remove tab character

## What type of PR is this? (check all applicable)

- [ ] Refactor
- [X] Feature
- [ ] Bug Fix
- [ ] Optimization
- [ ] Documentation Update
- [ ] Community Node Submission


## Have you discussed this change with the InvokeAI team?
- [X] Yes
- [ ] No, because:

      
## Have you updated all relevant documentation?
- [X] Yes
- [ ] No


## Description

This minor change adds the ability to filter the model lists returned by
V2 of the model manager using the model file format (e.g. "checkpoint").
Just thought this would be a useful feature.

## Related Tickets & Documents

<!--
For pull requests that relate or close an issue, please include them
below. 

For example having the text: "closes #1234" would connect the current
pull
request to issue 1234.  And when we merge the pull request, Github will
automatically close the issue.
-->

- Related Issue #
- Closes #

## QA Instructions, Screenshots, Recordings

<!-- 
Please provide steps on how to test changes, any hardware or 
software specifications as well as any other pertinent information. 
-->

## Merge Plan

This can be merged when approved without any adverse effects.

<!--
A merge plan describes how this PR should be handled after it is
approved.

Example merge plans:
- "This PR can be merged when approved"
- "This must be squash-merged when approved"
- "DO NOT MERGE - I will rebase and tidy commits before merging"
- "#dev-chat on discord needs to be advised of this change when it is
merged"

A merge plan is particularly important for large PRs or PRs that touch
the
database in any way.
-->

## Added/updated tests?

- [ ] Yes
- [X] No : minor feature - tested informally using the router API

## [optional] Are there any post deployment tasks we need to perform?

.github/pull_request_template.md

@@ -42,6 +42,21 @@ Please provide steps on how to test changes, any hardware or
 software specifications as well as any other pertinent information. 
 -->
 
+## Merge Plan
+
+<!--
+A merge plan describes how this PR should be handled after it is approved.
+
+Example merge plans:
+- "This PR can be merged when approved"
+- "This must be squash-merged when approved"
+- "DO NOT MERGE - I will rebase and tidy commits before merging"
+- "#dev-chat on discord needs to be advised of this change when it is merged"
+
+A merge plan is particularly important for large PRs or PRs that touch the
+database in any way.
+-->
+
 ## Added/updated tests?
 
 - [ ] Yes

.github/workflows/pypi-release.yml

@@ -15,19 +15,37 @@ jobs:
       TWINE_PASSWORD: ${{ secrets.PYPI_API_TOKEN }}
       TWINE_NON_INTERACTIVE: 1
     steps:
-      - name: checkout sources
-        uses: actions/checkout@v3
+      - name: Checkout sources
+        uses: actions/checkout@v4
 
-      - name: install deps
+      - name: Setup Node 20
+        uses: actions/setup-node@v4
+        with:
+          node-version: '20'
+
+      - name: Setup pnpm
+        uses: pnpm/action-setup@v2
+        with:
+          version: 8
+
+      - name: Install pnpm dependencies
+        working-directory: invokeai/frontend/web
+        run: 'pnpm install --prefer-frozen-lockfile'
+
+      - name: Build frontend
+        working-directory: invokeai/frontend/web
+        run: 'pnpm build'
+
+      - name: Install python deps
         run: pip install --upgrade build twine
 
-      - name: build package
+      - name: Build wheel package
         run: python3 -m build
 
-      - name: check distribution
+      - name: Check distribution
         run: twine check dist/*
 
-      - name: check PyPI versions
+      - name: Check PyPI versions
         if: github.ref == 'refs/heads/main' || startsWith(github.ref, 'refs/heads/release/')
         run: |
           pip install --upgrade requests
@@ -36,6 +54,6 @@ jobs:
           EXISTS=scripts.pypi_helper.local_on_pypi(); \
           print(f'PACKAGE_EXISTS={EXISTS}')" >> $GITHUB_ENV
 
-      - name: upload package
+      - name: Upload package
         if: env.PACKAGE_EXISTS == 'False' && env.TWINE_PASSWORD != ''
         run: twine upload dist/*

Makefile

@@ -1,6 +1,20 @@
 # simple Makefile with scripts that are otherwise hard to remember
 # to use, run from the repo root `make <command>`
 
+default: help
+
+help:
+	@echo Developer commands:
+	@echo
+	@echo "ruff           Run ruff, fixing any safely-fixable errors and formatting"
+	@echo "ruff-unsafe    Run ruff, fixing all fixable errors and formatting"
+	@echo "mypy           Run mypy using the config in pyproject.toml to identify type mismatches and other coding errors"
+	@echo "mypy-all       Run mypy ignoring the config in pyproject.tom but still ignoring missing imports"
+	@echo "frontend-build Build the frontend in order to run on localhost:9090"
+	@echo "frontend-dev   Run the frontend in developer mode on localhost:5173"
+	@echo "installer-zip  Build the installer .zip file for the current version"
+	@echo "tag-release    Tag the GitHub repository with the current version (use at release time only!)"
+
 # Runs ruff, fixing any safely-fixable errors and formatting
 ruff:
 		ruff check . --fix
@@ -18,4 +32,21 @@ mypy:
 # Runs mypy, ignoring the config in pyproject.toml but still ignoring missing (untyped) imports
 # (many files are ignored by the config, so this is useful for checking all files)
 mypy-all:
-		mypy scripts/invokeai-web.py --config-file= --ignore-missing-imports
+		mypy scripts/invokeai-web.py --config-file= --ignore-missing-imports
+
+# Build the frontend
+frontend-build:
+	cd invokeai/frontend/web && pnpm build
+
+# Run the frontend in dev mode
+frontend-dev:
+	cd invokeai/frontend/web && pnpm dev
+
+# Installer zip file
+installer-zip:
+	cd installer && ./create_installer.sh
+
+# Tag the release
+tag-release:
+	cd installer && ./tag_release.sh
+

