fix controlnets for latest diffusers v

Currently translated at 98.4% (2120 of 2153 strings)

translationBot(ui): update translation (Italian)

Currently translated at 97.3% (2097 of 2153 strings)

Co-authored-by: Riccardo Giovanetti <riccardo.giovanetti@gmail.com>
Translate-URL: https://hosted.weblate.org/projects/invokeai/web-ui/it/
Translation: InvokeAI/Web UI

.github/ISSUE_TEMPLATE/BUG_REPORT.yml

@@ -21,6 +21,20 @@ body:
         - label: I have searched the existing issues
           required: true
 
+  - type: dropdown
+    id: install_method
+    attributes:
+      label: Install method
+      description: How did you install Invoke?
+      multiple: false
+      options:
+        - "Invoke's Launcher"
+        - 'Stability Matrix'
+        - 'Pinokio'
+        - 'Manual'
+    validations:
+      required: true
+
   - type: markdown
     attributes:
       value: __Describe your environment__
@@ -76,8 +90,8 @@ body:
     attributes:
       label: Version number
       description: |
-        The version of Invoke you have installed. If it is not the latest version, please update and try again to confirm the issue still exists. If you are testing main, please include the commit hash instead.
-      placeholder: ex. 3.6.1
+        The version of Invoke you have installed. If it is not the [latest version](https://github.com/invoke-ai/InvokeAI/releases/latest), please update and try again to confirm the issue still exists. If you are testing main, please include the commit hash instead.
+      placeholder: ex. v6.0.2
     validations:
       required: true
 
@@ -85,17 +99,17 @@ body:
     id: browser-version
     attributes:
       label: Browser
-      description: Your web browser and version.
+      description: Your web browser and version, if you do not use the Launcher's provided GUI.
       placeholder: ex. Firefox 123.0b3
     validations:
-      required: true
+      required: false
 
   - type: textarea
     id: python-deps
     attributes:
-      label: Python dependencies
+      label: System Information
       description: |
-        If the problem occurred during image generation, click the gear icon at the bottom left corner, click "About", click the copy button and then paste here.
+        Click the gear icon at the bottom left corner, then click "About". Click the copy button and then paste here.
     validations:
       required: false
 

.github/actions/install-frontend-deps/action.yml

@@ -3,15 +3,15 @@ description: Installs frontend dependencies with pnpm, with caching
 runs:
   using: 'composite'
   steps:
-    - name: setup node 18
+    - name: setup node 20
       uses: actions/setup-node@v4
       with:
-        node-version: '18'
+        node-version: '20'
 
     - name: setup pnpm
       uses: pnpm/action-setup@v4
       with:
-        version: 8.15.6
+        version: 10
         run_install: false
 
     - name: get pnpm store directory

.github/pull_request_template.md

@@ -18,5 +18,6 @@
 
 - [ ] _The PR has a short but descriptive title, suitable for a changelog_
 - [ ] _Tests added / updated (if applicable)_
+- [ ] _❗Changes to a redux slice have a corresponding migration_
 - [ ] _Documentation added / updated (if applicable)_
 - [ ] _Updated `What's New` copy (if doing a release after this PR)_

.github/workflows/build-container.yml

@@ -45,6 +45,9 @@ jobs:
     steps:
       - name: Free up more disk space on the runner
         # https://github.com/actions/runner-images/issues/2840#issuecomment-1284059930
+        # the /mnt dir has 70GBs of free space
+        # /dev/sda1        74G   28K   70G   1% /mnt
+        # According to some online posts the /mnt is not always there, so checking before setting docker to use it
         run: |
           echo "----- Free space before cleanup"
           df -h
@@ -52,6 +55,11 @@ jobs:
           sudo rm -rf "$AGENT_TOOLSDIRECTORY"
           sudo swapoff /mnt/swapfile
           sudo rm -rf /mnt/swapfile
+          if [ -d /mnt ]; then
+            sudo chmod -R 777 /mnt
+            echo '{"data-root": "/mnt/docker-root"}' | sudo tee /etc/docker/daemon.json
+            sudo systemctl restart docker
+          fi
           echo "----- Free space after cleanup"
           df -h
 

.github/workflows/lfs-checks.yml

@@ -0,0 +1,30 @@
+# Checks that large files and LFS-tracked files are properly checked in with pointer format.
+# Uses https://github.com/ppremk/lfs-warning to detect LFS issues.
+
+name: 'lfs checks'
+
+on:
+  push:
+    branches:
+      - 'main'
+  pull_request:
+    types:
+      - 'ready_for_review'
+      - 'opened'
+      - 'synchronize'
+  merge_group:
+  workflow_dispatch:
+
+jobs:
+  lfs-check:
+    runs-on: ubuntu-latest
+    timeout-minutes: 5
+    permissions:
+      # Required to label and comment on the PRs
+      pull-requests: write
+    steps:
+      - name: checkout
+        uses: actions/checkout@v4
+
+      - name: check lfs files
+        uses: ppremk/lfs-warning@v3.3

.github/workflows/typegen-checks.yml

@@ -39,6 +39,18 @@ jobs:
       - name: checkout
         uses: actions/checkout@v4
 
+      - name: Free up more disk space on the runner
+        # https://github.com/actions/runner-images/issues/2840#issuecomment-1284059930
+        run: |
+          echo "----- Free space before cleanup"
+          df -h
+          sudo rm -rf /usr/share/dotnet
+          sudo rm -rf "$AGENT_TOOLSDIRECTORY"
+          sudo swapoff /mnt/swapfile
+          sudo rm -rf /mnt/swapfile
+          echo "----- Free space after cleanup"
+          df -h
+
       - name: check for changed files
         if: ${{ inputs.always_run != true }}
         id: changed-files

.gitignore

@@ -190,3 +190,5 @@ installer/update.bat
 installer/update.sh
 installer/InvokeAI-Installer/
 .aider*
+
+.claude/

docker/.env.sample

@@ -22,6 +22,10 @@
 ## GPU_DRIVER can be set to either `cuda` or `rocm` to enable GPU support in the container accordingly.
 # GPU_DRIVER=cuda #| rocm
 
+## If you are using ROCM, you will need to ensure that the render group within the container and the host system use the same group ID.
+## To obtain the group ID of the render group on the host system, run `getent group render` and grab the number.
+# RENDER_GROUP_ID=
+
 ## CONTAINER_UID can be set to the UID of the user on the host system that should own the files in the container.
 ## It is usually not necessary to change this. Use `id -u` on the host system to find the UID.
 # CONTAINER_UID=1000

docker/Dockerfile

@@ -5,8 +5,7 @@
 FROM docker.io/node:22-slim AS web-builder
 ENV PNPM_HOME="/pnpm"
 ENV PATH="$PNPM_HOME:$PATH"
-RUN corepack use pnpm@8.x
-RUN corepack enable
+RUN corepack use pnpm@10.x && corepack enable
 
 WORKDIR /build
 COPY invokeai/frontend/web/ ./
@@ -44,7 +43,6 @@ ENV \
     UV_MANAGED_PYTHON=1 \
     UV_LINK_MODE=copy \
     UV_PROJECT_ENVIRONMENT=/opt/venv \
-    UV_INDEX="https://download.pytorch.org/whl/cu124" \
     INVOKEAI_ROOT=/invokeai \
     INVOKEAI_HOST=0.0.0.0 \
     INVOKEAI_PORT=9090 \
@@ -75,19 +73,17 @@ RUN --mount=type=cache,target=/root/.cache/uv \
     --mount=type=bind,source=uv.lock,target=uv.lock \
     # this is just to get the package manager to recognize that the project exists, without making changes to the docker layer
     --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 [ "$GPU_DRIVER" = "rocm" ]; then UV_INDEX="https://download.pytorch.org/whl/rocm6.2"; \
-    fi && \
-    uv sync --frozen
-
-# build patchmatch
-RUN cd /usr/lib/$(uname -p)-linux-gnu/pkgconfig/ && ln -sf opencv4.pc opencv.pc
-RUN python -c "from patchmatch import patch_match"
+    ulimit -n 30000 && \
+    uv sync --extra $GPU_DRIVER --frozen
 
 # Link amdgpu.ids for ROCm builds
 # contributed by https://github.com/Rubonnek
 RUN mkdir -p "/opt/amdgpu/share/libdrm" &&\
-    ln -s "/usr/share/libdrm/amdgpu.ids" "/opt/amdgpu/share/libdrm/amdgpu.ids"
+    ln -s "/usr/share/libdrm/amdgpu.ids" "/opt/amdgpu/share/libdrm/amdgpu.ids" && groupadd render
+
+# build patchmatch
+RUN cd /usr/lib/$(uname -p)-linux-gnu/pkgconfig/ && ln -sf opencv4.pc opencv.pc
+RUN python -c "from patchmatch import patch_match"
 
 RUN mkdir -p ${INVOKEAI_ROOT} && chown -R ${CONTAINER_UID}:${CONTAINER_GID} ${INVOKEAI_ROOT}
 
@@ -106,8 +102,6 @@ COPY invokeai ${INVOKEAI_SRC}/invokeai
 RUN --mount=type=cache,target=/root/.cache/uv \
     --mount=type=bind,source=pyproject.toml,target=pyproject.toml \
     --mount=type=bind,source=uv.lock,target=uv.lock \
-    if [ "$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/rocm6.2"; \
-    fi && \
-    uv pip install -e .
+    ulimit -n 30000 && \
+    uv pip install -e .[$GPU_DRIVER]
 

docker/Dockerfile-rocm-full

@@ -0,0 +1,136 @@
+# syntax=docker/dockerfile:1.4
+
+#### Web UI ------------------------------------
+
+FROM docker.io/node:22-slim AS web-builder
+ENV PNPM_HOME="/pnpm"
+ENV PATH="$PNPM_HOME:$PATH"
+RUN corepack use pnpm@8.x
+RUN corepack enable
+
+WORKDIR /build
+COPY invokeai/frontend/web/ ./
+RUN --mount=type=cache,target=/pnpm/store \
+    pnpm install --frozen-lockfile
+RUN npx vite build
+
+## Backend ---------------------------------------
+
+FROM library/ubuntu:24.04
+
+ARG DEBIAN_FRONTEND=noninteractive
+RUN rm -f /etc/apt/apt.conf.d/docker-clean; echo 'Binary::apt::APT::Keep-Downloaded-Packages "true";' > /etc/apt/apt.conf.d/keep-cache
+RUN --mount=type=cache,target=/var/cache/apt \
+    --mount=type=cache,target=/var/lib/apt \
+    apt update && apt install -y --no-install-recommends \
+    ca-certificates \
+    git \
+    gosu \
+    libglib2.0-0 \
+    libgl1 \
+    libglx-mesa0 \
+    build-essential \
+    libopencv-dev \
+    libstdc++-10-dev \
+    wget
+
+ENV \
+    PYTHONUNBUFFERED=1 \
+    PYTHONDONTWRITEBYTECODE=1 \
+    VIRTUAL_ENV=/opt/venv \
+    INVOKEAI_SRC=/opt/invokeai \
+    PYTHON_VERSION=3.12 \
+    UV_PYTHON=3.12 \
+    UV_COMPILE_BYTECODE=1 \
+    UV_MANAGED_PYTHON=1 \
+    UV_LINK_MODE=copy \
+    UV_PROJECT_ENVIRONMENT=/opt/venv \
+    INVOKEAI_ROOT=/invokeai \
+    INVOKEAI_HOST=0.0.0.0 \
+    INVOKEAI_PORT=9090 \
+    PATH="/opt/venv/bin:$PATH" \
+    CONTAINER_UID=${CONTAINER_UID:-1000} \
+    CONTAINER_GID=${CONTAINER_GID:-1000}
+
+ARG GPU_DRIVER=cuda
+
+# Install `uv` for package management
+COPY --from=ghcr.io/astral-sh/uv:0.6.9 /uv /uvx /bin/
+
+# Install python & allow non-root user to use it by traversing the /root dir without read permissions
+RUN --mount=type=cache,target=/root/.cache/uv \
+    uv python install ${PYTHON_VERSION} && \
+    # chmod --recursive a+rX /root/.local/share/uv/python
+    chmod 711 /root
+
+WORKDIR ${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=/root/.cache/uv \
+    --mount=type=bind,source=pyproject.toml,target=pyproject.toml \
+    --mount=type=bind,source=uv.lock,target=uv.lock \
+    # this is just to get the package manager to recognize that the project exists, without making changes to the docker layer
+    --mount=type=bind,source=invokeai/version,target=invokeai/version \
+    ulimit -n 30000 && \
+    uv sync --extra $GPU_DRIVER --frozen
+
+RUN --mount=type=cache,target=/var/cache/apt \
+    --mount=type=cache,target=/var/lib/apt \
+    if [ "$GPU_DRIVER" = "rocm" ]; then \
+    wget -O /tmp/amdgpu-install.deb \
+    https://repo.radeon.com/amdgpu-install/6.3.4/ubuntu/noble/amdgpu-install_6.3.60304-1_all.deb && \
+    apt install -y /tmp/amdgpu-install.deb && \
+    apt update && \
+    amdgpu-install --usecase=rocm -y && \
+    apt-get autoclean && \
+    apt clean && \
+    rm -rf /tmp/* /var/tmp/* && \
+    usermod -a -G render ubuntu && \
+    usermod -a -G video ubuntu && \
+    echo "\\n/opt/rocm/lib\\n/opt/rocm/lib64" >> /etc/ld.so.conf.d/rocm.conf && \
+    ldconfig && \
+    update-alternatives --auto rocm; \
+    fi
+
+## Heathen711: Leaving this for review input, will remove before merge
+# RUN --mount=type=cache,target=/var/cache/apt \
+#     --mount=type=cache,target=/var/lib/apt \
+#     if [ "$GPU_DRIVER" = "rocm" ]; then \
+#     groupadd render && \
+#     usermod -a -G render ubuntu && \
+#     usermod -a -G video ubuntu; \
+#     fi
+
+## Link amdgpu.ids for ROCm builds
+## contributed by https://github.com/Rubonnek
+# RUN mkdir -p "/opt/amdgpu/share/libdrm" &&\
+#     ln -s "/usr/share/libdrm/amdgpu.ids" "/opt/amdgpu/share/libdrm/amdgpu.ids"
+
+# build patchmatch
+RUN cd /usr/lib/$(uname -p)-linux-gnu/pkgconfig/ && ln -sf opencv4.pc opencv.pc
+RUN python -c "from patchmatch import patch_match"
+
+RUN mkdir -p ${INVOKEAI_ROOT} && chown -R ${CONTAINER_UID}:${CONTAINER_GID} ${INVOKEAI_ROOT}
+
+COPY docker/docker-entrypoint.sh ./
+ENTRYPOINT ["/opt/invokeai/docker-entrypoint.sh"]
+CMD ["invokeai-web"]
+
+# --link requires buldkit w/ dockerfile syntax 1.4, does not work with podman
+COPY --link --from=web-builder /build/dist ${INVOKEAI_SRC}/invokeai/frontend/web/dist
+
+# add sources last to minimize image changes on code changes
+COPY invokeai ${INVOKEAI_SRC}/invokeai
+
+# this should not increase image size because we've already installed dependencies
+# in a previous layer
+RUN --mount=type=cache,target=/root/.cache/uv \
+    --mount=type=bind,source=pyproject.toml,target=pyproject.toml \
+    --mount=type=bind,source=uv.lock,target=uv.lock \
+    ulimit -n 30000 && \
+    uv pip install -e .[$GPU_DRIVER]
+

docker/docker-compose.yml

@@ -47,8 +47,9 @@ services:
 
   invokeai-rocm:
     <<: *invokeai
-    devices:
-      - /dev/kfd:/dev/kfd
-      - /dev/dri:/dev/dri
+    environment:
+      - AMD_VISIBLE_DEVICES=all
+      - RENDER_GROUP_ID=${RENDER_GROUP_ID}
+    runtime: amd
     profiles:
       - rocm

docker/docker-entrypoint.sh

@@ -21,6 +21,17 @@ _=$(id ${USER} 2>&1) || useradd -u ${USER_ID} ${USER}
 # ensure the UID is correct
 usermod -u ${USER_ID} ${USER} 1>/dev/null
 
+## ROCM specific configuration
+# render group within the container must match the host render group
+# otherwise the container will not be able to access the host GPU.
+if [[ -v "RENDER_GROUP_ID" ]] && [[ ! -z "${RENDER_GROUP_ID}" ]]; then
+  # ensure the render group exists
+  groupmod -g ${RENDER_GROUP_ID} render
+  usermod -a -G render ${USER}
+  usermod -a -G video ${USER}
+fi
+
+
 ### Set the $PUBLIC_KEY env var to enable SSH access.
 # We do not install openssh-server in the image by default to avoid bloat.
 # but it is useful to have the full SSH server e.g. on Runpod.

docker/run.sh

@@ -13,7 +13,7 @@ run() {
 
   # parse .env file for build args
   build_args=$(awk '$1 ~ /=[^$]/ && $0 !~ /^#/ {print "--build-arg " $0 " "}' .env) &&
-  profile="$(awk -F '=' '/GPU_DRIVER/ {print $2}' .env)"
+  profile="$(awk -F '=' '/GPU_DRIVER=/ {print $2}' .env)"
 
   # default to 'cuda' profile
   [[ -z "$profile" ]] && profile="cuda"
@@ -30,7 +30,7 @@ run() {
 
   printf "%s\n" "starting service $service_name"
   docker compose --profile "$profile" up -d "$service_name"
-  docker compose logs -f
+  docker compose --profile "$profile" logs -f
 }
 
 run

docs/contributing/MODEL_MANAGER.md

@@ -265,7 +265,7 @@ If the key is unrecognized, this call raises an
 
 #### exists(key) -> AnyModelConfig
 
-Returns True if a model with the given key exists in the databsae.
+Returns True if a model with the given key exists in the database.
 
 #### search_by_path(path) -> AnyModelConfig
 
@@ -718,7 +718,7 @@ When downloading remote models is implemented, additional
 configuration information, such as list of trigger terms, will be
 retrieved from the HuggingFace and Civitai model repositories.
 
-The probed values can be overriden by providing a dictionary in the
+The probed values can be overridden by providing a dictionary in the
 optional `config` argument passed to `import_model()`. You may provide
 overriding values for any of the model's configuration
 attributes. Here is an example of setting the
@@ -841,7 +841,7 @@ variable.
 
 #### installer.start(invoker)
 
-The `start` method is called by the API intialization routines when
+The `start` method is called by the API initialization routines when
 the API starts up. Its effect is to call `sync_to_config()` to
 synchronize the model record store database with what's currently on
 disk.

docs/contributing/contributors.md

@@ -16,7 +16,7 @@ We thank [all contributors](https://github.com/invoke-ai/InvokeAI/graphs/contrib
 - @psychedelicious (Spencer Mabrito) - Web Team Leader
 - @joshistoast (Josh Corbett) - Web Development
 - @cheerio (Mary Rogers) - Lead Engineer & Web App Development
-- @ebr (Eugene Brodsky) - Cloud/DevOps/Sofware engineer; your friendly neighbourhood cluster-autoscaler
+- @ebr (Eugene Brodsky) - Cloud/DevOps/Software engineer; your friendly neighbourhood cluster-autoscaler
 - @sunija - Standalone version
 - @brandon (Brandon Rising) - Platform, Infrastructure, Backend Systems
 - @ryanjdick (Ryan Dick) - Machine Learning & Training

docs/contributing/dev-environment.md

@@ -41,7 +41,7 @@ If you just want to use Invoke, you should use the [launcher][launcher link].
      With the modifications made, the install command should look something like this:
 
       ```sh
-      uv pip install -e ".[dev,test,docs,xformers]" --python 3.12 --python-preference only-managed --index=https://download.pytorch.org/whl/cu126 --reinstall
+      uv pip install -e ".[dev,test,docs,xformers]" --python 3.12 --python-preference only-managed --index=https://download.pytorch.org/whl/cu128 --reinstall
       ```
 
 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.
@@ -50,11 +50,11 @@ If you just want to use Invoke, you should use the [launcher][launcher link].
 
       If you only want to edit the docs, you can stop here and skip to the **Documentation** section below.
 
-7. Install the frontend dev toolchain:
+7. Install the frontend dev toolchain, paying attention to versions:
 
-      - [`nodejs`](https://nodejs.org/) (v20+)
+      - [`nodejs`](https://nodejs.org/) (tested on LTS, v22)
 
-      - [`pnpm`](https://pnpm.io/8.x/installation) (must be v8 - not v9!)
+      - [`pnpm`](https://pnpm.io/installation) (tested on v10)
 
 8. Do a production build of the frontend:
 

docs/contributing/frontend/workflows.md

@@ -297,7 +297,7 @@ Migration logic is in [migrations.ts].
 <!-- links -->
 
 [pydantic]: https://github.com/pydantic/pydantic 'pydantic'
-[zod]: https://github.com/colinhacks/zod 'zod/v4'
+[zod]: https://github.com/colinhacks/zod 'zod'
 [openapi-types]: https://github.com/kogosoftwarellc/open-api/tree/main/packages/openapi-types 'openapi-types'
 [reactflow]: https://github.com/xyflow/xyflow 'reactflow'
 [reactflow-concepts]: https://reactflow.dev/learn/concepts/terms-and-definitions

docs/installation/manual.md

@@ -69,34 +69,34 @@ The following commands vary depending on the version of Invoke being installed a
     - If you have an Nvidia 20xx series GPU or older, use `invokeai[xformers]`.
     - If you have an Nvidia 30xx series GPU or newer, or do not have an Nvidia GPU, use `invokeai`.
 
-7. Determine the `PyPI` index URL to use for installation, if any. This is necessary to get the right version of torch installed.
+7. Determine the torch backend to use for installation, if any. This is necessary to get the right version of torch installed. This is acheived by using [UV's built in torch support.](https://docs.astral.sh/uv/guides/integration/pytorch/#automatic-backend-selection)
 
     === "Invoke v5.12 and later"
 
-        - If you are on Windows or Linux with an Nvidia GPU, use `https://download.pytorch.org/whl/cu128`.
-        - If you are on Linux with no GPU, use `https://download.pytorch.org/whl/cpu`.
-        - If you are on Linux with an AMD GPU, use `https://download.pytorch.org/whl/rocm6.2.4`.
-        - **In all other cases, do not use an index.**
+        - If you are on Windows or Linux with an Nvidia GPU, use `--torch-backend=cu128`.
+        - If you are on Linux with no GPU, use `--torch-backend=cpu`.
+        - If you are on Linux with an AMD GPU, use `--torch-backend=rocm6.3`.
+        - **In all other cases, do not use a torch backend.**
 
     === "Invoke v5.10.0 to v5.11.0"
 
-        - If you are on Windows or Linux with an Nvidia GPU, use `https://download.pytorch.org/whl/cu126`.
-        - If you are on Linux with no GPU, use `https://download.pytorch.org/whl/cpu`.
-        - If you are on Linux with an AMD GPU, use `https://download.pytorch.org/whl/rocm6.2.4`.
+        - If you are on Windows or Linux with an Nvidia GPU, use `--torch-backend=cu126`.
+        - If you are on Linux with no GPU, use `--torch-backend=cpu`.
+        - If you are on Linux with an AMD GPU, use `--torch-backend=rocm6.2.4`.
         - **In all other cases, do not use an index.**
 
     === "Invoke v5.0.0 to v5.9.1"
 
-        - If you are on Windows with an Nvidia GPU, use `https://download.pytorch.org/whl/cu124`.
-        - If you are on Linux with no GPU, use `https://download.pytorch.org/whl/cpu`.
-        - If you are on Linux with an AMD GPU, use `https://download.pytorch.org/whl/rocm6.1`.
+        - If you are on Windows with an Nvidia GPU, use `--torch-backend=cu124`.
+        - If you are on Linux with no GPU, use `--torch-backend=cpu`.
+        - If you are on Linux with an AMD GPU, use `--torch-backend=rocm6.1`.
         - **In all other cases, do not use an index.**
 
     === "Invoke v4"
 
-        - If you are on Windows with an Nvidia GPU, use `https://download.pytorch.org/whl/cu124`.
-        - If you are on Linux with no GPU, use `https://download.pytorch.org/whl/cpu`.
-        - If you are on Linux with an AMD GPU, use `https://download.pytorch.org/whl/rocm5.2`.
+        - If you are on Windows with an Nvidia GPU, use `--torch-backend=cu124`.
+        - If you are on Linux with no GPU, use `--torch-backend=cpu`.
+        - If you are on Linux with an AMD GPU, use `--torch-backend=rocm5.2`.
         - **In all other cases, do not use an index.**
 
 8. Install the `invokeai` package. Substitute the package specifier and version.
@@ -105,10 +105,10 @@ The following commands vary depending on the version of Invoke being installed a
     uv pip install <PACKAGE_SPECIFIER>==<VERSION> --python 3.12 --python-preference only-managed --force-reinstall
     ```
 
-    If you determined you needed to use a `PyPI` index URL in the previous step, you'll need to add `--index=<INDEX_URL>` like this:
+    If you determined you needed to use a torch backend in the previous step, you'll need to set the backend like this:
 
     ```sh
-    uv pip install <PACKAGE_SPECIFIER>==<VERSION> --python 3.12 --python-preference only-managed --index=<INDEX_URL> --force-reinstall
+    uv pip install <PACKAGE_SPECIFIER>==<VERSION> --python 3.12 --python-preference only-managed --torch-backend=<VERSION> --force-reinstall
     ```
 
 9. Deactivate and reactivate your venv so that the invokeai-specific commands become available in the environment:

docs/installation/quick_start.md

@@ -33,30 +33,45 @@ Hardware requirements vary significantly depending on model and image output siz
 
 More detail on system requirements can be found [here](./requirements.md).
 
-## Step 2: Download
+## Step 2: Download and Set Up the Launcher
 
-Download the most recent launcher for your operating system:
+The Launcher manages your Invoke install. Follow these instructions to download and set up the Launcher.
 
-- [Download for Windows](https://download.invoke.ai/Invoke%20Community%20Edition.exe)
-- [Download for macOS](https://download.invoke.ai/Invoke%20Community%20Edition.dmg)
-- [Download for Linux](https://download.invoke.ai/Invoke%20Community%20Edition.AppImage)
+!!! info "Instructions for each OS"
 
-## Step 3: Install or Update
+    === "Windows"
 
-Run the launcher you just downloaded, click **Install** and follow the instructions to get set up.
+        - [Download for Windows](https://github.com/invoke-ai/launcher/releases/latest/download/Invoke.Community.Edition.Setup.latest.exe)
+        - Run the `EXE` to install the Launcher and start it.
+        - A desktop shortcut will be created; use this to run the Launcher in the future.
+        - You can delete the `EXE` file you downloaded.
+
+    === "macOS"
+
+        - [Download for macOS](https://github.com/invoke-ai/launcher/releases/latest/download/Invoke.Community.Edition-latest-arm64.dmg)
+        - Open the `DMG` and drag the app into `Applications`.
+        - Run the Launcher using its entry in `Applications`.
+        - You can delete the `DMG` file you downloaded.
+
+    === "Linux"
+
+        - [Download for Linux](https://github.com/invoke-ai/launcher/releases/latest/download/Invoke.Community.Edition-latest.AppImage)
+        - You may need to edit the `AppImage` file properties and make it executable.
+        - Optionally move the file to a location that does not require admin privileges and add a desktop shortcut for it.
+        - Run the Launcher by double-clicking the `AppImage` or the shortcut you made.
+
+## Step 3: Install Invoke
+
+Run the Launcher you just set up if you haven't already. Click **Install** and follow the instructions to install (or update) Invoke.
 
 If you have an existing Invoke installation, you can select it and let the launcher manage the install. You'll be able to update or launch the installation.
 
-!!! warning "Problem running the launcher on macOS"
+!!! tip "Updating"
 
-    macOS may not allow you to run the launcher. We are working to resolve this by signing the launcher executable. Until that is done, you can manually flag the launcher as safe:
+    The Launcher will check for updates for itself _and_ Invoke.
 
-    - Open the **Invoke Community Edition.dmg** file.
-    - Drag the launcher to **Applications**.
-    - Open a terminal.
-    - Run `xattr -d 'com.apple.quarantine' /Applications/Invoke\ Community\ Edition.app`.
-
-    You should now be able to run the launcher.
+    - When the Launcher detects an update is available for itself, you'll get a small popup window. Click through this and the Launcher will update itself.
+    - When the Launcher detects an update for Invoke, you'll see a small green alert in the Launcher. Click that and follow the instructions to update Invoke.
 
 ## Step 4: Launch
 

docs/nodes/NODES.md

@@ -41,7 +41,7 @@ Nodes have a "Use Cache" option in their footer. This allows for performance imp
 
 There are several node grouping concepts that can be examined with a narrow focus. These (and other) groupings can be pieced together to make up functional graph setups, and are important to understanding how groups of nodes work together as part of a whole. Note that the screenshots below aren't examples of complete functioning node graphs (see Examples).
 
-### Noise
+### Create Latent Noise
 
 An initial noise tensor is necessary for the latent diffusion process. As a result, the Denoising node requires a noise node input.  
 

invokeai/app/api/dependencies.py

@@ -10,6 +10,7 @@ from invokeai.app.services.board_images.board_images_default import BoardImagesS
 from invokeai.app.services.board_records.board_records_sqlite import SqliteBoardRecordStorage
 from invokeai.app.services.boards.boards_default import BoardService
 from invokeai.app.services.bulk_download.bulk_download_default import BulkDownloadService
+from invokeai.app.services.client_state_persistence.client_state_persistence_sqlite import ClientStatePersistenceSqlite
 from invokeai.app.services.config.config_default import InvokeAIAppConfig
 from invokeai.app.services.download.download_default import DownloadQueueService
 from invokeai.app.services.events.events_fastapievents import FastAPIEventService
@@ -151,6 +152,7 @@ class ApiDependencies:
         style_preset_records = SqliteStylePresetRecordsStorage(db=db)
         style_preset_image_files = StylePresetImageFileStorageDisk(style_presets_folder / "images")
         workflow_thumbnails = WorkflowThumbnailFileStorageDisk(workflow_thumbnails_folder)
+        client_state_persistence = ClientStatePersistenceSqlite(db=db)
 
         services = InvocationServices(
             board_image_records=board_image_records,
@@ -181,6 +183,7 @@ class ApiDependencies:
             style_preset_records=style_preset_records,
             style_preset_image_files=style_preset_image_files,
             workflow_thumbnails=workflow_thumbnails,
+            client_state_persistence=client_state_persistence,
         )
 
         ApiDependencies.invoker = Invoker(services)

invokeai/app/api/routers/board_videos.py

@@ -0,0 +1,39 @@
+from fastapi import Body, HTTPException
+from fastapi.routing import APIRouter
+
+from invokeai.app.services.videos_common import AddVideosToBoardResult, RemoveVideosFromBoardResult
+
+board_videos_router = APIRouter(prefix="/v1/board_videos", tags=["boards"])
+
+
+@board_videos_router.post(
+    "/batch",
+    operation_id="add_videos_to_board",
+    responses={
+        201: {"description": "Videos were added to board successfully"},
+    },
+    status_code=201,
+    response_model=AddVideosToBoardResult,
+)
+async def add_videos_to_board(
+    board_id: str = Body(description="The id of the board to add to"),
+    video_ids: list[str] = Body(description="The ids of the videos to add", embed=True),
+) -> AddVideosToBoardResult:
+    """Adds a list of videos to a board"""
+    raise HTTPException(status_code=501, detail="Not implemented")
+
+
+@board_videos_router.post(
+    "/batch/delete",
+    operation_id="remove_videos_from_board",
+    responses={
+        201: {"description": "Videos were removed from board successfully"},
+    },
+    status_code=201,
+    response_model=RemoveVideosFromBoardResult,
+)
+async def remove_videos_from_board(
+    video_ids: list[str] = Body(description="The ids of the videos to remove", embed=True),
+) -> RemoveVideosFromBoardResult:
+    """Removes a list of videos from their board, if they had one"""
+    raise HTTPException(status_code=501, detail="Not implemented")

invokeai/app/api/routers/client_state.py

@@ -0,0 +1,58 @@
+from fastapi import Body, HTTPException, Path, Query
+from fastapi.routing import APIRouter
+
+from invokeai.app.api.dependencies import ApiDependencies
+from invokeai.backend.util.logging import logging
+
+client_state_router = APIRouter(prefix="/v1/client_state", tags=["client_state"])
+
+
+@client_state_router.get(
+    "/{queue_id}/get_by_key",
+    operation_id="get_client_state_by_key",
+    response_model=str | None,
+)
+async def get_client_state_by_key(
+    queue_id: str = Path(description="The queue id to perform this operation on"),
+    key: str = Query(..., description="Key to get"),
+) -> str | None:
+    """Gets the client state"""
+    try:
+        return ApiDependencies.invoker.services.client_state_persistence.get_by_key(queue_id, key)
+    except Exception as e:
+        logging.error(f"Error getting client state: {e}")
+        raise HTTPException(status_code=500, detail="Error setting client state")
+
+
+@client_state_router.post(
+    "/{queue_id}/set_by_key",
+    operation_id="set_client_state",
+    response_model=str,
+)
+async def set_client_state(
+    queue_id: str = Path(description="The queue id to perform this operation on"),
+    key: str = Query(..., description="Key to set"),
+    value: str = Body(..., description="Stringified value to set"),
+) -> str:
+    """Sets the client state"""
+    try:
+        return ApiDependencies.invoker.services.client_state_persistence.set_by_key(queue_id, key, value)
+    except Exception as e:
+        logging.error(f"Error setting client state: {e}")
+        raise HTTPException(status_code=500, detail="Error setting client state")
+
+
+@client_state_router.post(
+    "/{queue_id}/delete",
+    operation_id="delete_client_state",
+    responses={204: {"description": "Client state deleted"}},
+)
+async def delete_client_state(
+    queue_id: str = Path(description="The queue id to perform this operation on"),
+) -> None:
+    """Deletes the client state"""
+    try:
+        ApiDependencies.invoker.services.client_state_persistence.delete(queue_id)
+    except Exception as e:
+        logging.error(f"Error deleting client state: {e}")
+        raise HTTPException(status_code=500, detail="Error deleting client state")

invokeai/app/api/routers/images.py

@@ -72,7 +72,7 @@ async def upload_image(
     resize_to: Optional[str] = Body(
         default=None,
         description=f"Dimensions to resize the image to, must be stringified tuple of 2 integers. Max total pixel count: {ResizeToDimensions.MAX_SIZE}",
-        example='"[1024,1024]"',
+        examples=['"[1024,1024]"'],
     ),
     metadata: Optional[str] = Body(
         default=None,

invokeai/app/api/routers/model_manager.py

@@ -292,7 +292,7 @@ async def get_hugging_face_models(
 )
 async def update_model_record(
     key: Annotated[str, Path(description="Unique key of model")],
-    changes: Annotated[ModelRecordChanges, Body(description="Model config", example=example_model_input)],
+    changes: Annotated[ModelRecordChanges, Body(description="Model config", examples=[example_model_input])],
 ) -> AnyModelConfig:
     """Update a model's config."""
     logger = ApiDependencies.invoker.services.logger
@@ -450,7 +450,7 @@ async def install_model(
     access_token: Optional[str] = Query(description="access token for the remote resource", default=None),
     config: ModelRecordChanges = Body(
         description="Object containing fields that override auto-probed values in the model config record, such as name, description and prediction_type ",
-        example={"name": "string", "description": "string"},
+        examples=[{"name": "string", "description": "string"}],
     ),
 ) -> ModelInstallJob:
     """Install a model using a string identifier.

invokeai/app/api/routers/session_queue.py

@@ -7,7 +7,6 @@ from pydantic import BaseModel, Field
 from invokeai.app.api.dependencies import ApiDependencies
 from invokeai.app.services.session_processor.session_processor_common import SessionProcessorStatus
 from invokeai.app.services.session_queue.session_queue_common import (
-    QUEUE_ITEM_STATUS,
     Batch,
     BatchStatus,
     CancelAllExceptCurrentResult,
@@ -18,6 +17,7 @@ from invokeai.app.services.session_queue.session_queue_common import (
     DeleteByDestinationResult,
     EnqueueBatchResult,
     FieldIdentifier,
+    ItemIdsResult,
     PruneResult,
     RetryItemsResult,
     SessionQueueCountsByDestination,
@@ -25,7 +25,7 @@ from invokeai.app.services.session_queue.session_queue_common import (
     SessionQueueItemNotFoundError,
     SessionQueueStatus,
 )
-from invokeai.app.services.shared.pagination import CursorPaginatedResults
+from invokeai.app.services.shared.sqlite.sqlite_common import SQLiteDirection
 
 session_queue_router = APIRouter(prefix="/v1/queue", tags=["queue"])
 
@@ -68,36 +68,6 @@ async def enqueue_batch(
         raise HTTPException(status_code=500, detail=f"Unexpected error while enqueuing batch: {e}")
 
 
-@session_queue_router.get(
-    "/{queue_id}/list",
-    operation_id="list_queue_items",
-    responses={
-        200: {"model": CursorPaginatedResults[SessionQueueItem]},
-    },
-)
-async def list_queue_items(
-    queue_id: str = Path(description="The queue id to perform this operation on"),
-    limit: int = Query(default=50, description="The number of items to fetch"),
-    status: Optional[QUEUE_ITEM_STATUS] = Query(default=None, description="The status of items to fetch"),
-    cursor: Optional[int] = Query(default=None, description="The pagination cursor"),
-    priority: int = Query(default=0, description="The pagination cursor priority"),
-    destination: Optional[str] = Query(default=None, description="The destination of queue items to fetch"),
-) -> CursorPaginatedResults[SessionQueueItem]:
-    """Gets cursor-paginated queue items"""
-
-    try:
-        return ApiDependencies.invoker.services.session_queue.list_queue_items(
-            queue_id=queue_id,
-            limit=limit,
-            status=status,
-            cursor=cursor,
-            priority=priority,
-            destination=destination,
-        )
-    except Exception as e:
-        raise HTTPException(status_code=500, detail=f"Unexpected error while listing all items: {e}")
-
-
 @session_queue_router.get(
     "/{queue_id}/list_all",
     operation_id="list_all_queue_items",
@@ -119,6 +89,56 @@ async def list_all_queue_items(
         raise HTTPException(status_code=500, detail=f"Unexpected error while listing all queue items: {e}")
 
