.to(TorchDevice.choose_torch_device())
             if self.fp32:
+                # FP32 mode: convert everything to float32 for maximum precision
                 vae.to(dtype=torch.float32)
-
-                use_torch_2_0_or_xformers = hasattr(vae.decoder, "mid_block") and isinstance(
-                    vae.decoder.mid_block.attentions[0].processor,
-                    (
-                        AttnProcessor2_0,
-                        XFormersAttnProcessor,
-                        LoRAXFormersAttnProcessor,
-                        LoRAAttnProcessor2_0,
-                    ),
-                )
-                # if xformers or torch_2_0 is used attention block does not need
-                # to be in float32 which can save lots of memory
-                if use_torch_2_0_or_xformers:
-                    vae.post_quant_conv.to(latents.dtype)
-                    vae.decoder.conv_in.to(latents.dtype)
-                    vae.decoder.mid_block.to(latents.dtype)
-                else:
-                    latents = latents.float()
-
+                latents = latents.float()
             else:
                 vae.to(dtype=torch.float16)
                 latents = latents.half()

invokeai/app/invocations/metadata.py

@@ -158,6 +158,10 @@ GENERATION_MODES = Literal[
     "cogview4_img2img",
     "cogview4_inpaint",
     "cogview4_outpaint",
+    "z_image_txt2img",
+    "z_image_img2img",
+    "z_image_inpaint",
+    "z_image_outpaint",
 ]
 
 
@@ -166,7 +170,7 @@ GENERATION_MODES = Literal[
     title="Core Metadata",
     tags=["metadata"],
     category="metadata",
-    version="2.0.0",
+    version="2.1.0",
     classification=Classification.Internal,
 )
 class CoreMetadataInvocation(BaseInvocation):
@@ -217,6 +221,10 @@ class CoreMetadataInvocation(BaseInvocation):
         default=None,
         description="The VAE used for decoding, if the main model's default was not used",
     )
+    qwen3_encoder: Optional[ModelIdentifierField] = InputField(
+        default=None,
+        description="The Qwen3 text encoder model used for Z-Image inference",
+    )
 
     # High resolution fix metadata.
     hrf_enabled: Optional[bool] = InputField(

invokeai/app/invocations/model.py

@@ -72,6 +72,14 @@ class GlmEncoderField(BaseModel):
     text_encoder: ModelIdentifierField = Field(description="Info to load text_encoder submodel")
 
 
+class Qwen3EncoderField(BaseModel):
+    """Field for Qwen3 text encoder used by Z-Image models."""
+
+    tokenizer: ModelIdentifierField = Field(description="Info to load tokenizer submodel")
+    text_encoder: ModelIdentifierField = Field(description="Info to load text_encoder submodel")
+    loras: List[LoRAField] = Field(default_factory=list, description="LoRAs to apply on model loading")
+
+
 class VAEField(BaseModel):
     vae: ModelIdentifierField = Field(description="Info to load vae submodel")
     seamless_axes: List[str] = Field(default_factory=list, description='Axes("x" and "y") to which apply seamless')

invokeai/app/invocations/primitives.py

@@ -27,6 +27,7 @@ from invokeai.app.invocations.fields import (
     SD3ConditioningField,
     TensorField,
     UIComponent,
+    ZImageConditioningField,
 )
 from invokeai.app.services.images.images_common import ImageDTO
 from invokeai.app.services.shared.invocation_context import InvocationContext
@@ -461,6 +462,17 @@ class CogView4ConditioningOutput(BaseInvocationOutput):
         return cls(conditioning=CogView4ConditioningField(conditioning_name=conditioning_name))
 
 
+@invocation_output("z_image_conditioning_output")
+class ZImageConditioningOutput(BaseInvocationOutput):
+    """Base class for nodes that output a Z-Image text conditioning tensor."""
+
+    conditioning: ZImageConditioningField = OutputField(description=FieldDescriptions.cond)
+
+    @classmethod
+    def build(cls, conditioning_name: str) -> "ZImageConditioningOutput":
+        return cls(conditioning=ZImageConditioningField(conditioning_name=conditioning_name))
+
+
 @invocation_output("conditioning_output")
 class ConditioningOutput(BaseInvocationOutput):
     """Base class for nodes that output a single conditioning tensor"""

