Merge branch 'main' into main

Author: simpletrontdip

.github/CODEOWNERS

@@ -1,12 +1,12 @@
 # continuous integration
-/.github/workflows/  @lstein @blessedcoolant @hipsterusername @ebr
+/.github/workflows/  @lstein @blessedcoolant @hipsterusername @ebr @jazzhaiku
 
 # documentation
 /docs/ @lstein @blessedcoolant @hipsterusername @Millu
 /mkdocs.yml @lstein  @blessedcoolant @hipsterusername @Millu
 
 # nodes
-/invokeai/app/ @Kyle0654 @blessedcoolant @psychedelicious @brandonrising @hipsterusername
+/invokeai/app/ @Kyle0654 @blessedcoolant @psychedelicious @brandonrising @hipsterusername @jazzhaiku
 
 # installation and configuration
 /pyproject.toml  @lstein @blessedcoolant @hipsterusername
@@ -22,7 +22,7 @@
 /invokeai/backend @blessedcoolant @psychedelicious @lstein @maryhipp @hipsterusername
 
 # generation, model management, postprocessing
-/invokeai/backend  @damian0815 @lstein @blessedcoolant @gregghelt2 @StAlKeR7779 @brandonrising @ryanjdick @hipsterusername
+/invokeai/backend  @damian0815 @lstein @blessedcoolant @gregghelt2 @StAlKeR7779 @brandonrising @ryanjdick @hipsterusername @jazzhaiku
 
 # front ends
 /invokeai/frontend/CLI @lstein @hipsterusername

.github/workflows/python-checks.yml

@@ -62,7 +62,7 @@ jobs:
 
       - name: install ruff
         if: ${{ steps.changed-files.outputs.python_any_changed == 'true' || inputs.always_run == true }}
-        run: pip install ruff==0.6.0
+        run: pip install ruff==0.9.9
         shell: bash
 
       - name: ruff check

docker/Dockerfile

@@ -13,48 +13,63 @@ RUN --mount=type=cache,target=/var/cache/apt,sharing=locked \
         git
 
 # Install `uv` for package management
-COPY --from=ghcr.io/astral-sh/uv:0.5.5 /uv /uvx /bin/
+COPY --from=ghcr.io/astral-sh/uv:0.6.0 /uv /uvx /bin/
 
 ENV VIRTUAL_ENV=/opt/venv
 ENV PATH="$VIRTUAL_ENV/bin:$PATH"
 ENV INVOKEAI_SRC=/opt/invokeai
 ENV PYTHON_VERSION=3.11
+ENV UV_PYTHON=3.11
 ENV UV_COMPILE_BYTECODE=1
 ENV UV_LINK_MODE=copy
+ENV UV_PROJECT_ENVIRONMENT="$VIRTUAL_ENV"
+ENV UV_INDEX="https://download.pytorch.org/whl/cu124"
 
 ARG GPU_DRIVER=cuda
-ARG TARGETPLATFORM="linux/amd64"
 # unused but available
 ARG BUILDPLATFORM
 
 # Switch to the `ubuntu` user to work around dependency issues with uv-installed python
 RUN mkdir -p ${VIRTUAL_ENV} && \
     mkdir -p ${INVOKEAI_SRC} && \
-    chmod -R a+w /opt
+    chmod -R a+w /opt && \
+    mkdir ~ubuntu/.cache && chown ubuntu: ~ubuntu/.cache
 USER ubuntu
 
-# Install python and create the venv
-RUN uv python install ${PYTHON_VERSION} && \
-    uv venv --relocatable --prompt "invoke" --python ${PYTHON_VERSION} ${VIRTUAL_ENV}
+# Install python
+RUN --mount=type=cache,target=/home/ubuntu/.cache/uv,uid=1000,gid=1000 \
+  uv python install ${PYTHON_VERSION}
 