 
+@session_queue_router.get(
+    "/{queue_id}/item_ids",
+    operation_id="get_queue_item_ids",
+    responses={
+        200: {"model": ItemIdsResult},
+    },
+)
+async def get_queue_item_ids(
+    queue_id: str = Path(description="The queue id to perform this operation on"),
+    order_dir: SQLiteDirection = Query(default=SQLiteDirection.Descending, description="The order of sort"),
+) -> ItemIdsResult:
+    """Gets all queue item ids that match the given parameters"""
+    try:
+        return ApiDependencies.invoker.services.session_queue.get_queue_item_ids(queue_id=queue_id, order_dir=order_dir)
+    except Exception as e:
+        raise HTTPException(status_code=500, detail=f"Unexpected error while listing all queue item ids: {e}")
+
+
+@session_queue_router.post(
+    "/{queue_id}/items_by_ids",
+    operation_id="get_queue_items_by_item_ids",
+    responses={200: {"model": list[SessionQueueItem]}},
+)
+async def get_queue_items_by_item_ids(
+    queue_id: str = Path(description="The queue id to perform this operation on"),
+    item_ids: list[int] = Body(
+        embed=True, description="Object containing list of queue item ids to fetch queue items for"
+    ),
+) -> list[SessionQueueItem]:
+    """Gets queue items for the specified queue item ids. Maintains order of item ids."""
+    try:
+        session_queue_service = ApiDependencies.invoker.services.session_queue
+
+        # Fetch queue items preserving the order of requested item ids
+        queue_items: list[SessionQueueItem] = []
+        for item_id in item_ids:
+            try:
+                queue_item = session_queue_service.get_queue_item(item_id=item_id)
+                if queue_item.queue_id != queue_id:  # Auth protection for items from other queues
+                    continue
+                queue_items.append(queue_item)
+            except Exception:
+                # Skip missing queue items - they may have been deleted between item id fetch and queue item fetch
+                continue
+
+        return queue_items
+    except Exception:
+        raise HTTPException(status_code=500, detail="Failed to get queue items")
+
+
 @session_queue_router.put(
     "/{queue_id}/processor/resume",
     operation_id="resume",
@@ -354,7 +374,10 @@ async def get_queue_item(
 ) -> SessionQueueItem:
     """Gets a queue item"""
     try:
-        return ApiDependencies.invoker.services.session_queue.get_queue_item(item_id)
+        queue_item = ApiDependencies.invoker.services.session_queue.get_queue_item(item_id=item_id)
+        if queue_item.queue_id != queue_id:
+            raise HTTPException(status_code=404, detail=f"Queue item with id {item_id} not found in queue {queue_id}")
+        return queue_item
     except SessionQueueItemNotFoundError:
         raise HTTPException(status_code=404, detail=f"Queue item with id {item_id} not found in queue {queue_id}")
     except Exception as e:

invokeai/app/api/routers/videos.py

@@ -0,0 +1,119 @@
+from typing import Optional
+
+from fastapi import Body, HTTPException, Path, Query
+from fastapi.routing import APIRouter
+
+from invokeai.app.services.shared.pagination import OffsetPaginatedResults
+from invokeai.app.services.shared.sqlite.sqlite_common import SQLiteDirection
+from invokeai.app.services.videos_common import (
+    DeleteVideosResult,
+    StarredVideosResult,
+    UnstarredVideosResult,
+    VideoDTO,
+    VideoIdsResult,
+    VideoRecordChanges,
+)
+
+videos_router = APIRouter(prefix="/v1/videos", tags=["videos"])
+
+
+@videos_router.patch(
+    "/i/{video_id}",
+    operation_id="update_video",
+    response_model=VideoDTO,
+)
+async def update_video(
+    video_id: str = Path(description="The id of the video to update"),
+    video_changes: VideoRecordChanges = Body(description="The changes to apply to the video"),
+) -> VideoDTO:
+    """Updates a video"""
+
+    raise HTTPException(status_code=501, detail="Not implemented")
+
+
+@videos_router.get(
+    "/i/{video_id}",
+    operation_id="get_video_dto",
+    response_model=VideoDTO,
+)
+async def get_video_dto(
+    video_id: str = Path(description="The id of the video to get"),
+) -> VideoDTO:
+    """Gets a video's DTO"""
+
+    raise HTTPException(status_code=501, detail="Not implemented")
+
+
+@videos_router.post("/delete", operation_id="delete_videos_from_list", response_model=DeleteVideosResult)
+async def delete_videos_from_list(
+    video_ids: list[str] = Body(description="The list of ids of videos to delete", embed=True),
+) -> DeleteVideosResult:
+    raise HTTPException(status_code=501, detail="Not implemented")
+
+
+@videos_router.post("/star", operation_id="star_videos_in_list", response_model=StarredVideosResult)
+async def star_videos_in_list(
+    video_ids: list[str] = Body(description="The list of ids of videos to star", embed=True),
+) -> StarredVideosResult:
+    raise HTTPException(status_code=501, detail="Not implemented")
+
+
+@videos_router.post("/unstar", operation_id="unstar_videos_in_list", response_model=UnstarredVideosResult)
+async def unstar_videos_in_list(
+    video_ids: list[str] = Body(description="The list of ids of videos to unstar", embed=True),
+) -> UnstarredVideosResult:
+    raise HTTPException(status_code=501, detail="Not implemented")
+
+
+@videos_router.delete("/uncategorized", operation_id="delete_uncategorized_videos", response_model=DeleteVideosResult)
+async def delete_uncategorized_videos() -> DeleteVideosResult:
+    """Deletes all videos that are uncategorized"""
+
+    raise HTTPException(status_code=501, detail="Not implemented")
+
+
+@videos_router.get("/", operation_id="list_video_dtos", response_model=OffsetPaginatedResults[VideoDTO])
+async def list_video_dtos(
+    is_intermediate: Optional[bool] = Query(default=None, description="Whether to list intermediate videos."),
+    board_id: Optional[str] = Query(
+        default=None,
+        description="The board id to filter by. Use 'none' to find videos without a board.",
+    ),
+    offset: int = Query(default=0, description="The page offset"),
+    limit: int = Query(default=10, description="The number of videos per page"),
+    order_dir: SQLiteDirection = Query(default=SQLiteDirection.Descending, description="The order of sort"),
+    starred_first: bool = Query(default=True, description="Whether to sort by starred videos first"),
+    search_term: Optional[str] = Query(default=None, description="The term to search for"),
+) -> OffsetPaginatedResults[VideoDTO]:
+    """Lists video DTOs"""
+
+    raise HTTPException(status_code=501, detail="Not implemented")
+
+
+@videos_router.get("/ids", operation_id="get_video_ids")
+async def get_video_ids(
+    is_intermediate: Optional[bool] = Query(default=None, description="Whether to list intermediate videos."),
+    board_id: Optional[str] = Query(
+        default=None,
+        description="The board id to filter by. Use 'none' to find videos without a board.",
+    ),
+    order_dir: SQLiteDirection = Query(default=SQLiteDirection.Descending, description="The order of sort"),
+    starred_first: bool = Query(default=True, description="Whether to sort by starred videos first"),
+    search_term: Optional[str] = Query(default=None, description="The term to search for"),
+) -> VideoIdsResult:
+    """Gets ordered list of video ids with metadata for optimistic updates"""
+
+    raise HTTPException(status_code=501, detail="Not implemented")
+
+
+@videos_router.post(
+    "/videos_by_ids",
+    operation_id="get_videos_by_ids",
+    responses={200: {"model": list[VideoDTO]}},
+)
+async def get_videos_by_ids(
+    video_ids: list[str] = Body(embed=True, description="Object containing list of video ids to fetch DTOs for"),
+) -> list[VideoDTO]:
+    """Gets video DTOs for the specified video ids. Maintains order of input ids."""
+
+    raise HTTPException(status_code=501, detail="Not implemented")

invokeai/app/api_app.py

@@ -18,7 +18,9 @@ from invokeai.app.api.no_cache_staticfiles import NoCacheStaticFiles
 from invokeai.app.api.routers import (
     app_info,
     board_images,
+    board_videos,
     boards,
+    client_state,
     download_queue,
     images,
     model_manager,
@@ -26,6 +28,7 @@ from invokeai.app.api.routers import (
     session_queue,
     style_presets,
     utilities,
+    videos,
     workflows,
 )
 from invokeai.app.api.sockets import SocketIO
@@ -124,13 +127,16 @@ app.include_router(utilities.utilities_router, prefix="/api")
 app.include_router(model_manager.model_manager_router, prefix="/api")
 app.include_router(download_queue.download_queue_router, prefix="/api")
 app.include_router(images.images_router, prefix="/api")
+app.include_router(videos.videos_router, prefix="/api")
 app.include_router(boards.boards_router, prefix="/api")
 app.include_router(board_images.board_images_router, prefix="/api")
+app.include_router(board_videos.board_videos_router, prefix="/api")
 app.include_router(model_relationships.model_relationships_router, prefix="/api")
 app.include_router(app_info.app_router, prefix="/api")
 app.include_router(session_queue.session_queue_router, prefix="/api")
 app.include_router(workflows.workflows_router, prefix="/api")
 app.include_router(style_presets.style_presets_router, prefix="/api")
+app.include_router(client_state.client_state_router, prefix="/api")
 
 app.openapi = get_openapi_func(app)
 
@@ -155,6 +161,12 @@ def overridden_redoc() -> HTMLResponse:
 
 web_root_path = Path(list(web_dir.__path__)[0])
 
+if app_config.unsafe_disable_picklescan:
+    logger.warning(
+        "The unsafe_disable_picklescan option is enabled. This disables malware scanning while installing and"
+        "loading models, which may allow malicious code to be executed. Use at your own risk."
+    )
+
 try:
     app.mount("/", NoCacheStaticFiles(directory=Path(web_root_path, "dist"), html=True), name="ui")
 except RuntimeError:

invokeai/app/invocations/cogview4_image_to_latents.py

@@ -17,6 +17,7 @@ from invokeai.app.services.shared.invocation_context import InvocationContext
 from invokeai.backend.model_manager.load.load_base import LoadedModel
 from invokeai.backend.stable_diffusion.diffusers_pipeline import image_resized_to_grid_as_tensor
 from invokeai.backend.util.devices import TorchDevice
+from invokeai.backend.util.vae_working_memory import estimate_vae_working_memory_cogview4
 
 # TODO(ryand): This is effectively a copy of SD3ImageToLatentsInvocation and a subset of ImageToLatentsInvocation. We
 # should refactor to avoid this duplication.
@@ -38,7 +39,11 @@ class CogView4ImageToLatentsInvocation(BaseInvocation, WithMetadata, WithBoard):
 
     @staticmethod
     def vae_encode(vae_info: LoadedModel, image_tensor: torch.Tensor) -> torch.Tensor:
-        with vae_info as vae:
+        assert isinstance(vae_info.model, AutoencoderKL)
+        estimated_working_memory = estimate_vae_working_memory_cogview4(
+            operation="encode", image_tensor=image_tensor, vae=vae_info.model
+        )
+        with vae_info.model_on_device(working_mem_bytes=estimated_working_memory) as (_, vae):
             assert isinstance(vae, AutoencoderKL)
 
             vae.disable_tiling()
@@ -62,6 +67,8 @@ class CogView4ImageToLatentsInvocation(BaseInvocation, WithMetadata, WithBoard):
             image_tensor = einops.rearrange(image_tensor, "c h w -> 1 c h w")
 
         vae_info = context.models.load(self.vae.vae)
+        assert isinstance(vae_info.model, AutoencoderKL)
+
         latents = self.vae_encode(vae_info=vae_info, image_tensor=image_tensor)
 
         latents = latents.to("cpu")

invokeai/app/invocations/cogview4_latents_to_image.py

@@ -6,7 +6,6 @@ from einops import rearrange
 from PIL import Image
 
 from invokeai.app.invocations.baseinvocation import BaseInvocation, Classification, invocation
-from invokeai.app.invocations.constants import LATENT_SCALE_FACTOR
 from invokeai.app.invocations.fields import (
     FieldDescriptions,
     Input,
@@ -20,6 +19,7 @@ from invokeai.app.invocations.primitives import ImageOutput
 from invokeai.app.services.shared.invocation_context import InvocationContext
 from invokeai.backend.stable_diffusion.extensions.seamless import SeamlessExt
 from invokeai.backend.util.devices import TorchDevice
+from invokeai.backend.util.vae_working_memory import estimate_vae_working_memory_cogview4
 
 # TODO(ryand): This is effectively a copy of SD3LatentsToImageInvocation and a subset of LatentsToImageInvocation. We
 # should refactor to avoid this duplication.
@@ -39,22 +39,15 @@ class CogView4LatentsToImageInvocation(BaseInvocation, WithMetadata, WithBoard):
     latents: LatentsField = InputField(description=FieldDescriptions.latents, input=Input.Connection)
     vae: VAEField = InputField(description=FieldDescriptions.vae, input=Input.Connection)
 
-    def _estimate_working_memory(self, latents: torch.Tensor, vae: AutoencoderKL) -> int:
-        """Estimate the working memory required by the invocation in bytes."""
-        out_h = LATENT_SCALE_FACTOR * latents.shape[-2]
-        out_w = LATENT_SCALE_FACTOR * latents.shape[-1]
-        element_size = next(vae.parameters()).element_size()
-        scaling_constant = 2200  # Determined experimentally.
-        working_memory = out_h * out_w * element_size * scaling_constant
-        return int(working_memory)
-
     @torch.no_grad()
     def invoke(self, context: InvocationContext) -> ImageOutput:
         latents = context.tensors.load(self.latents.latents_name)
 
         vae_info = context.models.load(self.vae.vae)
         assert isinstance(vae_info.model, (AutoencoderKL))
-        estimated_working_memory = self._estimate_working_memory(latents, vae_info.model)
+        estimated_working_memory = estimate_vae_working_memory_cogview4(
+            operation="decode", image_tensor=latents, vae=vae_info.model
+        )
         with (
             SeamlessExt.static_patch_model(vae_info.model, self.vae.seamless_axes),
             vae_info.model_on_device(working_mem_bytes=estimated_working_memory) as (_, vae),

invokeai/app/invocations/fields.py

@@ -1,11 +1,12 @@
 from enum import Enum
 from typing import Any, Callable, Optional, Tuple
 
-from pydantic import BaseModel, ConfigDict, Field, RootModel, TypeAdapter, model_validator
+from pydantic import BaseModel, ConfigDict, Field, RootModel, TypeAdapter
 from pydantic.fields import _Unset
 from pydantic_core import PydanticUndefined
 
 from invokeai.app.util.metaenum import MetaEnum
+from invokeai.backend.image_util.segment_anything.shared import BoundingBox
 from invokeai.backend.util.logging import InvokeAILogger
 
 logger = InvokeAILogger.get_logger()
@@ -64,12 +65,16 @@ class UIType(str, Enum, metaclass=MetaEnum):
     Imagen3Model = "Imagen3ModelField"
     Imagen4Model = "Imagen4ModelField"
     ChatGPT4oModel = "ChatGPT4oModelField"
+    Gemini2_5Model = "Gemini2_5ModelField"
     FluxKontextModel = "FluxKontextModelField"
+    Veo3Model = "Veo3ModelField"
+    RunwayModel = "RunwayModelField"
     # endregion
 
     # region Misc Field Types
     Scheduler = "SchedulerField"
     Any = "AnyField"
+    Video = "VideoField"
     # endregion
 
     # region Internal Field Types
@@ -224,6 +229,12 @@ class ImageField(BaseModel):
     image_name: str = Field(description="The name of the image")
 
 
+class VideoField(BaseModel):
+    """A video primitive field"""
+
+    video_id: str = Field(description="The id of the video")
+
+
 class BoardField(BaseModel):
     """A board primitive field"""
 
@@ -321,14 +332,9 @@ class ConditioningField(BaseModel):
     )
 
 
-class BoundingBoxField(BaseModel):
+class BoundingBoxField(BoundingBox):
     """A bounding box primitive value."""
 
-    x_min: int = Field(ge=0, description="The minimum x-coordinate of the bounding box (inclusive).")
-    x_max: int = Field(ge=0, description="The maximum x-coordinate of the bounding box (exclusive).")
-    y_min: int = Field(ge=0, description="The minimum y-coordinate of the bounding box (inclusive).")
-    y_max: int = Field(ge=0, description="The maximum y-coordinate of the bounding box (exclusive).")
-
     score: Optional[float] = Field(
         default=None,
         ge=0.0,
@@ -337,21 +343,6 @@ class BoundingBoxField(BaseModel):
         "when the bounding box was produced by a detector and has an associated confidence score.",
     )
 
-    @model_validator(mode="after")
-    def check_coords(self):
-        if self.x_min > self.x_max:
-            raise ValueError(f"x_min ({self.x_min}) is greater than x_max ({self.x_max}).")
-        if self.y_min > self.y_max:
-            raise ValueError(f"y_min ({self.y_min}) is greater than y_max ({self.y_max}).")
-        return self
-
-    def tuple(self) -> Tuple[int, int, int, int]:
-        """
-        Returns the bounding box as a tuple suitable for use with PIL's `Image.crop()` method.
-        This method returns a tuple of the form (left, upper, right, lower) == (x_min, y_min, x_max, y_max).
-        """
-        return (self.x_min, self.y_min, self.x_max, self.y_max)
-
 
 class MetadataField(RootModel[dict[str, Any]]):
     """

invokeai/app/invocations/flux_denoise.py

@@ -63,7 +63,7 @@ from invokeai.backend.util.devices import TorchDevice
     title="FLUX Denoise",
     tags=["image", "flux"],
     category="image",
-    version="4.0.0",
+    version="4.1.0",
 )
 class FluxDenoiseInvocation(BaseInvocation):
     """Run denoising process with a FLUX transformer model."""
@@ -153,7 +153,7 @@ class FluxDenoiseInvocation(BaseInvocation):
         description=FieldDescriptions.ip_adapter, title="IP-Adapter", default=None, input=Input.Connection
     )
 
-    kontext_conditioning: Optional[FluxKontextConditioningField] = InputField(
+    kontext_conditioning: FluxKontextConditioningField | list[FluxKontextConditioningField] | None = InputField(
         default=None,
         description="FLUX Kontext conditioning (reference image).",
         input=Input.Connection,
@@ -328,6 +328,21 @@ class FluxDenoiseInvocation(BaseInvocation):
             cfg_scale_end_step=self.cfg_scale_end_step,
         )
 
+        kontext_extension = None
+        if self.kontext_conditioning:
+            if not self.controlnet_vae:
+                raise ValueError("A VAE (e.g., controlnet_vae) must be provided to use Kontext conditioning.")
+
+            kontext_extension = KontextExtension(
+                context=context,
+                kontext_conditioning=self.kontext_conditioning
+                if isinstance(self.kontext_conditioning, list)
+                else [self.kontext_conditioning],
+                vae_field=self.controlnet_vae,
+                device=TorchDevice.choose_torch_device(),
+                dtype=inference_dtype,
+            )
+
         with ExitStack() as exit_stack:
             # Prepare ControlNet extensions.
             # Note: We do this before loading the transformer model to minimize peak memory (see implementation).
@@ -385,34 +400,22 @@ class FluxDenoiseInvocation(BaseInvocation):
                 dtype=inference_dtype,
             )
 
-            kontext_extension = None
-            if self.kontext_conditioning is not None:
-                if not self.controlnet_vae:
-                    raise ValueError("A VAE (e.g., controlnet_vae) must be provided to use Kontext conditioning.")
-
-                kontext_extension = KontextExtension(
-                    kontext_field=self.kontext_conditioning,
-                    context=context,
-                    vae_field=self.controlnet_vae,
-                    device=TorchDevice.choose_torch_device(),
-                    dtype=inference_dtype,
-                )
-
-            final_img, final_img_ids = x, img_ids
-            original_seq_len = x.shape[1]
+            # Prepare Kontext conditioning if provided
+            img_cond_seq = None
+            img_cond_seq_ids = None
             if kontext_extension is not None:
-                final_img, final_img_ids = kontext_extension.apply(final_img, final_img_ids)
+                # Ensure batch sizes match
+                kontext_extension.ensure_batch_size(x.shape[0])
+                img_cond_seq, img_cond_seq_ids = kontext_extension.kontext_latents, kontext_extension.kontext_ids
 
             x = denoise(
                 model=transformer,
-                img=final_img,
-                img_ids=final_img_ids,
+                img=x,
+                img_ids=img_ids,
                 pos_regional_prompting_extension=pos_regional_prompting_extension,
                 neg_regional_prompting_extension=neg_regional_prompting_extension,
                 timesteps=timesteps,
-                step_callback=self._build_step_callback(
-                    context, original_seq_len if kontext_extension is not None else None
-                ),
+                step_callback=self._build_step_callback(context),
                 guidance=self.guidance,
                 cfg_scale=cfg_scale,
                 inpaint_extension=inpaint_extension,
@@ -420,11 +423,10 @@ class FluxDenoiseInvocation(BaseInvocation):
                 pos_ip_adapter_extensions=pos_ip_adapter_extensions,
                 neg_ip_adapter_extensions=neg_ip_adapter_extensions,
                 img_cond=img_cond,
+                img_cond_seq=img_cond_seq,
+                img_cond_seq_ids=img_cond_seq_ids,
             )
 
-            if kontext_extension is not None:
-                x = x[:, :original_seq_len, :]  # Keep only the first original_seq_len tokens
-
         x = unpack(x.float(), self.height, self.width)
         return x
 
@@ -895,14 +897,11 @@ class FluxDenoiseInvocation(BaseInvocation):
             yield (lora_info.model, lora.weight)
             del lora_info
 
-    def _build_step_callback(
-        self, context: InvocationContext, original_seq_len: Optional[int] = None
-    ) -> Callable[[PipelineIntermediateState], None]:
+    def _build_step_callback(self, context: InvocationContext) -> Callable[[PipelineIntermediateState], None]:
         def step_callback(state: PipelineIntermediateState) -> None:
-            # Extract only main image tokens if Kontext conditioning was applied
+            # The denoise function now handles Kontext conditioning correctly,
+            # so we don't need to slice the latents here
             latents = state.latents.float()
-            if original_seq_len is not None:
-                latents = latents[:, :original_seq_len, :]
             state.latents = unpack(latents, self.height, self.width).squeeze()
             context.util.flux_step_callback(state)
 

invokeai/app/invocations/flux_vae_decode.py

@@ -3,7 +3,6 @@ from einops import rearrange
 from PIL import Image
 
 from invokeai.app.invocations.baseinvocation import BaseInvocation, invocation
-from invokeai.app.invocations.constants import LATENT_SCALE_FACTOR
 from invokeai.app.invocations.fields import (
     FieldDescriptions,
     Input,
@@ -18,6 +17,7 @@ from invokeai.app.services.shared.invocation_context import InvocationContext
 from invokeai.backend.flux.modules.autoencoder import AutoEncoder
 from invokeai.backend.model_manager.load.load_base import LoadedModel
 from invokeai.backend.util.devices import TorchDevice
+from invokeai.backend.util.vae_working_memory import estimate_vae_working_memory_flux
 
 
 @invocation(
@@ -39,17 +39,11 @@ class FluxVaeDecodeInvocation(BaseInvocation, WithMetadata, WithBoard):
         input=Input.Connection,
     )
 
-    def _estimate_working_memory(self, latents: torch.Tensor, vae: AutoEncoder) -> int:
-        """Estimate the working memory required by the invocation in bytes."""
-        out_h = LATENT_SCALE_FACTOR * latents.shape[-2]
-        out_w = LATENT_SCALE_FACTOR * latents.shape[-1]
-        element_size = next(vae.parameters()).element_size()
-        scaling_constant = 2200  # Determined experimentally.
-        working_memory = out_h * out_w * element_size * scaling_constant
-        return int(working_memory)
-
     def _vae_decode(self, vae_info: LoadedModel, latents: torch.Tensor) -> Image.Image:
-        estimated_working_memory = self._estimate_working_memory(latents, vae_info.model)
+        assert isinstance(vae_info.model, AutoEncoder)
+        estimated_working_memory = estimate_vae_working_memory_flux(
+            operation="decode", image_tensor=latents, vae=vae_info.model
+        )
         with vae_info.model_on_device(working_mem_bytes=estimated_working_memory) as (_, vae):
             assert isinstance(vae, AutoEncoder)
             vae_dtype = next(iter(vae.parameters())).dtype

invokeai/app/invocations/flux_vae_encode.py

@@ -15,6 +15,7 @@ from invokeai.backend.flux.modules.autoencoder import AutoEncoder
 from invokeai.backend.model_manager import LoadedModel
 from invokeai.backend.stable_diffusion.diffusers_pipeline import image_resized_to_grid_as_tensor
 from invokeai.backend.util.devices import TorchDevice
+from invokeai.backend.util.vae_working_memory import estimate_vae_working_memory_flux
 
 
 @invocation(
@@ -41,8 +42,12 @@ class FluxVaeEncodeInvocation(BaseInvocation):
         # TODO(ryand): Write a util function for generating random tensors that is consistent across devices / dtypes.
         # There's a starting point in get_noise(...), but it needs to be extracted and generalized. This function
         # should be used for VAE encode sampling.
+        assert isinstance(vae_info.model, AutoEncoder)
+        estimated_working_memory = estimate_vae_working_memory_flux(
+            operation="encode", image_tensor=image_tensor, vae=vae_info.model
+        )
         generator = torch.Generator(device=TorchDevice.choose_torch_device()).manual_seed(0)
-        with vae_info as vae:
+        with vae_info.model_on_device(working_mem_bytes=estimated_working_memory) as (_, vae):
             assert isinstance(vae, AutoEncoder)
             vae_dtype = next(iter(vae.parameters())).dtype
             image_tensor = image_tensor.to(device=TorchDevice.choose_torch_device(), dtype=vae_dtype)

invokeai/app/invocations/image.py

@@ -1347,3 +1347,96 @@ class PasteImageIntoBoundingBoxInvocation(BaseInvocation, WithMetadata, WithBoar
 
         image_dto = context.images.save(image=target_image)
         return ImageOutput.build(image_dto)
+
+
+@invocation(
+    "flux_kontext_image_prep",
+    title="FLUX Kontext Image Prep",
+    tags=["image", "concatenate", "flux", "kontext"],
+    category="image",
+    version="1.0.0",
+)
+class FluxKontextConcatenateImagesInvocation(BaseInvocation, WithMetadata, WithBoard):
+    """Prepares an image or images for use with FLUX Kontext. The first/single image is resized to the nearest
+    preferred Kontext resolution. All other images are concatenated horizontally, maintaining their aspect ratio."""
+
+    images: list[ImageField] = InputField(
+        description="The images to concatenate",
+        min_length=1,
+        max_length=10,
+    )
+
+    use_preferred_resolution: bool = InputField(
+        default=True, description="Use FLUX preferred resolutions for the first image"
+    )
+
+    def invoke(self, context: InvocationContext) -> ImageOutput:
+        from invokeai.backend.flux.util import PREFERED_KONTEXT_RESOLUTIONS
+
+        # Step 1: Load all images
+        pil_images = []
+        for image_field in self.images:
+            image = context.images.get_pil(image_field.image_name, mode="RGBA")
+            pil_images.append(image)
+
+        # Step 2: Determine target resolution for the first image
+        first_image = pil_images[0]
+        width, height = first_image.size
+
+        if self.use_preferred_resolution:
+            aspect_ratio = width / height
+
+            # Find the closest preferred resolution for the first image
+            _, target_width, target_height = min(
+                ((abs(aspect_ratio - w / h), w, h) for w, h in PREFERED_KONTEXT_RESOLUTIONS), key=lambda x: x[0]
+            )
+
+            # Apply BFL's scaling formula
+            scaled_height = 2 * int(target_height / 16)
+            final_height = 8 * scaled_height  # This will be consistent for all images
+            scaled_width = 2 * int(target_width / 16)
+            first_width = 8 * scaled_width
+        else:
+            # Use original dimensions of first image, ensuring divisibility by 16
+            final_height = 16 * (height // 16)
+            first_width = 16 * (width // 16)
+            # Ensure minimum dimensions
+            if final_height < 16:
+                final_height = 16
+            if first_width < 16:
+                first_width = 16
+
+        # Step 3: Process and resize all images with consistent height
+        processed_images = []
+        total_width = 0
+
+        for i, image in enumerate(pil_images):
+            if i == 0:
+                # First image uses the calculated dimensions
+                final_width = first_width
+            else:
+                # Subsequent images maintain aspect ratio with the same height
+                img_aspect_ratio = image.width / image.height
+                # Calculate width that maintains aspect ratio at the target height
+                calculated_width = int(final_height * img_aspect_ratio)
+                # Ensure width is divisible by 16 for proper VAE encoding
+                final_width = 16 * (calculated_width // 16)
+                # Ensure minimum width
+                if final_width < 16:
+                    final_width = 16
+
+            # Resize image to calculated dimensions
+            resized_image = image.resize((final_width, final_height), Image.Resampling.LANCZOS)
+            processed_images.append(resized_image)
+            total_width += final_width
+
+        # Step 4: Concatenate images horizontally
+        concatenated_image = Image.new("RGB", (total_width, final_height))
+        x_offset = 0
+        for img in processed_images:
+            concatenated_image.paste(img, (x_offset, 0))
+            x_offset += img.width
+
+        # Save the concatenated image
+        image_dto = context.images.save(image=concatenated_image)
+        return ImageOutput.build(image_dto)

invokeai/app/invocations/image_to_latents.py

@@ -27,6 +27,7 @@ from invokeai.backend.model_manager import LoadedModel
 from invokeai.backend.stable_diffusion.diffusers_pipeline import image_resized_to_grid_as_tensor
 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
 
 
 @invocation(
@@ -52,11 +53,24 @@ class ImageToLatentsInvocation(BaseInvocation):
     tile_size: int = InputField(default=0, multiple_of=8, description=FieldDescriptions.vae_tile_size)
     fp32: bool = InputField(default=False, description=FieldDescriptions.fp32)
 
-    @staticmethod
+    @classmethod
     def vae_encode(
-        vae_info: LoadedModel, upcast: bool, tiled: bool, image_tensor: torch.Tensor, tile_size: int = 0
+        cls,
+        vae_info: LoadedModel,
+        upcast: bool,
+        tiled: bool,
+        image_tensor: torch.Tensor,
+        tile_size: int = 0,
     ) -> torch.Tensor:
-        with vae_info as vae:
+        assert isinstance(vae_info.model, (AutoencoderKL, AutoencoderTiny))
+        estimated_working_memory = estimate_vae_working_memory_sd15_sdxl(
+            operation="encode",
+            image_tensor=image_tensor,
+            vae=vae_info.model,
+            tile_size=tile_size if tiled else None,
+            fp32=upcast,
+        )
+        with vae_info.model_on_device(working_mem_bytes=estimated_working_memory) as (_, vae):
             assert isinstance(vae, (AutoencoderKL, AutoencoderTiny))
             orig_dtype = vae.dtype
             if upcast:
@@ -113,6 +127,7 @@ 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))
 
         image_tensor = image_resized_to_grid_as_tensor(image.convert("RGB"))
         if image_tensor.dim() == 3:
@@ -120,7 +135,11 @@ class ImageToLatentsInvocation(BaseInvocation):
 
         context.util.signal_progress("Running VAE encoder")
         latents = self.vae_encode(
-            vae_info=vae_info, upcast=self.fp32, tiled=self.tiled, image_tensor=image_tensor, tile_size=self.tile_size
+            vae_info=vae_info,
+            upcast=self.fp32,
+            tiled=self.tiled or context.config.get().force_tiled_decode,
+            image_tensor=image_tensor,
+            tile_size=self.tile_size,
         )
 
         latents = latents.to("cpu")

invokeai/app/invocations/latents_to_image.py

@@ -27,6 +27,7 @@ from invokeai.app.services.shared.invocation_context import InvocationContext
 from invokeai.backend.stable_diffusion.extensions.seamless import SeamlessExt
 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
 
 
 @invocation(
@@ -53,39 +54,6 @@ class LatentsToImageInvocation(BaseInvocation, WithMetadata, WithBoard):
     tile_size: int = InputField(default=0, multiple_of=8, description=FieldDescriptions.vae_tile_size)
     fp32: bool = InputField(default=False, description=FieldDescriptions.fp32)
 
-    def _estimate_working_memory(
-        self, latents: torch.Tensor, use_tiling: bool, vae: AutoencoderKL | AutoencoderTiny
-    ) -> int:
-        """Estimate the working memory required by the invocation in bytes."""
-        # It was found experimentally that the peak working memory scales linearly with the number of pixels and the
-        # element size (precision). This estimate is accurate for both SD1 and SDXL.
-        element_size = 4 if self.fp32 else 2
-        scaling_constant = 2200  # Determined experimentally.
-
-        if use_tiling:
-            tile_size = self.tile_size
-            if tile_size == 0:
-                tile_size = vae.tile_sample_min_size
-                assert isinstance(tile_size, int)
-            out_h = tile_size
-            out_w = tile_size
-            working_memory = out_h * out_w * element_size * scaling_constant
-
-            # We add 25% to the working memory estimate when tiling is enabled to account for factors like tile overlap
-            # and number of tiles. We could make this more precise in the future, but this should be good enough for
-            # most use cases.
-            working_memory = working_memory * 1.25
-        else:
-            out_h = LATENT_SCALE_FACTOR * latents.shape[-2]
-            out_w = LATENT_SCALE_FACTOR * latents.shape[-1]
-            working_memory = out_h * out_w * element_size * scaling_constant
-
-        if self.fp32:
-            # If we are running in FP32, then we should account for the likely increase in model size (~250MB).
-            working_memory += 250 * 2**20
-
-        return int(working_memory)
-
     @torch.no_grad()
     def invoke(self, context: InvocationContext) -> ImageOutput:
         latents = context.tensors.load(self.latents.latents_name)
@@ -94,8 +62,13 @@ class LatentsToImageInvocation(BaseInvocation, WithMetadata, WithBoard):
 
         vae_info = context.models.load(self.vae.vae)
         assert isinstance(vae_info.model, (AutoencoderKL, AutoencoderTiny))
-
-        estimated_working_memory = self._estimate_working_memory(latents, use_tiling, vae_info.model)
+        estimated_working_memory = estimate_vae_working_memory_sd15_sdxl(
+            operation="decode",
+            image_tensor=latents,
+            vae=vae_info.model,
+            tile_size=self.tile_size if use_tiling else None,
+            fp32=self.fp32,
+        )
         with (
             SeamlessExt.static_patch_model(vae_info.model, self.vae.seamless_axes),
             vae_info.model_on_device(working_mem_bytes=estimated_working_memory) as (_, vae),

invokeai/app/invocations/primitives.py

@@ -27,6 +27,7 @@ from invokeai.app.invocations.fields import (
     SD3ConditioningField,
     TensorField,
     UIComponent,
+    VideoField,
 )
 from invokeai.app.services.images.images_common import ImageDTO
 from invokeai.app.services.shared.invocation_context import InvocationContext
@@ -287,6 +288,30 @@ class ImageCollectionInvocation(BaseInvocation):
         return ImageCollectionOutput(collection=self.collection)
 
 
+# endregion
+
+# region Video
+
+
+@invocation_output("video_output")
+class VideoOutput(BaseInvocationOutput):
+    """Base class for nodes that output a video"""
+
+    video: VideoField = OutputField(description="The output video")
+    width: int = OutputField(description="The width of the video in pixels")
+    height: int = OutputField(description="The height of the video in pixels")
+    duration_seconds: float = OutputField(description="The duration of the video in seconds")
+
+    @classmethod
+    def build(cls, video_id: str, width: int, height: int, duration_seconds: float) -> "VideoOutput":
+        return cls(
+            video=VideoField(video_id=video_id),
+            width=width,
+            height=height,
+            duration_seconds=duration_seconds,
+        )
+
+
 # endregion
 
 # region DenoiseMask
@@ -430,6 +455,15 @@ class FluxConditioningOutput(BaseInvocationOutput):
         return cls(conditioning=FluxConditioningField(conditioning_name=conditioning_name))
 
 
+@invocation_output("flux_conditioning_collection_output")
+class FluxConditioningCollectionOutput(BaseInvocationOutput):
+    """Base class for nodes that output a collection of conditioning tensors"""
+
+    collection: list[FluxConditioningField] = OutputField(
+        description="The output conditioning tensors",
+    )
+
+
 @invocation_output("sd3_conditioning_output")
 class SD3ConditioningOutput(BaseInvocationOutput):
     """Base class for nodes that output a single SD3 conditioning tensor"""

invokeai/app/invocations/sd3_image_to_latents.py

@@ -17,6 +17,7 @@ from invokeai.app.services.shared.invocation_context import InvocationContext
 from invokeai.backend.model_manager.load.load_base import LoadedModel
 from invokeai.backend.stable_diffusion.diffusers_pipeline import image_resized_to_grid_as_tensor
 from invokeai.backend.util.devices import TorchDevice
+from invokeai.backend.util.vae_working_memory import estimate_vae_working_memory_sd3
 
 
 @invocation(
@@ -34,7 +35,11 @@ class SD3ImageToLatentsInvocation(BaseInvocation, WithMetadata, WithBoard):
 
     @staticmethod
     def vae_encode(vae_info: LoadedModel, image_tensor: torch.Tensor) -> torch.Tensor:
-        with vae_info as vae:
+        assert isinstance(vae_info.model, AutoencoderKL)
+        estimated_working_memory = estimate_vae_working_memory_sd3(
+            operation="encode", image_tensor=image_tensor, vae=vae_info.model
+        )
+        with vae_info.model_on_device(working_mem_bytes=estimated_working_memory) as (_, vae):
             assert isinstance(vae, AutoencoderKL)
 
             vae.disable_tiling()
@@ -58,6 +63,8 @@ class SD3ImageToLatentsInvocation(BaseInvocation, WithMetadata, WithBoard):
             image_tensor = einops.rearrange(image_tensor, "c h w -> 1 c h w")
 
         vae_info = context.models.load(self.vae.vae)
+        assert isinstance(vae_info.model, AutoencoderKL)
+
         latents = self.vae_encode(vae_info=vae_info, image_tensor=image_tensor)
 
         latents = latents.to("cpu")

invokeai/app/invocations/sd3_latents_to_image.py

@@ -6,7 +6,6 @@ from einops import rearrange
 from PIL import Image
 
 from invokeai.app.invocations.baseinvocation import BaseInvocation, invocation
-from invokeai.app.invocations.constants import LATENT_SCALE_FACTOR
 from invokeai.app.invocations.fields import (
     FieldDescriptions,
     Input,
@@ -20,6 +19,7 @@ from invokeai.app.invocations.primitives import ImageOutput
 from invokeai.app.services.shared.invocation_context import InvocationContext
 from invokeai.backend.stable_diffusion.extensions.seamless import SeamlessExt
 from invokeai.backend.util.devices import TorchDevice
+from invokeai.backend.util.vae_working_memory import estimate_vae_working_memory_sd3
 
 
 @invocation(
@@ -41,22 +41,15 @@ class SD3LatentsToImageInvocation(BaseInvocation, WithMetadata, WithBoard):
         input=Input.Connection,
     )
 
-    def _estimate_working_memory(self, latents: torch.Tensor, vae: AutoencoderKL) -> int:
-        """Estimate the working memory required by the invocation in bytes."""
-        out_h = LATENT_SCALE_FACTOR * latents.shape[-2]
-        out_w = LATENT_SCALE_FACTOR * latents.shape[-1]
-        element_size = next(vae.parameters()).element_size()
-        scaling_constant = 2200  # Determined experimentally.
-        working_memory = out_h * out_w * element_size * scaling_constant
-        return int(working_memory)
-
     @torch.no_grad()
     def invoke(self, context: InvocationContext) -> ImageOutput:
         latents = context.tensors.load(self.latents.latents_name)
 
         vae_info = context.models.load(self.vae.vae)
         assert isinstance(vae_info.model, (AutoencoderKL))
-        estimated_working_memory = self._estimate_working_memory(latents, vae_info.model)
+        estimated_working_memory = estimate_vae_working_memory_sd3(
+            operation="decode", image_tensor=latents, vae=vae_info.model
+        )
         with (
             SeamlessExt.static_patch_model(vae_info.model, self.vae.seamless_axes),
             vae_info.model_on_device(working_mem_bytes=estimated_working_memory) as (_, vae),

invokeai/app/invocations/segment_anything.py

@@ -1,72 +1,75 @@
-from enum import Enum
+from itertools import zip_longest
 from pathlib import Path
 from typing import Literal
 
 import numpy as np
 import torch
 from PIL import Image
-from pydantic import BaseModel, Field
-from transformers import AutoProcessor
+from pydantic import BaseModel, Field, model_validator
 from transformers.models.sam import SamModel
 from transformers.models.sam.processing_sam import SamProcessor
+from transformers.models.sam2 import Sam2Model
+from transformers.models.sam2.processing_sam2 import Sam2Processor
 
 from invokeai.app.invocations.baseinvocation import BaseInvocation, invocation
 from invokeai.app.invocations.fields import BoundingBoxField, ImageField, InputField, TensorField
 from invokeai.app.invocations.primitives import MaskOutput
 from invokeai.app.services.shared.invocation_context import InvocationContext
 from invokeai.backend.image_util.segment_anything.mask_refinement import mask_to_polygon, polygon_to_mask
+from invokeai.backend.image_util.segment_anything.segment_anything_2_pipeline import SegmentAnything2Pipeline
 from invokeai.backend.image_util.segment_anything.segment_anything_pipeline import SegmentAnythingPipeline
+from invokeai.backend.image_util.segment_anything.shared import SAMInput, SAMPoint
 
-SegmentAnythingModelKey = Literal["segment-anything-base", "segment-anything-large", "segment-anything-huge"]
+SegmentAnythingModelKey = Literal[
+    "segment-anything-base",
+    "segment-anything-large",
+    "segment-anything-huge",
+    "segment-anything-2-tiny",
+    "segment-anything-2-small",
+    "segment-anything-2-base",
+    "segment-anything-2-large",
+]
 SEGMENT_ANYTHING_MODEL_IDS: dict[SegmentAnythingModelKey, str] = {
     "segment-anything-base": "facebook/sam-vit-base",
     "segment-anything-large": "facebook/sam-vit-large",
     "segment-anything-huge": "facebook/sam-vit-huge",
+    "segment-anything-2-tiny": "facebook/sam2.1-hiera-tiny",
+    "segment-anything-2-small": "facebook/sam2.1-hiera-small",
+    "segment-anything-2-base": "facebook/sam2.1-hiera-base-plus",
+    "segment-anything-2-large": "facebook/sam2.1-hiera-large",
 }
 
 
-class SAMPointLabel(Enum):
-    negative = -1
-    neutral = 0
-    positive = 1
-
-
-class SAMPoint(BaseModel):
-    x: int = Field(..., description="The x-coordinate of the point")
-    y: int = Field(..., description="The y-coordinate of the point")
-    label: SAMPointLabel = Field(..., description="The label of the point")
-
-
 class SAMPointsField(BaseModel):
-    points: list[SAMPoint] = Field(..., description="The points of the object")
+    points: list[SAMPoint] = Field(..., description="The points of the object", min_length=1)
 
-    def to_list(self) -> list[list[int]]:
+    def to_list(self) -> list[list[float]]:
         return [[point.x, point.y, point.label.value] for point in self.points]
 
 
 @invocation(
     "segment_anything",
     title="Segment Anything",
-    tags=["prompt", "segmentation"],
+    tags=["prompt", "segmentation", "sam", "sam2"],
     category="segmentation",
-    version="1.2.0",
+    version="1.3.0",
 )
 class SegmentAnythingInvocation(BaseInvocation):
-    """Runs a Segment Anything Model."""
+    """Runs a Segment Anything Model (SAM or SAM2)."""
 