docker/Dockerfile

@@ -100,6 +100,8 @@ ENV INVOKEAI_SRC=/opt/invokeai
 ENV VIRTUAL_ENV=/opt/venv/invokeai
 ENV INVOKEAI_ROOT=/invokeai
 ENV PATH="$VIRTUAL_ENV/bin:$INVOKEAI_SRC:$PATH"
+ENV CONTAINER_UID=${CONTAINER_UID:-1000}
+ENV CONTAINER_GID=${CONTAINER_GID:-1000}
 
 # --link requires buldkit w/ dockerfile syntax 1.4
 COPY --link --from=builder ${INVOKEAI_SRC} ${INVOKEAI_SRC}
@@ -117,7 +119,7 @@ WORKDIR ${INVOKEAI_SRC}
 RUN cd /usr/lib/$(uname -p)-linux-gnu/pkgconfig/ && ln -sf opencv4.pc opencv.pc
 RUN python3 -c "from patchmatch import patch_match"
 
-RUN mkdir -p ${INVOKEAI_ROOT} && chown -R 1000:1000 ${INVOKEAI_ROOT}
+RUN mkdir -p ${INVOKEAI_ROOT} && chown -R ${CONTAINER_UID}:${CONTAINER_GID} ${INVOKEAI_ROOT}
 
 COPY docker/docker-entrypoint.sh ./
 ENTRYPOINT ["/opt/invokeai/docker-entrypoint.sh"]

docs/contributing/MODEL_MANAGER.md

@@ -10,40 +10,36 @@ model. These are the:
   tracks the type of the model, its provenance, and where it can be
   found on disk.
   
-* _ModelLoadServiceBase_ Responsible for loading a model from disk
-  into RAM and VRAM and getting it ready for inference.
-  
-* _DownloadQueueServiceBase_ A multithreaded downloader responsible
-  for downloading models from a remote source to disk. The download
-  queue has special methods for downloading repo_id folders from
-  Hugging Face, as well as discriminating among model versions in
-  Civitai, but can be used for arbitrary content.
-  
 * _ModelInstallServiceBase_ A service for installing models to
   disk. It uses `DownloadQueueServiceBase` to download models and
   their metadata, and `ModelRecordServiceBase` to store that
   information. It is also responsible for managing the InvokeAI
   `models` directory and its contents.
 
+* _DownloadQueueServiceBase_ (**CURRENTLY UNDER DEVELOPMENT - NOT IMPLEMENTED**)
+  A multithreaded downloader responsible
+  for downloading models from a remote source to disk. The download
+  queue has special methods for downloading repo_id folders from
+  Hugging Face, as well as discriminating among model versions in
+  Civitai, but can be used for arbitrary content.
+  
+  * _ModelLoadServiceBase_ (**CURRENTLY UNDER DEVELOPMENT - NOT IMPLEMENTED**)
+  Responsible for loading a model from disk
+  into RAM and VRAM and getting it ready for inference.
+
+  
 ## Location of the Code
 
 All four of these services can be found in
 `invokeai/app/services` in the following directories:
 
 * `invokeai/app/services/model_records/`
-* `invokeai/app/services/downloads/`
-* `invokeai/app/services/model_loader/`
 * `invokeai/app/services/model_install/`
-
-With the exception of the install service, each of these is a thin
-shell around a corresponding implementation located in
-`invokeai/backend/model_manager`. The main difference between the
-modules found in app services and those in the backend folder is that
-the former add support for event reporting and are more tied to the
-needs of the InvokeAI API.
+* `invokeai/app/services/model_loader/` (**under development**)
+* `invokeai/app/services/downloads/`(**under development**)
 
 Code related to the FastAPI web API can be found in
-`invokeai/app/api/routers/models.py`.
+`invokeai/app/api/routers/model_records.py`.
 
 ***
 
@@ -165,10 +161,6 @@ of the fields, including `name`, `model_type` and `base_model`, are
 shared between `ModelConfigBase` and `ModelBase`, and this is a
 potential source of confusion.
 
-** TO DO: ** The `ModelBase` code needs to be revised to reduce the
-duplication of similar classes and to support using the `key` as the
-primary model identifier.
-
 ## Reading and Writing Model Configuration Records
 
 The `ModelRecordService` provides the ability to retrieve model
@@ -362,7 +354,7 @@ model and pass its key to `get_model()`.
 Several methods allow you to create and update stored model config
 records.
 
-#### add_model(key, config) -> ModelConfigBase:
+#### add_model(key, config) -> AnyModelConfig:
 
 Given a key and a configuration, this will add the model's
 configuration record to the database. `config` can either be a subclass of
@@ -386,27 +378,356 @@ fields to be updated. This will return an `AnyModelConfig` on success,
 or raise `InvalidModelConfigException` or `UnknownModelException`
 exceptions on failure.
 