 WORKDIR ${INVOKEAI_SRC}
-COPY invokeai ./invokeai
-COPY pyproject.toml ./
 
-# Editable mode helps use the same image for development:
-# the local working copy can be bind-mounted into the image
-# at path defined by ${INVOKEAI_SRC}
+# Install project's dependencies as a separate layer so they aren't rebuilt every commit.
+# bind-mount instead of copy to defer adding sources to the image until next layer.
+#
 # NOTE: there are no pytorch builds for arm64 + cuda, only cpu
 # x86_64/CUDA is the default
 RUN --mount=type=cache,target=/home/ubuntu/.cache/uv,uid=1000,gid=1000 \
+    --mount=type=bind,source=pyproject.toml,target=pyproject.toml \
+    --mount=type=bind,source=invokeai/version,target=invokeai/version \
+    if [ "$TARGETPLATFORM" = "linux/arm64" ] || [ "$GPU_DRIVER" = "cpu" ]; then \
+      UV_INDEX="https://download.pytorch.org/whl/cpu"; \
+    elif [ "$GPU_DRIVER" = "rocm" ]; then \
+      UV_INDEX="https://download.pytorch.org/whl/rocm6.1"; \
+    fi && \
+    uv sync --no-install-project
+
+# Now that the bulk of the dependencies have been installed, copy in the project files that change more frequently.
+COPY invokeai invokeai
+COPY pyproject.toml .
+
+RUN --mount=type=cache,target=/home/ubuntu/.cache/uv,uid=1000,gid=1000 \
+    --mount=type=bind,source=pyproject.toml,target=pyproject.toml \
     if [ "$TARGETPLATFORM" = "linux/arm64" ] || [ "$GPU_DRIVER" = "cpu" ]; then \
-        extra_index_url_arg="--extra-index-url https://download.pytorch.org/whl/cpu"; \
+      UV_INDEX="https://download.pytorch.org/whl/cpu"; \
     elif [ "$GPU_DRIVER" = "rocm" ]; then \
-        extra_index_url_arg="--extra-index-url https://download.pytorch.org/whl/rocm6.1"; \
-    else \
-        extra_index_url_arg="--extra-index-url https://download.pytorch.org/whl/cu124"; \
+      UV_INDEX="https://download.pytorch.org/whl/rocm6.1"; \
     fi && \
-    uv pip install --python ${PYTHON_VERSION} $extra_index_url_arg -e "."
+    uv sync
+
 
 #### Build the Web UI ------------------------------------
 
@@ -98,6 +113,7 @@ RUN apt update && apt install -y --no-install-recommends \
 
 ENV INVOKEAI_SRC=/opt/invokeai
 ENV VIRTUAL_ENV=/opt/venv
+ENV UV_PROJECT_ENVIRONMENT="$VIRTUAL_ENV"
 ENV PYTHON_VERSION=3.11
 ENV INVOKEAI_ROOT=/invokeai
 ENV INVOKEAI_HOST=0.0.0.0
@@ -109,7 +125,7 @@ ENV CONTAINER_GID=${CONTAINER_GID:-1000}
 # Install `uv` for package management
 # and install python for the ubuntu user (expected to exist on ubuntu >=24.x)
 # this is too tiny to optimize with multi-stage builds, but maybe we'll come back to it
-COPY --from=ghcr.io/astral-sh/uv:0.5.5 /uv /uvx /bin/
+COPY --from=ghcr.io/astral-sh/uv:0.6.0 /uv /uvx /bin/
 USER ubuntu
 RUN uv python install ${PYTHON_VERSION}
 USER root

invokeai/app/api/dependencies.py

@@ -36,6 +36,7 @@
 from invokeai.app.services.style_preset_records.style_preset_records_sqlite import SqliteStylePresetRecordsStorage
 from invokeai.app.services.urls.urls_default import LocalUrlService
 from invokeai.app.services.workflow_records.workflow_records_sqlite import SqliteWorkflowRecordsStorage
+from invokeai.app.services.workflow_thumbnails.workflow_thumbnails_disk import WorkflowThumbnailFileStorageDisk
 from invokeai.backend.stable_diffusion.diffusion.conditioning_data import ConditioningFieldData
 from invokeai.backend.util.logging import InvokeAILogger
 from invokeai.version.invokeai_version import __version__
@@ -83,6 +84,7 @@ def initialize(
 
         model_images_folder = config.models_path
         style_presets_folder = config.style_presets_path
+        workflow_thumbnails_folder = config.workflow_thumbnails_path
 
         db = init_db(config=config, logger=logger, image_files=image_files)
 
@@ -120,6 +122,7 @@ def initialize(
         workflow_records = SqliteWorkflowRecordsStorage(db=db)
         style_preset_records = SqliteStylePresetRecordsStorage(db=db)
         style_preset_image_files = StylePresetImageFileStorageDisk(style_presets_folder / "images")
+        workflow_thumbnails = WorkflowThumbnailFileStorageDisk(workflow_thumbnails_folder)
 
         services = InvocationServices(
             board_image_records=board_image_records,
@@ -147,6 +150,7 @@ def initialize(
             conditioning=conditioning,
             style_preset_records=style_preset_records,
             style_preset_image_files=style_preset_image_files,
+            workflow_thumbnails=workflow_thumbnails,
         )
 