     # Reference:
     # - https://arxiv.org/pdf/2304.02643
     # - https://huggingface.co/docs/transformers/v4.43.3/en/model_doc/grounding-dino#grounded-sam
     # - https://github.com/NielsRogge/Transformers-Tutorials/blob/a39f33ac1557b02ebfb191ea7753e332b5ca933f/Grounding%20DINO/GroundingDINO_with_Segment_Anything.ipynb
 
-    model: SegmentAnythingModelKey = InputField(description="The Segment Anything model to use.")
+    model: SegmentAnythingModelKey = InputField(description="The Segment Anything model to use (SAM or SAM2).")
     image: ImageField = InputField(description="The image to segment.")
     bounding_boxes: list[BoundingBoxField] | None = InputField(
-        default=None, description="The bounding boxes to prompt the SAM model with."
+        default=None, description="The bounding boxes to prompt the model with."
     )
     point_lists: list[SAMPointsField] | None = InputField(
         default=None,
-        description="The list of point lists to prompt the SAM model with. Each list of points represents a single object.",
+        description="The list of point lists to prompt the model with. Each list of points represents a single object.",
     )
     apply_polygon_refinement: bool = InputField(
         description="Whether to apply polygon refinement to the masks. This will smooth the edges of the masks slightly and ensure that each mask consists of a single closed polygon (before merging).",
@@ -77,14 +80,18 @@ class SegmentAnythingInvocation(BaseInvocation):
         default="all",
     )
 
+    @model_validator(mode="after")
+    def validate_points_and_boxes_len(self):
+        if self.point_lists is not None and self.bounding_boxes is not None:
+            if len(self.point_lists) != len(self.bounding_boxes):
+                raise ValueError("If both point_lists and bounding_boxes are provided, they must have the same length.")
+        return self
+
     @torch.no_grad()
     def invoke(self, context: InvocationContext) -> MaskOutput:
         # The models expect a 3-channel RGB image.
         image_pil = context.images.get_pil(self.image.image_name, mode="RGB")
 
-        if self.point_lists is not None and self.bounding_boxes is not None:
-            raise ValueError("Only one of point_lists or bounding_box can be provided.")
-
         if (not self.bounding_boxes or len(self.bounding_boxes) == 0) and (
             not self.point_lists or len(self.point_lists) == 0
         ):
@@ -111,26 +118,38 @@ class SegmentAnythingInvocation(BaseInvocation):
             # model, and figure out how to make it work in the pipeline.
             # torch_dtype=TorchDevice.choose_torch_dtype(),
         )
-
-        sam_processor = AutoProcessor.from_pretrained(model_path, local_files_only=True)
-        assert isinstance(sam_processor, SamProcessor)
+        sam_processor = SamProcessor.from_pretrained(model_path, local_files_only=True)
         return SegmentAnythingPipeline(sam_model=sam_model, sam_processor=sam_processor)
 
-    def _segment(self, context: InvocationContext, image: Image.Image) -> list[torch.Tensor]:
-        """Use Segment Anything (SAM) to generate masks given an image + a set of bounding boxes."""
-        # Convert the bounding boxes to the SAM input format.
-        sam_bounding_boxes = (
-            [[bb.x_min, bb.y_min, bb.x_max, bb.y_max] for bb in self.bounding_boxes] if self.bounding_boxes else None
-        )
-        sam_points = [p.to_list() for p in self.point_lists] if self.point_lists else None
+    @staticmethod
+    def _load_sam_2_model(model_path: Path):
+        sam2_model = Sam2Model.from_pretrained(model_path, local_files_only=True)
+        sam2_processor = Sam2Processor.from_pretrained(model_path, local_files_only=True)
+        return SegmentAnything2Pipeline(sam2_model=sam2_model, sam2_processor=sam2_processor)
 
-        with (
-            context.models.load_remote_model(
-                source=SEGMENT_ANYTHING_MODEL_IDS[self.model], loader=SegmentAnythingInvocation._load_sam_model
-            ) as sam_pipeline,
-        ):
-            assert isinstance(sam_pipeline, SegmentAnythingPipeline)
-            masks = sam_pipeline.segment(image=image, bounding_boxes=sam_bounding_boxes, point_lists=sam_points)
+    def _segment(self, context: InvocationContext, image: Image.Image) -> list[torch.Tensor]:
+        """Use Segment Anything (SAM or SAM2) to generate masks given an image + a set of bounding boxes."""
+
+        source = SEGMENT_ANYTHING_MODEL_IDS[self.model]
+        inputs: list[SAMInput] = []
+        for bbox_field, point_field in zip_longest(self.bounding_boxes or [], self.point_lists or [], fillvalue=None):
+            inputs.append(
+                SAMInput(
+                    bounding_box=bbox_field,
+                    points=point_field.points if point_field else None,
+                )
+            )
+
+        if "sam2" in source:
+            loader = SegmentAnythingInvocation._load_sam_2_model
+            with context.models.load_remote_model(source=source, loader=loader) as pipeline:
+                assert isinstance(pipeline, SegmentAnything2Pipeline)
+                masks = pipeline.segment(image=image, inputs=inputs)
+        else:
+            loader = SegmentAnythingInvocation._load_sam_model
+            with context.models.load_remote_model(source=source, loader=loader) as pipeline:
+                assert isinstance(pipeline, SegmentAnythingPipeline)
+                masks = pipeline.segment(image=image, inputs=inputs)
 
         masks = self._process_masks(masks)
         if self.apply_polygon_refinement:

invokeai/app/services/board_image_records/board_image_records_base.py

@@ -49,3 +49,11 @@ class BoardImageRecordStorageBase(ABC):
     ) -> int:
         """Gets the number of images for a board."""
         pass
+
+    @abstractmethod
+    def get_asset_count_for_board(
+        self,
+        board_id: str,
+    ) -> int:
+        """Gets the number of assets for a board."""
+        pass

invokeai/app/services/board_image_records/board_image_records_sqlite.py

@@ -3,6 +3,8 @@ from typing import Optional, cast
 
 from invokeai.app.services.board_image_records.board_image_records_base import BoardImageRecordStorageBase
 from invokeai.app.services.image_records.image_records_common import (
+    ASSETS_CATEGORIES,
+    IMAGE_CATEGORIES,
     ImageCategory,
     ImageRecord,
     deserialize_image_record,
@@ -14,15 +16,14 @@ from invokeai.app.services.shared.sqlite.sqlite_database import SqliteDatabase
 class SqliteBoardImageRecordStorage(BoardImageRecordStorageBase):
     def __init__(self, db: SqliteDatabase) -> None:
         super().__init__()
-        self._conn = db.conn
+        self._db = db
 
     def add_image_to_board(
         self,
         board_id: str,
         image_name: str,
     ) -> None:
-        try:
-            cursor = self._conn.cursor()
+        with self._db.transaction() as cursor:
             cursor.execute(
                 """--sql
                 INSERT INTO board_images (board_id, image_name)
@@ -31,17 +32,12 @@ class SqliteBoardImageRecordStorage(BoardImageRecordStorageBase):
                 """,
                 (board_id, image_name, board_id),
             )
-            self._conn.commit()
-        except sqlite3.Error as e:
-            self._conn.rollback()
-            raise e
 
     def remove_image_from_board(
         self,
         image_name: str,
     ) -> None:
-        try:
-            cursor = self._conn.cursor()
+        with self._db.transaction() as cursor:
             cursor.execute(
                 """--sql
                 DELETE FROM board_images
@@ -49,10 +45,6 @@ class SqliteBoardImageRecordStorage(BoardImageRecordStorageBase):
                 """,
                 (image_name,),
             )
-            self._conn.commit()
-        except sqlite3.Error as e:
-            self._conn.rollback()
-            raise e
 
     def get_images_for_board(
         self,
@@ -60,27 +52,26 @@ class SqliteBoardImageRecordStorage(BoardImageRecordStorageBase):
         offset: int = 0,
         limit: int = 10,
     ) -> OffsetPaginatedResults[ImageRecord]:
-        # TODO: this isn't paginated yet?
-        cursor = self._conn.cursor()
-        cursor.execute(
-            """--sql
-            SELECT images.*
-            FROM board_images
-            INNER JOIN images ON board_images.image_name = images.image_name
-            WHERE board_images.board_id = ?
-            ORDER BY board_images.updated_at DESC;
-            """,
-            (board_id,),
-        )
-        result = cast(list[sqlite3.Row], cursor.fetchall())
-        images = [deserialize_image_record(dict(r)) for r in result]
+        with self._db.transaction() as cursor:
+            cursor.execute(
+                """--sql
+                SELECT images.*
+                FROM board_images
+                INNER JOIN images ON board_images.image_name = images.image_name
+                WHERE board_images.board_id = ?
+                ORDER BY board_images.updated_at DESC;
+                """,
+                (board_id,),
+            )
+            result = cast(list[sqlite3.Row], cursor.fetchall())
+            images = [deserialize_image_record(dict(r)) for r in result]
 
-        cursor.execute(
-            """--sql
-            SELECT COUNT(*) FROM images WHERE 1=1;
-            """
-        )
-        count = cast(int, cursor.fetchone()[0])
+            cursor.execute(
+                """--sql
+                SELECT COUNT(*) FROM images WHERE 1=1;
+                """
+            )
+            count = cast(int, cursor.fetchone()[0])
 
         return OffsetPaginatedResults(items=images, offset=offset, limit=limit, total=count)
 
@@ -90,56 +81,55 @@ class SqliteBoardImageRecordStorage(BoardImageRecordStorageBase):
         categories: list[ImageCategory] | None,
         is_intermediate: bool | None,
     ) -> list[str]:
-        params: list[str | bool] = []
+        with self._db.transaction() as cursor:
+            params: list[str | bool] = []
 
-        # Base query is a join between images and board_images
-        stmt = """
-                SELECT images.image_name
-                FROM images
-                LEFT JOIN board_images ON board_images.image_name = images.image_name
-                WHERE 1=1
-                """
+            # Base query is a join between images and board_images
+            stmt = """
+                    SELECT images.image_name
+                    FROM images
+                    LEFT JOIN board_images ON board_images.image_name = images.image_name
+                    WHERE 1=1
+                    """
 
-        # Handle board_id filter
-        if board_id == "none":
-            stmt += """--sql
-                AND board_images.board_id IS NULL
-                """
-        else:
-            stmt += """--sql
-                AND board_images.board_id = ?
-                """
-            params.append(board_id)
+            # Handle board_id filter
+            if board_id == "none":
+                stmt += """--sql
+                    AND board_images.board_id IS NULL
+                    """
+            else:
+                stmt += """--sql
+                    AND board_images.board_id = ?
+                    """
+                params.append(board_id)
 
-        # Add the category filter
-        if categories is not None:
-            # Convert the enum values to unique list of strings
-            category_strings = [c.value for c in set(categories)]
-            # Create the correct length of placeholders
-            placeholders = ",".join("?" * len(category_strings))
-            stmt += f"""--sql
-                AND images.image_category IN ( {placeholders} )
-                """
+            # Add the category filter
+            if categories is not None:
+                # Convert the enum values to unique list of strings
+                category_strings = [c.value for c in set(categories)]
+                # Create the correct length of placeholders
+                placeholders = ",".join("?" * len(category_strings))
+                stmt += f"""--sql
+                    AND images.image_category IN ( {placeholders} )
+                    """
 
-            # Unpack the included categories into the query params
-            for c in category_strings:
-                params.append(c)
+                # Unpack the included categories into the query params
+                for c in category_strings:
+                    params.append(c)
 
-        # Add the is_intermediate filter
-        if is_intermediate is not None:
-            stmt += """--sql
-                AND images.is_intermediate = ?
-                """
-            params.append(is_intermediate)
+            # Add the is_intermediate filter
+            if is_intermediate is not None:
+                stmt += """--sql
+                    AND images.is_intermediate = ?
+                    """
+                params.append(is_intermediate)
 
-        # Put a ring on it
-        stmt += ";"
+            # Put a ring on it
+            stmt += ";"
 
-        # Execute the query
-        cursor = self._conn.cursor()
-        cursor.execute(stmt, params)
+            cursor.execute(stmt, params)
 
-        result = cast(list[sqlite3.Row], cursor.fetchall())
+            result = cast(list[sqlite3.Row], cursor.fetchall())
         image_names = [r[0] for r in result]
         return image_names
 
@@ -147,31 +137,54 @@ class SqliteBoardImageRecordStorage(BoardImageRecordStorageBase):
         self,
         image_name: str,
     ) -> Optional[str]:
-        cursor = self._conn.cursor()
-        cursor.execute(
-            """--sql
-                SELECT board_id
-                FROM board_images
-                WHERE image_name = ?;
-                """,
-            (image_name,),
-        )
-        result = cursor.fetchone()
+        with self._db.transaction() as cursor:
+            cursor.execute(
+                """--sql
+                    SELECT board_id
+                    FROM board_images
+                    WHERE image_name = ?;
+                    """,
+                (image_name,),
+            )
+            result = cursor.fetchone()
         if result is None:
             return None
         return cast(str, result[0])
 
     def get_image_count_for_board(self, board_id: str) -> int:
-        cursor = self._conn.cursor()
-        cursor.execute(
-            """--sql
-                SELECT COUNT(*)
-                FROM board_images
-                INNER JOIN images ON board_images.image_name = images.image_name
-                WHERE images.is_intermediate = FALSE
-                AND board_images.board_id = ?;
-                """,
-            (board_id,),
-        )
-        count = cast(int, cursor.fetchone()[0])
+        with self._db.transaction() as cursor:
+            # Convert the enum values to unique list of strings
+            category_strings = [c.value for c in set(IMAGE_CATEGORIES)]
+            # Create the correct length of placeholders
+            placeholders = ",".join("?" * len(category_strings))
+            cursor.execute(
+                f"""--sql
+                    SELECT COUNT(*)
+                    FROM board_images
+                    INNER JOIN images ON board_images.image_name = images.image_name
+                    WHERE images.is_intermediate = FALSE AND images.image_category IN ( {placeholders} )
+                    AND board_images.board_id = ?;
+                    """,
+                (*category_strings, board_id),
+            )
+            count = cast(int, cursor.fetchone()[0])
+        return count
+
+    def get_asset_count_for_board(self, board_id: str) -> int:
+        with self._db.transaction() as cursor:
+            # Convert the enum values to unique list of strings
+            category_strings = [c.value for c in set(ASSETS_CATEGORIES)]
+            # Create the correct length of placeholders
+            placeholders = ",".join("?" * len(category_strings))
+            cursor.execute(
+                f"""--sql
+                    SELECT COUNT(*)
+                    FROM board_images
+                    INNER JOIN images ON board_images.image_name = images.image_name
+                    WHERE images.is_intermediate = FALSE AND images.image_category IN ( {placeholders} )
+                    AND board_images.board_id = ?;
+                    """,
+                (*category_strings, board_id),
+            )
+            count = cast(int, cursor.fetchone()[0])
         return count

invokeai/app/services/board_records/board_records_sqlite.py

@@ -20,61 +20,57 @@ from invokeai.app.util.misc import uuid_string
 class SqliteBoardRecordStorage(BoardRecordStorageBase):
     def __init__(self, db: SqliteDatabase) -> None:
         super().__init__()
-        self._conn = db.conn
+        self._db = db
 
     def delete(self, board_id: str) -> None:
-        try:
-            cursor = self._conn.cursor()
-            cursor.execute(
-                """--sql
-                DELETE FROM boards
-                WHERE board_id = ?;
-                """,
-                (board_id,),
-            )
-            self._conn.commit()
-        except Exception as e:
-            self._conn.rollback()
-            raise BoardRecordDeleteException from e
+        with self._db.transaction() as cursor:
+            try:
+                cursor.execute(
+                    """--sql
+                    DELETE FROM boards
+                    WHERE board_id = ?;
+                    """,
+                    (board_id,),
+                )
+            except Exception as e:
+                raise BoardRecordDeleteException from e
 
     def save(
         self,
         board_name: str,
     ) -> BoardRecord:
-        try:
-            board_id = uuid_string()
-            cursor = self._conn.cursor()
-            cursor.execute(
-                """--sql
-                INSERT OR IGNORE INTO boards (board_id, board_name)
-                VALUES (?, ?);
-                """,
-                (board_id, board_name),
-            )
-            self._conn.commit()
-        except sqlite3.Error as e:
-            self._conn.rollback()
-            raise BoardRecordSaveException from e
+        with self._db.transaction() as cursor:
+            try:
+                board_id = uuid_string()
+                cursor.execute(
+                    """--sql
+                    INSERT OR IGNORE INTO boards (board_id, board_name)
+                    VALUES (?, ?);
+                    """,
+                    (board_id, board_name),
+                )
+            except sqlite3.Error as e:
+                raise BoardRecordSaveException from e
         return self.get(board_id)
 
     def get(
         self,
         board_id: str,
     ) -> BoardRecord:
-        try:
-            cursor = self._conn.cursor()
-            cursor.execute(
-                """--sql
-                SELECT *
-                FROM boards
-                WHERE board_id = ?;
-                """,
-                (board_id,),
-            )
+        with self._db.transaction() as cursor:
+            try:
+                cursor.execute(
+                    """--sql
+                    SELECT *
+                    FROM boards
+                    WHERE board_id = ?;
+                    """,
+                    (board_id,),
+                )
 
-            result = cast(Union[sqlite3.Row, None], cursor.fetchone())
-        except sqlite3.Error as e:
-            raise BoardRecordNotFoundException from e
+                result = cast(Union[sqlite3.Row, None], cursor.fetchone())
+            except sqlite3.Error as e:
+                raise BoardRecordNotFoundException from e
         if result is None:
             raise BoardRecordNotFoundException
         return BoardRecord(**dict(result))
@@ -84,45 +80,43 @@ class SqliteBoardRecordStorage(BoardRecordStorageBase):
         board_id: str,
         changes: BoardChanges,
     ) -> BoardRecord:
-        try:
-            cursor = self._conn.cursor()
-            # Change the name of a board
-            if changes.board_name is not None:
-                cursor.execute(
-                    """--sql
-                    UPDATE boards
-                    SET board_name = ?
-                    WHERE board_id = ?;
-                    """,
-                    (changes.board_name, board_id),
-                )
+        with self._db.transaction() as cursor:
+            try:
+                # Change the name of a board
+                if changes.board_name is not None:
+                    cursor.execute(
+                        """--sql
+                        UPDATE boards
+                        SET board_name = ?
+                        WHERE board_id = ?;
+                        """,
+                        (changes.board_name, board_id),
+                    )
 
-            # Change the cover image of a board
-            if changes.cover_image_name is not None:
-                cursor.execute(
-                    """--sql
-                    UPDATE boards
-                    SET cover_image_name = ?
-                    WHERE board_id = ?;
-                    """,
-                    (changes.cover_image_name, board_id),
-                )
+                # Change the cover image of a board
+                if changes.cover_image_name is not None:
+                    cursor.execute(
+                        """--sql
+                        UPDATE boards
+                        SET cover_image_name = ?
+                        WHERE board_id = ?;
+                        """,
+                        (changes.cover_image_name, board_id),
+                    )
 
-            # Change the archived status of a board
-            if changes.archived is not None:
-                cursor.execute(
-                    """--sql
-                    UPDATE boards
-                    SET archived = ?
-                    WHERE board_id = ?;
-                    """,
-                    (changes.archived, board_id),
-                )
+                # Change the archived status of a board
+                if changes.archived is not None:
+                    cursor.execute(
+                        """--sql
+                        UPDATE boards
+                        SET archived = ?
+                        WHERE board_id = ?;
+                        """,
+                        (changes.archived, board_id),
+                    )
 
-            self._conn.commit()
-        except sqlite3.Error as e:
-            self._conn.rollback()
-            raise BoardRecordSaveException from e
+            except sqlite3.Error as e:
+                raise BoardRecordSaveException from e
         return self.get(board_id)
 
     def get_many(
@@ -133,78 +127,77 @@ class SqliteBoardRecordStorage(BoardRecordStorageBase):
         limit: int = 10,
         include_archived: bool = False,
     ) -> OffsetPaginatedResults[BoardRecord]:
-        cursor = self._conn.cursor()
-
-        # Build base query
-        base_query = """
-                SELECT *
-                FROM boards
-                {archived_filter}
-                ORDER BY {order_by} {direction}
-                LIMIT ? OFFSET ?;
-            """
-
-        # Determine archived filter condition
-        archived_filter = "" if include_archived else "WHERE archived = 0"
-
-        final_query = base_query.format(
-            archived_filter=archived_filter, order_by=order_by.value, direction=direction.value
-        )
-
-        # Execute query to fetch boards
-        cursor.execute(final_query, (limit, offset))
-
-        result = cast(list[sqlite3.Row], cursor.fetchall())
-        boards = [deserialize_board_record(dict(r)) for r in result]
-
-        # Determine count query
-        if include_archived:
-            count_query = """
-                    SELECT COUNT(*)
-                    FROM boards;
-                """
-        else:
-            count_query = """
-                    SELECT COUNT(*)
+        with self._db.transaction() as cursor:
+            # Build base query
+            base_query = """
+                    SELECT *
                     FROM boards
-                    WHERE archived = 0;
+                    {archived_filter}
+                    ORDER BY {order_by} {direction}
+                    LIMIT ? OFFSET ?;
                 """
 
-        # Execute count query
-        cursor.execute(count_query)
+            # Determine archived filter condition
+            archived_filter = "" if include_archived else "WHERE archived = 0"
 
-        count = cast(int, cursor.fetchone()[0])
+            final_query = base_query.format(
+                archived_filter=archived_filter, order_by=order_by.value, direction=direction.value
+            )
+
+            # Execute query to fetch boards
+            cursor.execute(final_query, (limit, offset))
+
+            result = cast(list[sqlite3.Row], cursor.fetchall())
+            boards = [deserialize_board_record(dict(r)) for r in result]
+
+            # Determine count query
+            if include_archived:
+                count_query = """
+                        SELECT COUNT(*)
+                        FROM boards;
+                    """
+            else:
+                count_query = """
+                        SELECT COUNT(*)
+                        FROM boards
+                        WHERE archived = 0;
+                    """
+
+            # Execute count query
+            cursor.execute(count_query)
+
+            count = cast(int, cursor.fetchone()[0])
 
         return OffsetPaginatedResults[BoardRecord](items=boards, offset=offset, limit=limit, total=count)
 
     def get_all(
         self, order_by: BoardRecordOrderBy, direction: SQLiteDirection, include_archived: bool = False
     ) -> list[BoardRecord]:
-        cursor = self._conn.cursor()
-        if order_by == BoardRecordOrderBy.Name:
-            base_query = """
-                    SELECT *
-                    FROM boards
-                    {archived_filter}
-                    ORDER BY LOWER(board_name) {direction}
-                """
-        else:
-            base_query = """
-                    SELECT *
-                    FROM boards
-                    {archived_filter}
-                    ORDER BY {order_by} {direction}
-                """
+        with self._db.transaction() as cursor:
+            if order_by == BoardRecordOrderBy.Name:
+                base_query = """
+                        SELECT *
+                        FROM boards
+                        {archived_filter}
+                        ORDER BY LOWER(board_name) {direction}
+                    """
+            else:
+                base_query = """
+                        SELECT *
+                        FROM boards
+                        {archived_filter}
+                        ORDER BY {order_by} {direction}
+                    """
 
-        archived_filter = "" if include_archived else "WHERE archived = 0"
+            archived_filter = "" if include_archived else "WHERE archived = 0"
 
-        final_query = base_query.format(
-            archived_filter=archived_filter, order_by=order_by.value, direction=direction.value
-        )
+            final_query = base_query.format(
+                archived_filter=archived_filter, order_by=order_by.value, direction=direction.value
+            )
 
-        cursor.execute(final_query)
+            cursor.execute(final_query)
 
-        result = cast(list[sqlite3.Row], cursor.fetchall())
+            result = cast(list[sqlite3.Row], cursor.fetchall())
         boards = [deserialize_board_record(dict(r)) for r in result]
 
         return boards

invokeai/app/services/boards/boards_common.py

@@ -12,12 +12,20 @@ class BoardDTO(BoardRecord):
     """The URL of the thumbnail of the most recent image in the board."""
     image_count: int = Field(description="The number of images in the board.")
     """The number of images in the board."""
+    asset_count: int = Field(description="The number of assets in the board.")
+    """The number of assets in the board."""
+    video_count: int = Field(description="The number of videos in the board.")
+    """The number of videos in the board."""
 
 
-def board_record_to_dto(board_record: BoardRecord, cover_image_name: Optional[str], image_count: int) -> BoardDTO:
+def board_record_to_dto(
+    board_record: BoardRecord, cover_image_name: Optional[str], image_count: int, asset_count: int, video_count: int
+) -> BoardDTO:
     """Converts a board record to a board DTO."""
     return BoardDTO(
         **board_record.model_dump(exclude={"cover_image_name"}),
         cover_image_name=cover_image_name,
         image_count=image_count,
+        asset_count=asset_count,
+        video_count=video_count,
     )

invokeai/app/services/boards/boards_default.py

@@ -17,7 +17,7 @@ class BoardService(BoardServiceABC):
         board_name: str,
     ) -> BoardDTO:
         board_record = self.__invoker.services.board_records.save(board_name)
-        return board_record_to_dto(board_record, None, 0)
+        return board_record_to_dto(board_record, None, 0, 0, 0)
 
     def get_dto(self, board_id: str) -> BoardDTO:
         board_record = self.__invoker.services.board_records.get(board_id)
@@ -27,7 +27,9 @@ class BoardService(BoardServiceABC):
         else:
             cover_image_name = None
         image_count = self.__invoker.services.board_image_records.get_image_count_for_board(board_id)
-        return board_record_to_dto(board_record, cover_image_name, image_count)
+        asset_count = self.__invoker.services.board_image_records.get_asset_count_for_board(board_id)
+        video_count = 0  # noop for OSS
+        return board_record_to_dto(board_record, cover_image_name, image_count, asset_count, video_count)
 
     def update(
         self,
@@ -42,7 +44,9 @@ class BoardService(BoardServiceABC):
             cover_image_name = None
 
         image_count = self.__invoker.services.board_image_records.get_image_count_for_board(board_id)
-        return board_record_to_dto(board_record, cover_image_name, image_count)
+        asset_count = self.__invoker.services.board_image_records.get_asset_count_for_board(board_id)
+        video_count = 0  # noop for OSS
+        return board_record_to_dto(board_record, cover_image_name, image_count, asset_count, video_count)
 
     def delete(self, board_id: str) -> None:
         self.__invoker.services.board_records.delete(board_id)
@@ -67,7 +71,9 @@ class BoardService(BoardServiceABC):
                 cover_image_name = None
 
             image_count = self.__invoker.services.board_image_records.get_image_count_for_board(r.board_id)
-            board_dtos.append(board_record_to_dto(r, cover_image_name, image_count))
+            asset_count = self.__invoker.services.board_image_records.get_asset_count_for_board(r.board_id)
+            video_count = 0  # noop for OSS
+            board_dtos.append(board_record_to_dto(r, cover_image_name, image_count, asset_count, video_count))
 
         return OffsetPaginatedResults[BoardDTO](items=board_dtos, offset=offset, limit=limit, total=len(board_dtos))
 
@@ -84,6 +90,8 @@ class BoardService(BoardServiceABC):
                 cover_image_name = None
 
             image_count = self.__invoker.services.board_image_records.get_image_count_for_board(r.board_id)
-            board_dtos.append(board_record_to_dto(r, cover_image_name, image_count))
+            asset_count = self.__invoker.services.board_image_records.get_asset_count_for_board(r.board_id)
+            video_count = 0  # noop for OSS
+            board_dtos.append(board_record_to_dto(r, cover_image_name, image_count, asset_count, video_count))
 
         return board_dtos

invokeai/app/services/bulk_download/bulk_download_default.py

@@ -150,4 +150,15 @@ class BulkDownloadService(BulkDownloadBase):
     def _is_valid_path(self, path: Union[str, Path]) -> bool:
         """Validates the path given for a bulk download."""
         path = path if isinstance(path, Path) else Path(path)
-        return path.exists()
+
+        # Resolve the path to handle any path traversal attempts (e.g., ../)
+        resolved_path = path.resolve()
+
+        # The path may not traverse out of the bulk downloads folder or its subfolders
+        does_not_traverse = resolved_path.parent == self._bulk_downloads_folder.resolve()
+
+        # The path must exist and be a .zip file
+        does_exist = resolved_path.exists()
+        is_zip_file = resolved_path.suffix == ".zip"
+
+        return does_exist and is_zip_file and does_not_traverse

invokeai/app/services/client_state_persistence/client_state_persistence_base.py

@@ -0,0 +1,42 @@
+from abc import ABC, abstractmethod
+
+
+class ClientStatePersistenceABC(ABC):
+    """
+    Base class for client persistence implementations.
+    This class defines the interface for persisting client data.
+    """
+
+    @abstractmethod
+    def set_by_key(self, queue_id: str, key: str, value: str) -> str:
+        """
+        Set a key-value pair for the client.
+
+        Args:
+            key (str): The key to set.
+            value (str): The value to set for the key.
+
+        Returns:
+            str: The value that was set.
+        """
+        pass
+
+    @abstractmethod
+    def get_by_key(self, queue_id: str, key: str) -> str | None:
+        """
+        Get the value for a specific key of the client.
+
+        Args:
+            key (str): The key to retrieve the value for.
+
+        Returns:
+            str | None: The value associated with the key, or None if the key does not exist.
+        """
+        pass
+
+    @abstractmethod
+    def delete(self, queue_id: str) -> None:
+        """
+        Delete all client state.
+        """
+        pass

invokeai/app/services/client_state_persistence/client_state_persistence_sqlite.py

@@ -0,0 +1,65 @@
+import json
+
+from invokeai.app.services.client_state_persistence.client_state_persistence_base import ClientStatePersistenceABC
+from invokeai.app.services.invoker import Invoker
+from invokeai.app.services.shared.sqlite.sqlite_database import SqliteDatabase
+
+
+class ClientStatePersistenceSqlite(ClientStatePersistenceABC):
+    """
+    Base class for client persistence implementations.
+    This class defines the interface for persisting client data.
+    """
+
+    def __init__(self, db: SqliteDatabase) -> None:
+        super().__init__()
+        self._db = db
+        self._default_row_id = 1
+
+    def start(self, invoker: Invoker) -> None:
+        self._invoker = invoker
+
+    def _get(self) -> dict[str, str] | None:
+        with self._db.transaction() as cursor:
+            cursor.execute(
+                f"""
+                SELECT data FROM client_state
+                WHERE id = {self._default_row_id}
+                """
+            )
+            row = cursor.fetchone()
+            if row is None:
+                return None
+            return json.loads(row[0])
+
+    def set_by_key(self, queue_id: str, key: str, value: str) -> str:
+        state = self._get() or {}
+        state.update({key: value})
+
+        with self._db.transaction() as cursor:
+            cursor.execute(
+                f"""
+                INSERT INTO client_state (id, data)
+                VALUES ({self._default_row_id}, ?)
+                ON CONFLICT(id) DO UPDATE
+                  SET data = excluded.data;
+                """,
+                (json.dumps(state),),
+            )
+
+        return value
+
+    def get_by_key(self, queue_id: str, key: str) -> str | None:
+        state = self._get()
+        if state is None:
+            return None
+        return state.get(key, None)
+
+    def delete(self, queue_id: str) -> None:
+        with self._db.transaction() as cursor:
+            cursor.execute(
+                f"""
+                DELETE FROM client_state
+                WHERE id = {self._default_row_id}
+                """
+            )

invokeai/app/services/config/config_default.py

@@ -107,6 +107,7 @@ class InvokeAIAppConfig(BaseSettings):
         hashing_algorithm: Model hashing algorthim for model installs. 'blake3_multi' is best for SSDs. 'blake3_single' is best for spinning disk HDDs. 'random' disables hashing, instead assigning a UUID to models. Useful when using a memory db to reduce model installation time, or if you don't care about storing stable hashes for models. Alternatively, any other hashlib algorithm is accepted, though these are not nearly as performant as blake3.<br>Valid values: `blake3_multi`, `blake3_single`, `random`, `md5`, `sha1`, `sha224`, `sha256`, `sha384`, `sha512`, `blake2b`, `blake2s`, `sha3_224`, `sha3_256`, `sha3_384`, `sha3_512`, `shake_128`, `shake_256`
         remote_api_tokens: List of regular expression and token pairs used when downloading models from URLs. The download URL is tested against the regex, and if it matches, the token is provided in as a Bearer token.
         scan_models_on_startup: Scan the models directory on startup, registering orphaned models. This is typically only used in conjunction with `use_memory_db` for testing purposes.
+        unsafe_disable_picklescan: UNSAFE. Disable the picklescan security check during model installation. Recommended only for development and testing purposes. This will allow arbitrary code execution during model installation, so should never be used in production.
     """
 
     _root: Optional[Path] = PrivateAttr(default=None)
@@ -196,6 +197,7 @@ class InvokeAIAppConfig(BaseSettings):
     hashing_algorithm: HASHING_ALGORITHMS = Field(default="blake3_single",  description="Model hashing algorthim for model installs. 'blake3_multi' is best for SSDs. 'blake3_single' is best for spinning disk HDDs. 'random' disables hashing, instead assigning a UUID to models. Useful when using a memory db to reduce model installation time, or if you don't care about storing stable hashes for models. Alternatively, any other hashlib algorithm is accepted, though these are not nearly as performant as blake3.")
     remote_api_tokens: Optional[list[URLRegexTokenPair]] = Field(default=None, description="List of regular expression and token pairs used when downloading models from URLs. The download URL is tested against the regex, and if it matches, the token is provided in as a Bearer token.")
     scan_models_on_startup:        bool = Field(default=False,              description="Scan the models directory on startup, registering orphaned models. This is typically only used in conjunction with `use_memory_db` for testing purposes.")
+    unsafe_disable_picklescan:     bool = Field(default=False,              description="UNSAFE. Disable the picklescan security check during model installation. Recommended only for development and testing purposes. This will allow arbitrary code execution during model installation, so should never be used in production.")
 