-***TO DO:*** Investigate why `update_model()` returns an
-`AnyModelConfig` while `add_model()` returns a `ModelConfigBase`.
-
-### rename_model(key, new_name) -> ModelConfigBase:
-
-This is a special case of `update_model()` for the use case of
-changing the model's name. It is broken out because there are cases in
-which the InvokeAI application wants to synchronize the model's name
-with its path in the `models` directory after changing the name, type
-or base. However, when using the ModelRecordService directly, the call
-is equivalent to:
-
-```
-store.rename_model(key, {'name': 'new_name'})
-```
-
-***TO DO:*** Investigate why `rename_model()` is returning a
-`ModelConfigBase` while `update_model()` returns a `AnyModelConfig`.
-
 ***
 
+## Model installation
+
+The `ModelInstallService` class implements the
+`ModelInstallServiceBase` abstract base class, and provides a one-stop
+shop for all your model install needs. It provides the following
+functionality:
+
+- Registering a model config record for a model already located on the
+  local filesystem, without moving it or changing its path.
+  
+- Installing a model alreadiy located on the local filesystem, by
+  moving it into the InvokeAI root directory under the
+  `models` folder (or wherever config parameter `models_dir`
+  specifies).
+	
+- Probing of models to determine their type, base type and other key
+  information.
+  
+- Interface with the InvokeAI event bus to provide status updates on
+  the download, installation and registration process.
+  
+- Downloading a model from an arbitrary URL and installing it in
+  `models_dir` (_implementation pending_).
+  
+- Special handling for Civitai model URLs which allow the user to
+  paste in a model page's URL or download link (_implementation pending_).
+
+  
+- Special handling for HuggingFace repo_ids to recursively download
+  the contents of the repository, paying attention to alternative
+  variants such as fp16. (_implementation pending_)
+  
+### Initializing the installer
+
+A default installer is created at InvokeAI api startup time and stored
+in `ApiDependencies.invoker.services.model_install` and can
+also be retrieved from an invocation's `context` argument with
+`context.services.model_install`.
+
+In the event you wish to create a new installer, you may use the
+following initialization pattern:
+
+```
+from invokeai.app.services.config import InvokeAIAppConfig
+from invokeai.app.services.model_records import ModelRecordServiceSQL
+from invokeai.app.services.model_install import ModelInstallService
+from invokeai.app.services.shared.sqlite import SqliteDatabase
+from invokeai.backend.util.logging import InvokeAILogger
+
+config = InvokeAIAppConfig.get_config()
+config.parse_args()
+logger = InvokeAILogger.get_logger(config=config)
+db = SqliteDatabase(config, logger)
+
+store = ModelRecordServiceSQL(db)
+installer = ModelInstallService(config, store)
+```
+
+The full form of `ModelInstallService()` takes the following
+required parameters:
+
+| **Argument**     | **Type**                     | **Description**              |
+|------------------|------------------------------|------------------------------|
+| `config`         | InvokeAIAppConfig       | InvokeAI app configuration object |
+| `record_store`   | ModelRecordServiceBase  | Config record storage database |
+| `event_bus`      | EventServiceBase        | Optional event bus to send download/install progress events to |
+
+Once initialized, the installer will provide the following methods:
+
+#### install_job = installer.import_model()
+
+The `import_model()` method is the core of the installer. The
+following illustrates basic usage:
+
+```
+from invokeai.app.services.model_install import (
+	LocalModelSource,
+	HFModelSource,
+	URLModelSource,
+)
+
+source1 = LocalModelSource(path='/opt/models/sushi.safetensors')   # a local safetensors file
+source2 = LocalModelSource(path='/opt/models/sushi_diffusers')     # a local diffusers folder
+
+source3 = HFModelSource(repo_id='runwayml/stable-diffusion-v1-5')  # a repo_id
+source4 = HFModelSource(repo_id='runwayml/stable-diffusion-v1-5', subfolder='vae')  # a subfolder within a repo_id
+source5 = HFModelSource(repo_id='runwayml/stable-diffusion-v1-5', variant='fp16')   # a named variant of a HF model
+
+source6 = URLModelSource(url='https://civitai.com/api/download/models/63006')       # model located at a URL
+source7 = URLModelSource(url='https://civitai.com/api/download/models/63006', access_token='letmein') # with an access token
+
+for source in [source1, source2, source3, source4, source5, source6, source7]:
+   install_job = installer.install_model(source)
+   
+source2job = installer.wait_for_installs()
+for source in sources:
+    job = source2job[source]
+	if job.status == "completed":
+		model_config = job.config_out
+		model_key = model_config.key
+		print(f"{source} installed as {model_key}")
+	elif job.status == "error":
+	    print(f"{source}: {job.error_type}.\nStack trace:\n{job.error}")
+	
+```
+
+As shown here, the `import_model()` method accepts a variety of
+sources, including local safetensors files, local diffusers folders,
+HuggingFace repo_ids with and without a subfolder designation,
+Civitai model URLs and arbitrary URLs that point to checkpoint files
+(but not to folders).
+
+Each call to `import_model()` return a `ModelInstallJob` job, 
+an object which tracks the progress of the install.
+
+If a remote model is requested, the model's files are downloaded in
+parallel across a multiple set of threads using the download
+queue. During the download process, the `ModelInstallJob` is updated
+to provide status and progress information. After the files (if any)
+are downloaded, the remainder of the installation runs in a single
+serialized background thread. These are the model probing, file
+copying, and config record database update steps.
+
+Multiple install jobs can be queued up. You may block until all
+install jobs are completed (or errored) by calling the
+`wait_for_installs()` method as shown in the code
+example. `wait_for_installs()` will return a `dict` that maps the
+requested source to its job. This object can be interrogated
+to determine its status. If the job errored out, then the error type
+and details can be recovered from `job.error_type` and `job.error`.
+
+The full list of arguments to `import_model()` is as follows:
+
+| **Argument**     | **Type**                     | **Default** | **Description**                           |
+|------------------|------------------------------|-------------|-------------------------------------------|
+| `source`         | Union[str, Path, AnyHttpUrl] |             | The source of the model, Path, URL or repo_id |
+| `inplace`        | bool                         | True        | Leave a local model in its current location |
+| `variant`        | str                          | None        | Desired variant, such as 'fp16' or 'onnx' (HuggingFace only) |
+| `subfolder`      | str                          | None        | Repository subfolder (HuggingFace only)   |
+| `config`         | Dict[str, Any]               | None        | Override all or a portion of model's probed attributes |
+| `access_token`   | str                          | None        | Provide authorization information needed to download |
+
+
+The `inplace` field controls how local model Paths are handled. If
+True (the default), then the model is simply registered in its current
+location by the installer's `ModelConfigRecordService`. Otherwise, a
+copy of the model put into the location specified by the `models_dir`
+application configuration parameter.
+
+The `variant` field is used for HuggingFace repo_ids only. If
+provided, the repo_id download handler will look for and download