invokeai/app/invocations/prompt_template.py

@@ -0,0 +1,57 @@
+from invokeai.app.invocations.baseinvocation import BaseInvocation, BaseInvocationOutput, invocation, invocation_output
+from invokeai.app.invocations.fields import InputField, OutputField, StylePresetField, UIComponent
+from invokeai.app.services.shared.invocation_context import InvocationContext
+
+
+@invocation_output("prompt_template_output")
+class PromptTemplateOutput(BaseInvocationOutput):
+    """Output for the Prompt Template node"""
+
+    positive_prompt: str = OutputField(description="The positive prompt with the template applied")
+    negative_prompt: str = OutputField(description="The negative prompt with the template applied")
+
+
+@invocation(
+    "prompt_template",
+    title="Prompt Template",
+    tags=["prompt", "template", "style", "preset"],
+    category="prompt",
+    version="1.0.0",
+)
+class PromptTemplateInvocation(BaseInvocation):
+    """Applies a Style Preset template to positive and negative prompts.
+
+    Select a Style Preset and provide positive/negative prompts. The node replaces
+    {prompt} placeholders in the template with your input prompts.
+    """
+
+    style_preset: StylePresetField = InputField(
+        description="The Style Preset to use as a template",
+    )
+    positive_prompt: str = InputField(
+        default="",
+        description="The positive prompt to insert into the template's {prompt} placeholder",
+        ui_component=UIComponent.Textarea,
+    )
+    negative_prompt: str = InputField(
+        default="",
+        description="The negative prompt to insert into the template's {prompt} placeholder",
+        ui_component=UIComponent.Textarea,
+    )
+
+    def invoke(self, context: InvocationContext) -> PromptTemplateOutput:
+        # Fetch the style preset from the database
+        style_preset = context._services.style_preset_records.get(self.style_preset.style_preset_id)
+
+        # Get the template prompts
+        positive_template = style_preset.preset_data.positive_prompt
+        negative_template = style_preset.preset_data.negative_prompt
+
+        # Replace {prompt} placeholder with the input prompts
+        rendered_positive = positive_template.replace("{prompt}", self.positive_prompt)
+        rendered_negative = negative_template.replace("{prompt}", self.negative_prompt)
+
+        return PromptTemplateOutput(
+            positive_prompt=rendered_positive,
+            negative_prompt=rendered_negative,
+        )

invokeai/app/invocations/z_image_control.py

@@ -0,0 +1,112 @@
+# Copyright (c) 2024, Lincoln D. Stein and the InvokeAI Development Team
+"""Z-Image Control invocation for spatial conditioning."""
+
+from pydantic import BaseModel, Field
+
+from invokeai.app.invocations.baseinvocation import (
+    BaseInvocation,
+    BaseInvocationOutput,
+    Classification,
+    invocation,
+    invocation_output,
+)
+from invokeai.app.invocations.fields import (
+    FieldDescriptions,
+    ImageField,
+    InputField,
+    OutputField,
+)
+from invokeai.app.invocations.model import ModelIdentifierField
+from invokeai.app.services.shared.invocation_context import InvocationContext
+from invokeai.backend.model_manager.taxonomy import BaseModelType, ModelType
+
+
+class ZImageControlField(BaseModel):
+    """A Z-Image control conditioning field for spatial control (Canny, HED, Depth, Pose, MLSD)."""
+
+    image_name: str = Field(description="The name of the preprocessed control image")
+    control_model: ModelIdentifierField = Field(description="The Z-Image ControlNet adapter model")
+    control_context_scale: float = Field(
+        default=0.75,
+        ge=0.0,
+        le=2.0,
+        description="The strength of the control signal. Recommended range: 0.65-0.80.",
+    )
+    begin_step_percent: float = Field(
+        default=0.0,
+        ge=0.0,
+        le=1.0,
+        description="When the control is first applied (% of total steps)",
+    )
+    end_step_percent: float = Field(
+        default=1.0,
+        ge=0.0,
+        le=1.0,
+        description="When the control is last applied (% of total steps)",
+    )
+
+
+@invocation_output("z_image_control_output")
+class ZImageControlOutput(BaseInvocationOutput):
+    """Z-Image Control output containing control configuration."""
+
+    control: ZImageControlField = OutputField(description="Z-Image control conditioning")
+
+
+@invocation(
+    "z_image_control",
+    title="Z-Image ControlNet",
+    tags=["image", "z-image", "control", "controlnet"],
+    category="control",
+    version="1.1.0",
+    classification=Classification.Prototype,
+)
+class ZImageControlInvocation(BaseInvocation):
+    """Configure Z-Image ControlNet for spatial conditioning.
+
+    Takes a preprocessed control image (e.g., Canny edges, depth map, pose)
+    and a Z-Image ControlNet adapter model to enable spatial control.
+
+    Supports 5 control modes: Canny, HED, Depth, Pose, MLSD.
+    Recommended control_context_scale: 0.65-0.80.
+    """
+
+    image: ImageField = InputField(
+        description="The preprocessed control image (Canny, HED, Depth, Pose, or MLSD)",
+    )
+    control_model: ModelIdentifierField = InputField(
+        description=FieldDescriptions.controlnet_model,
+        title="Control Model",
+        ui_model_base=BaseModelType.ZImage,
+        ui_model_type=ModelType.ControlNet,
+    )
+    control_context_scale: float = InputField(
+        default=0.75,
+        ge=0.0,
+        le=2.0,
+        description="Strength of the control signal. Recommended range: 0.65-0.80.",
+        title="Control Scale",
+    )
+    begin_step_percent: float = InputField(
+        default=0.0,
+        ge=0.0,
+        le=1.0,
+        description="When the control is first applied (% of total steps)",
+    )
+    end_step_percent: float = InputField(
+        default=1.0,
+        ge=0.0,
+        le=1.0,
+        description="When the control is last applied (% of total steps)",
+    )
+
+    def invoke(self, context: InvocationContext) -> ZImageControlOutput:
+        return ZImageControlOutput(
+            control=ZImageControlField(
+                image_name=self.image.image_name,
+                control_model=self.control_model,
+                control_context_scale=self.control_context_scale,
+                begin_step_percent=self.begin_step_percent,
+                end_step_percent=self.end_step_percent,
+            )
+        )