     # fmt: on
 

invokeai/app/services/download/download_default.py

@@ -8,6 +8,7 @@ import time
 import traceback
 from pathlib import Path
 from queue import Empty, PriorityQueue
+from shutil import disk_usage
 from typing import TYPE_CHECKING, Any, Dict, List, Literal, Optional, Set
 
 import requests
@@ -335,6 +336,14 @@ class DownloadQueueService(DownloadQueueServiceBase):
 
         assert job.download_path
 
+        free_space = disk_usage(job.download_path.parent).free
+        GB = 2**30
+        self._logger.debug(f"Download is {job.total_bytes / GB:.2f} GB of {free_space / GB:.2f} GB free.")
+        if free_space < job.total_bytes:
+            raise RuntimeError(
+                f"Free disk space {free_space / GB:.2f} GB is not enough for download of {job.total_bytes / GB:.2f} GB."
+            )
+
         # Don't clobber an existing file. See commit 82c2c85202f88c6d24ff84710f297cfc6ae174af
         # for code that instead resumes an interrupted download.
         if job.download_path.exists():

invokeai/app/services/events/events_common.py

@@ -234,8 +234,8 @@ class QueueItemStatusChangedEvent(QueueItemEventBase):
     error_type: Optional[str] = Field(default=None, description="The error type, if any")
     error_message: Optional[str] = Field(default=None, description="The error message, if any")
     error_traceback: Optional[str] = Field(default=None, description="The error traceback, if any")
-    created_at: Optional[str] = Field(default=None, description="The timestamp when the queue item was created")
-    updated_at: Optional[str] = Field(default=None, description="The timestamp when the queue item was last updated")
+    created_at: str = Field(description="The timestamp when the queue item was created")
+    updated_at: str = Field(description="The timestamp when the queue item was last updated")
     started_at: Optional[str] = Field(default=None, description="The timestamp when the queue item was started")
     completed_at: Optional[str] = Field(default=None, description="The timestamp when the queue item was completed")
     batch_status: BatchStatus = Field(description="The status of the batch")
@@ -258,8 +258,8 @@ class QueueItemStatusChangedEvent(QueueItemEventBase):
             error_type=queue_item.error_type,
             error_message=queue_item.error_message,
             error_traceback=queue_item.error_traceback,
-            created_at=str(queue_item.created_at) if queue_item.created_at else None,
-            updated_at=str(queue_item.updated_at) if queue_item.updated_at else None,
+            created_at=str(queue_item.created_at),
+            updated_at=str(queue_item.updated_at),
             started_at=str(queue_item.started_at) if queue_item.started_at else None,
             completed_at=str(queue_item.completed_at) if queue_item.completed_at else None,
             batch_status=batch_status,

invokeai/app/services/image_records/image_records_common.py

@@ -58,6 +58,15 @@ class ImageCategory(str, Enum, metaclass=MetaEnum):
     """OTHER: The image is some other type of image with a specialized purpose. To be used by external nodes."""
 
 
+IMAGE_CATEGORIES: list[ImageCategory] = [ImageCategory.GENERAL]
+ASSETS_CATEGORIES: list[ImageCategory] = [
+    ImageCategory.CONTROL,
+    ImageCategory.MASK,
+    ImageCategory.USER,
+    ImageCategory.OTHER,
+]
+
+
 class InvalidImageCategoryException(ValueError):
     """Raised when a provided value is not a valid ImageCategory.
 

invokeai/app/services/image_records/image_records_sqlite.py

@@ -24,22 +24,22 @@ from invokeai.app.services.shared.sqlite.sqlite_database import SqliteDatabase
 class SqliteImageRecordStorage(ImageRecordStorageBase):
     def __init__(self, db: SqliteDatabase) -> None:
         super().__init__()
-        self._conn = db.conn
+        self._db = db
 
     def get(self, image_name: str) -> ImageRecord:
-        try:
-            cursor = self._conn.cursor()
-            cursor.execute(
-                f"""--sql
-                SELECT {IMAGE_DTO_COLS} FROM images
-                WHERE image_name = ?;
-                """,
-                (image_name,),
-            )
+        with self._db.transaction() as cursor:
+            try:
+                cursor.execute(
+                    f"""--sql
+                    SELECT {IMAGE_DTO_COLS} FROM images
+                    WHERE image_name = ?;
+                    """,
+                    (image_name,),
+                )
 
-            result = cast(Optional[sqlite3.Row], cursor.fetchone())
-        except sqlite3.Error as e:
-            raise ImageRecordNotFoundException from e
+                result = cast(Optional[sqlite3.Row], cursor.fetchone())
+            except sqlite3.Error as e:
+                raise ImageRecordNotFoundException from e
 
         if not result:
             raise ImageRecordNotFoundException
@@ -47,17 +47,20 @@ class SqliteImageRecordStorage(ImageRecordStorageBase):
         return deserialize_image_record(dict(result))
 
     def get_metadata(self, image_name: str) -> Optional[MetadataField]:
-        try:
-            cursor = self._conn.cursor()
-            cursor.execute(
-                """--sql
-                SELECT metadata FROM images
-                WHERE image_name = ?;
-                """,
-                (image_name,),
-            )
+        with self._db.transaction() as cursor:
+            try:
+                cursor.execute(
+                    """--sql
+                    SELECT metadata FROM images
+                    WHERE image_name = ?;
+                    """,
+                    (image_name,),
+                )
 
-            result = cast(Optional[sqlite3.Row], cursor.fetchone())
+                result = cast(Optional[sqlite3.Row], cursor.fetchone())
+
+            except sqlite3.Error as e:
+                raise ImageRecordNotFoundException from e
 
             if not result:
                 raise ImageRecordNotFoundException
@@ -65,64 +68,60 @@ class SqliteImageRecordStorage(ImageRecordStorageBase):
             as_dict = dict(result)
             metadata_raw = cast(Optional[str], as_dict.get("metadata", None))
             return MetadataFieldValidator.validate_json(metadata_raw) if metadata_raw is not None else None
-        except sqlite3.Error as e:
-            raise ImageRecordNotFoundException from e
 
     def update(
         self,
         image_name: str,
         changes: ImageRecordChanges,
     ) -> None:
-        try:
-            cursor = self._conn.cursor()
-            # Change the category of the image
-            if changes.image_category is not None:
-                cursor.execute(
-                    """--sql
-                    UPDATE images
-                    SET image_category = ?
-                    WHERE image_name = ?;
-                    """,
-                    (changes.image_category, image_name),
-                )
+        with self._db.transaction() as cursor:
+            try:
+                # Change the category of the image
+                if changes.image_category is not None:
+                    cursor.execute(
+                        """--sql
+                        UPDATE images
+                        SET image_category = ?
+                        WHERE image_name = ?;
+                        """,
+                        (changes.image_category, image_name),
+                    )
 
-            # Change the session associated with the image
-            if changes.session_id is not None:
-                cursor.execute(
-                    """--sql
-                    UPDATE images
-                    SET session_id = ?
-                    WHERE image_name = ?;
-                    """,
-                    (changes.session_id, image_name),
-                )
+                # Change the session associated with the image
+                if changes.session_id is not None:
+                    cursor.execute(
+                        """--sql
+                        UPDATE images
+                        SET session_id = ?
+                        WHERE image_name = ?;
+                        """,
+                        (changes.session_id, image_name),
+                    )
 
-            # Change the image's `is_intermediate`` flag
-            if changes.is_intermediate is not None:
-                cursor.execute(
-                    """--sql
-                    UPDATE images
-                    SET is_intermediate = ?
-                    WHERE image_name = ?;
-                    """,
-                    (changes.is_intermediate, image_name),
-                )
+                # Change the image's `is_intermediate`` flag
+                if changes.is_intermediate is not None:
+                    cursor.execute(
+                        """--sql
+                        UPDATE images
+                        SET is_intermediate = ?
+                        WHERE image_name = ?;
+                        """,
+                        (changes.is_intermediate, image_name),
+                    )
 
-            # Change the image's `starred`` state
-            if changes.starred is not None:
-                cursor.execute(
-                    """--sql
-                    UPDATE images
-                    SET starred = ?
-                    WHERE image_name = ?;
-                    """,
-                    (changes.starred, image_name),
-                )
+                # Change the image's `starred`` state
+                if changes.starred is not None:
+                    cursor.execute(
+                        """--sql
+                        UPDATE images
+                        SET starred = ?
+                        WHERE image_name = ?;
+                        """,
+                        (changes.starred, image_name),
+                    )
 
-            self._conn.commit()
-        except sqlite3.Error as e:
-            self._conn.rollback()
-            raise ImageRecordSaveException from e
+            except sqlite3.Error as e:
+                raise ImageRecordSaveException from e
 
     def get_many(
         self,
@@ -136,170 +135,162 @@ class SqliteImageRecordStorage(ImageRecordStorageBase):
         board_id: Optional[str] = None,
         search_term: Optional[str] = None,
     ) -> OffsetPaginatedResults[ImageRecord]:
-        cursor = self._conn.cursor()
-
-        # Manually build two queries - one for the count, one for the records
-        count_query = """--sql
-        SELECT COUNT(*)
-        FROM images
-        LEFT JOIN board_images ON board_images.image_name = images.image_name
-        WHERE 1=1
-        """
-
-        images_query = f"""--sql
-        SELECT {IMAGE_DTO_COLS}
-        FROM images
-        LEFT JOIN board_images ON board_images.image_name = images.image_name
-        WHERE 1=1
-        """
-
-        query_conditions = ""
-        query_params: list[Union[int, str, bool]] = []
-
-        if image_origin is not None:
-            query_conditions += """--sql
-            AND images.image_origin = ?
-            """
-            query_params.append(image_origin.value)
-
-        if categories is not None:
-            # Convert the enum values to unique list of strings
-            category_strings = [c.value for c in set(categories)]
-            # Create the correct length of placeholders
-            placeholders = ",".join("?" * len(category_strings))
-
-            query_conditions += f"""--sql
-            AND images.image_category IN ( {placeholders} )
+        with self._db.transaction() as cursor:
+            # Manually build two queries - one for the count, one for the records
+            count_query = """--sql
+            SELECT COUNT(*)
+            FROM images
+            LEFT JOIN board_images ON board_images.image_name = images.image_name
+            WHERE 1=1
             """
 
-            # Unpack the included categories into the query params
-            for c in category_strings:
-                query_params.append(c)
-
-        if is_intermediate is not None:
-            query_conditions += """--sql
-            AND images.is_intermediate = ?
+            images_query = f"""--sql
+            SELECT {IMAGE_DTO_COLS}
+            FROM images
+            LEFT JOIN board_images ON board_images.image_name = images.image_name
+            WHERE 1=1
             """
 
-            query_params.append(is_intermediate)
+            query_conditions = ""
+            query_params: list[Union[int, str, bool]] = []
 
-        # board_id of "none" is reserved for images without a board
-        if board_id == "none":
-            query_conditions += """--sql
-            AND board_images.board_id IS NULL
-            """
-        elif board_id is not None:
-            query_conditions += """--sql
-            AND board_images.board_id = ?
-            """
-            query_params.append(board_id)
+            if image_origin is not None:
+                query_conditions += """--sql
+                AND images.image_origin = ?
+                """
+                query_params.append(image_origin.value)
 
-        # Search term condition
-        if search_term:
-            query_conditions += """--sql
-            AND (
-                images.metadata LIKE ?
-                OR images.created_at LIKE ?
-            )
-            """
-            query_params.append(f"%{search_term.lower()}%")
-            query_params.append(f"%{search_term.lower()}%")
+            if categories is not None:
+                # Convert the enum values to unique list of strings
+                category_strings = [c.value for c in set(categories)]
+                # Create the correct length of placeholders
+                placeholders = ",".join("?" * len(category_strings))
 
-        if starred_first:
-            query_pagination = f"""--sql
-            ORDER BY images.starred DESC, images.created_at {order_dir.value} LIMIT ? OFFSET ?
-            """
-        else:
-            query_pagination = f"""--sql
-            ORDER BY images.created_at {order_dir.value} LIMIT ? OFFSET ?
-            """
+                query_conditions += f"""--sql
+                AND images.image_category IN ( {placeholders} )
+                """
 
-        # Final images query with pagination
-        images_query += query_conditions + query_pagination + ";"
-        # Add all the parameters
-        images_params = query_params.copy()
-        # Add the pagination parameters
-        images_params.extend([limit, offset])
+                # Unpack the included categories into the query params
+                for c in category_strings:
+                    query_params.append(c)
 
-        # Build the list of images, deserializing each row
-        cursor.execute(images_query, images_params)
-        result = cast(list[sqlite3.Row], cursor.fetchall())
-        images = [deserialize_image_record(dict(r)) for r in result]
+            if is_intermediate is not None:
+                query_conditions += """--sql
+                AND images.is_intermediate = ?
+                """
 
-        # Set up and execute the count query, without pagination
-        count_query += query_conditions + ";"
-        count_params = query_params.copy()
-        cursor.execute(count_query, count_params)
-        count = cast(int, cursor.fetchone()[0])
+                query_params.append(is_intermediate)
+
+            # board_id of "none" is reserved for images without a board
+            if board_id == "none":
+                query_conditions += """--sql
+                AND board_images.board_id IS NULL
+                """
+            elif board_id is not None:
+                query_conditions += """--sql
+                AND board_images.board_id = ?
+                """
+                query_params.append(board_id)
+
+            # Search term condition
+            if search_term:
+                query_conditions += """--sql
+                AND (
+                    images.metadata LIKE ?
+                    OR images.created_at LIKE ?
+                )
+                """
+                query_params.append(f"%{search_term.lower()}%")
+                query_params.append(f"%{search_term.lower()}%")
+
+            if starred_first:
+                query_pagination = f"""--sql
+                ORDER BY images.starred DESC, images.created_at {order_dir.value} LIMIT ? OFFSET ?
+                """
+            else:
+                query_pagination = f"""--sql
+                ORDER BY images.created_at {order_dir.value} LIMIT ? OFFSET ?
+                """
+
+            # Final images query with pagination
+            images_query += query_conditions + query_pagination + ";"
+            # Add all the parameters
+            images_params = query_params.copy()
+            # Add the pagination parameters
+            images_params.extend([limit, offset])
+
+            # Build the list of images, deserializing each row
+            cursor.execute(images_query, images_params)
+            result = cast(list[sqlite3.Row], cursor.fetchall())
+
+            images = [deserialize_image_record(dict(r)) for r in result]
+
+            # Set up and execute the count query, without pagination
+            count_query += query_conditions + ";"
+            count_params = query_params.copy()
+            cursor.execute(count_query, count_params)
+            count = cast(int, cursor.fetchone()[0])
 
         return OffsetPaginatedResults(items=images, offset=offset, limit=limit, total=count)
 
     def delete(self, image_name: str) -> None:
-        try:
-            cursor = self._conn.cursor()
-            cursor.execute(
-                """--sql
-                DELETE FROM images
-                WHERE image_name = ?;
-                """,
-                (image_name,),
-            )
-            self._conn.commit()
-        except sqlite3.Error as e:
-            self._conn.rollback()
-            raise ImageRecordDeleteException from e
+        with self._db.transaction() as cursor:
+            try:
+                cursor.execute(
+                    """--sql
+                    DELETE FROM images
+                    WHERE image_name = ?;
+                    """,
+                    (image_name,),
+                )
+            except sqlite3.Error as e:
+                raise ImageRecordDeleteException from e
 
     def delete_many(self, image_names: list[str]) -> None:
-        try:
-            cursor = self._conn.cursor()
+        with self._db.transaction() as cursor:
+            try:
+                placeholders = ",".join("?" for _ in image_names)
 
-            placeholders = ",".join("?" for _ in image_names)
+                # Construct the SQLite query with the placeholders
+                query = f"DELETE FROM images WHERE image_name IN ({placeholders})"
 
-            # Construct the SQLite query with the placeholders
-            query = f"DELETE FROM images WHERE image_name IN ({placeholders})"
+                # Execute the query with the list of IDs as parameters
+                cursor.execute(query, image_names)
 
-            # Execute the query with the list of IDs as parameters
-            cursor.execute(query, image_names)
-
-            self._conn.commit()
-        except sqlite3.Error as e:
-            self._conn.rollback()
-            raise ImageRecordDeleteException from e
+            except sqlite3.Error as e:
+                raise ImageRecordDeleteException from e
 
     def get_intermediates_count(self) -> int:
-        cursor = self._conn.cursor()
-        cursor.execute(
-            """--sql
-            SELECT COUNT(*) FROM images
-            WHERE is_intermediate = TRUE;
-            """
-        )
-        count = cast(int, cursor.fetchone()[0])
-        self._conn.commit()
+        with self._db.transaction() as cursor:
+            cursor.execute(
+                """--sql
+                SELECT COUNT(*) FROM images
+                WHERE is_intermediate = TRUE;
+                """
+            )
+            count = cast(int, cursor.fetchone()[0])
         return count
 
     def delete_intermediates(self) -> list[str]:
-        try:
-            cursor = self._conn.cursor()
-            cursor.execute(
-                """--sql
-                SELECT image_name FROM images
-                WHERE is_intermediate = TRUE;
-                """
-            )
-            result = cast(list[sqlite3.Row], cursor.fetchall())
-            image_names = [r[0] for r in result]
-            cursor.execute(
-                """--sql
-                DELETE FROM images
-                WHERE is_intermediate = TRUE;
-                """
-            )
-            self._conn.commit()
-            return image_names
-        except sqlite3.Error as e:
-            self._conn.rollback()
-            raise ImageRecordDeleteException from e
+        with self._db.transaction() as cursor:
+            try:
+                cursor.execute(
+                    """--sql
+                    SELECT image_name FROM images
+                    WHERE is_intermediate = TRUE;
+                    """
+                )
+                result = cast(list[sqlite3.Row], cursor.fetchall())
+                image_names = [r[0] for r in result]
+                cursor.execute(
+                    """--sql
+                    DELETE FROM images
+                    WHERE is_intermediate = TRUE;
+                    """
+                )
+            except sqlite3.Error as e:
+                raise ImageRecordDeleteException from e
+        return image_names
 
     def save(
         self,
@@ -315,73 +306,71 @@ class SqliteImageRecordStorage(ImageRecordStorageBase):
         node_id: Optional[str] = None,
         metadata: Optional[str] = None,
     ) -> datetime:
-        try:
-            cursor = self._conn.cursor()
-            cursor.execute(
-                """--sql
-                INSERT OR IGNORE INTO images (
-                    image_name,
-                    image_origin,
-                    image_category,
-                    width,
-                    height,
-                    node_id,
-                    session_id,
-                    metadata,
-                    is_intermediate,
-                    starred,
-                    has_workflow
-                    )
-                VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?);
-                """,
-                (
-                    image_name,
-                    image_origin.value,
-                    image_category.value,
-                    width,
-                    height,
-                    node_id,
-                    session_id,
-                    metadata,
-                    is_intermediate,
-                    starred,
-                    has_workflow,
-                ),
-            )
-            self._conn.commit()
+        with self._db.transaction() as cursor:
+            try:
+                cursor.execute(
+                    """--sql
+                    INSERT OR IGNORE INTO images (
+                        image_name,
+                        image_origin,
+                        image_category,
+                        width,
+                        height,
+                        node_id,
+                        session_id,
+                        metadata,
+                        is_intermediate,
+                        starred,
+                        has_workflow
+                        )
+                    VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?);
+                    """,
+                    (
+                        image_name,
+                        image_origin.value,
+                        image_category.value,
+                        width,
+                        height,
+                        node_id,
+                        session_id,
+                        metadata,
+                        is_intermediate,
+                        starred,
+                        has_workflow,
+                    ),
+                )
 
-            cursor.execute(
-                """--sql
-                SELECT created_at
-                FROM images
-                WHERE image_name = ?;
-                """,
-                (image_name,),
-            )
+                cursor.execute(
+                    """--sql
+                    SELECT created_at
+                    FROM images
+                    WHERE image_name = ?;
+                    """,
+                    (image_name,),
+                )
 
-            created_at = datetime.fromisoformat(cursor.fetchone()[0])
+                created_at = datetime.fromisoformat(cursor.fetchone()[0])
 
-            return created_at
-        except sqlite3.Error as e:
-            self._conn.rollback()
-            raise ImageRecordSaveException from e
+            except sqlite3.Error as e:
+                raise ImageRecordSaveException from e
+        return created_at
 
     def get_most_recent_image_for_board(self, board_id: str) -> Optional[ImageRecord]:
-        cursor = self._conn.cursor()
-        cursor.execute(
-            """--sql
-            SELECT images.*
-            FROM images
-            JOIN board_images ON images.image_name = board_images.image_name
-            WHERE board_images.board_id = ?
-            AND images.is_intermediate = FALSE
-            ORDER BY images.starred DESC, images.created_at DESC
-            LIMIT 1;
-            """,
-            (board_id,),
-        )
+        with self._db.transaction() as cursor:
+            cursor.execute(
+                """--sql
+                SELECT images.*
+                FROM images
+                JOIN board_images ON images.image_name = board_images.image_name
+                WHERE board_images.board_id = ?
+                AND images.is_intermediate = FALSE
+                ORDER BY images.starred DESC, images.created_at DESC
+                LIMIT 1;
+                """,
+                (board_id,),
+            )
 
-        result = cast(Optional[sqlite3.Row], cursor.fetchone())
+            result = cast(Optional[sqlite3.Row], cursor.fetchone())
 
         if result is None:
             return None
@@ -398,85 +387,84 @@ class SqliteImageRecordStorage(ImageRecordStorageBase):
         board_id: Optional[str] = None,
         search_term: Optional[str] = None,
     ) -> ImageNamesResult:
-        cursor = self._conn.cursor()
+        with self._db.transaction() as cursor:
+            # Build query conditions (reused for both starred count and image names queries)
+            query_conditions = ""
+            query_params: list[Union[int, str, bool]] = []
 
-        # Build query conditions (reused for both starred count and image names queries)
-        query_conditions = ""
-        query_params: list[Union[int, str, bool]] = []
+            if image_origin is not None:
+                query_conditions += """--sql
+                AND images.image_origin = ?
+                """
+                query_params.append(image_origin.value)
 
-        if image_origin is not None:
-            query_conditions += """--sql
-            AND images.image_origin = ?
-            """
-            query_params.append(image_origin.value)
+            if categories is not None:
+                category_strings = [c.value for c in set(categories)]
+                placeholders = ",".join("?" * len(category_strings))
+                query_conditions += f"""--sql
+                AND images.image_category IN ( {placeholders} )
+                """
+                for c in category_strings:
+                    query_params.append(c)
 
-        if categories is not None:
-            category_strings = [c.value for c in set(categories)]
-            placeholders = ",".join("?" * len(category_strings))
-            query_conditions += f"""--sql
-            AND images.image_category IN ( {placeholders} )
-            """
-            for c in category_strings:
-                query_params.append(c)
+            if is_intermediate is not None:
+                query_conditions += """--sql
+                AND images.is_intermediate = ?
+                """
+                query_params.append(is_intermediate)
 
-        if is_intermediate is not None:
-            query_conditions += """--sql
-            AND images.is_intermediate = ?
-            """
-            query_params.append(is_intermediate)
+            if board_id == "none":
+                query_conditions += """--sql
+                AND board_images.board_id IS NULL
+                """
+            elif board_id is not None:
+                query_conditions += """--sql
+                AND board_images.board_id = ?
+                """
+                query_params.append(board_id)
 
-        if board_id == "none":
-            query_conditions += """--sql
-            AND board_images.board_id IS NULL
-            """
-        elif board_id is not None:
-            query_conditions += """--sql
-            AND board_images.board_id = ?
-            """
-            query_params.append(board_id)
+            if search_term:
+                query_conditions += """--sql
+                AND (
+                    images.metadata LIKE ?
+                    OR images.created_at LIKE ?
+                )
+                """
+                query_params.append(f"%{search_term.lower()}%")
+                query_params.append(f"%{search_term.lower()}%")
 
-        if search_term:
-            query_conditions += """--sql
-            AND (
-                images.metadata LIKE ?
-                OR images.created_at LIKE ?
-            )
-            """
-            query_params.append(f"%{search_term.lower()}%")
-            query_params.append(f"%{search_term.lower()}%")
+            # Get starred count if starred_first is enabled
+            starred_count = 0
+            if starred_first:
+                starred_count_query = f"""--sql
+                SELECT COUNT(*)
+                FROM images
+                LEFT JOIN board_images ON board_images.image_name = images.image_name
+                WHERE images.starred = TRUE AND (1=1{query_conditions})
+                """
+                cursor.execute(starred_count_query, query_params)
+                starred_count = cast(int, cursor.fetchone()[0])
 
-        # Get starred count if starred_first is enabled
-        starred_count = 0
-        if starred_first:
-            starred_count_query = f"""--sql
-            SELECT COUNT(*)
-            FROM images
-            LEFT JOIN board_images ON board_images.image_name = images.image_name
-            WHERE images.starred = TRUE AND (1=1{query_conditions})
-            """
-            cursor.execute(starred_count_query, query_params)
-            starred_count = cast(int, cursor.fetchone()[0])
+            # Get all image names with proper ordering
+            if starred_first:
+                names_query = f"""--sql
+                SELECT images.image_name
+                FROM images
+                LEFT JOIN board_images ON board_images.image_name = images.image_name
+                WHERE 1=1{query_conditions}
+                ORDER BY images.starred DESC, images.created_at {order_dir.value}
+                """
+            else:
+                names_query = f"""--sql
+                SELECT images.image_name
+                FROM images
+                LEFT JOIN board_images ON board_images.image_name = images.image_name
+                WHERE 1=1{query_conditions}
+                ORDER BY images.created_at {order_dir.value}
+                """
 
-        # Get all image names with proper ordering
-        if starred_first:
-            names_query = f"""--sql
-            SELECT images.image_name
-            FROM images
-            LEFT JOIN board_images ON board_images.image_name = images.image_name
-            WHERE 1=1{query_conditions}
-            ORDER BY images.starred DESC, images.created_at {order_dir.value}
-            """
-        else:
-            names_query = f"""--sql
-            SELECT images.image_name
-            FROM images
-            LEFT JOIN board_images ON board_images.image_name = images.image_name
-            WHERE 1=1{query_conditions}
-            ORDER BY images.created_at {order_dir.value}
-            """
-
-        cursor.execute(names_query, query_params)
-        result = cast(list[sqlite3.Row], cursor.fetchall())
+            cursor.execute(names_query, query_params)
+            result = cast(list[sqlite3.Row], cursor.fetchall())
         image_names = [row[0] for row in result]
 
         return ImageNamesResult(image_names=image_names, starred_count=starred_count, total_count=len(image_names))

invokeai/app/services/invocation_services.py

@@ -17,6 +17,7 @@ if TYPE_CHECKING:
     from invokeai.app.services.board_records.board_records_base import BoardRecordStorageBase
     from invokeai.app.services.boards.boards_base import BoardServiceABC
     from invokeai.app.services.bulk_download.bulk_download_base import BulkDownloadBase
+    from invokeai.app.services.client_state_persistence.client_state_persistence_base import ClientStatePersistenceABC
     from invokeai.app.services.config import InvokeAIAppConfig
     from invokeai.app.services.download import DownloadQueueServiceBase
     from invokeai.app.services.events.events_base import EventServiceBase
@@ -73,6 +74,7 @@ class InvocationServices:
         style_preset_records: "StylePresetRecordsStorageBase",
         style_preset_image_files: "StylePresetImageFileStorageBase",
         workflow_thumbnails: "WorkflowThumbnailServiceBase",
+        client_state_persistence: "ClientStatePersistenceABC",
     ):
         self.board_images = board_images
         self.board_image_records = board_image_records
@@ -102,3 +104,4 @@ class InvocationServices:
         self.style_preset_records = style_preset_records
         self.style_preset_image_files = style_preset_image_files
         self.workflow_thumbnails = workflow_thumbnails
+        self.client_state_persistence = client_state_persistence

invokeai/app/services/model_install/model_install_default.py

@@ -7,7 +7,7 @@ import threading
 import time
 from pathlib import Path
 from queue import Empty, Queue
-from shutil import copyfile, copytree, move, rmtree
+from shutil import move, rmtree
 from tempfile import mkdtemp
 from typing import TYPE_CHECKING, Any, Dict, List, Optional, Tuple, Type, Union
 
@@ -51,6 +51,7 @@ from invokeai.backend.model_manager.metadata import (
 from invokeai.backend.model_manager.metadata.metadata_base import HuggingFaceMetadata
 from invokeai.backend.model_manager.search import ModelSearch
 from invokeai.backend.model_manager.taxonomy import ModelRepoVariant, ModelSourceType
+from invokeai.backend.model_manager.util.lora_metadata_extractor import apply_lora_metadata
 from invokeai.backend.util import InvokeAILogger
 from invokeai.backend.util.catch_sigint import catch_sigint
 from invokeai.backend.util.devices import TorchDevice
@@ -185,13 +186,15 @@ class ModelInstallService(ModelInstallServiceBase):
         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)
+            if Path(model_path).is_file():
+                # Careful! Don't use pathlib.Path(...).with_suffix - it can will strip everything after the first dot.
+                preferred_name = f"{preferred_name}{model_path.suffix}"
 
         dest_path = (
             self.app_config.models_path / info.base.value / info.type.value / (preferred_name or model_path.name)
         )
         try:
-            new_path = self._copy_model(model_path, dest_path)
+            new_path = self._move_model(model_path, dest_path)
         except FileExistsError as excp:
             raise DuplicateModelException(
                 f"A model named {model_path.name} is already installed at {dest_path.as_posix()}"
@@ -616,30 +619,17 @@ class ModelInstallService(ModelInstallServiceBase):
         self.record_store.update_model(key, ModelRecordChanges(path=model.path))
         return model
 
-    def _copy_model(self, old_path: Path, new_path: Path) -> Path:
-        if old_path == new_path:
-            return old_path
-        new_path.parent.mkdir(parents=True, exist_ok=True)
-        if old_path.is_dir():
-            copytree(old_path, new_path)
-        else:
-            copyfile(old_path, new_path)
-        return new_path
-
     def _move_model(self, old_path: Path, new_path: Path) -> Path:
         if old_path == new_path:
             return old_path
 
+        if new_path.exists():
+            raise FileExistsError(f"Cannot move {old_path} to {new_path}: destination already exists")
+
         new_path.parent.mkdir(parents=True, exist_ok=True)
 
-        # if path already exists then we jigger the name to make it unique
-        counter: int = 1
-        while new_path.exists():
-            path = new_path.with_stem(new_path.stem + f"_{counter:02d}")
-            if not path.exists():
-                new_path = path
-            counter += 1
         move(old_path, new_path)
+
         return new_path
 
     def _probe(self, model_path: Path, config: Optional[ModelRecordChanges] = None):
@@ -667,6 +657,10 @@ class ModelInstallService(ModelInstallServiceBase):
 
         info = info or self._probe(model_path, config)
 
+        # Apply LoRA metadata if applicable
+        model_images_path = self.app_config.models_path / "model_images"
+        apply_lora_metadata(info, model_path.resolve(), model_images_path)
+
         model_path = model_path.resolve()
 
         # Models in the Invoke-managed models dir should use relative paths.

invokeai/app/services/model_load/model_load_default.py

@@ -87,9 +87,21 @@ class ModelLoadService(ModelLoadServiceBase):
         def torch_load_file(checkpoint: Path) -> AnyModel:
             scan_result = scan_file_path(checkpoint)
             if scan_result.infected_files != 0:
-                raise Exception(f"The model at {checkpoint} is potentially infected by malware. Aborting load.")
+                if self._app_config.unsafe_disable_picklescan:
+                    self._logger.warning(
+                        f"Model at {checkpoint} is potentially infected by malware, but picklescan is disabled. "
+                        "Proceeding with caution."
+                    )
+                else:
+                    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.")
+                if self._app_config.unsafe_disable_picklescan:
+                    self._logger.warning(
+                        f"Error scanning model at {checkpoint} for malware, but picklescan is disabled. "
+                        "Proceeding with caution."
+                    )
+                else:
+                    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_records/model_records_base.py

@@ -15,6 +15,7 @@ from invokeai.app.util.model_exclude_null import BaseModelExcludeNull
 from invokeai.backend.model_manager.config import (
     AnyModelConfig,
     ControlAdapterDefaultSettings,
+    LoraModelDefaultSettings,
     MainModelDefaultSettings,
 )
 from invokeai.backend.model_manager.taxonomy import (
@@ -83,8 +84,8 @@ class ModelRecordChanges(BaseModelExcludeNull):
     file_size: Optional[int] = Field(description="Size of model file", default=None)
     format: Optional[str] = Field(description="format of model file", default=None)
     trigger_phrases: Optional[set[str]] = Field(description="Set of trigger phrases for this model", default=None)
-    default_settings: Optional[MainModelDefaultSettings | ControlAdapterDefaultSettings] = Field(
-        description="Default settings for this model", default=None
+    default_settings: Optional[MainModelDefaultSettings | LoraModelDefaultSettings | ControlAdapterDefaultSettings] = (
+        Field(description="Default settings for this model", default=None)
     )
 
     # Checkpoint-specific changes

invokeai/app/services/model_records/model_records_sql.py

@@ -78,11 +78,6 @@ class ModelRecordServiceSQL(ModelRecordServiceBase):
         self._db = db
         self._logger = logger
 
-    @property
-    def db(self) -> SqliteDatabase:
-        """Return the underlying database."""
-        return self._db
-
     def add_model(self, config: AnyModelConfig) -> AnyModelConfig:
         """
         Add a model to the database.
@@ -93,38 +88,33 @@ class ModelRecordServiceSQL(ModelRecordServiceBase):
 
         Can raise DuplicateModelException and InvalidModelConfigException exceptions.
         """
-        try:
-            cursor = self._db.conn.cursor()
-            cursor.execute(
-                """--sql
-                INSERT INTO models (
-                    id,
-                    config
-                    )
-                VALUES (?,?);
-                """,
-                (
-                    config.key,
-                    config.model_dump_json(),
-                ),
-            )
-            self._db.conn.commit()
+        with self._db.transaction() as cursor:
+            try:
+                cursor.execute(
+                    """--sql
+                    INSERT INTO models (
+                        id,
+                        config
+                        )
+                    VALUES (?,?);
+                    """,
+                    (
+                        config.key,
+                        config.model_dump_json(),
+                    ),
+                )
 
-        except sqlite3.IntegrityError as e:
-            self._db.conn.rollback()
-            if "UNIQUE constraint failed" in str(e):
-                if "models.path" in str(e):
-                    msg = f"A model with path '{config.path}' is already installed"
-                elif "models.name" in str(e):
-                    msg = f"A model with name='{config.name}', type='{config.type}', base='{config.base}' is already installed"
+            except sqlite3.IntegrityError as e:
+                if "UNIQUE constraint failed" in str(e):
+                    if "models.path" in str(e):
+                        msg = f"A model with path '{config.path}' is already installed"
+                    elif "models.name" in str(e):
+                        msg = f"A model with name='{config.name}', type='{config.type}', base='{config.base}' is already installed"
+                    else:
+                        msg = f"A model with key '{config.key}' is already installed"
+                    raise DuplicateModelException(msg) from e
                 else:
-                    msg = f"A model with key '{config.key}' is already installed"
-                raise DuplicateModelException(msg) from e
-            else:
-                raise e
-        except sqlite3.Error as e:
-            self._db.conn.rollback()
-            raise e
+                    raise e
 
         return self.get_model(config.key)
 
@@ -136,8 +126,7 @@ class ModelRecordServiceSQL(ModelRecordServiceBase):
 
         Can raise an UnknownModelException
         """
-        try:
-            cursor = self._db.conn.cursor()
+        with self._db.transaction() as cursor:
             cursor.execute(
                 """--sql
                 DELETE FROM models
@@ -147,22 +136,17 @@ class ModelRecordServiceSQL(ModelRecordServiceBase):
             )
             if cursor.rowcount == 0:
                 raise UnknownModelException("model not found")
-            self._db.conn.commit()
-        except sqlite3.Error as e:
-            self._db.conn.rollback()
-            raise e
 
     def update_model(self, key: str, changes: ModelRecordChanges) -> AnyModelConfig:
-        record = self.get_model(key)
+        with self._db.transaction() as cursor:
+            record = self.get_model(key)
 
-        # Model configs use pydantic's `validate_assignment`, so each change is validated by pydantic.
-        for field_name in changes.model_fields_set:
-            setattr(record, field_name, getattr(changes, field_name))
+            # Model configs use pydantic's `validate_assignment`, so each change is validated by pydantic.
+            for field_name in changes.model_fields_set:
+                setattr(record, field_name, getattr(changes, field_name))
 
-        json_serialized = record.model_dump_json()
+            json_serialized = record.model_dump_json()
 
-        try:
-            cursor = self._db.conn.cursor()
             cursor.execute(
                 """--sql
                 UPDATE models
@@ -174,10 +158,6 @@ class ModelRecordServiceSQL(ModelRecordServiceBase):
             )
             if cursor.rowcount == 0:
                 raise UnknownModelException("model not found")
-            self._db.conn.commit()
-        except sqlite3.Error as e:
-            self._db.conn.rollback()
-            raise e
 
         return self.get_model(key)
 
@@ -189,30 +169,30 @@ class ModelRecordServiceSQL(ModelRecordServiceBase):
 
         Exceptions: UnknownModelException
         """
-        cursor = self._db.conn.cursor()
-        cursor.execute(
-            """--sql
-            SELECT config, strftime('%s',updated_at) FROM models
-            WHERE id=?;
-            """,
-            (key,),
-        )
-        rows = cursor.fetchone()
+        with self._db.transaction() as cursor:
+            cursor.execute(
+                """--sql
+                SELECT config, strftime('%s',updated_at) FROM models
+                WHERE id=?;
+                """,
+                (key,),
+            )
+            rows = cursor.fetchone()
         if not rows:
             raise UnknownModelException("model not found")
         model = ModelConfigFactory.make_config(json.loads(rows[0]), timestamp=rows[1])
         return model
 
     def get_model_by_hash(self, hash: str) -> AnyModelConfig:
-        cursor = self._db.conn.cursor()
-        cursor.execute(
-            """--sql
-            SELECT config, strftime('%s',updated_at) FROM models
-            WHERE hash=?;
-            """,
-            (hash,),
-        )
-        rows = cursor.fetchone()
+        with self._db.transaction() as cursor:
+            cursor.execute(
+                """--sql
+                SELECT config, strftime('%s',updated_at) FROM models
+                WHERE hash=?;
+                """,
+                (hash,),
+            )
+            rows = cursor.fetchone()
         if not rows:
             raise UnknownModelException("model not found")
         model = ModelConfigFactory.make_config(json.loads(rows[0]), timestamp=rows[1])
@@ -224,15 +204,15 @@ class ModelRecordServiceSQL(ModelRecordServiceBase):
 
         :param key: Unique key for the model to be deleted
         """
-        cursor = self._db.conn.cursor()
-        cursor.execute(
-            """--sql
-            select count(*) FROM models
-            WHERE id=?;
-            """,
-            (key,),
-        )
-        count = cursor.fetchone()[0]
+        with self._db.transaction() as cursor:
+            cursor.execute(
+                """--sql
+                select count(*) FROM models
+                WHERE id=?;
+                """,
+                (key,),
+            )
+            count = cursor.fetchone()[0]
         return count > 0
 
     def search_by_attr(
@@ -255,43 +235,42 @@ class ModelRecordServiceSQL(ModelRecordServiceBase):
         If none of the optional filters are passed, will return all
         models in the database.
         """
+        with self._db.transaction() as cursor:
+            assert isinstance(order_by, ModelRecordOrderBy)
+            ordering = {
+                ModelRecordOrderBy.Default: "type, base, name, format",
+                ModelRecordOrderBy.Type: "type",
+                ModelRecordOrderBy.Base: "base",
+                ModelRecordOrderBy.Name: "name",
+                ModelRecordOrderBy.Format: "format",
+            }
 