+tensors files that follow the convention for the selected variant:
+
+- "fp16" will select files named "*model.fp16.{safetensors,bin}"
+- "onnx" will select files ending with the suffix ".onnx"
+- "openvino" will select files beginning with "openvino_model"
+
+In the special case of the "fp16" variant, the installer will select
+the 32-bit version of the files if the 16-bit version is unavailable.
+
+`subfolder` is used for HuggingFace repo_ids only. If provided, the
+model will be downloaded from the designated subfolder rather than the
+top-level repository folder. If a subfolder is attached to the repo_id
+using the format `repo_owner/repo_name:subfolder`, then the subfolder
+specified by the repo_id will override the subfolder argument.
+
+`config` can be used to override all or a portion of the configuration
+attributes returned by the model prober. See the section below for
+details.
+
+`access_token` is passed to the download queue and used to access
+repositories that require it.
+
+#### Monitoring the install job process
+
+When you create an install job with `import_model()`, it launches the
+download and installation process in the background and returns a
+`ModelInstallJob` object for monitoring the process.
+
+The `ModelInstallJob` class has the following structure:
+
+| **Attribute** | **Type**        |  **Description** |
+|----------------|-----------------|------------------|
+| `status`       | `InstallStatus`  | An enum of ["waiting", "running", "completed" and "error" |
+| `config_in`    | `dict`          | Overriding configuration values provided by the caller |
+| `config_out`   | `AnyModelConfig`| After successful completion, contains the configuration record written to the database | 
+| `inplace`      | `boolean`       | True if the caller asked to install the model in place using its local path | 
+| `source`       | `ModelSource`   | The local path, remote URL or repo_id of the model to be installed | 
+| `local_path`   | `Path`          | If a remote model, holds the path of the model after it is downloaded; if a local model, same as `source` |
+| `error_type`   | `str`           | Name of the exception that led to an error status |
+| `error`        | `str`           | Traceback of the error |
+
+
+If the `event_bus` argument was provided, events will also be
+broadcast to the InvokeAI event bus. The events will appear on the bus
+as an event of type `EventServiceBase.model_event`, a timestamp and
+the following event names:
+
+- `model_install_started`
+
+The payload will contain the keys `timestamp` and `source`. The latter
+indicates the requested model source for installation.
+
+- `model_install_progress`
+
+Emitted at regular intervals when downloading a remote model, the
+payload will contain the keys `timestamp`, `source`, `current_bytes`
+and `total_bytes`. These events are _not_ emitted when a local model
+already on the filesystem is imported.
+
+- `model_install_completed`
+
+Issued once at the end of a successful installation. The payload will
+contain the keys `timestamp`, `source` and `key`, where `key` is the
+ID under which the model has been registered.
+
+- `model_install_error`
+
+Emitted if the installation process fails for some reason. The payload
+will contain the keys `timestamp`, `source`, `error_type` and
+`error`. `error_type` is a short message indicating the nature of the
+error, and `error` is the long traceback to help debug the problem.
+
+#### Model confguration and probing
+
+The install service uses the `invokeai.backend.model_manager.probe`
+module during import to determine the model's type, base type, and
+other configuration parameters. Among other things, it assigns a
+default name and description for the model based on probed
+fields. 
+
+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
+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
+`SchedulerPredictionType` and `name` for an sd-2 model:
+
+This is typically used to set
+the model's name and description, but can also be used to overcome
+cases in which automatic probing is unable to (correctly) determine
+the model's attribute. The most common situation is the
+`prediction_type` field for sd-2 (and rare sd-1) models. Here is an
+example of how it works:
+
+```
+install_job = installer.import_model(
+               source='stabilityai/stable-diffusion-2-1',
+			   variant='fp16',
+			   config=dict(
+			         prediction_type=SchedulerPredictionType('v_prediction')
+					 name='stable diffusion 2 base model',
+	           )
+	      )
+```
+
+### Other installer methods
+
+This section describes additional methods provided by the installer class.
+
+#### jobs = installer.wait_for_installs()
+
+Block until all pending installs are completed or errored and then
+returns a list of completed jobs.
+
+#### jobs = installer.list_jobs([source])
+
+Return a list of all active and complete `ModelInstallJobs`. An
+optional `source` argument allows you to filter the returned list by a
+model source string pattern using a partial string match.
+
+#### jobs = installer.get_job(source)
+
+Return a list of `ModelInstallJob` corresponding to the indicated
+model source.
+
+#### installer.prune_jobs
+
+Remove non-pending jobs (completed or errored) from the job list
+returned by `list_jobs()` and `get_job()`.
+
+#### installer.app_config, installer.record_store,
+installer.event_bus
+
+Properties that provide access to the installer's `InvokeAIAppConfig`,
+`ModelRecordServiceBase` and `EventServiceBase` objects.
+
+#### key = installer.register_path(model_path, config), key = installer.install_path(model_path, config)
+
+These methods bypass the download queue and directly register or
+install the model at the indicated path, returning the unique ID for
+the installed model.
+
+Both methods accept a Path object corresponding to a checkpoint or
+diffusers folder, and an optional dict of config attributes to use to
+override the values derived from model probing.
+
+The difference between `register_path()` and `install_path()` is that
+the former creates a model configuration record without changing the
+location of the model in the filesystem. The latter makes a copy of
+the model inside the InvokeAI models directory before registering
+it.
+
+#### installer.unregister(key)
+
+This will remove the model config record for the model at key, and is
+equivalent to `installer.record_store.del_model(key)`
+
+#### installer.delete(key)
+
+This is similar to `unregister()` but has the additional effect of
+conditionally deleting the underlying model file(s) if they reside
+within the InvokeAI models directory
+
+#### installer.unconditionally_delete(key)
+
+This method is similar to `unregister()`, but also unconditionally
+deletes the corresponding model weights file(s), regardless of whether
+they are inside or outside the InvokeAI models hierarchy.
+
+#### List[str]=installer.scan_directory(scan_dir: Path, install: bool)
+
+This method will recursively scan the directory indicated in
+`scan_dir` for new models and either install them in the models
+directory or register them in place, depending on the setting of
+`install` (default False).
+
+The return value is the list of keys of the new installed/registered
+models.
+
+#### installer.sync_to_config()
+
+This method synchronizes models in the models directory and autoimport
+directory to those in the `ModelConfigRecordService` database. New
+models are registered and orphan models are unregistered.
+
+#### installer.start(invoker)
+
+The `start` method is called by the API intialization 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.
+
+# The remainder of this documentation is provisional, pending implementation of the Download and Load services
+
 ## Let's get loaded, the lowdown on ModelLoadService
 