         ApiDependencies.invoker = Invoker(services)

invokeai/app/api/extract_metadata_from_image.py

@@ -0,0 +1,124 @@
+import json
+import logging
+from dataclasses import dataclass
+
+from PIL import Image
+
+from invokeai.app.services.workflow_records.workflow_records_common import WorkflowWithoutIDValidator
+
+
+@dataclass
+class ExtractedMetadata:
+    invokeai_metadata: str | None
+    invokeai_workflow: str | None
+    invokeai_graph: str | None
+
+
+def extract_metadata_from_image(
+    pil_image: Image.Image,
+    invokeai_metadata_override: str | None,
+    invokeai_workflow_override: str | None,
+    invokeai_graph_override: str | None,
+    logger: logging.Logger,
+) -> ExtractedMetadata:
+    """
+    Extracts the "invokeai_metadata", "invokeai_workflow", and "invokeai_graph" data embedded in the PIL Image.
+
+    These items are stored as stringified JSON in the image file's metadata, so we need to do some parsing to validate
+    them. Once parsed, the values are returned as they came (as strings), or None if they are not present or invalid.
+
+    In some situations, we may prefer to override the values extracted from the image file with some other values.
+
+    For example, when uploading an image via API, the client can optionally provide the metadata directly in the request,
+    as opposed to embedding it in the image file. In this case, the client-provided metadata will be used instead of the
+    metadata embedded in the image file.
+
+    Args:
+        pil_image: The PIL Image object.
+        invokeai_metadata_override: The metadata override provided by the client.
+        invokeai_workflow_override: The workflow override provided by the client.
+        invokeai_graph_override: The graph override provided by the client.
+        logger: The logger to use for debug logging.
+
+    Returns:
+        ExtractedMetadata: The extracted metadata, workflow, and graph.
+    """
+
+    # The fallback value for metadata is None.
+    stringified_metadata: str | None = None
+
+    # Use the metadata override if provided, else attempt to extract it from the image file.
+    metadata_raw = invokeai_metadata_override or pil_image.info.get("invokeai_metadata", None)
+
+    # If the metadata is present in the image file, we will attempt to parse it as JSON. When we create images,
+    # we always store metadata as a stringified JSON dict. So, we expect it to be a string here.
+    if isinstance(metadata_raw, str):
+        try:
+            # Must be a JSON string
+            metadata_parsed = json.loads(metadata_raw)
+            # Must be a dict
+            if isinstance(metadata_parsed, dict):
+                # Looks good, overwrite the fallback value
+                stringified_metadata = metadata_raw
+        except Exception as e:
+            logger.debug(f"Failed to parse metadata for uploaded image, {e}")
+            pass
+
+    # We expect the workflow, if embedded in the image, to be a JSON-stringified WorkflowWithoutID. We will store it
+    # as a string.
+    workflow_raw: str | None = invokeai_workflow_override or pil_image.info.get("invokeai_workflow", None)
+
+    # The fallback value for workflow is None.
+    stringified_workflow: str | None = None
+
+    # If the workflow is present in the image file, we will attempt to parse it as JSON. When we create images, we
+    # always store workflows as a stringified JSON WorkflowWithoutID. So, we expect it to be a string here.
+    if isinstance(workflow_raw, str):
+        try:
+            # Validate the workflow JSON before storing it
+            WorkflowWithoutIDValidator.validate_json(workflow_raw)
+            # Looks good, overwrite the fallback value
+            stringified_workflow = workflow_raw
+        except Exception:
+            logger.debug("Failed to parse workflow for uploaded image")
+            pass
+
+    # We expect the workflow, if embedded in the image, to be a JSON-stringified Graph. We will store it as a
+    # string.
+    graph_raw: str | None = invokeai_graph_override or pil_image.info.get("invokeai_graph", None)
+
+    # The fallback value for graph is None.
+    stringified_graph: str | None = None
+
+    # If the graph is present in the image file, we will attempt to parse it as JSON. When we create images, we
+    # always store graphs as a stringified JSON Graph. So, we expect it to be a string here.
+    if isinstance(graph_raw, str):
+        try:
+            # TODO(psyche): Due to pydantic's handling of None values, it is possible for the graph to fail validation,
+            # even if it is a direct dump of a valid graph. Node fields in the graph are allowed to have be unset if
+            # they have incoming connections, but something about the ser/de process cannot adequately handle this.
+            #
+            # In lieu of fixing the graph validation, we will just do a simple check here to see if the graph is dict
+            # with the correct keys. This is not a perfect solution, but it should be good enough for now.
+
+            # FIX ME: Validate the graph JSON before storing it
+            # Graph.model_validate_json(graph_raw)
+
+            # Crappy workaround to validate JSON
+            graph_parsed = json.loads(graph_raw)
+            if not isinstance(graph_parsed, dict):
+                raise ValueError("Not a dict")
+            if not isinstance(graph_parsed.get("nodes", None), dict):
+                raise ValueError("'nodes' is not a dict")
+            if not isinstance(graph_parsed.get("edges", None), list):
+                raise ValueError("'edges' is not a list")
+
+            # Looks good, overwrite the fallback value
+            stringified_graph = graph_raw
+        except Exception as e:
+            logger.debug(f"Failed to parse graph for uploaded image, {e}")
+            pass
+
+    return ExtractedMetadata(
+        invokeai_metadata=stringified_metadata, invokeai_workflow=stringified_workflow, invokeai_graph=stringified_graph
+    )