-        assert isinstance(order_by, ModelRecordOrderBy)
-        ordering = {
-            ModelRecordOrderBy.Default: "type, base, name, format",
-            ModelRecordOrderBy.Type: "type",
-            ModelRecordOrderBy.Base: "base",
-            ModelRecordOrderBy.Name: "name",
-            ModelRecordOrderBy.Format: "format",
-        }
+            where_clause: list[str] = []
+            bindings: list[str] = []
+            if model_name:
+                where_clause.append("name=?")
+                bindings.append(model_name)
+            if base_model:
+                where_clause.append("base=?")
+                bindings.append(base_model)
+            if model_type:
+                where_clause.append("type=?")
+                bindings.append(model_type)
+            if model_format:
+                where_clause.append("format=?")
+                bindings.append(model_format)
+            where = f"WHERE {' AND '.join(where_clause)}" if where_clause else ""
 
-        where_clause: list[str] = []
-        bindings: list[str] = []
-        if model_name:
-            where_clause.append("name=?")
-            bindings.append(model_name)
-        if base_model:
-            where_clause.append("base=?")
-            bindings.append(base_model)
-        if model_type:
-            where_clause.append("type=?")
-            bindings.append(model_type)
-        if model_format:
-            where_clause.append("format=?")
-            bindings.append(model_format)
-        where = f"WHERE {' AND '.join(where_clause)}" if where_clause else ""
-
-        cursor = self._db.conn.cursor()
-        cursor.execute(
-            f"""--sql
-            SELECT config, strftime('%s',updated_at)
-            FROM models
-            {where}
-            ORDER BY {ordering[order_by]} -- using ? to bind doesn't work here for some reason;
-            """,
-            tuple(bindings),
-        )
-        result = cursor.fetchall()
+            cursor.execute(
+                f"""--sql
+                SELECT config, strftime('%s',updated_at)
+                FROM models
+                {where}
+                ORDER BY {ordering[order_by]} -- using ? to bind doesn't work here for some reason;
+                """,
+                tuple(bindings),
+            )
+            result = cursor.fetchall()
 
         # Parse the model configs.
         results: list[AnyModelConfig] = []
@@ -313,69 +292,68 @@ class ModelRecordServiceSQL(ModelRecordServiceBase):
 
     def search_by_path(self, path: Union[str, Path]) -> List[AnyModelConfig]:
         """Return models with the indicated path."""
-        cursor = self._db.conn.cursor()
-        cursor.execute(
-            """--sql
-            SELECT config, strftime('%s',updated_at) FROM models
-            WHERE path=?;
-            """,
-            (str(path),),
-        )
-        results = [ModelConfigFactory.make_config(json.loads(x[0]), timestamp=x[1]) for x in cursor.fetchall()]
+        with self._db.transaction() as cursor:
+            cursor.execute(
+                """--sql
+                SELECT config, strftime('%s',updated_at) FROM models
+                WHERE path=?;
+                """,
+                (str(path),),
+            )
+            results = [ModelConfigFactory.make_config(json.loads(x[0]), timestamp=x[1]) for x in cursor.fetchall()]
         return results
 
     def search_by_hash(self, hash: str) -> List[AnyModelConfig]:
         """Return models with the indicated hash."""
-        cursor = self._db.conn.cursor()
-        cursor.execute(
-            """--sql
-            SELECT config, strftime('%s',updated_at) FROM models
-            WHERE hash=?;
-            """,
-            (hash,),
-        )
-        results = [ModelConfigFactory.make_config(json.loads(x[0]), timestamp=x[1]) for x in cursor.fetchall()]
+        with self._db.transaction() as cursor:
+            cursor.execute(
+                """--sql
+                SELECT config, strftime('%s',updated_at) FROM models
+                WHERE hash=?;
+                """,
+                (hash,),
+            )
+            results = [ModelConfigFactory.make_config(json.loads(x[0]), timestamp=x[1]) for x in cursor.fetchall()]
         return results
 
     def list_models(
         self, page: int = 0, per_page: int = 10, order_by: ModelRecordOrderBy = ModelRecordOrderBy.Default
     ) -> PaginatedResults[ModelSummary]:
         """Return a paginated summary listing of each model in the database."""
-        assert isinstance(order_by, ModelRecordOrderBy)
-        ordering = {
-            ModelRecordOrderBy.Default: "type, base, name, format",
-            ModelRecordOrderBy.Type: "type",
-            ModelRecordOrderBy.Base: "base",
-            ModelRecordOrderBy.Name: "name",
-            ModelRecordOrderBy.Format: "format",
-        }
+        with self._db.transaction() as cursor:
+            assert isinstance(order_by, ModelRecordOrderBy)
+            ordering = {
+                ModelRecordOrderBy.Default: "type, base, name, format",
+                ModelRecordOrderBy.Type: "type",
+                ModelRecordOrderBy.Base: "base",
+                ModelRecordOrderBy.Name: "name",
+                ModelRecordOrderBy.Format: "format",
+            }
 
-        cursor = self._db.conn.cursor()
+            # Lock so that the database isn't updated while we're doing the two queries.
+            # query1: get the total number of model configs
+            cursor.execute(
+                """--sql
+                select count(*) from models;
+                """,
+                (),
+            )
+            total = int(cursor.fetchone()[0])
 
-        # Lock so that the database isn't updated while we're doing the two queries.
-        # query1: get the total number of model configs
-        cursor.execute(
-            """--sql
-            select count(*) from models;
-            """,
-            (),
-        )
-        total = int(cursor.fetchone()[0])
-
-        # query2: fetch key fields
-        cursor.execute(
-            f"""--sql
-            SELECT config
-            FROM models
-            ORDER BY {ordering[order_by]} -- using ? to bind doesn't work here for some reason
-            LIMIT ?
-            OFFSET ?;
-            """,
-            (
-                per_page,
-                page * per_page,
-            ),
-        )
-        rows = cursor.fetchall()
+            # query2: fetch key fields
+            cursor.execute(
+                f"""--sql
+                SELECT config
+                FROM models
+                ORDER BY {ordering[order_by]} -- using ? to bind doesn't work here for some reason
+                LIMIT ?
+                OFFSET ?;
+                """,
+                (
+                    per_page,
+                    page * per_page,
+                ),
+            )
+            rows = cursor.fetchall()
         items = [ModelSummary.model_validate(dict(x)) for x in rows]
         return PaginatedResults(page=page, pages=ceil(total / per_page), per_page=per_page, total=total, items=items)

invokeai/app/services/model_relationship_records/model_relationship_records_sqlite.py

@@ -1,5 +1,3 @@
-import sqlite3
-
 from invokeai.app.services.model_relationship_records.model_relationship_records_base import (
     ModelRelationshipRecordStorageBase,
 )
@@ -9,58 +7,49 @@ from invokeai.app.services.shared.sqlite.sqlite_database import SqliteDatabase
 class SqliteModelRelationshipRecordStorage(ModelRelationshipRecordStorageBase):
     def __init__(self, db: SqliteDatabase) -> None:
         super().__init__()
-        self._conn = db.conn
+        self._db = db
 
     def add_model_relationship(self, model_key_1: str, model_key_2: str) -> None:
-        if model_key_1 == model_key_2:
-            raise ValueError("Cannot relate a model to itself.")
-        a, b = sorted([model_key_1, model_key_2])
-        try:
-            cursor = self._conn.cursor()
+        with self._db.transaction() as cursor:
+            if model_key_1 == model_key_2:
+                raise ValueError("Cannot relate a model to itself.")
+            a, b = sorted([model_key_1, model_key_2])
             cursor.execute(
                 "INSERT OR IGNORE INTO model_relationships (model_key_1, model_key_2) VALUES (?, ?)",
                 (a, b),
             )
-            self._conn.commit()
-        except sqlite3.Error as e:
-            self._conn.rollback()
-            raise e
 
     def remove_model_relationship(self, model_key_1: str, model_key_2: str) -> None:
-        a, b = sorted([model_key_1, model_key_2])
-        try:
-            cursor = self._conn.cursor()
+        with self._db.transaction() as cursor:
+            a, b = sorted([model_key_1, model_key_2])
             cursor.execute(
                 "DELETE FROM model_relationships WHERE model_key_1 = ? AND model_key_2 = ?",
                 (a, b),
             )
-            self._conn.commit()
-        except sqlite3.Error as e:
-            self._conn.rollback()
-            raise e
 
     def get_related_model_keys(self, model_key: str) -> list[str]:
-        cursor = self._conn.cursor()
-        cursor.execute(
-            """
-            SELECT model_key_2 FROM model_relationships WHERE model_key_1 = ?
-            UNION
-            SELECT model_key_1 FROM model_relationships WHERE model_key_2 = ?
-            """,
-            (model_key, model_key),
-        )
-        return [row[0] for row in cursor.fetchall()]
+        with self._db.transaction() as cursor:
+            cursor.execute(
+                """
+                SELECT model_key_2 FROM model_relationships WHERE model_key_1 = ?
+                UNION
+                SELECT model_key_1 FROM model_relationships WHERE model_key_2 = ?
+                """,
+                (model_key, model_key),
+            )
+            result = [row[0] for row in cursor.fetchall()]
+        return result
 
     def get_related_model_keys_batch(self, model_keys: list[str]) -> list[str]:
-        cursor = self._conn.cursor()
-
-        key_list = ",".join("?" for _ in model_keys)
-        cursor.execute(
-            f"""
-            SELECT model_key_2 FROM model_relationships WHERE model_key_1 IN ({key_list})
-            UNION
-            SELECT model_key_1 FROM model_relationships WHERE model_key_2 IN ({key_list})
-            """,
-            model_keys + model_keys,
-        )
-        return [row[0] for row in cursor.fetchall()]
+        with self._db.transaction() as cursor:
+            key_list = ",".join("?" for _ in model_keys)
+            cursor.execute(
+                f"""
+                SELECT model_key_2 FROM model_relationships WHERE model_key_1 IN ({key_list})
+                UNION
+                SELECT model_key_1 FROM model_relationships WHERE model_key_2 IN ({key_list})
+                """,
+                model_keys + model_keys,
+            )
+            result = [row[0] for row in cursor.fetchall()]
+        return result

invokeai/app/services/session_queue/session_queue_base.py

@@ -2,7 +2,6 @@ from abc import ABC, abstractmethod
 from typing import Any, Coroutine, Optional
 
 from invokeai.app.services.session_queue.session_queue_common import (
-    QUEUE_ITEM_STATUS,
     Batch,
     BatchStatus,
     CancelAllExceptCurrentResult,
@@ -15,6 +14,7 @@ from invokeai.app.services.session_queue.session_queue_common import (
     EnqueueBatchResult,
     IsEmptyResult,
     IsFullResult,
+    ItemIdsResult,
     PruneResult,
     RetryItemsResult,
     SessionQueueCountsByDestination,
@@ -22,7 +22,7 @@ from invokeai.app.services.session_queue.session_queue_common import (
     SessionQueueStatus,
 )
 from invokeai.app.services.shared.graph import GraphExecutionState
-from invokeai.app.services.shared.pagination import CursorPaginatedResults
+from invokeai.app.services.shared.sqlite.sqlite_common import SQLiteDirection
 
 
 class SessionQueueBase(ABC):
@@ -135,19 +135,6 @@ class SessionQueueBase(ABC):
         """Deletes all queue items except in-progress items"""
         pass
 
-    @abstractmethod
-    def list_queue_items(
-        self,
-        queue_id: str,
-        limit: int,
-        priority: int,
-        cursor: Optional[int] = None,
-        status: Optional[QUEUE_ITEM_STATUS] = None,
-        destination: Optional[str] = None,
-    ) -> CursorPaginatedResults[SessionQueueItem]:
-        """Gets a page of session queue items"""
-        pass
-
     @abstractmethod
     def list_all_queue_items(
         self,
@@ -157,9 +144,18 @@ class SessionQueueBase(ABC):
         """Gets all queue items that match the given parameters"""
         pass
 
+    @abstractmethod
+    def get_queue_item_ids(
+        self,
+        queue_id: str,
+        order_dir: SQLiteDirection = SQLiteDirection.Descending,
+    ) -> ItemIdsResult:
+        """Gets all queue item ids that match the given parameters"""
+        pass
+
     @abstractmethod
     def get_queue_item(self, item_id: int) -> SessionQueueItem:
-        """Gets a session queue item by ID"""
+        """Gets a session queue item by ID for a given queue"""
         pass
 
     @abstractmethod

invokeai/app/services/session_queue/session_queue_common.py

@@ -176,6 +176,14 @@ DEFAULT_QUEUE_ID = "default"
 
 QUEUE_ITEM_STATUS = Literal["pending", "in_progress", "completed", "failed", "canceled"]
 
+
+class ItemIdsResult(BaseModel):
+    """Response containing ordered item ids with metadata for optimistic updates."""
+
+    item_ids: list[int] = Field(description="Ordered list of item ids")
+    total_count: int = Field(description="Total number of queue items matching the query")
+
+
 NodeFieldValueValidator = TypeAdapter(list[NodeFieldValue])
 
 

invokeai/app/services/session_queue/session_queue_sqlite.py

@@ -22,6 +22,7 @@ from invokeai.app.services.session_queue.session_queue_common import (
     EnqueueBatchResult,
     IsEmptyResult,
     IsFullResult,
+    ItemIdsResult,
     PruneResult,
     RetryItemsResult,
     SessionQueueCountsByDestination,
@@ -33,7 +34,7 @@ from invokeai.app.services.session_queue.session_queue_common import (
     prepare_values_to_insert,
 )
 from invokeai.app.services.shared.graph import GraphExecutionState
-from invokeai.app.services.shared.pagination import CursorPaginatedResults
+from invokeai.app.services.shared.sqlite.sqlite_common import SQLiteDirection
 from invokeai.app.services.shared.sqlite.sqlite_database import SqliteDatabase
 
 
@@ -50,15 +51,14 @@ class SqliteSessionQueue(SessionQueueBase):
 
     def __init__(self, db: SqliteDatabase) -> None:
         super().__init__()
-        self._conn = db.conn
+        self._db = db
 
     def _set_in_progress_to_canceled(self) -> None:
         """
         Sets all in_progress queue items to canceled. Run on app startup, not associated with any queue.
         This is necessary because the invoker may have been killed while processing a queue item.
         """
-        try:
-            cursor = self._conn.cursor()
+        with self._db.transaction() as cursor:
             cursor.execute(
                 """--sql
                 UPDATE session_queue
@@ -66,87 +66,79 @@ class SqliteSessionQueue(SessionQueueBase):
                 WHERE status = 'in_progress';
                 """
             )
-        except Exception:
-            self._conn.rollback()
-            raise
 
     def _get_current_queue_size(self, queue_id: str) -> int:
         """Gets the current number of pending queue items"""
-        cursor = self._conn.cursor()
-        cursor.execute(
-            """--sql
-            SELECT count(*)
-            FROM session_queue
-            WHERE
-              queue_id = ?
-              AND status = 'pending'
-            """,
-            (queue_id,),
-        )
-        return cast(int, cursor.fetchone()[0])
+        with self._db.transaction() as cursor:
+            cursor.execute(
+                """--sql
+                SELECT count(*)
+                FROM session_queue
+                WHERE
+                queue_id = ?
+                AND status = 'pending'
+                """,
+                (queue_id,),
+            )
+            count = cast(int, cursor.fetchone()[0])
+        return count
 
     def _get_highest_priority(self, queue_id: str) -> int:
         """Gets the highest priority value in the queue"""
-        cursor = self._conn.cursor()
-        cursor.execute(
-            """--sql
-            SELECT MAX(priority)
-            FROM session_queue
-            WHERE
-              queue_id = ?
-              AND status = 'pending'
-            """,
-            (queue_id,),
-        )
-        return cast(Union[int, None], cursor.fetchone()[0]) or 0
+        with self._db.transaction() as cursor:
+            cursor.execute(
+                """--sql
+                SELECT MAX(priority)
+                FROM session_queue
+                WHERE
+                queue_id = ?
+                AND status = 'pending'
+                """,
+                (queue_id,),
+            )
+            priority = cast(Union[int, None], cursor.fetchone()[0]) or 0
+        return priority
 
     async def enqueue_batch(self, queue_id: str, batch: Batch, prepend: bool) -> EnqueueBatchResult:
-        try:
-            # TODO: how does this work in a multi-user scenario?
-            current_queue_size = self._get_current_queue_size(queue_id)
-            max_queue_size = self.__invoker.services.configuration.max_queue_size
-            max_new_queue_items = max_queue_size - current_queue_size
+        current_queue_size = self._get_current_queue_size(queue_id)
+        max_queue_size = self.__invoker.services.configuration.max_queue_size
+        max_new_queue_items = max_queue_size - current_queue_size
 
-            priority = 0
-            if prepend:
-                priority = self._get_highest_priority(queue_id) + 1
+        priority = 0
+        if prepend:
+            priority = self._get_highest_priority(queue_id) + 1
 
-            requested_count = await asyncio.to_thread(
-                calc_session_count,
-                batch=batch,
-            )
-            values_to_insert = await asyncio.to_thread(
-                prepare_values_to_insert,
-                queue_id=queue_id,
-                batch=batch,
-                priority=priority,
-                max_new_queue_items=max_new_queue_items,
-            )
-            enqueued_count = len(values_to_insert)
+        requested_count = await asyncio.to_thread(
+            calc_session_count,
+            batch=batch,
+        )
+        values_to_insert = await asyncio.to_thread(
+            prepare_values_to_insert,
+            queue_id=queue_id,
+            batch=batch,
+            priority=priority,
+            max_new_queue_items=max_new_queue_items,
+        )
+        enqueued_count = len(values_to_insert)
 
-            with self._conn:
-                cursor = self._conn.cursor()
-                cursor.executemany(
-                    """--sql
+        with self._db.transaction() as cursor:
+            cursor.executemany(
+                """--sql
                     INSERT INTO session_queue (queue_id, session, session_id, batch_id, field_values, priority, workflow, origin, destination, retried_from_item_id)
                     VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?)
                     """,
-                    values_to_insert,
-                )
-            with self._conn:
-                cursor = self._conn.cursor()
-                cursor.execute(
-                    """--sql
+                values_to_insert,
+            )
+            cursor.execute(
+                """--sql
                     SELECT item_id
                     FROM session_queue
                     WHERE batch_id = ?
                     ORDER BY item_id DESC;
                     """,
-                    (batch.batch_id,),
-                )
-                item_ids = [row[0] for row in cursor.fetchall()]
-        except Exception:
-            raise
+                (batch.batch_id,),
+            )
+            item_ids = [row[0] for row in cursor.fetchall()]
         enqueue_result = EnqueueBatchResult(
             queue_id=queue_id,
             requested=requested_count,
@@ -159,19 +151,19 @@ class SqliteSessionQueue(SessionQueueBase):
         return enqueue_result
 
     def dequeue(self) -> Optional[SessionQueueItem]:
-        cursor = self._conn.cursor()
-        cursor.execute(
-            """--sql
-            SELECT *
-            FROM session_queue
-            WHERE status = 'pending'
-            ORDER BY
-                priority DESC,
-                item_id ASC
-            LIMIT 1
-            """
-        )
-        result = cast(Union[sqlite3.Row, None], cursor.fetchone())
+        with self._db.transaction() as cursor:
+            cursor.execute(
+                """--sql
+                SELECT *
+                FROM session_queue
+                WHERE status = 'pending'
+                ORDER BY
+                    priority DESC,
+                    item_id ASC
+                LIMIT 1
+                """
+            )
+            result = cast(Union[sqlite3.Row, None], cursor.fetchone())
         if result is None:
             return None
         queue_item = SessionQueueItem.queue_item_from_dict(dict(result))
@@ -179,40 +171,40 @@ class SqliteSessionQueue(SessionQueueBase):
         return queue_item
 
     def get_next(self, queue_id: str) -> Optional[SessionQueueItem]:
-        cursor = self._conn.cursor()
-        cursor.execute(
-            """--sql
-            SELECT *
-            FROM session_queue
-            WHERE
-                queue_id = ?
-                AND status = 'pending'
-            ORDER BY
-                priority DESC,
-                created_at ASC
-            LIMIT 1
-            """,
-            (queue_id,),
-        )
-        result = cast(Union[sqlite3.Row, None], cursor.fetchone())
+        with self._db.transaction() as cursor:
+            cursor.execute(
+                """--sql
+                SELECT *
+                FROM session_queue
+                WHERE
+                    queue_id = ?
+                    AND status = 'pending'
+                ORDER BY
+                    priority DESC,
+                    created_at ASC
+                LIMIT 1
+                """,
+                (queue_id,),
+            )
+            result = cast(Union[sqlite3.Row, None], cursor.fetchone())
         if result is None:
             return None
         return SessionQueueItem.queue_item_from_dict(dict(result))
 
     def get_current(self, queue_id: str) -> Optional[SessionQueueItem]:
-        cursor = self._conn.cursor()
-        cursor.execute(
-            """--sql
-            SELECT *
-            FROM session_queue
-            WHERE
-                queue_id = ?
-                AND status = 'in_progress'
-            LIMIT 1
-            """,
-            (queue_id,),
-        )
-        result = cast(Union[sqlite3.Row, None], cursor.fetchone())
+        with self._db.transaction() as cursor:
+            cursor.execute(
+                """--sql
+                SELECT *
+                FROM session_queue
+                WHERE
+                    queue_id = ?
+                    AND status = 'in_progress'
+                LIMIT 1
+                """,
+                (queue_id,),
+            )
+            result = cast(Union[sqlite3.Row, None], cursor.fetchone())
         if result is None:
             return None
         return SessionQueueItem.queue_item_from_dict(dict(result))
@@ -225,8 +217,7 @@ class SqliteSessionQueue(SessionQueueBase):
         error_message: Optional[str] = None,
         error_traceback: Optional[str] = None,
     ) -> SessionQueueItem:
-        try:
-            cursor = self._conn.cursor()
+        with self._db.transaction() as cursor:
             cursor.execute(
                 """--sql
                 SELECT status FROM session_queue WHERE item_id = ?
@@ -234,12 +225,15 @@ class SqliteSessionQueue(SessionQueueBase):
                 (item_id,),
             )
             row = cursor.fetchone()
-            if row is None:
-                raise SessionQueueItemNotFoundError(f"No queue item with id {item_id}")
-            current_status = row[0]
-            # Only update if not already finished (completed, failed or canceled)
-            if current_status in ("completed", "failed", "canceled"):
-                return self.get_queue_item(item_id)
+        if row is None:
+            raise SessionQueueItemNotFoundError(f"No queue item with id {item_id}")
+        current_status = row[0]
+
+        # Only update if not already finished (completed, failed or canceled)
+        if current_status in ("completed", "failed", "canceled"):
+            return self.get_queue_item(item_id)
+
+        with self._db.transaction() as cursor:
             cursor.execute(
                 """--sql
                 UPDATE session_queue
@@ -248,10 +242,7 @@ class SqliteSessionQueue(SessionQueueBase):
                 """,
                 (status, error_type, error_message, error_traceback, item_id),
             )
-            self._conn.commit()
-        except Exception:
-            self._conn.rollback()
-            raise
+
         queue_item = self.get_queue_item(item_id)
         batch_status = self.get_batch_status(queue_id=queue_item.queue_id, batch_id=queue_item.batch_id)
         queue_status = self.get_queue_status(queue_id=queue_item.queue_id)
@@ -259,35 +250,34 @@ class SqliteSessionQueue(SessionQueueBase):
         return queue_item
 
     def is_empty(self, queue_id: str) -> IsEmptyResult:
-        cursor = self._conn.cursor()
-        cursor.execute(
-            """--sql
-            SELECT count(*)
-            FROM session_queue
-            WHERE queue_id = ?
-            """,
-            (queue_id,),
-        )
-        is_empty = cast(int, cursor.fetchone()[0]) == 0
+        with self._db.transaction() as cursor:
+            cursor.execute(
+                """--sql
+                SELECT count(*)
+                FROM session_queue
+                WHERE queue_id = ?
+                """,
+                (queue_id,),
+            )
+            is_empty = cast(int, cursor.fetchone()[0]) == 0
         return IsEmptyResult(is_empty=is_empty)
 
     def is_full(self, queue_id: str) -> IsFullResult:
-        cursor = self._conn.cursor()
-        cursor.execute(
-            """--sql
-            SELECT count(*)
-            FROM session_queue
-            WHERE queue_id = ?
-            """,
-            (queue_id,),
-        )
-        max_queue_size = self.__invoker.services.configuration.max_queue_size
-        is_full = cast(int, cursor.fetchone()[0]) >= max_queue_size
+        with self._db.transaction() as cursor:
+            cursor.execute(
+                """--sql
+                SELECT count(*)
+                FROM session_queue
+                WHERE queue_id = ?
+                """,
+                (queue_id,),
+            )
+            max_queue_size = self.__invoker.services.configuration.max_queue_size
+            is_full = cast(int, cursor.fetchone()[0]) >= max_queue_size
         return IsFullResult(is_full=is_full)
 
     def clear(self, queue_id: str) -> ClearResult:
-        try:
-            cursor = self._conn.cursor()
+        with self._db.transaction() as cursor:
             cursor.execute(
                 """--sql
                 SELECT COUNT(*)
@@ -305,24 +295,19 @@ class SqliteSessionQueue(SessionQueueBase):
                 """,
                 (queue_id,),
             )
-            self._conn.commit()
-        except Exception:
-            self._conn.rollback()
-            raise
         self.__invoker.services.events.emit_queue_cleared(queue_id)
         return ClearResult(deleted=count)
 
     def prune(self, queue_id: str) -> PruneResult:
-        try:
-            cursor = self._conn.cursor()
+        with self._db.transaction() as cursor:
             where = """--sql
                 WHERE
-                  queue_id = ?
-                  AND (
+                queue_id = ?
+                AND (
                     status = 'completed'
                     OR status = 'failed'
                     OR status = 'canceled'
-                  )
+                )
                 """
             cursor.execute(
                 f"""--sql
@@ -341,10 +326,6 @@ class SqliteSessionQueue(SessionQueueBase):
                 """,
                 (queue_id,),
             )
-            self._conn.commit()
-        except Exception:
-            self._conn.rollback()
-            raise
         return PruneResult(deleted=count)
 
     def cancel_queue_item(self, item_id: int) -> SessionQueueItem:
@@ -357,8 +338,7 @@ class SqliteSessionQueue(SessionQueueBase):
             self.cancel_queue_item(item_id)
         except SessionQueueItemNotFoundError:
             pass
-        try:
-            cursor = self._conn.cursor()
+        with self._db.transaction() as cursor:
             cursor.execute(
                 """--sql
                 DELETE
@@ -367,10 +347,6 @@ class SqliteSessionQueue(SessionQueueBase):
                 """,
                 (item_id,),
             )
-            self._conn.commit()
-        except Exception:
-            self._conn.rollback()
-            raise
 
     def complete_queue_item(self, item_id: int) -> SessionQueueItem:
         queue_item = self._set_queue_item_status(item_id=item_id, status="completed")
@@ -393,8 +369,7 @@ class SqliteSessionQueue(SessionQueueBase):
         return queue_item
 
     def cancel_by_batch_ids(self, queue_id: str, batch_ids: list[str]) -> CancelByBatchIDsResult:
-        try:
-            cursor = self._conn.cursor()
+        with self._db.transaction() as cursor:
             current_queue_item = self.get_current(queue_id)
             placeholders = ", ".join(["?" for _ in batch_ids])
             where = f"""--sql
@@ -404,6 +379,8 @@ class SqliteSessionQueue(SessionQueueBase):
                   AND status != 'canceled'
                   AND status != 'completed'
                   AND status != 'failed'
+                  -- We will cancel the current item separately below - skip it here
+                  AND status != 'in_progress'
                 """
             params = [queue_id] + batch_ids
             cursor.execute(
@@ -423,17 +400,14 @@ class SqliteSessionQueue(SessionQueueBase):
                 """,
                 tuple(params),
             )
-            self._conn.commit()
-            if current_queue_item is not None and current_queue_item.batch_id in batch_ids:
-                self._set_queue_item_status(current_queue_item.item_id, "canceled")
-        except Exception:
-            self._conn.rollback()
-            raise
+
+        if current_queue_item is not None and current_queue_item.batch_id in batch_ids:
+            self._set_queue_item_status(current_queue_item.item_id, "canceled")
+
         return CancelByBatchIDsResult(canceled=count)
 
     def cancel_by_destination(self, queue_id: str, destination: str) -> CancelByDestinationResult:
-        try:
-            cursor = self._conn.cursor()
+        with self._db.transaction() as cursor:
             current_queue_item = self.get_current(queue_id)
             where = """--sql
                 WHERE
@@ -442,6 +416,8 @@ class SqliteSessionQueue(SessionQueueBase):
                   AND status != 'canceled'
                   AND status != 'completed'
                   AND status != 'failed'
+                  -- We will cancel the current item separately below - skip it here
+                  AND status != 'in_progress'
                 """
             params = (queue_id, destination)
             cursor.execute(
@@ -461,17 +437,12 @@ class SqliteSessionQueue(SessionQueueBase):
                 """,
                 params,
             )
-            self._conn.commit()
-            if current_queue_item is not None and current_queue_item.destination == destination:
-                self._set_queue_item_status(current_queue_item.item_id, "canceled")
-        except Exception:
-            self._conn.rollback()
-            raise
+        if current_queue_item is not None and current_queue_item.destination == destination:
+            self._set_queue_item_status(current_queue_item.item_id, "canceled")
         return CancelByDestinationResult(canceled=count)
 
     def delete_by_destination(self, queue_id: str, destination: str) -> DeleteByDestinationResult:
-        try:
-            cursor = self._conn.cursor()
+        with self._db.transaction() as cursor:
             current_queue_item = self.get_current(queue_id)
             if current_queue_item is not None and current_queue_item.destination == destination:
                 self.cancel_queue_item(current_queue_item.item_id)
@@ -497,15 +468,10 @@ class SqliteSessionQueue(SessionQueueBase):
                 """,
                 params,
             )
-            self._conn.commit()
-        except Exception:
-            self._conn.rollback()
-            raise
         return DeleteByDestinationResult(deleted=count)
 
     def delete_all_except_current(self, queue_id: str) -> DeleteAllExceptCurrentResult:
-        try:
-            cursor = self._conn.cursor()
+        with self._db.transaction() as cursor:
             where = """--sql
                 WHERE
                   queue_id == ?
@@ -528,15 +494,10 @@ class SqliteSessionQueue(SessionQueueBase):
                 """,
                 (queue_id,),
             )
-            self._conn.commit()
-        except Exception:
-            self._conn.rollback()
-            raise
         return DeleteAllExceptCurrentResult(deleted=count)
 
     def cancel_by_queue_id(self, queue_id: str) -> CancelByQueueIDResult:
-        try:
-            cursor = self._conn.cursor()
+        with self._db.transaction() as cursor:
             current_queue_item = self.get_current(queue_id)
             where = """--sql
                 WHERE
@@ -544,6 +505,8 @@ class SqliteSessionQueue(SessionQueueBase):
                   AND status != 'canceled'
                   AND status != 'completed'
                   AND status != 'failed'
+                  -- We will cancel the current item separately below - skip it here
+                  AND status != 'in_progress'
                 """
             params = [queue_id]
             cursor.execute(
@@ -563,21 +526,13 @@ class SqliteSessionQueue(SessionQueueBase):
                 """,
                 tuple(params),
             )
-            self._conn.commit()
-            if current_queue_item is not None and current_queue_item.queue_id == queue_id:
-                batch_status = self.get_batch_status(queue_id=queue_id, batch_id=current_queue_item.batch_id)
-                queue_status = self.get_queue_status(queue_id=queue_id)
-                self.__invoker.services.events.emit_queue_item_status_changed(
-                    current_queue_item, batch_status, queue_status
-                )
-        except Exception:
-            self._conn.rollback()
-            raise
+
+        if current_queue_item is not None and current_queue_item.queue_id == queue_id:
+            self._set_queue_item_status(current_queue_item.item_id, "canceled")
         return CancelByQueueIDResult(canceled=count)
 
     def cancel_all_except_current(self, queue_id: str) -> CancelAllExceptCurrentResult:
-        try:
-            cursor = self._conn.cursor()
+        with self._db.transaction() as cursor:
             where = """--sql
                 WHERE
                   queue_id == ?
@@ -600,30 +555,25 @@ class SqliteSessionQueue(SessionQueueBase):
                 """,
                 (queue_id,),
             )
-            self._conn.commit()
-        except Exception:
-            self._conn.rollback()
-            raise
         return CancelAllExceptCurrentResult(canceled=count)
 
     def get_queue_item(self, item_id: int) -> SessionQueueItem:
-        cursor = self._conn.cursor()
-        cursor.execute(
-            """--sql
-            SELECT * FROM session_queue
-            WHERE
-                item_id = ?
-            """,
-            (item_id,),
-        )
-        result = cast(Union[sqlite3.Row, None], cursor.fetchone())
+        with self._db.transaction() as cursor:
+            cursor.execute(
+                """--sql
+                SELECT * FROM session_queue
+                WHERE
+                    item_id = ?
+                """,
+                (item_id,),
+            )
+            result = cast(Union[sqlite3.Row, None], cursor.fetchone())
         if result is None:
             raise SessionQueueItemNotFoundError(f"No queue item with id {item_id}")
         return SessionQueueItem.queue_item_from_dict(dict(result))
 
     def set_queue_item_session(self, item_id: int, session: GraphExecutionState) -> SessionQueueItem:
-        try:
-            cursor = self._conn.cursor()
+        with self._db.transaction() as cursor:
             # Use exclude_none so we don't end up with a bunch of nulls in the graph - this can cause validation errors
             # when the graph is loaded. Graph execution occurs purely in memory - the session saved here is not referenced
             # during execution.
@@ -636,111 +586,74 @@ class SqliteSessionQueue(SessionQueueBase):
                 """,
                 (session_json, item_id),
             )
-            self._conn.commit()
-        except Exception:
-            self._conn.rollback()
-            raise
         return self.get_queue_item(item_id)
 
-    def list_queue_items(
-        self,
-        queue_id: str,
-        limit: int,
-        priority: int,
-        cursor: Optional[int] = None,
-        status: Optional[QUEUE_ITEM_STATUS] = None,
-        destination: Optional[str] = None,
-    ) -> CursorPaginatedResults[SessionQueueItem]:
-        cursor_ = self._conn.cursor()
-        item_id = cursor
-        query = """--sql
-            SELECT *
-            FROM session_queue
-            WHERE queue_id = ?
-        """
-        params: list[Union[str, int]] = [queue_id]
-
-        if status is not None:
-            query += """--sql
-                AND status = ?
-                """
-            params.append(status)
-
-        if destination is not None:
-            query += """---sql
-                AND destination = ?
-            """
-            params.append(destination)
-
-        if item_id is not None:
-            query += """--sql
-                AND (priority < ?) OR (priority = ? AND item_id > ?)
-                """
-            params.extend([priority, priority, item_id])
-
-        query += """--sql
-            ORDER BY
-                priority DESC,
-                item_id ASC
-            LIMIT ?
-            """
-        params.append(limit + 1)
-        cursor_.execute(query, params)
-        results = cast(list[sqlite3.Row], cursor_.fetchall())
-        items = [SessionQueueItem.queue_item_from_dict(dict(result)) for result in results]
-        has_more = False
-        if len(items) > limit:
-            # remove the extra item
-            items.pop()
-            has_more = True
-        return CursorPaginatedResults(items=items, limit=limit, has_more=has_more)
-
     def list_all_queue_items(
         self,
         queue_id: str,
         destination: Optional[str] = None,
     ) -> list[SessionQueueItem]:
         """Gets all queue items that match the given parameters"""
-        cursor_ = self._conn.cursor()
-        query = """--sql
-            SELECT *
-            FROM session_queue
-            WHERE queue_id = ?
-        """
-        params: list[Union[str, int]] = [queue_id]
-
-        if destination is not None:
-            query += """---sql
-                AND destination = ?
+        with self._db.transaction() as cursor:
+            query = """--sql
+                SELECT *
+                FROM session_queue
+                WHERE queue_id = ?
             """
-            params.append(destination)
+            params: list[Union[str, int]] = [queue_id]
 
-        query += """--sql
-            ORDER BY
-                priority DESC,
-                item_id ASC
-            ;
-            """
-        cursor_.execute(query, params)
-        results = cast(list[sqlite3.Row], cursor_.fetchall())
+            if destination is not None:
+                query += """---sql
+                    AND destination = ?
+                """
+                params.append(destination)
+
+            query += """--sql
+                ORDER BY
+                    priority DESC,
+                    item_id ASC
+                ;
+                """
+            cursor.execute(query, params)
+            results = cast(list[sqlite3.Row], cursor.fetchall())
         items = [SessionQueueItem.queue_item_from_dict(dict(result)) for result in results]
         return items
 