 The `ModelLoadService` is responsible for loading a named model into
@@ -863,351 +1184,3 @@ other resources that it might have been using.
 This will start/pause/cancel all jobs that have been submitted to the
 queue and have not yet reached a terminal state.
 
-## Model installation
-
-The `ModelInstallService` class implements the
-`ModelInstallServiceBase` abstract base class, and provides a one-stop
-shop for all your model install needs. It provides the following
-functionality:
-
-- Registering a model config record for a model already located on the
-  local filesystem, without moving it or changing its path.
-  
-- Installing a model alreadiy located on the local filesystem, by
-  moving it into the InvokeAI root directory under the
-  `models` folder (or wherever config parameter `models_dir`
-  specifies).
-	
-- Downloading a model from an arbitrary URL and installing it in
-  `models_dir`.
-  
-- Special handling for Civitai model URLs which allow the user to
-  paste in a model page's URL or download link. Any metadata provided
-  by Civitai, such as trigger terms, are captured and placed in the
-  model config record.
-  
-- Special handling for HuggingFace repo_ids to recursively download
-  the contents of the repository, paying attention to alternative
-  variants such as fp16.
-  
-- Probing of models to determine their type, base type and other key
-  information.
-  
-- Interface with the InvokeAI event bus to provide status updates on
-  the download, installation and registration process.
-  
-### Initializing the installer
-
-A default installer is created at InvokeAI api startup time and stored
-in `ApiDependencies.invoker.services.model_install_service` and can
-also be retrieved from an invocation's `context` argument with
-`context.services.model_install_service`.
-
-In the event you wish to create a new installer, you may use the
-following initialization pattern:
-
-```
-from invokeai.app.services.config import InvokeAIAppConfig
-from invokeai.app.services.download_manager import DownloadQueueServive
-from invokeai.app.services.model_record_service import ModelRecordServiceBase
-
-config = InvokeAI.get_config()
-queue = DownloadQueueService()
-store = ModelRecordServiceBase.open(config)
-installer = ModelInstallService(config=config, queue=queue, store=store)
-```
-
-The full form of `ModelInstallService()` takes the following
-parameters. Each parameter will default to a reasonable value, but it
-is recommended that you set them explicitly as shown in the above example.
-
-| **Argument**     | **Type**                     | **Default** | **Description**                           |
-|------------------|------------------------------|-------------|-------------------------------------------|
-| `config`         | InvokeAIAppConfig        | Use system-wide config  | InvokeAI app configuration object |
-| `queue`          | DownloadQueueServiceBase  | Create a new download queue for internal use  | Download queue |
-| `store`          | ModelRecordServiceBase  | Use config to select the database to open  | Config storage database |
-| `event_bus`      | EventServiceBase        | None | An event bus to send download/install progress events to |
-| `event_handlers` | List[DownloadEventHandler]  | None | Event handlers for the download queue |
-
-Note that if `store` is not provided, then the class will use
-`ModelRecordServiceBase.open(config)` to select the database to use.
-
-Once initialized, the installer will provide the following methods:
-
-#### install_job = installer.install_model()
-
-The `install_model()` method is the core of the installer. The
-following illustrates basic usage:
-
-```
-sources = [
-	Path('/opt/models/sushi.safetensors'),   # a local safetensors file
-	Path('/opt/models/sushi_diffusers/'),    # a local diffusers folder
-	'runwayml/stable-diffusion-v1-5',        # a repo_id
-    'runwayml/stable-diffusion-v1-5:vae',    # a subfolder within a repo_id
-	'https://civitai.com/api/download/models/63006', # a civitai direct download link
-	'https://civitai.com/models/8765?modelVersionId=10638',  # civitai model page
-	'https://s3.amazon.com/fjacks/sd-3.safetensors', # arbitrary URL
-]
-
-for source in sources:
-   install_job = installer.install_model(source)
-   
-source2key = installer.wait_for_installs()
-for source in sources:
-    model_key = source2key[source]
-	print(f"{source} installed as {model_key}")
-```
-
-As shown here, the `install_model()` method accepts a variety of
-sources, including local safetensors files, local diffusers folders,
-HuggingFace repo_ids with and without a subfolder designation,
-Civitai model URLs and arbitrary URLs that point to checkpoint files
-(but not to folders).
-
-Each call to `install_model()` will return a `ModelInstallJob` job, a
-subclass of `DownloadJobBase`. The install job has additional
-install-specific fields described in the next section.
-
-Each install job will run in a series of background threads using
-the object's download queue. You may block until all install jobs are
-completed (or errored) by calling the `wait_for_installs()` method as
-shown in the code example. `wait_for_installs()` will return a `dict`
-that maps the requested source to the key of the installed model. In
-the case that a model fails to download or install, its value in the
-dict will be None. The actual cause of the error will be reported in
-the corresponding job's `error` field.
-
-Alternatively you may install event handlers and/or listen for events
-on the InvokeAI event bus in order to monitor the progress of the
-requested installs.
-
-The full list of arguments to `model_install()` is as follows:
-
-| **Argument**     | **Type**                     | **Default** | **Description**                           |
-|------------------|------------------------------|-------------|-------------------------------------------|
-| `source`         | Union[str, Path, AnyHttpUrl] |             | The source of the model, Path, URL or repo_id |
-| `inplace`        | bool                         | True        | Leave a local model in its current location |
-| `variant`        | str                          | None        | Desired variant, such as 'fp16' or 'onnx' (HuggingFace only) |
-| `subfolder`      | str                          | None        | Repository subfolder (HuggingFace only)   |
-| `probe_override` | Dict[str, Any]               | None        | Override all or a portion of model's probed attributes |
-| `metadata`       | ModelSourceMetadata          | None        | Provide metadata that will be added to model's config |
-| `access_token`   | str                          | None        | Provide authorization information needed to download |
-| `priority`       | int                          | 10          | Download queue priority for the job       |
-
-
-The `inplace` field controls how local model Paths are handled. If
-True (the default), then the model is simply registered in its current
-location by the installer's `ModelConfigRecordService`. Otherwise, the
-model will be moved into the location specified by the  `models_dir`
-application configuration parameter. 
-
-The `variant` field is used for HuggingFace repo_ids only. If
-provided, the repo_id download handler will look for and download
-tensors files that follow the convention for the selected variant:
-
-- "fp16" will select files named "*model.fp16.{safetensors,bin}"
-- "onnx" will select files ending with the suffix ".onnx"