invokeai/app/api/routers/images.py

@@ -6,9 +6,10 @@
 from fastapi.responses import FileResponse
 from fastapi.routing import APIRouter
 from PIL import Image
-from pydantic import BaseModel, Field, JsonValue
+from pydantic import BaseModel, Field
 
 from invokeai.app.api.dependencies import ApiDependencies
+from invokeai.app.api.extract_metadata_from_image import extract_metadata_from_image
 from invokeai.app.invocations.fields import MetadataField
 from invokeai.app.services.image_records.image_records_common import (
     ImageCategory,
@@ -45,18 +46,16 @@ async def upload_image(
     board_id: Optional[str] = Query(default=None, description="The board to add this image to, if any"),
     session_id: Optional[str] = Query(default=None, description="The session ID associated with this upload, if any"),
     crop_visible: Optional[bool] = Query(default=False, description="Whether to crop the image"),
-    metadata: Optional[JsonValue] = Body(
-        default=None, description="The metadata to associate with the image", embed=True
+    metadata: Optional[str] = Body(
+        default=None,
+        description="The metadata to associate with the image, must be a stringified JSON dict",
+        embed=True,
     ),
 ) -> ImageDTO:
     """Uploads an image"""
     if not file.content_type or not file.content_type.startswith("image"):
         raise HTTPException(status_code=415, detail="Not an image")
 
-    _metadata = None
-    _workflow = None
-    _graph = None
-
     contents = await file.read()
     try:
         pil_image = Image.open(io.BytesIO(contents))
@@ -67,30 +66,13 @@ async def upload_image(
         ApiDependencies.invoker.services.logger.error(traceback.format_exc())
         raise HTTPException(status_code=415, detail="Failed to read image")
 
-    # TODO: retain non-invokeai metadata on upload?
-    # attempt to parse metadata from image
-    metadata_raw = metadata if isinstance(metadata, str) else pil_image.info.get("invokeai_metadata", None)
-    if isinstance(metadata_raw, str):
-        _metadata = metadata_raw
-    else:
-        ApiDependencies.invoker.services.logger.debug("Failed to parse metadata for uploaded image")
-        pass
-
-    # attempt to parse workflow from image
-    workflow_raw = pil_image.info.get("invokeai_workflow", None)
-    if isinstance(workflow_raw, str):
-        _workflow = workflow_raw
-    else:
-        ApiDependencies.invoker.services.logger.debug("Failed to parse workflow for uploaded image")
-        pass
-
-    # attempt to extract graph from image
-    graph_raw = pil_image.info.get("invokeai_graph", None)
-    if isinstance(graph_raw, str):
-        _graph = graph_raw
-    else:
-        ApiDependencies.invoker.services.logger.debug("Failed to parse graph for uploaded image")
-        pass
+    extracted_metadata = extract_metadata_from_image(
+        pil_image=pil_image,
+        invokeai_metadata_override=metadata,
+        invokeai_workflow_override=None,
+        invokeai_graph_override=None,
+        logger=ApiDependencies.invoker.services.logger,
+    )
 
     try:
         image_dto = ApiDependencies.invoker.services.images.create(
@@ -99,9 +81,9 @@ async def upload_image(
             image_category=image_category,
             session_id=session_id,
             board_id=board_id,
-            metadata=_metadata,
-            workflow=_workflow,
-            graph=_graph,
+            metadata=extracted_metadata.invokeai_metadata,
+            workflow=extracted_metadata.invokeai_workflow,
+            graph=extracted_metadata.invokeai_graph,
             is_intermediate=is_intermediate,
         )