+    def get_queue_item_ids(
+        self,
+        queue_id: str,
+        order_dir: SQLiteDirection = SQLiteDirection.Descending,
+    ) -> ItemIdsResult:
+        with self._db.transaction() as cursor_:
+            query = f"""--sql
+                SELECT item_id
+                FROM session_queue
+                WHERE queue_id = ?
+                ORDER BY created_at {order_dir.value}
+                """
+            query_params = [queue_id]
+
+            cursor_.execute(query, query_params)
+            result = cast(list[sqlite3.Row], cursor_.fetchall())
+        item_ids = [row[0] for row in result]
+
+        return ItemIdsResult(item_ids=item_ids, total_count=len(item_ids))
+
     def get_queue_status(self, queue_id: str) -> SessionQueueStatus:
-        cursor = self._conn.cursor()
-        cursor.execute(
-            """--sql
-            SELECT status, count(*)
-            FROM session_queue
-            WHERE queue_id = ?
-            GROUP BY status
-            """,
-            (queue_id,),
-        )
-        counts_result = cast(list[sqlite3.Row], cursor.fetchall())
+        with self._db.transaction() as cursor:
+            cursor.execute(
+                """--sql
+                SELECT status, count(*)
+                FROM session_queue
+                WHERE queue_id = ?
+                GROUP BY status
+                """,
+                (queue_id,),
+            )
+            counts_result = cast(list[sqlite3.Row], cursor.fetchall())
 
         current_item = self.get_current(queue_id=queue_id)
-        total = sum(row[1] for row in counts_result)
+        total = sum(row[1] or 0 for row in counts_result)
         counts: dict[str, int] = {row[0]: row[1] for row in counts_result}
         return SessionQueueStatus(
             queue_id=queue_id,
@@ -756,20 +669,20 @@ class SqliteSessionQueue(SessionQueueBase):
         )
 
     def get_batch_status(self, queue_id: str, batch_id: str) -> BatchStatus:
-        cursor = self._conn.cursor()
-        cursor.execute(
-            """--sql
-            SELECT status, count(*), origin, destination
-            FROM session_queue
-            WHERE
-                queue_id = ?
-                AND batch_id = ?
-            GROUP BY status
-            """,
-            (queue_id, batch_id),
-        )
-        result = cast(list[sqlite3.Row], cursor.fetchall())
-        total = sum(row[1] for row in result)
+        with self._db.transaction() as cursor:
+            cursor.execute(
+                """--sql
+                SELECT status, count(*), origin, destination
+                FROM session_queue
+                WHERE
+                    queue_id = ?
+                    AND batch_id = ?
+                GROUP BY status
+                """,
+                (queue_id, batch_id),
+            )
+            result = cast(list[sqlite3.Row], cursor.fetchall())
+        total = sum(row[1] or 0 for row in result)
         counts: dict[str, int] = {row[0]: row[1] for row in result}
         origin = result[0]["origin"] if result else None
         destination = result[0]["destination"] if result else None
@@ -788,20 +701,20 @@ class SqliteSessionQueue(SessionQueueBase):
         )
 
     def get_counts_by_destination(self, queue_id: str, destination: str) -> SessionQueueCountsByDestination:
-        cursor = self._conn.cursor()
-        cursor.execute(
-            """--sql
-            SELECT status, count(*)
-            FROM session_queue
-            WHERE queue_id = ?
-            AND destination = ?
-            GROUP BY status
-            """,
-            (queue_id, destination),
-        )
-        counts_result = cast(list[sqlite3.Row], cursor.fetchall())
+        with self._db.transaction() as cursor:
+            cursor.execute(
+                """--sql
+                SELECT status, count(*)
+                FROM session_queue
+                WHERE queue_id = ?
+                AND destination = ?
+                GROUP BY status
+                """,
+                (queue_id, destination),
+            )
+            counts_result = cast(list[sqlite3.Row], cursor.fetchall())
 
-        total = sum(row[1] for row in counts_result)
+        total = sum(row[1] or 0 for row in counts_result)
         counts: dict[str, int] = {row[0]: row[1] for row in counts_result}
 
         return SessionQueueCountsByDestination(
@@ -817,8 +730,7 @@ class SqliteSessionQueue(SessionQueueBase):
 
     def retry_items_by_id(self, queue_id: str, item_ids: list[int]) -> RetryItemsResult:
         """Retries the given queue items"""
-        try:
-            cursor = self._conn.cursor()
+        with self._db.transaction() as cursor:
             values_to_insert: list[ValueToInsertTuple] = []
             retried_item_ids: list[int] = []
 
@@ -869,10 +781,6 @@ class SqliteSessionQueue(SessionQueueBase):
                 values_to_insert,
             )
 
-            self._conn.commit()
-        except Exception:
-            self._conn.rollback()
-            raise
         retry_result = RetryItemsResult(
             queue_id=queue_id,
             retried_item_ids=retried_item_ids,

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

@@ -1,4 +1,7 @@
 import sqlite3
+import threading
+from collections.abc import Generator
+from contextlib import contextmanager
 from logging import Logger
 from pathlib import Path
 
@@ -26,46 +29,65 @@ class SqliteDatabase:
 
     def __init__(self, db_path: Path | None, logger: Logger, verbose: bool = False) -> None:
         """Initializes the database. This is used internally by the class constructor."""
-        self.logger = logger
-        self.db_path = db_path
-        self.verbose = verbose
+        self._logger = logger
+        self._db_path = db_path
+        self._verbose = verbose
+        self._lock = threading.RLock()
 
-        if not self.db_path:
+        if not self._db_path:
             logger.info("Initializing in-memory database")
         else:
-            self.db_path.parent.mkdir(parents=True, exist_ok=True)
-            self.logger.info(f"Initializing database at {self.db_path}")
+            self._db_path.parent.mkdir(parents=True, exist_ok=True)
+            self._logger.info(f"Initializing database at {self._db_path}")
 
-        self.conn = sqlite3.connect(database=self.db_path or sqlite_memory, check_same_thread=False)
-        self.conn.row_factory = sqlite3.Row
+        self._conn = sqlite3.connect(database=self._db_path or sqlite_memory, check_same_thread=False)
+        self._conn.row_factory = sqlite3.Row
 
-        if self.verbose:
-            self.conn.set_trace_callback(self.logger.debug)
+        if self._verbose:
+            self._conn.set_trace_callback(self._logger.debug)
 
         # Enable foreign key constraints
-        self.conn.execute("PRAGMA foreign_keys = ON;")
+        self._conn.execute("PRAGMA foreign_keys = ON;")
 
         # Enable Write-Ahead Logging (WAL) mode for better concurrency
-        self.conn.execute("PRAGMA journal_mode = WAL;")
+        self._conn.execute("PRAGMA journal_mode = WAL;")
 
         # Set a busy timeout to prevent database lockups during writes
-        self.conn.execute("PRAGMA busy_timeout = 5000;")  # 5 seconds
+        self._conn.execute("PRAGMA busy_timeout = 5000;")  # 5 seconds
 
     def clean(self) -> None:
         """
         Cleans the database by running the VACUUM command, reporting on the freed space.
         """
         # No need to clean in-memory database
-        if not self.db_path:
+        if not self._db_path:
             return
         try:
-            initial_db_size = Path(self.db_path).stat().st_size
-            self.conn.execute("VACUUM;")
-            self.conn.commit()
-            final_db_size = Path(self.db_path).stat().st_size
-            freed_space_in_mb = round((initial_db_size - final_db_size) / 1024 / 1024, 2)
-            if freed_space_in_mb > 0:
-                self.logger.info(f"Cleaned database (freed {freed_space_in_mb}MB)")
+            with self._conn as conn:
+                initial_db_size = Path(self._db_path).stat().st_size
+                conn.execute("VACUUM;")
+                conn.commit()
+                final_db_size = Path(self._db_path).stat().st_size
+                freed_space_in_mb = round((initial_db_size - final_db_size) / 1024 / 1024, 2)
+                if freed_space_in_mb > 0:
+                    self._logger.info(f"Cleaned database (freed {freed_space_in_mb}MB)")
         except Exception as e:
-            self.logger.error(f"Error cleaning database: {e}")
+            self._logger.error(f"Error cleaning database: {e}")
             raise
+
+    @contextmanager
+    def transaction(self) -> Generator[sqlite3.Cursor, None, None]:
+        """
+        Thread-safe context manager for DB work.
+        Acquires the RLock, yields a Cursor, then commits or rolls back.
+        """
+        with self._lock:
+            cursor = self._conn.cursor()
+            try:
+                yield cursor
+                self._conn.commit()
+            except:
+                self._conn.rollback()
+                raise
+            finally:
+                cursor.close()

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

@@ -23,6 +23,7 @@ from invokeai.app.services.shared.sqlite_migrator.migrations.migration_17 import
 from invokeai.app.services.shared.sqlite_migrator.migrations.migration_18 import build_migration_18
 from invokeai.app.services.shared.sqlite_migrator.migrations.migration_19 import build_migration_19
 from invokeai.app.services.shared.sqlite_migrator.migrations.migration_20 import build_migration_20
+from invokeai.app.services.shared.sqlite_migrator.migrations.migration_21 import build_migration_21
 from invokeai.app.services.shared.sqlite_migrator.sqlite_migrator_impl import SqliteMigrator
 
 
@@ -63,6 +64,7 @@ def init_db(config: InvokeAIAppConfig, logger: Logger, image_files: ImageFileSto
     migrator.register_migration(build_migration_18())
     migrator.register_migration(build_migration_19(app_config=config))
     migrator.register_migration(build_migration_20())
+    migrator.register_migration(build_migration_21())
     migrator.run_migrations()
 
     return db

invokeai/app/services/shared/sqlite_migrator/migrations/migration_21.py

@@ -0,0 +1,40 @@
+import sqlite3
+
+from invokeai.app.services.shared.sqlite_migrator.sqlite_migrator_common import Migration
+
+
+class Migration21Callback:
+    def __call__(self, cursor: sqlite3.Cursor) -> None:
+        cursor.execute(
+            """
+            CREATE TABLE client_state (
+              id          INTEGER PRIMARY KEY CHECK(id = 1),
+              data        TEXT    NOT NULL, -- Frontend will handle the shape of this data
+              updated_at  DATETIME NOT NULL DEFAULT (CURRENT_TIMESTAMP)
+            );
+            """
+        )
+        cursor.execute(
+            """
+            CREATE TRIGGER tg_client_state_updated_at
+            AFTER UPDATE ON client_state
+            FOR EACH ROW
+            BEGIN
+              UPDATE client_state
+                SET updated_at = CURRENT_TIMESTAMP
+              WHERE id = OLD.id;
+            END;
+            """
+        )
+
+
+def build_migration_21() -> Migration:
+    """Builds the migration object for migrating from version 20 to version 21. This includes:
+    - Creating the `client_state` table.
+    - Adding a trigger to update the `updated_at` field on updates.
+    """
+    return Migration(
+        from_version=20,
+        to_version=21,
+        callback=Migration21Callback(),
+    )

invokeai/app/services/shared/sqlite_migrator/sqlite_migrator_impl.py

@@ -32,7 +32,7 @@ class SqliteMigrator:
 
     def __init__(self, db: SqliteDatabase) -> None:
         self._db = db
-        self._logger = db.logger
+        self._logger = db._logger
         self._migration_set = MigrationSet()
         self._backup_path: Optional[Path] = None
 
@@ -45,7 +45,7 @@ class SqliteMigrator:
         """Migrates the database to the latest version."""
         # This throws if there is a problem.
         self._migration_set.validate_migration_chain()
-        cursor = self._db.conn.cursor()
+        cursor = self._db._conn.cursor()
         self._create_migrations_table(cursor=cursor)
 
         if self._migration_set.count == 0:
@@ -59,13 +59,13 @@ class SqliteMigrator:
         self._logger.info("Database update needed")
 
         # Make a backup of the db if it needs to be updated and is a file db
-        if self._db.db_path is not None:
+        if self._db._db_path is not None:
             timestamp = datetime.now().strftime("%Y%m%d-%H%M%S")
-            self._backup_path = self._db.db_path.parent / f"{self._db.db_path.stem}_backup_{timestamp}.db"
+            self._backup_path = self._db._db_path.parent / f"{self._db._db_path.stem}_backup_{timestamp}.db"
             self._logger.info(f"Backing up database to {str(self._backup_path)}")
             # Use SQLite to do the backup
             with closing(sqlite3.connect(self._backup_path)) as backup_conn:
-                self._db.conn.backup(backup_conn)
+                self._db._conn.backup(backup_conn)
         else:
             self._logger.info("Using in-memory database, no backup needed")
 
@@ -81,7 +81,7 @@ class SqliteMigrator:
         try:
             # Using sqlite3.Connection as a context manager commits a the transaction on exit, or rolls it back if an
             # exception is raised.
-            with self._db.conn as conn:
+            with self._db._conn as conn:
                 cursor = conn.cursor()
                 if self._get_current_version(cursor) != migration.from_version:
                     raise MigrationError(

invokeai/app/services/style_preset_records/style_preset_records_sqlite.py

@@ -17,7 +17,7 @@ from invokeai.app.util.misc import uuid_string
 class SqliteStylePresetRecordsStorage(StylePresetRecordsStorageBase):
     def __init__(self, db: SqliteDatabase) -> None:
         super().__init__()
-        self._conn = db.conn
+        self._db = db
 
     def start(self, invoker: Invoker) -> None:
         self._invoker = invoker
@@ -25,24 +25,23 @@ class SqliteStylePresetRecordsStorage(StylePresetRecordsStorageBase):
 
     def get(self, style_preset_id: str) -> StylePresetRecordDTO:
         """Gets a style preset by ID."""
-        cursor = self._conn.cursor()
-        cursor.execute(
-            """--sql
-            SELECT *
-            FROM style_presets
-            WHERE id = ?;
-            """,
-            (style_preset_id,),
-        )
-        row = cursor.fetchone()
+        with self._db.transaction() as cursor:
+            cursor.execute(
+                """--sql
+                SELECT *
+                FROM style_presets
+                WHERE id = ?;
+                """,
+                (style_preset_id,),
+            )
+            row = cursor.fetchone()
         if row is None:
             raise StylePresetNotFoundError(f"Style preset with id {style_preset_id} not found")
         return StylePresetRecordDTO.from_dict(dict(row))
 
     def create(self, style_preset: StylePresetWithoutId) -> StylePresetRecordDTO:
         style_preset_id = uuid_string()
-        try:
-            cursor = self._conn.cursor()
+        with self._db.transaction() as cursor:
             cursor.execute(
                 """--sql
                 INSERT OR IGNORE INTO style_presets (
@@ -60,16 +59,11 @@ class SqliteStylePresetRecordsStorage(StylePresetRecordsStorageBase):
                     style_preset.type,
                 ),
             )
-            self._conn.commit()
-        except Exception:
-            self._conn.rollback()
-            raise
         return self.get(style_preset_id)
 
     def create_many(self, style_presets: list[StylePresetWithoutId]) -> None:
         style_preset_ids = []
-        try:
-            cursor = self._conn.cursor()
+        with self._db.transaction() as cursor:
             for style_preset in style_presets:
                 style_preset_id = uuid_string()
                 style_preset_ids.append(style_preset_id)
@@ -90,16 +84,11 @@ class SqliteStylePresetRecordsStorage(StylePresetRecordsStorageBase):
                         style_preset.type,
                     ),
                 )
-            self._conn.commit()
-        except Exception:
-            self._conn.rollback()
-            raise
 
         return None
 
     def update(self, style_preset_id: str, changes: StylePresetChanges) -> StylePresetRecordDTO:
-        try:
-            cursor = self._conn.cursor()
+        with self._db.transaction() as cursor:
             # Change the name of a style preset
             if changes.name is not None:
                 cursor.execute(
@@ -122,15 +111,10 @@ class SqliteStylePresetRecordsStorage(StylePresetRecordsStorageBase):
                     (changes.preset_data.model_dump_json(), style_preset_id),
                 )
 
-            self._conn.commit()
-        except Exception:
-            self._conn.rollback()
-            raise
         return self.get(style_preset_id)
 
     def delete(self, style_preset_id: str) -> None:
-        try:
-            cursor = self._conn.cursor()
+        with self._db.transaction() as cursor:
             cursor.execute(
                 """--sql
                 DELETE from style_presets
@@ -138,51 +122,41 @@ class SqliteStylePresetRecordsStorage(StylePresetRecordsStorageBase):
                 """,
                 (style_preset_id,),
             )
-            self._conn.commit()
-        except Exception:
-            self._conn.rollback()
-            raise
         return None
 
     def get_many(self, type: PresetType | None = None) -> list[StylePresetRecordDTO]:
-        main_query = """
-            SELECT
-                *
-            FROM style_presets
-            """
+        with self._db.transaction() as cursor:
+            main_query = """
+                SELECT
+                    *
+                FROM style_presets
+                """
 
-        if type is not None:
-            main_query += "WHERE type = ? "
+            if type is not None:
+                main_query += "WHERE type = ? "
 
-        main_query += "ORDER BY LOWER(name) ASC"
+            main_query += "ORDER BY LOWER(name) ASC"
 
-        cursor = self._conn.cursor()
-        if type is not None:
-            cursor.execute(main_query, (type,))
-        else:
-            cursor.execute(main_query)
+            if type is not None:
+                cursor.execute(main_query, (type,))
+            else:
+                cursor.execute(main_query)
 
-        rows = cursor.fetchall()
+            rows = cursor.fetchall()
         style_presets = [StylePresetRecordDTO.from_dict(dict(row)) for row in rows]
 
         return style_presets
 
     def _sync_default_style_presets(self) -> None:
         """Syncs default style presets to the database. Internal use only."""
-
-        # First delete all existing default style presets
-        try:
-            cursor = self._conn.cursor()
+        with self._db.transaction() as cursor:
+            # First delete all existing default style presets
             cursor.execute(
                 """--sql
                 DELETE FROM style_presets
                 WHERE type = "default";
                 """
             )
-            self._conn.commit()
-        except Exception:
-            self._conn.rollback()
-            raise
         # Next, parse and create the default style presets
         with open(Path(__file__).parent / Path("default_style_presets.json"), "r") as file:
             presets = json.load(file)

invokeai/app/services/videos_common.py

@@ -0,0 +1,179 @@
+import datetime
+from typing import Optional, Union
+
+from pydantic import BaseModel, Field, StrictBool, StrictStr
+
+from invokeai.app.util.misc import get_iso_timestamp
+from invokeai.app.util.model_exclude_null import BaseModelExcludeNull
+
+VIDEO_DTO_COLS = ", ".join(
+    [
+        "videos." + c
+        for c in [
+            "video_id",
+            "width",
+            "height",
+            "session_id",
+            "node_id",
+            "is_intermediate",
+            "created_at",
+            "updated_at",
+            "deleted_at",
+            "starred",
+        ]
+    ]
+)
+
+
+class VideoRecord(BaseModelExcludeNull):
+    """Deserialized video record without metadata."""
+
+    video_id: str = Field(description="The unique id of the video.")
+    """The unique id of the video."""
+    width: int = Field(description="The width of the video in px.")
+    """The actual width of the video in px. This may be different from the width in metadata."""
+    height: int = Field(description="The height of the video in px.")
+    """The actual height of the video in px. This may be different from the height in metadata."""
+    created_at: Union[datetime.datetime, str] = Field(description="The created timestamp of the video.")
+    """The created timestamp of the video."""
+    updated_at: Union[datetime.datetime, str] = Field(description="The updated timestamp of the video.")
+    """The updated timestamp of the video."""
+    deleted_at: Optional[Union[datetime.datetime, str]] = Field(
+        default=None, description="The deleted timestamp of the video."
+    )
+    """The deleted timestamp of the video."""
+    is_intermediate: bool = Field(description="Whether this is an intermediate video.")
+    """Whether this is an intermediate video."""
+    session_id: Optional[str] = Field(
+        default=None,
+        description="The session ID that generated this video, if it is a generated video.",
+    )
+    """The session ID that generated this video, if it is a generated video."""
+    node_id: Optional[str] = Field(
+        default=None,
+        description="The node ID that generated this video, if it is a generated video.",
+    )
+    """The node ID that generated this video, if it is a generated video."""
+    starred: bool = Field(description="Whether this video is starred.")
+    """Whether this video is starred."""
+
+
+class VideoRecordChanges(BaseModelExcludeNull):
+    """A set of changes to apply to a video record.
+
+    Only limited changes are valid:
+      - `session_id`: change the session associated with a video
+      - `is_intermediate`: change the video's `is_intermediate` flag
+      - `starred`: change whether the video is starred
+    """
+
+    session_id: Optional[StrictStr] = Field(
+        default=None,
+        description="The video's new session ID.",
+    )
+    """The video's new session ID."""
+    is_intermediate: Optional[StrictBool] = Field(default=None, description="The video's new `is_intermediate` flag.")
+    """The video's new `is_intermediate` flag."""
+    starred: Optional[StrictBool] = Field(default=None, description="The video's new `starred` state")
+    """The video's new `starred` state."""
+
+
+def deserialize_video_record(video_dict: dict) -> VideoRecord:
+    """Deserializes a video record."""
+
+    # Retrieve all the values, setting "reasonable" defaults if they are not present.
+    video_id = video_dict.get("video_id", "unknown")
+    width = video_dict.get("width", 0)
+    height = video_dict.get("height", 0)
+    session_id = video_dict.get("session_id", None)
+    node_id = video_dict.get("node_id", None)
+    created_at = video_dict.get("created_at", get_iso_timestamp())
+    updated_at = video_dict.get("updated_at", get_iso_timestamp())
+    deleted_at = video_dict.get("deleted_at", get_iso_timestamp())
+    is_intermediate = video_dict.get("is_intermediate", False)
+    starred = video_dict.get("starred", False)
+
+    return VideoRecord(
+        video_id=video_id,
+        width=width,
+        height=height,
+        session_id=session_id,
+        node_id=node_id,
+        created_at=created_at,
+        updated_at=updated_at,
+        deleted_at=deleted_at,
+        is_intermediate=is_intermediate,
+        starred=starred,
+    )
+
+
+class VideoCollectionCounts(BaseModel):
+    starred_count: int = Field(description="The number of starred videos in the collection.")
+    unstarred_count: int = Field(description="The number of unstarred videos in the collection.")
+
+
+class VideoIdsResult(BaseModel):
+    """Response containing ordered video ids with metadata for optimistic updates."""
+
+    video_ids: list[str] = Field(description="Ordered list of video ids")
+    starred_count: int = Field(description="Number of starred videos (when starred_first=True)")
+    total_count: int = Field(description="Total number of videos matching the query")
+
+
+class VideoUrlsDTO(BaseModelExcludeNull):
+    """The URLs for an image and its thumbnail."""
+
+    video_id: str = Field(description="The unique id of the video.")
+    """The unique id of the video."""
+    video_url: str = Field(description="The URL of the video.")
+    """The URL of the video."""
+    thumbnail_url: str = Field(description="The URL of the video's thumbnail.")
+    """The URL of the video's thumbnail."""
+
+
+class VideoDTO(VideoRecord, VideoUrlsDTO):
+    """Deserialized video record, enriched for the frontend."""
+
+    board_id: Optional[str] = Field(
+        default=None, description="The id of the board the image belongs to, if one exists."
+    )
+    """The id of the board the image belongs to, if one exists."""
+
+
+def video_record_to_dto(
+    video_record: VideoRecord,
+    video_url: str,
+    thumbnail_url: str,
+    board_id: Optional[str],
+) -> VideoDTO:
+    """Converts a video record to a video DTO."""
+    return VideoDTO(
+        **video_record.model_dump(),
+        video_url=video_url,
+        thumbnail_url=thumbnail_url,
+        board_id=board_id,
+    )
+
+
+class ResultWithAffectedBoards(BaseModel):
+    affected_boards: list[str] = Field(description="The ids of boards affected by the delete operation")
+
+
+class DeleteVideosResult(ResultWithAffectedBoards):
+    deleted_videos: list[str] = Field(description="The ids of the videos that were deleted")
+
+
+class StarredVideosResult(ResultWithAffectedBoards):
+    starred_videos: list[str] = Field(description="The ids of the videos that were starred")
+
+
+class UnstarredVideosResult(ResultWithAffectedBoards):
+    unstarred_videos: list[str] = Field(description="The ids of the videos that were unstarred")
+
+
+class AddVideosToBoardResult(ResultWithAffectedBoards):
+    added_videos: list[str] = Field(description="The video ids that were added to the board")
+
+
+class RemoveVideosFromBoardResult(ResultWithAffectedBoards):
+    removed_videos: list[str] = Field(description="The video ids that were removed from their board")

invokeai/app/services/workflow_records/workflow_records_sqlite.py

@@ -25,7 +25,7 @@ SQL_TIME_FORMAT = "%Y-%m-%d %H:%M:%f"
 class SqliteWorkflowRecordsStorage(WorkflowRecordsStorageBase):
     def __init__(self, db: SqliteDatabase) -> None:
         super().__init__()
-        self._conn = db.conn
+        self._db = db
 
     def start(self, invoker: Invoker) -> None:
         self._invoker = invoker
@@ -33,16 +33,16 @@ class SqliteWorkflowRecordsStorage(WorkflowRecordsStorageBase):
 
     def get(self, workflow_id: str) -> WorkflowRecordDTO:
         """Gets a workflow by ID. Updates the opened_at column."""
-        cursor = self._conn.cursor()
-        cursor.execute(
-            """--sql
-            SELECT workflow_id, workflow, name, created_at, updated_at, opened_at
-            FROM workflow_library
-            WHERE workflow_id = ?;
-            """,
-            (workflow_id,),
-        )
-        row = cursor.fetchone()
+        with self._db.transaction() as cursor:
+            cursor.execute(
+                """--sql
+                SELECT workflow_id, workflow, name, created_at, updated_at, opened_at
+                FROM workflow_library
+                WHERE workflow_id = ?;
+                """,
+                (workflow_id,),
+            )
+            row = cursor.fetchone()
         if row is None:
             raise WorkflowNotFoundError(f"Workflow with id {workflow_id} not found")
         return WorkflowRecordDTO.from_dict(dict(row))
@@ -51,9 +51,8 @@ class SqliteWorkflowRecordsStorage(WorkflowRecordsStorageBase):
         if workflow.meta.category is WorkflowCategory.Default:
             raise ValueError("Default workflows cannot be created via this method")
 
-        try:
+        with self._db.transaction() as cursor:
             workflow_with_id = Workflow(**workflow.model_dump(), id=uuid_string())
-            cursor = self._conn.cursor()
             cursor.execute(
                 """--sql
                 INSERT OR IGNORE INTO workflow_library (
@@ -64,18 +63,13 @@ class SqliteWorkflowRecordsStorage(WorkflowRecordsStorageBase):
                 """,
                 (workflow_with_id.id, workflow_with_id.model_dump_json()),
             )
-            self._conn.commit()
-        except Exception:
-            self._conn.rollback()
-            raise
         return self.get(workflow_with_id.id)
 
     def update(self, workflow: Workflow) -> WorkflowRecordDTO:
         if workflow.meta.category is WorkflowCategory.Default:
             raise ValueError("Default workflows cannot be updated")
 
-        try:
-            cursor = self._conn.cursor()
+        with self._db.transaction() as cursor:
             cursor.execute(
                 """--sql
                 UPDATE workflow_library
@@ -84,18 +78,13 @@ class SqliteWorkflowRecordsStorage(WorkflowRecordsStorageBase):
                 """,
                 (workflow.model_dump_json(), workflow.id),
             )
-            self._conn.commit()
-        except Exception:
-            self._conn.rollback()
-            raise
         return self.get(workflow.id)
 
     def delete(self, workflow_id: str) -> None:
         if self.get(workflow_id).workflow.meta.category is WorkflowCategory.Default:
             raise ValueError("Default workflows cannot be deleted")
 
-        try:
-            cursor = self._conn.cursor()
+        with self._db.transaction() as cursor:
             cursor.execute(
                 """--sql
                 DELETE from workflow_library
@@ -103,10 +92,6 @@ class SqliteWorkflowRecordsStorage(WorkflowRecordsStorageBase):
                 """,
                 (workflow_id,),
             )
-            self._conn.commit()
-        except Exception:
-            self._conn.rollback()
-            raise
         return None
 
     def get_many(
@@ -121,108 +106,108 @@ class SqliteWorkflowRecordsStorage(WorkflowRecordsStorageBase):
         has_been_opened: Optional[bool] = None,
         is_published: Optional[bool] = None,
     ) -> PaginatedResults[WorkflowRecordListItemDTO]:
-        # sanitize!
-        assert order_by in WorkflowRecordOrderBy
-        assert direction in SQLiteDirection
+        with self._db.transaction() as cursor:
+            # sanitize!
+            assert order_by in WorkflowRecordOrderBy
+            assert direction in SQLiteDirection
 
-        # We will construct the query dynamically based on the query params
+            # We will construct the query dynamically based on the query params
 
-        # The main query to get the workflows / counts
-        main_query = """
-                SELECT
-                    workflow_id,
-                    category,
-                    name,
-                    description,
-                    created_at,
-                    updated_at,
-                    opened_at,
-                    tags
-                FROM workflow_library
-                """
-        count_query = "SELECT COUNT(*) FROM workflow_library"
+            # The main query to get the workflows / counts
+            main_query = """
+                    SELECT
+                        workflow_id,
+                        category,
+                        name,
+                        description,
+                        created_at,
+                        updated_at,
+                        opened_at,
+                        tags
+                    FROM workflow_library
+                    """
+            count_query = "SELECT COUNT(*) FROM workflow_library"
 
-        # Start with an empty list of conditions and params
-        conditions: list[str] = []
-        params: list[str | int] = []
+            # Start with an empty list of conditions and params
+            conditions: list[str] = []
+            params: list[str | int] = []
 
-        if categories:
-            # Categories is a list of WorkflowCategory enum values, and a single string in the DB
+            if categories:
+                # Categories is a list of WorkflowCategory enum values, and a single string in the DB
 
-            # Ensure all categories are valid (is this necessary?)
-            assert all(c in WorkflowCategory for c in categories)
+                # Ensure all categories are valid (is this necessary?)
+                assert all(c in WorkflowCategory for c in categories)
 
-            # Construct a placeholder string for the number of categories
-            placeholders = ", ".join("?" for _ in categories)
+                # Construct a placeholder string for the number of categories
+                placeholders = ", ".join("?" for _ in categories)
 
-            # Construct the condition string & params
-            category_condition = f"category IN ({placeholders})"
-            category_params = [category.value for category in categories]
+                # Construct the condition string & params
+                category_condition = f"category IN ({placeholders})"
+                category_params = [category.value for category in categories]
 
-            conditions.append(category_condition)
-            params.extend(category_params)
+                conditions.append(category_condition)
+                params.extend(category_params)
 
-        if tags:
-            # Tags is a list of strings, and a single string in the DB
-            # The string in the DB has no guaranteed format
+            if tags:
+                # Tags is a list of strings, and a single string in the DB
+                # The string in the DB has no guaranteed format
 
-            # Construct a list of conditions for each tag
-            tags_conditions = ["tags LIKE ?" for _ in tags]
-            tags_conditions_joined = " OR ".join(tags_conditions)
-            tags_condition = f"({tags_conditions_joined})"
+                # Construct a list of conditions for each tag
+                tags_conditions = ["tags LIKE ?" for _ in tags]
+                tags_conditions_joined = " OR ".join(tags_conditions)
+                tags_condition = f"({tags_conditions_joined})"
 
-            # And the params for the tags, case-insensitive
-            tags_params = [f"%{t.strip()}%" for t in tags]
+                # And the params for the tags, case-insensitive
+                tags_params = [f"%{t.strip()}%" for t in tags]
 
-            conditions.append(tags_condition)
-            params.extend(tags_params)
+                conditions.append(tags_condition)
+                params.extend(tags_params)
 
-        if has_been_opened:
-            conditions.append("opened_at IS NOT NULL")
-        elif has_been_opened is False:
-            conditions.append("opened_at IS NULL")
+            if has_been_opened:
+                conditions.append("opened_at IS NOT NULL")
+            elif has_been_opened is False:
+                conditions.append("opened_at IS NULL")
 
-        # Ignore whitespace in the query
-        stripped_query = query.strip() if query else None
-        if stripped_query:
-            # Construct a wildcard query for the name, description, and tags
-            wildcard_query = "%" + stripped_query + "%"
-            query_condition = "(name LIKE ? OR description LIKE ? OR tags LIKE ?)"
+            # Ignore whitespace in the query
+            stripped_query = query.strip() if query else None
+            if stripped_query:
+                # Construct a wildcard query for the name, description, and tags
+                wildcard_query = "%" + stripped_query + "%"
+                query_condition = "(name LIKE ? OR description LIKE ? OR tags LIKE ?)"
 
-            conditions.append(query_condition)
-            params.extend([wildcard_query, wildcard_query, wildcard_query])
+                conditions.append(query_condition)
+                params.extend([wildcard_query, wildcard_query, wildcard_query])
 
-        if conditions:
-            # If there are conditions, add a WHERE clause and then join the conditions
-            main_query += " WHERE "
-            count_query += " WHERE "
+            if conditions:
+                # If there are conditions, add a WHERE clause and then join the conditions
+                main_query += " WHERE "
+                count_query += " WHERE "
 
-            all_conditions = " AND ".join(conditions)
-            main_query += all_conditions
-            count_query += all_conditions
+                all_conditions = " AND ".join(conditions)
+                main_query += all_conditions
+                count_query += all_conditions
 
-        # After this point, the query and params differ for the main query and the count query
-        main_params = params.copy()
-        count_params = params.copy()
+            # After this point, the query and params differ for the main query and the count query
+            main_params = params.copy()
+            count_params = params.copy()
 
-        # Main query also gets ORDER BY and LIMIT/OFFSET
-        main_query += f" ORDER BY {order_by.value} {direction.value}"
+            # Main query also gets ORDER BY and LIMIT/OFFSET
+            main_query += f" ORDER BY {order_by.value} {direction.value}"
 
-        if per_page:
-            main_query += " LIMIT ? OFFSET ?"
-            main_params.extend([per_page, page * per_page])
+            if per_page:
+                main_query += " LIMIT ? OFFSET ?"
+                main_params.extend([per_page, page * per_page])
 
-        # Put a ring on it
-        main_query += ";"
-        count_query += ";"
+            # Put a ring on it
+            main_query += ";"
+            count_query += ";"
 
-        cursor = self._conn.cursor()
-        cursor.execute(main_query, main_params)
-        rows = cursor.fetchall()
-        workflows = [WorkflowRecordListItemDTOValidator.validate_python(dict(row)) for row in rows]
+            cursor.execute(main_query, main_params)
+            rows = cursor.fetchall()
+            workflows = [WorkflowRecordListItemDTOValidator.validate_python(dict(row)) for row in rows]
 
-        cursor.execute(count_query, count_params)
-        total = cursor.fetchone()[0]
+            cursor.execute(count_query, count_params)
+            total = cursor.fetchone()[0]
 
         if per_page:
             pages = total // per_page + (total % per_page > 0)
@@ -247,46 +232,46 @@ class SqliteWorkflowRecordsStorage(WorkflowRecordsStorageBase):
         if not tags:
             return {}
 
-        cursor = self._conn.cursor()
-        result: dict[str, int] = {}
-        # Base conditions for categories and selected tags
-        base_conditions: list[str] = []
-        base_params: list[str | int] = []
+        with self._db.transaction() as cursor:
+            result: dict[str, int] = {}
+            # Base conditions for categories and selected tags
+            base_conditions: list[str] = []
+            base_params: list[str | int] = []
 
-        # Add category conditions
-        if categories:
-            assert all(c in WorkflowCategory for c in categories)
-            placeholders = ", ".join("?" for _ in categories)
-            base_conditions.append(f"category IN ({placeholders})")
-            base_params.extend([category.value for category in categories])
+            # Add category conditions
+            if categories:
+                assert all(c in WorkflowCategory for c in categories)
+                placeholders = ", ".join("?" for _ in categories)
+                base_conditions.append(f"category IN ({placeholders})")
+                base_params.extend([category.value for category in categories])
 
-        if has_been_opened:
-            base_conditions.append("opened_at IS NOT NULL")
-        elif has_been_opened is False:
-            base_conditions.append("opened_at IS NULL")
+            if has_been_opened:
+                base_conditions.append("opened_at IS NOT NULL")
+            elif has_been_opened is False:
+                base_conditions.append("opened_at IS NULL")
 
-        # For each tag to count, run a separate query
-        for tag in tags:
-            # Start with the base conditions
-            conditions = base_conditions.copy()
-            params = base_params.copy()
+            # For each tag to count, run a separate query
+            for tag in tags:
+                # Start with the base conditions
+                conditions = base_conditions.copy()
+                params = base_params.copy()
 
-            # Add this specific tag condition
-            conditions.append("tags LIKE ?")
-            params.append(f"%{tag.strip()}%")
+                # Add this specific tag condition
+                conditions.append("tags LIKE ?")
+                params.append(f"%{tag.strip()}%")
 
-            # Construct the full query
-            stmt = """--sql
-                SELECT COUNT(*)
-                FROM workflow_library
-                """
+                # Construct the full query
+                stmt = """--sql
+                    SELECT COUNT(*)
+                    FROM workflow_library
+                    """
 
-            if conditions:
-                stmt += " WHERE " + " AND ".join(conditions)
+                if conditions:
+                    stmt += " WHERE " + " AND ".join(conditions)
 
-            cursor.execute(stmt, params)
-            count = cursor.fetchone()[0]
-            result[tag] = count
+                cursor.execute(stmt, params)
+                count = cursor.fetchone()[0]
+                result[tag] = count
 
         return result
 
@@ -296,52 +281,51 @@ class SqliteWorkflowRecordsStorage(WorkflowRecordsStorageBase):
         has_been_opened: Optional[bool] = None,
         is_published: Optional[bool] = None,
     ) -> dict[str, int]:
-        cursor = self._conn.cursor()
-        result: dict[str, int] = {}
-        # Base conditions for categories
-        base_conditions: list[str] = []
-        base_params: list[str | int] = []
+        with self._db.transaction() as cursor:
+            result: dict[str, int] = {}
+            # Base conditions for categories
+            base_conditions: list[str] = []
+            base_params: list[str | int] = []
 
-        # Add category conditions
-        if categories:
-            assert all(c in WorkflowCategory for c in categories)
-            placeholders = ", ".join("?" for _ in categories)
-            base_conditions.append(f"category IN ({placeholders})")
-            base_params.extend([category.value for category in categories])
+            # Add category conditions
+            if categories:
+                assert all(c in WorkflowCategory for c in categories)
+                placeholders = ", ".join("?" for _ in categories)
+                base_conditions.append(f"category IN ({placeholders})")
+                base_params.extend([category.value for category in categories])
 
-        if has_been_opened:
-            base_conditions.append("opened_at IS NOT NULL")
-        elif has_been_opened is False:
-            base_conditions.append("opened_at IS NULL")
+            if has_been_opened:
+                base_conditions.append("opened_at IS NOT NULL")
+            elif has_been_opened is False:
+                base_conditions.append("opened_at IS NULL")
 
-        # For each category to count, run a separate query
-        for category in categories:
-            # Start with the base conditions
-            conditions = base_conditions.copy()
-            params = base_params.copy()
+            # For each category to count, run a separate query
+            for category in categories:
+                # Start with the base conditions
+                conditions = base_conditions.copy()
+                params = base_params.copy()
 
-            # Add this specific category condition
-            conditions.append("category = ?")
-            params.append(category.value)
+                # Add this specific category condition
+                conditions.append("category = ?")
+                params.append(category.value)
 
-            # Construct the full query
-            stmt = """--sql
-                SELECT COUNT(*)
-                FROM workflow_library
-                """
+                # Construct the full query
+                stmt = """--sql
+                    SELECT COUNT(*)
+                    FROM workflow_library
+                    """
 
-            if conditions:
-                stmt += " WHERE " + " AND ".join(conditions)
+                if conditions:
+                    stmt += " WHERE " + " AND ".join(conditions)
 
-            cursor.execute(stmt, params)
-            count = cursor.fetchone()[0]
-            result[category.value] = count
+                cursor.execute(stmt, params)
+                count = cursor.fetchone()[0]
+                result[category.value] = count
 
         return result
 
     def update_opened_at(self, workflow_id: str) -> None:
-        try:
-            cursor = self._conn.cursor()
+        with self._db.transaction() as cursor:
             cursor.execute(
                 f"""--sql
                 UPDATE workflow_library
@@ -350,10 +334,6 @@ class SqliteWorkflowRecordsStorage(WorkflowRecordsStorageBase):
                 """,
                 (workflow_id,),
             )
-            self._conn.commit()
-        except Exception:
-            self._conn.rollback()
-            raise
 
     def _sync_default_workflows(self) -> None:
         """Syncs default workflows to the database. Internal use only."""
@@ -368,8 +348,7 @@ class SqliteWorkflowRecordsStorage(WorkflowRecordsStorageBase):
         meaningless, as they are overwritten every time the server starts.
         """
 
-        try:
-            cursor = self._conn.cursor()
+        with self._db.transaction() as cursor:
             workflows_from_file: list[Workflow] = []
             workflows_to_update: list[Workflow] = []
             workflows_to_add: list[Workflow] = []
@@ -449,8 +428,3 @@ class SqliteWorkflowRecordsStorage(WorkflowRecordsStorageBase):
                     """,
                     (w.model_dump_json(), w.id),
                 )
-
-            self._conn.commit()
-        except Exception:
-            self._conn.rollback()
-            raise

invokeai/backend/flux/denoise.py

@@ -30,8 +30,11 @@ def denoise(
     controlnet_extensions: list[XLabsControlNetExtension | InstantXControlNetExtension],
     pos_ip_adapter_extensions: list[XLabsIPAdapterExtension],
     neg_ip_adapter_extensions: list[XLabsIPAdapterExtension],
-    # extra img tokens
+    # extra img tokens (channel-wise)
     img_cond: torch.Tensor | None,
+    # extra img tokens (sequence-wise) - for Kontext conditioning
+    img_cond_seq: torch.Tensor | None = None,
+    img_cond_seq_ids: torch.Tensor | None = None,
 ):
     # step 0 is the initial state
     total_steps = len(timesteps) - 1
@@ -46,6 +49,10 @@ def denoise(
     )
     # guidance_vec is ignored for schnell.
     guidance_vec = torch.full((img.shape[0],), guidance, device=img.device, dtype=img.dtype)
+
+    # Store original sequence length for slicing predictions
+    original_seq_len = img.shape[1]
+
     for step_index, (t_curr, t_prev) in tqdm(list(enumerate(zip(timesteps[:-1], timesteps[1:], strict=True)))):
         t_vec = torch.full((img.shape[0],), t_curr, dtype=img.dtype, device=img.device)
 
@@ -71,10 +78,26 @@ def denoise(
         # controlnet_residuals datastructure is efficient in that it likely contains multiple references to the same
         # tensors. Calculating the sum materializes each tensor into its own instance.
         merged_controlnet_residuals = sum_controlnet_flux_outputs(controlnet_residuals)
-        pred_img = torch.cat((img, img_cond), dim=-1) if img_cond is not None else img
+
+        # Prepare input for model - concatenate fresh each step
+        img_input = img
+        img_input_ids = img_ids
+
+        # Add channel-wise conditioning (for ControlNet, FLUX Fill, etc.)
+        if img_cond is not None:
+            img_input = torch.cat((img_input, img_cond), dim=-1)
+
+        # Add sequence-wise conditioning (for Kontext)
+        if img_cond_seq is not None:
+            assert img_cond_seq_ids is not None, (
+                "You need to provide either both or neither of the sequence conditioning"
+            )
+            img_input = torch.cat((img_input, img_cond_seq), dim=1)
+            img_input_ids = torch.cat((img_input_ids, img_cond_seq_ids), dim=1)
+
         pred = model(
-            img=pred_img,
-            img_ids=img_ids,
+            img=img_input,
+            img_ids=img_input_ids,
             txt=pos_regional_prompting_extension.regional_text_conditioning.t5_embeddings,
             txt_ids=pos_regional_prompting_extension.regional_text_conditioning.t5_txt_ids,
             y=pos_regional_prompting_extension.regional_text_conditioning.clip_embeddings,
@@ -88,6 +111,10 @@ def denoise(
             regional_prompting_extension=pos_regional_prompting_extension,
         )
 
+        # Slice prediction to only include the main image tokens
+        if img_cond_seq is not None:
+            pred = pred[:, :original_seq_len]
+
         step_cfg_scale = cfg_scale[step_index]
 
         # If step_cfg_scale, is 1.0, then we don't need to run the negative prediction.
@@ -98,9 +125,26 @@ def denoise(
             if neg_regional_prompting_extension is None:
                 raise ValueError("Negative text conditioning is required when cfg_scale is not 1.0.")
 