-- "openvino" will select files beginning with "openvino_model"
-
-In the special case of the "fp16" variant, the installer will select
-the 32-bit version of the files if the 16-bit version is unavailable.
-
-`subfolder` is used for HuggingFace repo_ids only. If provided, the
-model will be downloaded from the designated subfolder rather than the
-top-level repository folder. If a subfolder is attached to the repo_id
-using the format `repo_owner/repo_name:subfolder`, then the subfolder
-specified by the repo_id will override the subfolder argument.
-
-`probe_override` can be used to override all or a portion of the
-attributes returned by the model prober. This can be used to overcome
-cases in which automatic probing is unable to (correctly) determine
-the model's attribute. The most common situation is the
-`prediction_type` field for sd-2 (and rare sd-1) models. Here is an
-example of how it works:
-
-```
-install_job = installer.install_model(
-               source='stabilityai/stable-diffusion-2-1',
-			   variant='fp16',
-			   probe_override=dict(
-			         prediction_type=SchedulerPredictionType('v_prediction')
-	           )
-	      )
-```
-
-`metadata` allows you to attach custom metadata to the installed
-model. See the next section for details.
-
-`priority` and `access_token` are passed to the download queue and
-have the same effect as they do for the DownloadQueueServiceBase.
-
-#### Monitoring the install job process
-
-When you create an install job with `model_install()`, events will be
-passed to the list of `DownloadEventHandlers` provided at installer
-initialization time. Event handlers can also be added to individual
-model install jobs by calling their `add_handler()` method as
-described earlier for the `DownloadQueueService`.
-
-If the `event_bus` argument was provided, events will also be
-broadcast to the InvokeAI event bus. The events will appear on the bus
-as a singular event type named `model_event` with a payload of
-`job`. You can then retrieve the job and check its status.
-
-** TO DO: ** consider breaking `model_event` into
-`model_install_started`, `model_install_completed`, etc. The event bus
-features have not yet been tested with FastAPI/websockets, and it may
-turn out that the job object is not serializable.
-
-#### Model metadata and probing
-
-The install service has special handling for HuggingFace and Civitai
-URLs that capture metadata from the source and include it in the model
-configuration record. For example, fetching the Civitai model 8765
-will produce a config record similar to this (using YAML
-representation):
-
-```
-5abc3ef8600b6c1cc058480eaae3091e:
-  path: sd-1/lora/to8contrast-1-5.safetensors
-  name: to8contrast-1-5
-  base_model: sd-1
-  model_type: lora
-  model_format: lycoris
-  key: 5abc3ef8600b6c1cc058480eaae3091e
-  hash: 5abc3ef8600b6c1cc058480eaae3091e
-  description: 'Trigger terms: to8contrast style'
-  author: theovercomer8
-  license: allowCommercialUse=Sell; allowDerivatives=True; allowNoCredit=True
-  source: https://civitai.com/models/8765?modelVersionId=10638
-  thumbnail_url: null
-  tags:
-  - model
-  - style
-  - portraits
-```
-
-For sources that do not provide model metadata, you can attach custom
-fields by providing a `metadata` argument to `model_install()` using
-an initialized `ModelSourceMetadata` object (available for import from
-`model_install_service.py`):
-
-```
-from invokeai.app.services.model_install_service import ModelSourceMetadata
-meta = ModelSourceMetadata(
-                name="my model",
-				author="Sushi Chef",
-				description="Highly customized model; trigger with 'sushi',"
-				license="mit",
-				thumbnail_url="http://s3.amazon.com/ljack/pics/sushi.png",
-				tags=list('sfw', 'food')
-			)
-install_job = installer.install_model(
-               source='sushi_chef/model3',
-			   variant='fp16',
-			   metadata=meta,
-	      )
-```
-
-It is not currently recommended to provide custom metadata when
-installing from Civitai or HuggingFace source, as the metadata
-provided by the source will overwrite the fields you provide. Instead,
-after the model is installed you can use
-`ModelRecordService.update_model()` to change the desired fields.
-
-** TO DO: ** Change the logic so that the caller's metadata fields take
-precedence over those provided by the source.
-
-
-#### Other installer methods
-
-This section describes additional, less-frequently-used attributes and
-methods provided by the installer class.
-
-##### installer.wait_for_installs()
-
-This is equivalent to the `DownloadQueue` `join()` method. It will
-block until all the active jobs in the install queue have reached a
-terminal state (completed, errored or cancelled).
-
-##### installer.queue, installer.store, installer.config
-
-These attributes provide access to the `DownloadQueueServiceBase`,
-`ModelConfigRecordServiceBase`, and `InvokeAIAppConfig` objects that
-the installer uses.
-
-For example, to temporarily pause all pending installations, you can
-do this:
-
-```
-installer.queue.pause_all_jobs()
-```
-##### key = installer.register_path(model_path, overrides), key = installer.install_path(model_path, overrides)
-
-These methods bypass the download queue and directly register or
-install the model at the indicated path, returning the unique ID for
-the installed model.
-
-Both methods accept a Path object corresponding to a checkpoint or
-diffusers folder, and an optional dict of attributes to use to
-override the values derived from model probing.
-
-The difference between `register_path()` and `install_path()` is that
-the former will not move the model from its current position, while
-the latter will move it into the `models_dir` hierarchy.
-
-##### installer.unregister(key)
-
-This will remove the model config record for the model at key, and is
-equivalent to `installer.store.unregister(key)`
-
-##### installer.delete(key)
-
-This is similar to `unregister()` but has the additional effect of
-deleting the underlying model file(s) -- even if they were outside the
-`models_dir` directory!
-
-##### installer.conditionally_delete(key)
-
-This method will call `unregister()` if the model identified by `key`
-is outside the `models_dir` hierarchy, and call `delete()` if the
-model is inside.
-
-#### List[str]=installer.scan_directory(scan_dir: Path, install: bool)
-
-This method will recursively scan the directory indicated in
-`scan_dir` for new models and either install them in the models
-directory or register them in place, depending on the setting of
-`install` (default False).
-
-The return value is the list of keys of the new installed/registered
-models.
-
-#### installer.scan_models_directory()
-
-This method scans the models directory for new models and registers
-them in place. Models that are present in the
-`ModelConfigRecordService` database whose paths are not found will be
-unregistered.
-
-#### installer.sync_to_config()
-
-This method synchronizes models in the models directory and autoimport
-directory to those in the `ModelConfigRecordService` database. New
-models are registered and orphan models are unregistered.
-
-#### hash=installer.hash(model_path)
-
-This method is calls the fasthash algorithm on a model's Path
-(either a file or a folder) to generate a unique ID based on the
-contents of the model.
-
-##### installer.start(invoker)
-
-The `start` method is called by the API intialization 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.
-
-This method should not ordinarily be called manually.