+            # For negative prediction with Kontext, we need to include the reference images
+            # to maintain consistency between positive and negative passes. Without this,
+            # CFG would create artifacts as the attention mechanism would see different
+            # spatial structures in each pass
+            neg_img_input = img
+            neg_img_input_ids = img_ids
+
+            # Add channel-wise conditioning for negative pass if present
+            if img_cond is not None:
+                neg_img_input = torch.cat((neg_img_input, img_cond), dim=-1)
+
+            # Add sequence-wise conditioning (Kontext) for negative pass
+            # This ensures reference images are processed consistently
+            if img_cond_seq is not None:
+                neg_img_input = torch.cat((neg_img_input, img_cond_seq), dim=1)
+                neg_img_input_ids = torch.cat((neg_img_input_ids, img_cond_seq_ids), dim=1)
+
             neg_pred = model(
-                img=img,
-                img_ids=img_ids,
+                img=neg_img_input,
+                img_ids=neg_img_input_ids,
                 txt=neg_regional_prompting_extension.regional_text_conditioning.t5_embeddings,
                 txt_ids=neg_regional_prompting_extension.regional_text_conditioning.t5_txt_ids,
                 y=neg_regional_prompting_extension.regional_text_conditioning.clip_embeddings,
@@ -113,6 +157,10 @@ def denoise(
                 ip_adapter_extensions=neg_ip_adapter_extensions,
                 regional_prompting_extension=neg_regional_prompting_extension,
             )
+
+            # Slice negative prediction to match main image tokens
+            if img_cond_seq is not None:
+                neg_pred = neg_pred[:, :original_seq_len]
             pred = neg_pred + step_cfg_scale * (pred - neg_pred)
 
         preview_img = img - t_curr * pred

invokeai/backend/flux/extensions/kontext_extension.py

@@ -1,13 +1,14 @@
-import einops
 import torch
+import torch.nn.functional as F
+import torchvision.transforms as T
 from einops import repeat
 
 from invokeai.app.invocations.fields import FluxKontextConditioningField
-from invokeai.app.invocations.flux_vae_encode import FluxVaeEncodeInvocation
 from invokeai.app.invocations.model import VAEField
 from invokeai.app.services.shared.invocation_context import InvocationContext
+from invokeai.backend.flux.modules.autoencoder import AutoEncoder
 from invokeai.backend.flux.sampling_utils import pack
-from invokeai.backend.stable_diffusion.diffusers_pipeline import image_resized_to_grid_as_tensor
+from invokeai.backend.util.devices import TorchDevice
 
 
 def generate_img_ids_with_offset(
@@ -17,8 +18,10 @@ def generate_img_ids_with_offset(
     device: torch.device,
     dtype: torch.dtype,
     idx_offset: int = 0,
+    h_offset: int = 0,
+    w_offset: int = 0,
 ) -> torch.Tensor:
-    """Generate tensor of image position ids with an optional offset.
+    """Generate tensor of image position ids with optional index and spatial offsets.
 
     Args:
         latent_height (int): Height of image in latent space (after packing, this becomes h//2).
@@ -26,7 +29,9 @@ def generate_img_ids_with_offset(
         batch_size (int): Number of images in the batch.
         device (torch.device): Device to create tensors on.
         dtype (torch.dtype): Data type for the tensors.
-        idx_offset (int): Offset to add to the first dimension of the image ids.
+        idx_offset (int): Offset to add to the first dimension of the image ids (default: 0).
+        h_offset (int): Spatial offset for height/y-coordinates in latent space (default: 0).
+        w_offset (int): Spatial offset for width/x-coordinates in latent space (default: 0).
 
     Returns:
         torch.Tensor: Image position ids with shape [batch_size, (latent_height//2 * latent_width//2), 3].
@@ -40,6 +45,10 @@ def generate_img_ids_with_offset(
     packed_height = latent_height // 2
     packed_width = latent_width // 2
 
+    # Convert spatial offsets from latent space to packed space
+    packed_h_offset = h_offset // 2
+    packed_w_offset = w_offset // 2
+
     # Create base tensor for position IDs with shape [packed_height, packed_width, 3]
     # The 3 channels represent: [batch_offset, y_position, x_position]
     img_ids = torch.zeros(packed_height, packed_width, 3, device=device, dtype=dtype)
@@ -47,13 +56,13 @@ def generate_img_ids_with_offset(
     # Set the batch offset for all positions
     img_ids[..., 0] = idx_offset
 
-    # Create y-coordinate indices (vertical positions)
-    y_indices = torch.arange(packed_height, device=device, dtype=dtype)
+    # Create y-coordinate indices (vertical positions) with spatial offset
+    y_indices = torch.arange(packed_height, device=device, dtype=dtype) + packed_h_offset
     # Broadcast y_indices to match the spatial dimensions [packed_height, 1]
     img_ids[..., 1] = y_indices[:, None]
 
-    # Create x-coordinate indices (horizontal positions)
-    x_indices = torch.arange(packed_width, device=device, dtype=dtype)
+    # Create x-coordinate indices (horizontal positions) with spatial offset
+    x_indices = torch.arange(packed_width, device=device, dtype=dtype) + packed_w_offset
     # Broadcast x_indices to match the spatial dimensions [1, packed_width]
     img_ids[..., 2] = x_indices[None, :]
 
@@ -71,69 +80,139 @@ class KontextExtension:
 
     def __init__(
         self,
-        kontext_field: FluxKontextConditioningField,
+        kontext_conditioning: list[FluxKontextConditioningField],
         context: InvocationContext,
         vae_field: VAEField,
         device: torch.device,
         dtype: torch.dtype,
     ):
         """
-        Initializes the KontextExtension, pre-processing the reference image
+        Initializes the KontextExtension, pre-processing the reference images
         into latents and positional IDs.
         """
         self._context = context
         self._device = device
         self._dtype = dtype
         self._vae_field = vae_field
-        self.kontext_field = kontext_field
+        self.kontext_conditioning = kontext_conditioning
 
         # Pre-process and cache the kontext latents and ids upon initialization.
         self.kontext_latents, self.kontext_ids = self._prepare_kontext()
 
     def _prepare_kontext(self) -> tuple[torch.Tensor, torch.Tensor]:
-        """Encodes the reference image and prepares its latents and IDs."""
-        image = self._context.images.get_pil(self.kontext_field.image.image_name)
+        """Encodes the reference images and prepares their concatenated latents and IDs with spatial tiling."""
+        all_latents = []
+        all_ids = []
+
+        # Track cumulative dimensions for spatial tiling
+        # These track the running extent of the virtual canvas in latent space
+        canvas_h = 0  # Running canvas height
+        canvas_w = 0  # Running canvas width
 
-        # Reuse VAE encoding logic from FluxVaeEncodeInvocation
         vae_info = self._context.models.load(self._vae_field.vae)
-        image_tensor = image_resized_to_grid_as_tensor(image.convert("RGB"))
-        if image_tensor.dim() == 3:
-            image_tensor = einops.rearrange(image_tensor, "c h w -> 1 c h w")
-        image_tensor = image_tensor.to(self._device)
 
-        kontext_latents_unpacked = FluxVaeEncodeInvocation.vae_encode(vae_info=vae_info, image_tensor=image_tensor)
+        for idx, kontext_field in enumerate(self.kontext_conditioning):
+            image = self._context.images.get_pil(kontext_field.image.image_name)
 
-        # Extract tensor dimensions with descriptive names
-        # Latent tensor shape: [batch_size, channels, latent_height, latent_width]
-        batch_size, _, latent_height, latent_width = kontext_latents_unpacked.shape
+            # Convert to RGB
+            image = image.convert("RGB")
 
-        # Pack the latents and generate IDs. The idx_offset distinguishes these
-        # tokens from the main image's tokens, which have an index of 0.
-        kontext_latents_packed = pack(kontext_latents_unpacked).to(self._device, self._dtype)
-        kontext_ids = generate_img_ids_with_offset(
-            latent_height=latent_height,
-            latent_width=latent_width,
-            batch_size=batch_size,
-            device=self._device,
-            dtype=self._dtype,
-            idx_offset=1,  # Distinguishes reference tokens from main image tokens
-        )
+            # Convert to tensor using torchvision transforms for consistency
+            transformation = T.Compose(
+                [
+                    T.ToTensor(),  # Converts PIL image to tensor and scales to [0, 1]
+                ]
+            )
+            image_tensor = transformation(image)
+            # Convert from [0, 1] to [-1, 1] range expected by VAE
+            image_tensor = image_tensor * 2.0 - 1.0
+            image_tensor = image_tensor.unsqueeze(0)  # Add batch dimension
+            image_tensor = image_tensor.to(self._device)
 
-        return kontext_latents_packed, kontext_ids
+            # Continue with VAE encoding
+            # Don't sample from the distribution for reference images - use the mean (matching ComfyUI)
+            # Estimate working memory for encode operation (50% of decode memory requirements)
+            img_h = image_tensor.shape[-2]
+            img_w = image_tensor.shape[-1]
+            element_size = next(vae_info.model.parameters()).element_size()
+            scaling_constant = 1100  # 50% of decode scaling constant (2200)
+            estimated_working_memory = int(img_h * img_w * element_size * scaling_constant)
 
-    def apply(
-        self,
-        img: torch.Tensor,
-        img_ids: torch.Tensor,
-    ) -> tuple[torch.Tensor, torch.Tensor]:
-        """Concatenates the pre-processed kontext data to the main image sequence."""
-        # Ensure batch sizes match, repeating kontext data if necessary for batch operations.
-        if img.shape[0] != self.kontext_latents.shape[0]:
-            self.kontext_latents = self.kontext_latents.repeat(img.shape[0], 1, 1)
-            self.kontext_ids = self.kontext_ids.repeat(img.shape[0], 1, 1)
+            with vae_info.model_on_device(working_mem_bytes=estimated_working_memory) as (_, vae):
+                assert isinstance(vae, AutoEncoder)
+                vae_dtype = next(iter(vae.parameters())).dtype
+                image_tensor = image_tensor.to(device=TorchDevice.choose_torch_device(), dtype=vae_dtype)
+                # Use sample=False to get the distribution mean without noise
+                kontext_latents_unpacked = vae.encode(image_tensor, sample=False)
+                TorchDevice.empty_cache()
 
-        # Concatenate along the sequence dimension (dim=1)
-        combined_img = torch.cat([img, self.kontext_latents], dim=1)
-        combined_img_ids = torch.cat([img_ids, self.kontext_ids], dim=1)
+            # Extract tensor dimensions
+            batch_size, _, latent_height, latent_width = kontext_latents_unpacked.shape
 
-        return combined_img, combined_img_ids
+            # Pad latents to be compatible with patch_size=2
+            # This ensures dimensions are even for the pack() function
+            pad_h = (2 - latent_height % 2) % 2
+            pad_w = (2 - latent_width % 2) % 2
+            if pad_h > 0 or pad_w > 0:
+                kontext_latents_unpacked = F.pad(kontext_latents_unpacked, (0, pad_w, 0, pad_h), mode="circular")
+                # Update dimensions after padding
+                _, _, latent_height, latent_width = kontext_latents_unpacked.shape
+
+            # Pack the latents
+            kontext_latents_packed = pack(kontext_latents_unpacked).to(self._device, self._dtype)
+
+            # Determine spatial offsets for this reference image
+            h_offset = 0
+            w_offset = 0
+
+            if idx > 0:  # First image starts at (0, 0)
+                # Calculate potential canvas dimensions for each tiling option
+                # Option 1: Tile vertically (below existing content)
+                potential_h_vertical = canvas_h + latent_height
+
+                # Option 2: Tile horizontally (to the right of existing content)
+                potential_w_horizontal = canvas_w + latent_width
+
+                # Choose arrangement that minimizes the maximum dimension
+                # This keeps the canvas closer to square, optimizing attention computation
+                if potential_h_vertical > potential_w_horizontal:
+                    # Tile horizontally (to the right of existing images)
+                    w_offset = canvas_w
+                    canvas_w = canvas_w + latent_width
+                    canvas_h = max(canvas_h, latent_height)
+                else:
+                    # Tile vertically (below existing images)
+                    h_offset = canvas_h
+                    canvas_h = canvas_h + latent_height
+                    canvas_w = max(canvas_w, latent_width)
+            else:
+                # First image - just set canvas dimensions
+                canvas_h = latent_height
+                canvas_w = latent_width
+
+            # Generate IDs with both index offset and spatial offsets
+            kontext_ids = generate_img_ids_with_offset(
+                latent_height=latent_height,
+                latent_width=latent_width,
+                batch_size=batch_size,
+                device=self._device,
+                dtype=self._dtype,
+                idx_offset=1,  # All reference images use index=1 (matching ComfyUI implementation)
+                h_offset=h_offset,
+                w_offset=w_offset,
+            )
+
+            all_latents.append(kontext_latents_packed)
+            all_ids.append(kontext_ids)
+
+        # Concatenate all latents and IDs along the sequence dimension
+        concatenated_latents = torch.cat(all_latents, dim=1)  # Concatenate along sequence dimension
+        concatenated_ids = torch.cat(all_ids, dim=1)  # Concatenate along sequence dimension
+
+        return concatenated_latents, concatenated_ids
+
+    def ensure_batch_size(self, target_batch_size: int) -> None:
+        """Ensures the kontext latents and IDs match the target batch size by repeating if necessary."""
+        if self.kontext_latents.shape[0] != target_batch_size:
+            self.kontext_latents = self.kontext_latents.repeat(target_batch_size, 1, 1)
+            self.kontext_ids = self.kontext_ids.repeat(target_batch_size, 1, 1)

invokeai/backend/flux/sampling_utils.py

@@ -174,11 +174,13 @@ def generate_img_ids(h: int, w: int, batch_size: int, device: torch.device, dtyp
         dtype = torch.float16
 
     img_ids = torch.zeros(h // 2, w // 2, 3, device=device, dtype=dtype)
+    # Set batch offset to 0 for main image tokens
+    img_ids[..., 0] = 0
     img_ids[..., 1] = img_ids[..., 1] + torch.arange(h // 2, device=device, dtype=dtype)[:, None]
     img_ids[..., 2] = img_ids[..., 2] + torch.arange(w // 2, device=device, dtype=dtype)[None, :]
     img_ids = repeat(img_ids, "h w c -> b (h w) c", b=batch_size)
 
     if device.type == "mps":
-        img_ids.to(orig_dtype)
+        img_ids = img_ids.to(orig_dtype)
 
     return img_ids

invokeai/backend/flux/util.py

@@ -18,6 +18,29 @@ class ModelSpec:
     repo_ae: str | None
 
 
+# Preferred resolutions for Kontext models to avoid tiling artifacts
+# These are the specific resolutions the model was trained on
+PREFERED_KONTEXT_RESOLUTIONS = [
+    (672, 1568),
+    (688, 1504),
+    (720, 1456),
+    (752, 1392),
+    (800, 1328),
+    (832, 1248),
+    (880, 1184),
+    (944, 1104),
+    (1024, 1024),
+    (1104, 944),
+    (1184, 880),
+    (1248, 832),
+    (1328, 800),
+    (1392, 752),
+    (1456, 720),
+    (1504, 688),
+    (1568, 672),
+]
+
+
 max_seq_lengths: Dict[str, Literal[256, 512]] = {
     "flux-dev": 512,
     "flux-dev-fill": 512,

invokeai/backend/image_util/imwatermark/vendor.py

@@ -0,0 +1,304 @@
+# This file is vendored from https://github.com/ShieldMnt/invisible-watermark
+#
+# `invisible-watermark` is MIT licensed as of August 23, 2025, when the code was copied into this repo.
+#
+# Why we vendored it in:
+# `invisible-watermark` has a dependency on `opencv-python`, which conflicts with Invoke's dependency on
+# `opencv-contrib-python`. It's easier to copy the code over than complicate the installation process by
+# requiring an extra post-install step of removing `opencv-python` and installing `opencv-contrib-python`.
+
+import struct
+import uuid
+import base64
+import cv2
+import numpy as np
+import pywt
+
+
+class WatermarkEncoder(object):
+    def __init__(self, content=b""):
+        seq = np.array([n for n in content], dtype=np.uint8)
+        self._watermarks = list(np.unpackbits(seq))
+        self._wmLen = len(self._watermarks)
+        self._wmType = "bytes"
+
+    def set_by_ipv4(self, addr):
+        bits = []
+        ips = addr.split(".")
+        for ip in ips:
+            bits += list(np.unpackbits(np.array([ip % 255], dtype=np.uint8)))
+        self._watermarks = bits
+        self._wmLen = len(self._watermarks)
+        self._wmType = "ipv4"
+        assert self._wmLen == 32
+
+    def set_by_uuid(self, uid):
+        u = uuid.UUID(uid)
+        self._wmType = "uuid"
+        seq = np.array([n for n in u.bytes], dtype=np.uint8)
+        self._watermarks = list(np.unpackbits(seq))
+        self._wmLen = len(self._watermarks)
+
+    def set_by_bytes(self, content):
+        self._wmType = "bytes"
+        seq = np.array([n for n in content], dtype=np.uint8)
+        self._watermarks = list(np.unpackbits(seq))
+        self._wmLen = len(self._watermarks)
+
+    def set_by_b16(self, b16):
+        content = base64.b16decode(b16)
+        self.set_by_bytes(content)
+        self._wmType = "b16"
+
+    def set_by_bits(self, bits=[]):
+        self._watermarks = [int(bit) % 2 for bit in bits]
+        self._wmLen = len(self._watermarks)
+        self._wmType = "bits"
+
+    def set_watermark(self, wmType="bytes", content=""):
+        if wmType == "ipv4":
+            self.set_by_ipv4(content)
+        elif wmType == "uuid":
+            self.set_by_uuid(content)
+        elif wmType == "bits":
+            self.set_by_bits(content)
+        elif wmType == "bytes":
+            self.set_by_bytes(content)
+        elif wmType == "b16":
+            self.set_by_b16(content)
+        else:
+            raise NameError("%s is not supported" % wmType)
+
+    def get_length(self):
+        return self._wmLen
+
+    # @classmethod
+    # def loadModel(cls):
+    #     RivaWatermark.loadModel()
+
+    def encode(self, cv2Image, method="dwtDct", **configs):
+        (r, c, channels) = cv2Image.shape
+        if r * c < 256 * 256:
+            raise RuntimeError("image too small, should be larger than 256x256")
+
+        if method == "dwtDct":
+            embed = EmbedMaxDct(self._watermarks, wmLen=self._wmLen, **configs)
+            return embed.encode(cv2Image)
+        # elif method == 'dwtDctSvd':
+        #     embed = EmbedDwtDctSvd(self._watermarks, wmLen=self._wmLen, **configs)
+        #     return embed.encode(cv2Image)
+        # elif method == 'rivaGan':
+        #     embed = RivaWatermark(self._watermarks, self._wmLen)
+        #     return embed.encode(cv2Image)
+        else:
+            raise NameError("%s is not supported" % method)
+
+
+class WatermarkDecoder(object):
+    def __init__(self, wm_type="bytes", length=0):
+        self._wmType = wm_type
+        if wm_type == "ipv4":
+            self._wmLen = 32
+        elif wm_type == "uuid":
+            self._wmLen = 128
+        elif wm_type == "bytes":
+            self._wmLen = length
+        elif wm_type == "bits":
+            self._wmLen = length
+        elif wm_type == "b16":
+            self._wmLen = length
+        else:
+            raise NameError("%s is unsupported" % wm_type)
+
+    def reconstruct_ipv4(self, bits):
+        ips = [str(ip) for ip in list(np.packbits(bits))]
+        return ".".join(ips)
+
+    def reconstruct_uuid(self, bits):
+        nums = np.packbits(bits)
+        bstr = b""
+        for i in range(16):
+            bstr += struct.pack(">B", nums[i])
+
+        return str(uuid.UUID(bytes=bstr))
+
+    def reconstruct_bits(self, bits):
+        # return ''.join([str(b) for b in bits])
+        return bits
+
+    def reconstruct_b16(self, bits):
+        bstr = self.reconstruct_bytes(bits)
+        return base64.b16encode(bstr)
+
+    def reconstruct_bytes(self, bits):
+        nums = np.packbits(bits)
+        bstr = b""
+        for i in range(self._wmLen // 8):
+            bstr += struct.pack(">B", nums[i])
+        return bstr
+
+    def reconstruct(self, bits):
+        if len(bits) != self._wmLen:
+            raise RuntimeError("bits are not matched with watermark length")
+
+        if self._wmType == "ipv4":
+            return self.reconstruct_ipv4(bits)
+        elif self._wmType == "uuid":
+            return self.reconstruct_uuid(bits)
+        elif self._wmType == "bits":
+            return self.reconstruct_bits(bits)
+        elif self._wmType == "b16":
+            return self.reconstruct_b16(bits)
+        else:
+            return self.reconstruct_bytes(bits)
+
+    def decode(self, cv2Image, method="dwtDct", **configs):
+        (r, c, channels) = cv2Image.shape
+        if r * c < 256 * 256:
+            raise RuntimeError("image too small, should be larger than 256x256")
+
+        bits = []
+        if method == "dwtDct":
+            embed = EmbedMaxDct(watermarks=[], wmLen=self._wmLen, **configs)
+            bits = embed.decode(cv2Image)
+        # elif method == 'dwtDctSvd':
+        #     embed = EmbedDwtDctSvd(watermarks=[], wmLen=self._wmLen, **configs)
+        #     bits = embed.decode(cv2Image)
+        # elif method == 'rivaGan':
+        #     embed = RivaWatermark(watermarks=[], wmLen=self._wmLen, **configs)
+        #     bits = embed.decode(cv2Image)
+        else:
+            raise NameError("%s is not supported" % method)
+        return self.reconstruct(bits)
+
+    # @classmethod
+    # def loadModel(cls):
+    #     RivaWatermark.loadModel()
+
+
+class EmbedMaxDct(object):
+    def __init__(self, watermarks=[], wmLen=8, scales=[0, 36, 36], block=4):
+        self._watermarks = watermarks
+        self._wmLen = wmLen
+        self._scales = scales
+        self._block = block
+
+    def encode(self, bgr):
+        (row, col, channels) = bgr.shape
+
+        yuv = cv2.cvtColor(bgr, cv2.COLOR_BGR2YUV)
+
+        for channel in range(2):
+            if self._scales[channel] <= 0:
+                continue
+
+            ca1, (h1, v1, d1) = pywt.dwt2(yuv[: row // 4 * 4, : col // 4 * 4, channel], "haar")
+            self.encode_frame(ca1, self._scales[channel])
+
+            yuv[: row // 4 * 4, : col // 4 * 4, channel] = pywt.idwt2((ca1, (v1, h1, d1)), "haar")
+
+        bgr_encoded = cv2.cvtColor(yuv, cv2.COLOR_YUV2BGR)
+        return bgr_encoded
+
+    def decode(self, bgr):
+        (row, col, channels) = bgr.shape
+
+        yuv = cv2.cvtColor(bgr, cv2.COLOR_BGR2YUV)
+
+        scores = [[] for i in range(self._wmLen)]
+        for channel in range(2):
+            if self._scales[channel] <= 0:
+                continue
+
+            ca1, (h1, v1, d1) = pywt.dwt2(yuv[: row // 4 * 4, : col // 4 * 4, channel], "haar")
+
+            scores = self.decode_frame(ca1, self._scales[channel], scores)
+
+        avgScores = list(map(lambda l: np.array(l).mean(), scores))
+
+        bits = np.array(avgScores) * 255 > 127
+        return bits
+
+    def decode_frame(self, frame, scale, scores):
+        (row, col) = frame.shape
+        num = 0
+
+        for i in range(row // self._block):
+            for j in range(col // self._block):
+                block = frame[
+                    i * self._block : i * self._block + self._block, j * self._block : j * self._block + self._block
+                ]
+
+                score = self.infer_dct_matrix(block, scale)
+                # score = self.infer_dct_svd(block, scale)
+                wmBit = num % self._wmLen
+                scores[wmBit].append(score)
+                num = num + 1
+
+        return scores
+
+    def diffuse_dct_svd(self, block, wmBit, scale):
+        u, s, v = np.linalg.svd(cv2.dct(block))
+
+        s[0] = (s[0] // scale + 0.25 + 0.5 * wmBit) * scale
+        return cv2.idct(np.dot(u, np.dot(np.diag(s), v)))
+
+    def infer_dct_svd(self, block, scale):
+        u, s, v = np.linalg.svd(cv2.dct(block))
+
+        score = 0
+        score = int((s[0] % scale) > scale * 0.5)
+        return score
+        if score >= 0.5:
+            return 1.0
+        else:
+            return 0.0
+
+    def diffuse_dct_matrix(self, block, wmBit, scale):
+        pos = np.argmax(abs(block.flatten()[1:])) + 1
+        i, j = pos // self._block, pos % self._block
+        val = block[i][j]
+        if val >= 0.0:
+            block[i][j] = (val // scale + 0.25 + 0.5 * wmBit) * scale
+        else:
+            val = abs(val)
+            block[i][j] = -1.0 * (val // scale + 0.25 + 0.5 * wmBit) * scale
+        return block
+
+    def infer_dct_matrix(self, block, scale):
+        pos = np.argmax(abs(block.flatten()[1:])) + 1
+        i, j = pos // self._block, pos % self._block
+
+        val = block[i][j]
+        if val < 0:
+            val = abs(val)
+
+        if (val % scale) > 0.5 * scale:
+            return 1
+        else:
+            return 0
+
+    def encode_frame(self, frame, scale):
+        """
+        frame is a matrix (M, N)
+
+        we get K (watermark bits size) blocks (self._block x self._block)
+
+        For i-th block, we encode watermark[i] bit into it
+        """
+        (row, col) = frame.shape
+        num = 0
+        for i in range(row // self._block):
+            for j in range(col // self._block):
+                block = frame[
+                    i * self._block : i * self._block + self._block, j * self._block : j * self._block + self._block
+                ]
+                wmBit = self._watermarks[(num % self._wmLen)]
+
+                diffusedBlock = self.diffuse_dct_matrix(block, wmBit, scale)
+                # diffusedBlock = self.diffuse_dct_svd(block, wmBit, scale)
+                frame[
+                    i * self._block : i * self._block + self._block, j * self._block : j * self._block + self._block
+                ] = diffusedBlock
+
+                num = num + 1

invokeai/backend/image_util/invisible_watermark.py

@@ -6,13 +6,10 @@ configuration variable, that allows the watermarking to be supressed.
 
 import cv2
 import numpy as np
-from imwatermark import WatermarkEncoder
 from PIL import Image
 
 import invokeai.backend.util.logging as logger
-from invokeai.app.services.config.config_default import get_config
-
-config = get_config()
+from invokeai.backend.image_util.imwatermark.vendor import WatermarkEncoder
 
 
 class InvisibleWatermark:

invokeai/backend/image_util/segment_anything/segment_anything_2_pipeline.py

@@ -0,0 +1,109 @@
+from typing import Optional
+
+import torch
+from PIL import Image
+
+# Import SAM2 components - these should be available in transformers 4.56.0+
+from transformers.models.sam2 import Sam2Model
+from transformers.models.sam2.processing_sam2 import Sam2Processor
+
+from invokeai.backend.image_util.segment_anything.shared import SAMInput
+from invokeai.backend.raw_model import RawModel
+
+
+class SegmentAnything2Pipeline(RawModel):
+    """A wrapper class for the transformers SAM2 model and processor that makes it compatible with the model manager."""
+
+    def __init__(self, sam2_model: Sam2Model, sam2_processor: Sam2Processor):
+        """Initialize the SAM2 pipeline.
+
+        Args:
+            sam2_model: The SAM2 model
+            sam2_processor: The SAM2 processor (can be Sam2Processor or Sam2VideoProcessor)
+        """
+        self._sam2_model = sam2_model
+        self._sam2_processor = sam2_processor
+
+    def to(self, device: Optional[torch.device] = None, dtype: Optional[torch.dtype] = None):
+        # HACK: The SAM2 pipeline may not work on MPS devices. We only allow it to be moved to CPU or CUDA.
+        if device is not None and device.type not in {"cpu", "cuda"}:
+            device = None
+        self._sam2_model.to(device=device, dtype=dtype)
+
+    def calc_size(self) -> int:
+        # HACK: Fix the circular import issue.
+        from invokeai.backend.model_manager.load.model_util import calc_module_size
+
+        return calc_module_size(self._sam2_model)
+
+    def segment(
+        self,
+        image: Image.Image,
+        inputs: list[SAMInput],
+    ) -> torch.Tensor:
+        """Segment the image using the provided inputs.
+
+        Args:
+            image: The image to segment.
+            inputs: A list of SAMInput objects containing bounding boxes and/or point lists.
+
+        Returns:
+            torch.Tensor: The segmentation masks. dtype: torch.bool. shape: [num_masks, channels, height, width].
+        """
+
+        input_boxes: list[list[float]] = []
+        input_points: list[list[list[float]]] = []
+        input_labels: list[list[int]] = []
+
+        for i in inputs:
+            box: list[float] | None = None
+            points: list[list[float]] | None = None
+            labels: list[int] | None = None
+
+            if i.bounding_box is not None:
+                box: list[float] | None = [
+                    i.bounding_box.x_min,
+                    i.bounding_box.y_min,
+                    i.bounding_box.x_max,
+                    i.bounding_box.y_max,
+                ]
+
+            if i.points is not None:
+                points = []
+                labels = []
+                for point in i.points:
+                    points.append([point.x, point.y])
+                    labels.append(point.label.value)
+
+            if box is not None:
+                input_boxes.append(box)
+            if points is not None:
+                input_points.append(points)
+            if labels is not None:
+                input_labels.append(labels)
+
+        batched_input_boxes = [input_boxes] if input_boxes else None
+        batched_input_points = [input_points] if input_points else None
+        batched_input_labels = [input_labels] if input_labels else None
+
+        processed_inputs = self._sam2_processor(
+            images=image,
+            input_boxes=batched_input_boxes,
+            input_points=batched_input_points,
+            input_labels=batched_input_labels,
+            return_tensors="pt",
+        ).to(self._sam2_model.device)
+
+        # Generate masks using the SAM2 model
+        outputs = self._sam2_model(**processed_inputs)
+
+        # Post-process the masks to get the final segmentation
+        masks = self._sam2_processor.post_process_masks(
+            masks=outputs.pred_masks,
+            original_sizes=processed_inputs.original_sizes,
+            reshaped_input_sizes=processed_inputs.reshaped_input_sizes,
+        )
+
+        # There should be only one batch.
+        assert len(masks) == 1
+        return masks[0]

invokeai/backend/image_util/segment_anything/segment_anything_pipeline.py

@@ -1,20 +1,13 @@
-from typing import Optional, TypeAlias
+from typing import Optional
 
 import torch
 from PIL import Image
 from transformers.models.sam import SamModel
 from transformers.models.sam.processing_sam import SamProcessor
 
+from invokeai.backend.image_util.segment_anything.shared import SAMInput
 from invokeai.backend.raw_model import RawModel
 
-# Type aliases for the inputs to the SAM model.
-ListOfBoundingBoxes: TypeAlias = list[list[int]]
-"""A list of bounding boxes. Each bounding box is in the format [xmin, ymin, xmax, ymax]."""
-ListOfPoints: TypeAlias = list[list[int]]
-"""A list of points. Each point is in the format [x, y]."""
-ListOfPointLabels: TypeAlias = list[int]
-"""A list of SAM point labels. Each label is an integer where -1 is background, 0 is neutral, and 1 is foreground."""
-
 
 class SegmentAnythingPipeline(RawModel):
     """A wrapper class for the transformers SAM model and processor that makes it compatible with the model manager."""
@@ -38,55 +31,65 @@ class SegmentAnythingPipeline(RawModel):
     def segment(
         self,
         image: Image.Image,
-        bounding_boxes: list[list[int]] | None = None,
-        point_lists: list[list[list[int]]] | None = None,
+        inputs: list[SAMInput],
     ) -> torch.Tensor:
-        """Run the SAM model.
-
-        Either bounding_boxes or point_lists must be provided. If both are provided, bounding_boxes will be used and
-        point_lists will be ignored.
+        """Segment the image using the provided inputs.
 
         Args:
-            image (Image.Image): The image to segment.
-            bounding_boxes (list[list[int]]): The bounding box prompts. Each bounding box is in the format
-                [xmin, ymin, xmax, ymax].
-            point_lists (list[list[list[int]]]): The points prompts. Each point is in the format [x, y, label].
-                `label` is an integer where -1 is background, 0 is neutral, and 1 is foreground.
+            image: The image to segment.
+            inputs: A list of SAMInput objects containing bounding boxes and/or point lists.
 