docs/features/CONFIGURATION.md

@@ -154,14 +154,16 @@ groups in `invokeia.yaml`:
 
 ### Web Server
 
-| Setting  | Default Value  |  Description |
-|----------|----------------|--------------|
-| `host`     | `localhost`      | Name or IP address of the network interface that the web server will listen on  |
-| `port`     | `9090`           | Network port number that the web server will listen on  |
-| `allow_origins`  | `[]`       | A list of host names or IP addresses that are allowed to connect to the InvokeAI API in the format `['host1','host2',...]` |
-| `allow_credentials` | `true`   | Require credentials for a foreign host to access the InvokeAI API (don't change this) |
-| `allow_methods` | `*`         | List of HTTP methods ("GET", "POST") that the web server is allowed to use when accessing the API  |
-| `allow_headers` | `*`         | List of HTTP headers that the web server will accept when accessing the API  |
+| Setting             | Default Value | Description                                                                                                                |
+|---------------------|---------------|----------------------------------------------------------------------------------------------------------------------------|
+| `host`              | `localhost`   | Name or IP address of the network interface that the web server will listen on                                             |
+| `port`              | `9090`        | Network port number that the web server will listen on                                                                     |
+| `allow_origins`     | `[]`          | A list of host names or IP addresses that are allowed to connect to the InvokeAI API in the format `['host1','host2',...]` |
+| `allow_credentials` | `true`        | Require credentials for a foreign host to access the InvokeAI API (don't change this)                                      |
+| `allow_methods`     | `*`           | List of HTTP methods ("GET", "POST") that the web server is allowed to use when accessing the API                          |
+| `allow_headers`     | `*`           | List of HTTP headers that the web server will accept when accessing the API                                                |
+| `ssl_certfile`      | null          | Path to an SSL certificate file, used to enable HTTPS.                                                                     |
+| `ssl_keyfile`       | null          | Path to an SSL keyfile, if the key is not included in the certificate file.                                                |
 