         Returns:
             torch.Tensor: The segmentation masks. dtype: torch.bool. shape: [num_masks, channels, height, width].
         """
 
-        # Prep the inputs:
-        # - Create a list of bounding boxes or points and labels.
-        # - Add a batch dimension of 1 to the inputs.
-        if bounding_boxes:
-            input_boxes: list[ListOfBoundingBoxes] | None = [bounding_boxes]
-            input_points: list[ListOfPoints] | None = None
-            input_labels: list[ListOfPointLabels] | None = None
-        elif point_lists:
-            input_boxes: list[ListOfBoundingBoxes] | None = None
-            input_points: list[ListOfPoints] | None = []
-            input_labels: list[ListOfPointLabels] | None = []
-            for point_list in point_lists:
-                input_points.append([[p[0], p[1]] for p in point_list])
-                input_labels.append([p[2] for p in point_list])
+        input_boxes: list[list[float]] = []
+        input_points: list[list[list[float]]] = []
+        input_labels: list[list[int]] = []
 
-        else:
-            raise ValueError("Either bounding_boxes or points and labels must be provided.")
+        for i in inputs:
+            box: list[float] | None = None
+            points: list[list[float]] | None = None
+            labels: list[int] | None = None
 
-        inputs = self._sam_processor(
+            if i.bounding_box is not None:
+                box: list[float] | None = [
+                    i.bounding_box.x_min,
+                    i.bounding_box.y_min,
+                    i.bounding_box.x_max,
+                    i.bounding_box.y_max,
+                ]
+
+            if i.points is not None:
+                points = []
+                labels = []
+                for point in i.points:
+                    points.append([point.x, point.y])
+                    labels.append(point.label.value)
+
+            if box is not None:
+                input_boxes.append(box)
+            if points is not None:
+                input_points.append(points)
+            if labels is not None:
+                input_labels.append(labels)
+
+        batched_input_boxes = [input_boxes] if input_boxes else None
+        batched_input_points = input_points if input_points else None
+        batched_input_labels = input_labels if input_labels else None
+
+        processed_inputs = self._sam_processor(
             images=image,
-            input_boxes=input_boxes,
-            input_points=input_points,
-            input_labels=input_labels,
+            input_boxes=batched_input_boxes,
+            input_points=batched_input_points,
+            input_labels=batched_input_labels,
             return_tensors="pt",
         ).to(self._sam_model.device)
-        outputs = self._sam_model(**inputs)
+        outputs = self._sam_model(**processed_inputs)
         masks = self._sam_processor.post_process_masks(
             masks=outputs.pred_masks,
-            original_sizes=inputs.original_sizes,
-            reshaped_input_sizes=inputs.reshaped_input_sizes,
+            original_sizes=processed_inputs.original_sizes,
+            reshaped_input_sizes=processed_inputs.reshaped_input_sizes,
         )
 
         # There should be only one batch.

invokeai/backend/image_util/segment_anything/shared.py

@@ -0,0 +1,49 @@
+from enum import Enum
+
+from pydantic import BaseModel, model_validator
+from pydantic.fields import Field
+
+
+class BoundingBox(BaseModel):
+    x_min: int = Field(..., description="The minimum x-coordinate of the bounding box (inclusive).")
+    x_max: int = Field(..., description="The maximum x-coordinate of the bounding box (exclusive).")
+    y_min: int = Field(..., description="The minimum y-coordinate of the bounding box (inclusive).")
+    y_max: int = Field(..., description="The maximum y-coordinate of the bounding box (exclusive).")
+
+    @model_validator(mode="after")
+    def check_coords(self):
+        if self.x_min > self.x_max:
+            raise ValueError(f"x_min ({self.x_min}) is greater than x_max ({self.x_max}).")
+        if self.y_min > self.y_max:
+            raise ValueError(f"y_min ({self.y_min}) is greater than y_max ({self.y_max}).")
+        return self
+
+    def tuple(self) -> tuple[int, int, int, int]:
+        """
+        Returns the bounding box as a tuple suitable for use with PIL's `Image.crop()` method.
+        This method returns a tuple of the form (left, upper, right, lower) == (x_min, y_min, x_max, y_max).
+        """
+        return (self.x_min, self.y_min, self.x_max, self.y_max)
+
+
+class SAMPointLabel(Enum):
+    negative = -1
+    neutral = 0
+    positive = 1
+
+
+class SAMPoint(BaseModel):
+    x: int = Field(..., description="The x-coordinate of the point")
+    y: int = Field(..., description="The y-coordinate of the point")
+    label: SAMPointLabel = Field(..., description="The label of the point")
+
+
+class SAMInput(BaseModel):
+    bounding_box: BoundingBox | None = Field(None, description="The bounding box to use for segmentation")
+    points: list[SAMPoint] | None = Field(None, description="The points to use for segmentation")
+
+    @model_validator(mode="after")
+    def check_input(self):
+        if not self.bounding_box and not self.points:
+            raise ValueError("Either bounding_box or points must be provided")
+        return self

invokeai/backend/model_manager/config.py

@@ -90,6 +90,11 @@ class MainModelDefaultSettings(BaseModel):
     model_config = ConfigDict(extra="forbid")
 
 
+class LoraModelDefaultSettings(BaseModel):
+    weight: float | None = Field(default=None, ge=-1, le=2, description="Default weight for this model")
+    model_config = ConfigDict(extra="forbid")
+
+
 class ControlAdapterDefaultSettings(BaseModel):
     # This could be narrowed to controlnet processor nodes, but they change. Leaving this a string is safer.
     preprocessor: str | None
@@ -187,7 +192,7 @@ class ModelConfigBase(ABC, BaseModel):
             else:
                 return config_cls.from_model_on_disk(mod, **overrides)
 
-        raise InvalidModelConfigException("No valid config found")
+        raise InvalidModelConfigException("Unable to determine model type")
 
     @classmethod
     def get_tag(cls) -> Tag:
@@ -287,6 +292,9 @@ class LoRAConfigBase(ABC, BaseModel):
 
     type: Literal[ModelType.LoRA] = ModelType.LoRA
     trigger_phrases: Optional[set[str]] = Field(description="Set of trigger phrases for this model", default=None)
+    default_settings: Optional[LoraModelDefaultSettings] = Field(
+        description="Default settings for this model", default=None
+    )
 
     @classmethod
     def flux_lora_format(cls, mod: ModelOnDisk):
@@ -492,6 +500,15 @@ class MainConfigBase(ABC, BaseModel):
     variant: AnyVariant = ModelVariantType.Normal
 
 
+class VideoConfigBase(ABC, BaseModel):
+    type: Literal[ModelType.Video] = ModelType.Video
+    trigger_phrases: Optional[set[str]] = Field(description="Set of trigger phrases for this model", default=None)
+    default_settings: Optional[MainModelDefaultSettings] = Field(
+        description="Default settings for this model", default=None
+    )
+    variant: AnyVariant = ModelVariantType.Normal
+
+
 class MainCheckpointConfig(CheckpointConfigBase, MainConfigBase, LegacyProbeMixin, ModelConfigBase):
     """Model config for main checkpoint models."""
 
@@ -649,6 +666,21 @@ class ApiModelConfig(MainConfigBase, ModelConfigBase):
         raise NotImplementedError("API models are not parsed from disk.")
 
 
+class VideoApiModelConfig(VideoConfigBase, ModelConfigBase):
+    """Model config for API-based video models."""
+
+    format: Literal[ModelFormat.Api] = ModelFormat.Api
+
+    @classmethod
+    def matches(cls, mod: ModelOnDisk) -> bool:
+        # API models are not stored on disk, so we can't match them.
+        return False
+
+    @classmethod
+    def parse(cls, mod: ModelOnDisk) -> dict[str, Any]:
+        raise NotImplementedError("API models are not parsed from disk.")
+
+
 def get_model_discriminator_value(v: Any) -> str:
     """
     Computes the discriminator value for a model config.
@@ -718,12 +750,13 @@ AnyModelConfig = Annotated[
         Annotated[FluxReduxConfig, FluxReduxConfig.get_tag()],
         Annotated[LlavaOnevisionConfig, LlavaOnevisionConfig.get_tag()],
         Annotated[ApiModelConfig, ApiModelConfig.get_tag()],
+        Annotated[VideoApiModelConfig, VideoApiModelConfig.get_tag()],
     ],
     Discriminator(get_model_discriminator_value),
 ]
 
 AnyModelConfigValidator = TypeAdapter(AnyModelConfig)
-AnyDefaultSettings: TypeAlias = Union[MainModelDefaultSettings, ControlAdapterDefaultSettings]
+AnyDefaultSettings: TypeAlias = Union[MainModelDefaultSettings, LoraModelDefaultSettings, ControlAdapterDefaultSettings]
 
 
 class ModelConfigFactory:

invokeai/backend/model_manager/legacy_probe.py

@@ -9,6 +9,7 @@ import spandrel
 import torch
 
 import invokeai.backend.util.logging as logger
+from invokeai.app.services.config.config_default import get_config
 from invokeai.app.util.misc import uuid_string
 from invokeai.backend.flux.controlnet.state_dict_utils import (
     is_state_dict_instantx_controlnet,
@@ -22,6 +23,7 @@ from invokeai.backend.model_manager.config import (
     AnyModelConfig,
     ControlAdapterDefaultSettings,
     InvalidModelConfigException,
+    LoraModelDefaultSettings,
     MainModelDefaultSettings,
     ModelConfigFactory,
     SubmodelDefinition,
@@ -216,6 +218,8 @@ class ModelProbe(object):
         if not fields["default_settings"]:
             if fields["type"] in {ModelType.ControlNet, ModelType.T2IAdapter, ModelType.ControlLoRa}:
                 fields["default_settings"] = get_default_settings_control_adapters(fields["name"])
+            if fields["type"] in {ModelType.LoRA}:
+                fields["default_settings"] = get_default_settings_lora()
             elif fields["type"] is ModelType.Main:
                 fields["default_settings"] = get_default_settings_main(fields["base"])
 
@@ -493,9 +497,21 @@ class ModelProbe(object):
         # scan model
         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 get_config().unsafe_disable_picklescan:
+                logger.warning(
+                    f"The model {model_name} is potentially infected by malware, but picklescan is disabled. "
+                    "Proceeding with caution."
+                )
+            else:
+                raise RuntimeError(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.")
+            if get_config().unsafe_disable_picklescan:
+                logger.warning(
+                    f"Error scanning the model at {model_name} for malware, but picklescan is disabled. "
+                    "Proceeding with caution."
+                )
+            else:
+                raise RuntimeError(f"Error scanning the model at {model_name} for malware. Aborting import.")
 
 
 # Probing utilities
@@ -530,6 +546,10 @@ def get_default_settings_control_adapters(model_name: str) -> Optional[ControlAd
     return None
 
 
+def get_default_settings_lora() -> LoraModelDefaultSettings:
+    return LoraModelDefaultSettings()
+
+
 def get_default_settings_main(model_base: BaseModelType) -> Optional[MainModelDefaultSettings]:
     if model_base is BaseModelType.StableDiffusion1 or model_base is BaseModelType.StableDiffusion2:
         return MainModelDefaultSettings(width=512, height=512)

invokeai/backend/model_manager/model_on_disk.py

@@ -6,13 +6,17 @@ import torch
 from picklescan.scanner import scan_file_path
 from safetensors import safe_open
 
+from invokeai.app.services.config.config_default import get_config
 from invokeai.backend.model_hash.model_hash import HASHING_ALGORITHMS, ModelHash
 from invokeai.backend.model_manager.taxonomy import ModelRepoVariant
 from invokeai.backend.quantization.gguf.loaders import gguf_sd_loader
+from invokeai.backend.util.logging import InvokeAILogger
 from invokeai.backend.util.silence_warnings import SilenceWarnings
 
 StateDict: TypeAlias = dict[str | int, Any]  # When are the keys int?
 
+logger = InvokeAILogger.get_logger()
+
 
 class ModelOnDisk:
     """A utility class representing a model stored on disk."""
@@ -79,8 +83,24 @@ class ModelOnDisk:
         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.")
+                if scan_result.infected_files != 0:
+                    if get_config().unsafe_disable_picklescan:
+                        logger.warning(
+                            f"The model {path.stem} is potentially infected by malware, but picklescan is disabled. "
+                            "Proceeding with caution."
+                        )
+                    else:
+                        raise RuntimeError(
+                            f"The model {path.stem} is potentially infected by malware. Aborting import."
+                        )
+                if scan_result.scan_err:
+                    if get_config().unsafe_disable_picklescan:
+                        logger.warning(
+                            f"Error scanning the model at {path.stem} for malware, but picklescan is disabled. "
+                            "Proceeding with caution."
+                        )
+                    else:
+                        raise RuntimeError(f"Error scanning the model at {path.stem} for malware. Aborting import.")
                 checkpoint = torch.load(path, map_location="cpu")
                 assert isinstance(checkpoint, dict)
             elif path.suffix.endswith(".gguf"):

invokeai/backend/model_manager/starter_models.py

@@ -143,11 +143,35 @@ flux_dev = StarterModel(
 flux_kontext = StarterModel(
     name="FLUX.1 Kontext dev",
     base=BaseModelType.Flux,
-    source="black-forest-labs/FLUX.1-Kontext-dev::flux1-kontext-dev.safetensors",
+    source="https://huggingface.co/black-forest-labs/FLUX.1-Kontext-dev/resolve/main/flux1-kontext-dev.safetensors",
     description="FLUX.1 Kontext dev transformer in bfloat16. Total size with dependencies: ~33GB",
     type=ModelType.Main,
     dependencies=[t5_base_encoder, flux_vae, clip_l_encoder],
 )
+flux_kontext_quantized = StarterModel(
+    name="FLUX.1 Kontext dev (quantized)",
+    base=BaseModelType.Flux,
+    source="https://huggingface.co/unsloth/FLUX.1-Kontext-dev-GGUF/resolve/main/flux1-kontext-dev-Q4_K_M.gguf",
+    description="FLUX.1 Kontext dev quantized (q4_k_m). Total size with dependencies: ~14GB",
+    type=ModelType.Main,
+    dependencies=[t5_8b_quantized_encoder, flux_vae, clip_l_encoder],
+)
+flux_krea = StarterModel(
+    name="FLUX.1 Krea dev",
+    base=BaseModelType.Flux,
+    source="https://huggingface.co/InvokeAI/FLUX.1-Krea-dev/resolve/main/flux1-krea-dev.safetensors",
+    description="FLUX.1 Krea dev. Total size with dependencies: ~33GB",
+    type=ModelType.Main,
+    dependencies=[t5_8b_quantized_encoder, flux_vae, clip_l_encoder],
+)
+flux_krea_quantized = StarterModel(
+    name="FLUX.1 Krea dev (quantized)",
+    base=BaseModelType.Flux,
+    source="https://huggingface.co/InvokeAI/FLUX.1-Krea-dev-GGUF/resolve/main/flux1-krea-dev-Q4_K_M.gguf",
+    description="FLUX.1 Krea dev quantized (q4_k_m). Total size with dependencies: ~14GB",
+    type=ModelType.Main,
+    dependencies=[t5_8b_quantized_encoder, flux_vae, clip_l_encoder],
+)
 sd35_medium = StarterModel(
     name="SD3.5 Medium",
     base=BaseModelType.StableDiffusion3,
@@ -572,13 +596,14 @@ t2i_sketch_sdxl = StarterModel(
 )
 # endregion
 # region SpandrelImageToImage
-realesrgan_anime = StarterModel(
-    name="RealESRGAN_x4plus_anime_6B",
+animesharp_v4_rcan = StarterModel(
+    name="2x-AnimeSharpV4_RCAN",
     base=BaseModelType.Any,
-    source="https://github.com/xinntao/Real-ESRGAN/releases/download/v0.2.2.4/RealESRGAN_x4plus_anime_6B.pth",
-    description="A Real-ESRGAN 4x upscaling model (optimized for anime images).",
+    source="https://github.com/Kim2091/Kim2091-Models/releases/download/2x-AnimeSharpV4/2x-AnimeSharpV4_RCAN.safetensors",
+    description="A 2x upscaling model (optimized for anime images).",
     type=ModelType.SpandrelImageToImage,
 )
+
 realesrgan_x4 = StarterModel(
     name="RealESRGAN_x4plus",
     base=BaseModelType.Any,
@@ -664,7 +689,7 @@ flux_fill = StarterModel(
 # 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] = [
-    flux_kontext,
+    flux_kontext_quantized,
     flux_schnell_quantized,
     flux_dev_quantized,
     flux_schnell,
@@ -724,7 +749,7 @@ STARTER_MODELS: list[StarterModel] = [
     t2i_lineart_sdxl,
     t2i_sketch_sdxl,
     realesrgan_x4,
-    realesrgan_anime,
+    animesharp_v4_rcan,
     realesrgan_x2,
     swinir,
     t5_base_encoder,
@@ -735,6 +760,8 @@ STARTER_MODELS: list[StarterModel] = [
     llava_onevision,
     flux_fill,
     cogview4,
+    flux_krea,
+    flux_krea_quantized,
 ]
 
 sd1_bundle: list[StarterModel] = [
@@ -785,7 +812,8 @@ flux_bundle: list[StarterModel] = [
     flux_depth_control_lora,
     flux_redux,
     flux_fill,
-    flux_kontext,
+    flux_kontext_quantized,
+    flux_krea_quantized,
 ]
 
 STARTER_BUNDLES: dict[str, StarterModelBundle] = {

invokeai/backend/model_manager/taxonomy.py

@@ -28,8 +28,11 @@ class BaseModelType(str, Enum):
     CogView4 = "cogview4"
     Imagen3 = "imagen3"
     Imagen4 = "imagen4"
+    Gemini2_5 = "gemini-2.5"
     ChatGPT4o = "chatgpt-4o"
     FluxKontext = "flux-kontext"
+    Veo3 = "veo3"
+    Runway = "runway"
 
 
 class ModelType(str, Enum):
@@ -51,6 +54,7 @@ class ModelType(str, Enum):
     SigLIP = "siglip"
     FluxRedux = "flux_redux"
     LlavaOnevision = "llava_onevision"
+    Video = "video"
 
 
 class SubModelType(str, Enum):

invokeai/backend/model_manager/util/lora_metadata_extractor.py

@@ -0,0 +1,145 @@
+"""Utility functions for extracting metadata from LoRA model files."""
+
+import json
+import logging
+from pathlib import Path
+from typing import Any, Dict, Optional, Set, Tuple
+
+from PIL import Image
+
+from invokeai.app.util.thumbnails import make_thumbnail
+from invokeai.backend.model_manager.config import AnyModelConfig, ModelType
+
+logger = logging.getLogger(__name__)
+
+
+def extract_lora_metadata(
+    model_path: Path, model_key: str, model_images_path: Path
+) -> Tuple[Optional[str], Optional[Set[str]]]:
+    """
+    Extract metadata for a LoRA model from associated JSON and image files.
+
+    Args:
+        model_path: Path to the LoRA model file
+        model_key: Unique key for the model
+        model_images_path: Path to the model images directory
+
+    Returns:
+        Tuple of (description, trigger_phrases)
+    """
+    model_stem = model_path.stem
+    model_dir = model_path.parent
+
+    # Find and process preview image
+    _process_preview_image(model_stem, model_dir, model_key, model_images_path)
+
+    # Extract metadata from JSON
+    description, trigger_phrases = _extract_json_metadata(model_stem, model_dir)
+
+    return description, trigger_phrases
+
+
+def _process_preview_image(model_stem: str, model_dir: Path, model_key: str, model_images_path: Path) -> bool:
+    """Find and process a preview image for the model, saving it to the model images store."""
+    image_extensions = [".png", ".jpg", ".jpeg", ".webp"]
+
+    for ext in image_extensions:
+        image_path = model_dir / f"{model_stem}{ext}"
+        if image_path.exists():
+            try:
+                # Open the image
+                with Image.open(image_path) as img:
+                    # Create thumbnail and save to model images directory
+                    thumbnail = make_thumbnail(img, 256)
+                    thumbnail_path = model_images_path / f"{model_key}.webp"
+                    thumbnail.save(thumbnail_path, format="webp")
+
+                logger.info(f"Processed preview image {image_path.name} for model {model_key}")
+                return True
+
+            except Exception as e:
+                logger.warning(f"Failed to process preview image {image_path.name}: {e}")
+                return False
+
+    return False
+
+
+def _extract_json_metadata(model_stem: str, model_dir: Path) -> Tuple[Optional[str], Optional[Set[str]]]:
+    """Extract metadata from a JSON file with the same name as the model."""
+    json_path = model_dir / f"{model_stem}.json"
+
+    if not json_path.exists():
+        return None, None
+
+    try:
+        with open(json_path, "r", encoding="utf-8") as f:
+            metadata = json.load(f)
+
+        # Extract description
+        description = _build_description(metadata)
+
+        # Extract trigger phrases
+        trigger_phrases = _extract_trigger_phrases(metadata)
+
+        if description or trigger_phrases:
+            logger.info(f"Applied metadata from {json_path.name}")
+
+        return description, trigger_phrases
+
+    except (json.JSONDecodeError, IOError, Exception) as e:
+        logger.warning(f"Failed to read metadata from {json_path}: {e}")
+        return None, None
+
+
+def _build_description(metadata: Dict[str, Any]) -> Optional[str]:
+    """Build a description from metadata fields."""
+    description_parts = []
+
+    if description := metadata.get("description"):
+        description_parts.append(str(description).strip())
+
+    if notes := metadata.get("notes"):
+        description_parts.append(str(notes).strip())
+
+    return " | ".join(description_parts) if description_parts else None
+
+
+def _extract_trigger_phrases(metadata: Dict[str, Any]) -> Optional[Set[str]]:
+    """Extract trigger phrases from metadata."""
+    if not (activation_text := metadata.get("activation text")):
+        return None
+
+    activation_text = str(activation_text).strip()
+    if not activation_text:
+        return None
+
+    # Split on commas and clean up each phrase
+    phrases = [phrase.strip() for phrase in activation_text.split(",") if phrase.strip()]
+
+    return set(phrases) if phrases else None
+
+
+def apply_lora_metadata(info: AnyModelConfig, model_path: Path, model_images_path: Path) -> None:
+    """
+    Apply extracted metadata to a LoRA model configuration.
+
+    Args:
+        info: The model configuration to update
+        model_path: Path to the LoRA model file
+        model_images_path: Path to the model images directory
+    """
+    # Only process LoRA models
+    if info.type != ModelType.LoRA:
+        return
+
+    # Extract and apply metadata
+    description, trigger_phrases = extract_lora_metadata(model_path, info.key, model_images_path)
+
+    # We don't set cover_image path in the config anymore since images are stored
+    # separately in the model images store by model key
+
+    if description:
+        info.description = description
+
+    if trigger_phrases:
+        info.trigger_phrases = trigger_phrases

invokeai/backend/model_manager/util/model_util.py

@@ -8,8 +8,12 @@ import picklescan.scanner as pscan
 import safetensors
 import torch
 
+from invokeai.app.services.config.config_default import get_config
 from invokeai.backend.model_manager.taxonomy import ClipVariantType
 from invokeai.backend.quantization.gguf.loaders import gguf_sd_loader
+from invokeai.backend.util.logging import InvokeAILogger
+
+logger = InvokeAILogger.get_logger()
 
 
 def _fast_safetensors_reader(path: str) -> Dict[str, torch.Tensor]:
@@ -59,9 +63,21 @@ def read_checkpoint_meta(path: Union[str, Path], scan: bool = True) -> Dict[str,
         if scan:
             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 get_config().unsafe_disable_picklescan:
+                    logger.warning(
+                        f"The model {path} is potentially infected by malware, but picklescan is disabled. "
+                        "Proceeding with caution."
+                    )
+                else:
+                    raise RuntimeError(f"The model {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.")
+                if get_config().unsafe_disable_picklescan:
+                    logger.warning(
+                        f"Error scanning the model at {path} for malware, but picklescan is disabled. "
+                        "Proceeding with caution."
+                    )
+                else:
+                    raise RuntimeError(f"Error scanning the model at {path} for malware. Aborting import.")
 
         checkpoint = torch.load(path, map_location=torch.device("meta"))
     return checkpoint

invokeai/backend/patches/lora_conversions/flux_diffusers_lora_conversion_utils.py

@@ -18,16 +18,25 @@ def is_state_dict_likely_in_flux_diffusers_format(state_dict: Dict[str, torch.Te
     # First, check that all keys end in "lora_A.weight" or "lora_B.weight" (i.e. are in PEFT format).
     all_keys_in_peft_format = all(k.endswith(("lora_A.weight", "lora_B.weight")) for k in state_dict.keys())
 
-    # Next, check that this is likely a FLUX model by spot-checking a few keys.
-    expected_keys = [
+    # Check if keys use transformer prefix
+    transformer_prefix_keys = [
         "transformer.single_transformer_blocks.0.attn.to_q.lora_A.weight",
         "transformer.single_transformer_blocks.0.attn.to_q.lora_B.weight",
         "transformer.transformer_blocks.0.attn.add_q_proj.lora_A.weight",
         "transformer.transformer_blocks.0.attn.add_q_proj.lora_B.weight",
     ]
-    all_expected_keys_present = all(k in state_dict for k in expected_keys)
+    transformer_keys_present = all(k in state_dict for k in transformer_prefix_keys)
 
-    return all_keys_in_peft_format and all_expected_keys_present
+    # Check if keys use base_model.model prefix
+    base_model_prefix_keys = [
+        "base_model.model.single_transformer_blocks.0.attn.to_q.lora_A.weight",
+        "base_model.model.single_transformer_blocks.0.attn.to_q.lora_B.weight",
+        "base_model.model.transformer_blocks.0.attn.add_q_proj.lora_A.weight",
+        "base_model.model.transformer_blocks.0.attn.add_q_proj.lora_B.weight",
+    ]
+    base_model_keys_present = all(k in state_dict for k in base_model_prefix_keys)
+
+    return all_keys_in_peft_format and (transformer_keys_present or base_model_keys_present)
 
 
 def lora_model_from_flux_diffusers_state_dict(
@@ -49,8 +58,16 @@ def lora_layers_from_flux_diffusers_grouped_state_dict(
     https://github.com/huggingface/diffusers/blob/55ac421f7bb12fd00ccbef727be4dc2f3f920abb/scripts/convert_flux_to_diffusers.py
     """
 
-    # Remove the "transformer." prefix from all keys.
-    grouped_state_dict = {k.replace("transformer.", ""): v for k, v in grouped_state_dict.items()}
+    # Determine which prefix is used and remove it from all keys.
+    # Check if any key starts with "base_model.model." prefix
+    has_base_model_prefix = any(k.startswith("base_model.model.") for k in grouped_state_dict.keys())
+
+    if has_base_model_prefix:
+        # Remove the "base_model.model." prefix from all keys.
+        grouped_state_dict = {k.replace("base_model.model.", ""): v for k, v in grouped_state_dict.items()}
+    else:
+        # Remove the "transformer." prefix from all keys.
+        grouped_state_dict = {k.replace("transformer.", ""): v for k, v in grouped_state_dict.items()}
 
     # Constants for FLUX.1
     num_double_layers = 19

invokeai/backend/quantization/scripts/load_flux_model_bnb_llm_int8.py

@@ -20,7 +20,7 @@ def main():
         "/data/invokeai/models/.download_cache/https__huggingface.co_black-forest-labs_flux.1-schnell_resolve_main_flux1-schnell.safetensors/flux1-schnell.safetensors"
     )
 
-    with log_time("Intialize FLUX transformer on meta device"):
+    with log_time("Initialize FLUX transformer on meta device"):
         # TODO(ryand): Determine if this is a schnell model or a dev model and load the appropriate config.
         p = params["flux-schnell"]
 

invokeai/backend/quantization/scripts/load_flux_model_bnb_nf4.py

@@ -33,7 +33,7 @@ def main():
     )
 
     # inference_dtype = torch.bfloat16
-    with log_time("Intialize FLUX transformer on meta device"):
+    with log_time("Initialize FLUX transformer on meta device"):
         # TODO(ryand): Determine if this is a schnell model or a dev model and load the appropriate config.
         p = params["flux-schnell"]
 

invokeai/backend/quantization/scripts/quantize_t5_xxl_bnb_llm_int8.py

@@ -27,7 +27,7 @@ def main():
     """
     model_path = Path("/data/misc/text_encoder_2")
 
-    with log_time("Intialize T5 on meta device"):
+    with log_time("Initialize T5 on meta device"):
         model_config = AutoConfig.from_pretrained(model_path)
         with accelerate.init_empty_weights():
             model = AutoModelForTextEncoding.from_config(model_config)

invokeai/backend/util/hotfixes.py

@@ -5,7 +5,11 @@ import torch
 from diffusers.configuration_utils import ConfigMixin, register_to_config
 from diffusers.loaders.single_file_model import FromOriginalModelMixin
 from diffusers.models.attention_processor import AttentionProcessor, AttnProcessor
-from diffusers.models.controlnet import ControlNetConditioningEmbedding, ControlNetOutput, zero_module
+from diffusers.models.controlnets.controlnet import (
+    ControlNetConditioningEmbedding,
+    ControlNetOutput,
+    zero_module,
+)
 from diffusers.models.embeddings import (
     TextImageProjection,
     TextImageTimeEmbedding,
@@ -775,7 +779,15 @@ class ControlNetModel(ModelMixin, ConfigMixin, FromOriginalModelMixin):
 
 
 diffusers.ControlNetModel = ControlNetModel
-diffusers.models.controlnet.ControlNetModel = ControlNetModel
+# Patch both the new and legacy module paths for compatibility
+try:
+    diffusers.models.controlnets.controlnet.ControlNetModel = ControlNetModel
+except Exception:
+    # Fallback for environments still exposing the legacy path
+    try:
+        diffusers.models.controlnet.ControlNetModel = ControlNetModel
+    except Exception:
+        pass
 
 
 # patch LoRACompatibleConv to use original Conv2D forward function

invokeai/backend/util/vae_working_memory.py

@@ -0,0 +1,117 @@
+from typing import Literal
+
+import torch
+from diffusers.models.autoencoders.autoencoder_kl import AutoencoderKL
+from diffusers.models.autoencoders.autoencoder_tiny import AutoencoderTiny
+
+from invokeai.app.invocations.constants import LATENT_SCALE_FACTOR
+from invokeai.backend.flux.modules.autoencoder import AutoEncoder
+
+
+def estimate_vae_working_memory_sd15_sdxl(
+    operation: Literal["encode", "decode"],
+    image_tensor: torch.Tensor,
+    vae: AutoencoderKL | AutoencoderTiny,
+    tile_size: int | None,
+    fp32: bool,
+) -> int:
+    """Estimate the working memory required to encode or decode the given tensor."""
+    # It was found experimentally that the peak working memory scales linearly with the number of pixels and the
+    # element size (precision). This estimate is accurate for both SD1 and SDXL.
+    element_size = 4 if fp32 else 2
+
+    # This constant is determined experimentally and takes into consideration both allocated and reserved memory. See #8414
+    # Encoding uses ~45% the working memory as decoding.
+    scaling_constant = 2200 if operation == "decode" else 1100
+
+    latent_scale_factor_for_operation = LATENT_SCALE_FACTOR if operation == "decode" else 1
+
+    if tile_size is not None:
+        if tile_size == 0:
+            tile_size = vae.tile_sample_min_size
+            assert isinstance(tile_size, int)
+        h = tile_size
+        w = tile_size
+        working_memory = h * w * element_size * scaling_constant
+
+        # We add 25% to the working memory estimate when tiling is enabled to account for factors like tile overlap
+        # and number of tiles. We could make this more precise in the future, but this should be good enough for
+        # most use cases.
+        working_memory = working_memory * 1.25
+    else:
+        h = latent_scale_factor_for_operation * image_tensor.shape[-2]
+        w = latent_scale_factor_for_operation * image_tensor.shape[-1]
+        working_memory = h * w * element_size * scaling_constant
+
+    if fp32:
+        # If we are running in FP32, then we should account for the likely increase in model size (~250MB).
+        working_memory += 250 * 2**20
+
+    print(f"estimate_vae_working_memory_sd15_sdxl: {int(working_memory)}")
+
+    return int(working_memory)
+
+
+def estimate_vae_working_memory_cogview4(
+    operation: Literal["encode", "decode"], image_tensor: torch.Tensor, vae: AutoencoderKL
+) -> int:
+    """Estimate the working memory required by the invocation in bytes."""
+    latent_scale_factor_for_operation = LATENT_SCALE_FACTOR if operation == "decode" else 1
+
+    h = latent_scale_factor_for_operation * image_tensor.shape[-2]
+    w = latent_scale_factor_for_operation * image_tensor.shape[-1]
+    element_size = next(vae.parameters()).element_size()
+
+    # This constant is determined experimentally and takes into consideration both allocated and reserved memory. See #8414
+    # Encoding uses ~45% the working memory as decoding.
+    scaling_constant = 2200 if operation == "decode" else 1100
+    working_memory = h * w * element_size * scaling_constant
+
+    print(f"estimate_vae_working_memory_cogview4: {int(working_memory)}")
+
+    return int(working_memory)
+
+
+def estimate_vae_working_memory_flux(
+    operation: Literal["encode", "decode"], image_tensor: torch.Tensor, vae: AutoEncoder
+) -> int:
+    """Estimate the working memory required by the invocation in bytes."""
+
+    latent_scale_factor_for_operation = LATENT_SCALE_FACTOR if operation == "decode" else 1
+
+    out_h = latent_scale_factor_for_operation * image_tensor.shape[-2]
+    out_w = latent_scale_factor_for_operation * image_tensor.shape[-1]
+    element_size = next(vae.parameters()).element_size()
+
+    # This constant is determined experimentally and takes into consideration both allocated and reserved memory. See #8414
+    # Encoding uses ~45% the working memory as decoding.
+    scaling_constant = 2200 if operation == "decode" else 1100
+
+    working_memory = out_h * out_w * element_size * scaling_constant
+
+    print(f"estimate_vae_working_memory_flux: {int(working_memory)}")
+
+    return int(working_memory)
+
+
+def estimate_vae_working_memory_sd3(
+    operation: Literal["encode", "decode"], image_tensor: torch.Tensor, vae: AutoencoderKL
+) -> int:
+    """Estimate the working memory required by the invocation in bytes."""
+    # Encode operations use approximately 50% of the memory required for decode operations
+
+    latent_scale_factor_for_operation = LATENT_SCALE_FACTOR if operation == "decode" else 1
+
+    h = latent_scale_factor_for_operation * image_tensor.shape[-2]
+    w = latent_scale_factor_for_operation * image_tensor.shape[-1]
+    element_size = next(vae.parameters()).element_size()
+
+    # This constant is determined experimentally and takes into consideration both allocated and reserved memory. See #8414
+    # Encoding uses ~45% the working memory as decoding.
+    scaling_constant = 2200 if operation == "decode" else 1100
+
+    working_memory = h * w * element_size * scaling_constant
+
+    print(f"estimate_vae_working_memory_sd3: {int(working_memory)}")
+
+    return int(working_memory)

invokeai/frontend/web/.eslintignore

@@ -1,10 +0,0 @@
-dist/
-static/
-.husky/
-node_modules/
-patches/
-stats.html
-index.html
-.yarn/
-*.scss
-src/services/api/schema.ts

invokeai/frontend/web/.eslintrc.js

@@ -1,70 +0,0 @@
-module.exports = {
-  extends: ['@invoke-ai/eslint-config-react'],
-  plugins: ['path', 'i18next'],
-  rules: {
-    // TODO(psyche): Enable this rule. Requires no default exports in components - many changes.
-    'react-refresh/only-export-components': 'off',
-    // TODO(psyche): Enable this rule. Requires a lot of eslint-disable-next-line comments.
-    '@typescript-eslint/consistent-type-assertions': 'off',
-    // https://github.com/qdanik/eslint-plugin-path
-    'path/no-relative-imports': ['error', { maxDepth: 0 }],
-    // https://github.com/edvardchen/eslint-plugin-i18next/blob/HEAD/docs/rules/no-literal-string.md
-    // TODO: ENABLE THIS RULE BEFORE v6.0.0
-    // 'i18next/no-literal-string': 'error',
-    // https://eslint.org/docs/latest/rules/no-console
-    'no-console': 'warn',
-    // https://eslint.org/docs/latest/rules/no-promise-executor-return
-    'no-promise-executor-return': 'error',
-    // https://eslint.org/docs/latest/rules/require-await
-    'require-await': 'error',
-    // TODO: ENABLE THIS RULE BEFORE v6.0.0
-    'react/display-name': 'off',
-    'no-restricted-properties': [
-      'error',
-      {
-        object: 'crypto',
-        property: 'randomUUID',
-        message: 'Use of crypto.randomUUID is not allowed as it is not available in all browsers.',
-      },
-      {
-        object: 'navigator',
-        property: 'clipboard',
-        message:
-          'The Clipboard API is not available by default in Firefox. Use the `useClipboard` hook instead, which wraps clipboard access to prevent errors.',
-      },
-    ],
-    'no-restricted-imports': [
-      'error',
-      {
-        paths: [
-          {
-            name: 'lodash-es',
-            importNames: ['isEqual'],
-            message: 'Please use objectEquals from @observ33r/object-equals instead.',
-          },
-          {
-            name: 'lodash-es',
-            message: 'Please use es-toolkit instead.',
-          },
-          {
-            name: 'es-toolkit',
-            importNames: ['isEqual'],
-            message: 'Please use objectEquals from @observ33r/object-equals instead.',
-          },
-        ],
-      },
-    ],
-  },
-  overrides: [
-    /**
-     * Overrides for stories
-     */
-    {
-      files: ['*.stories.tsx'],
-      rules: {
-        // We may not have i18n available in stories.
-        'i18next/no-literal-string': 'off',
-      },
-    },
-  ],
-};

invokeai/frontend/web/.gitignore

@@ -44,4 +44,5 @@ yalc.lock
 
 # vitest
 tsconfig.vitest-temp.json
-coverage/
+coverage/
+*.tgz

invokeai/frontend/web/.prettierignore

@@ -14,3 +14,4 @@ static/
 src/theme/css/overlayscrollbars.css
 src/theme_/css/overlayscrollbars.css
 pnpm-lock.yaml
+.claude

invokeai/frontend/web/.prettierrc.js

@@ -1,11 +0,0 @@
-module.exports = {
-  ...require('@invoke-ai/prettier-config-react'),
-  overrides: [
-    {
-      files: ['public/locales/*.json'],
-      options: {
-        tabWidth: 4,
-      },
-    },
-  ],
-};

invokeai/frontend/web/.prettierrc.json

@@ -0,0 +1,17 @@
+{
+  "$schema": "http://json.schemastore.org/prettierrc",
+  "trailingComma": "es5",
+  "printWidth": 120,
+  "tabWidth": 2,
+  "semi": true,
+  "singleQuote": true,
+  "endOfLine": "auto",
+  "overrides": [
+    {
+      "files": ["public/locales/*.json"],
+      "options": {
+        "tabWidth": 4
+      }
+    }
+  ]
+}

invokeai/frontend/web/.storybook/ReduxInit.tsx

@@ -1,21 +1,23 @@
-import { PropsWithChildren, memo, useEffect } from 'react';
-import { modelChanged } from '../src/features/controlLayers/store/paramsSlice';
-import { useAppDispatch } from '../src/app/store/storeHooks';
 import { useGlobalModifiersInit } from '@invoke-ai/ui-library';
+import type { PropsWithChildren } from 'react';
+import { memo, useEffect } from 'react';
+
+import { useAppDispatch } from '../src/app/store/storeHooks';
+import { modelChanged } from '../src/features/controlLayers/store/paramsSlice';
 /**
  * Initializes some state for storybook. Must be in a different component
  * so that it is run inside the redux context.
  */
-export const ReduxInit = memo((props: PropsWithChildren) => {
+export const ReduxInit = memo(({ children }: PropsWithChildren) => {
   const dispatch = useAppDispatch();
   useGlobalModifiersInit();
   useEffect(() => {
     dispatch(
       modelChanged({ model: { key: 'test_model', hash: 'some_hash', name: 'some name', base: 'sd-1', type: 'main' } })
     );
-  }, []);
+  }, [dispatch]);
 
-  return props.children;
+  return children;
 });
 
 ReduxInit.displayName = 'ReduxInit';

invokeai/frontend/web/.storybook/main.ts

@@ -2,19 +2,13 @@ import type { StorybookConfig } from '@storybook/react-vite';
 
 const config: StorybookConfig = {
   stories: ['../src/**/*.mdx', '../src/**/*.stories.@(js|jsx|mjs|ts|tsx)'],
-  addons: [
-    '@storybook/addon-links',
-    '@storybook/addon-essentials',
-    '@storybook/addon-interactions',
-    '@storybook/addon-storysource',
-  ],
+  addons: ['@storybook/addon-links', '@storybook/addon-docs'],
+
   framework: {
     name: '@storybook/react-vite',
     options: {},
   },
-  docs: {
-    autodocs: 'tag',
-  },
+
   core: {
     disableTelemetry: true,
   },