 The documentation for InvokeAI's API can be accessed by browsing to the following URL: [http://localhost:9090/docs].
 

docs/installation/020_INSTALL_MANUAL.md

@@ -293,6 +293,19 @@ manager, please follow these steps:
 
 ## Developer Install
 
+!!! warning
+
+    InvokeAI uses a SQLite database. By running on `main`, you accept responsibility for your database. This
+    means making regular backups (especially before pulling) and/or fixing it yourself in the event that a
+    PR introduces a schema change.
+    
+    If you don't need persistent backend storage, you can use an ephemeral in-memory database by setting
+    `use_memory_db: true` under `Path:` in your `invokeai.yaml` file.
+
+    If this is untenable, you should run the application via the official installer or a manual install of the
+    python package from pypi. These releases will not break your database.
+
+
 If you have an interest in how InvokeAI works, or you would like to
 add features or bugfixes, you are encouraged to install the source
 code for InvokeAI. For this to work, you will need to install the
@@ -388,3 +401,5 @@ environment variable INVOKEAI_ROOT to point to the installation directory.
 
 Note that if you run into problems with the Conda installation, the InvokeAI
 staff will **not** be able to help you out. Caveat Emptor!
+
+[dev-chat]: https://discord.com/channels/1020123559063990373/1049495067846524939

docs/javascripts/init_kapa_widget.js

@@ -0,0 +1,10 @@
+document.addEventListener("DOMContentLoaded", function () {
+  var script = document.createElement("script");
+  script.src = "https://widget.kapa.ai/kapa-widget.bundle.js";
+  script.setAttribute("data-website-id", "b5973bb1-476b-451e-8cf4-98de86745a10");
+  script.setAttribute("data-project-name", "Invoke.AI");
+  script.setAttribute("data-project-color", "#11213C");
+  script.setAttribute("data-project-logo", "https://avatars.githubusercontent.com/u/113954515?s=280&v=4");
+  script.async = true;
+  document.head.appendChild(script);
+});

installer/create_installer.sh

@@ -13,14 +13,6 @@ function is_bin_in_path {
     builtin type -P "$1" &>/dev/null
 }
 
-function does_tag_exist {
-    git rev-parse --quiet --verify "refs/tags/$1" >/dev/null
-}
-
-function git_show_ref {
-    git show-ref --dereference $1 --abbrev 7
-}
-
 function git_show {
     git show -s --format='%h %s' $1
 }
@@ -53,50 +45,11 @@ VERSION=$(
 )
 PATCH=""
 VERSION="v${VERSION}${PATCH}"
-LATEST_TAG="v3-latest"
-
-echo "Building installer for version $VERSION..."
-echo
-
-if does_tag_exist $VERSION; then
-    echo -e "${BCYAN}${VERSION}${RESET} already exists:"
-    git_show_ref tags/$VERSION
-    echo
-fi
-if does_tag_exist $LATEST_TAG; then
-    echo -e "${BCYAN}${LATEST_TAG}${RESET} already exists:"
-    git_show_ref tags/$LATEST_TAG
-    echo
-fi
 
 echo -e "${BGREEN}HEAD${RESET}:"
 git_show
 echo
 
-echo -e -n "Create tags ${BCYAN}${VERSION}${RESET} and ${BCYAN}${LATEST_TAG}${RESET} @ ${BGREEN}HEAD${RESET}, ${RED}deleting existing tags on remote${RESET}? "
-read -e -p 'y/n [n]: ' input
-RESPONSE=${input:='n'}
-if [ "$RESPONSE" == 'y' ]; then
-    echo
-    echo -e "Deleting ${BCYAN}${VERSION}${RESET} tag on remote..."
-    git push origin :refs/tags/$VERSION
-
-    echo -e "Tagging ${BGREEN}HEAD${RESET} with ${BCYAN}${VERSION}${RESET} locally..."
-    if ! git tag -fa $VERSION; then
-        echo "Existing/invalid tag"
-        exit -1
-    fi
-
-    echo -e "Deleting ${BCYAN}${LATEST_TAG}${RESET} tag on remote..."
-    git push origin :refs/tags/$LATEST_TAG
-
-    echo -e "Tagging ${BGREEN}HEAD${RESET} with ${BCYAN}${LATEST_TAG}${RESET} locally..."
-    git tag -fa $LATEST_TAG
-
-    echo
-    echo -e "${BYELLOW}Remember to 'git push origin --tags'!${RESET}"
-fi
-
 # ---------------------- FRONTEND ----------------------
 
 pushd ../invokeai/frontend/web >/dev/null

installer/lib/installer.py

@@ -244,9 +244,9 @@ class InvokeAiInstance:
                 "numpy~=1.24.0",  # choose versions that won't be uninstalled during phase 2
                 "urllib3~=1.26.0",
                 "requests~=2.28.0",
-                "torch==2.1.0",
+                "torch==2.1.1",
                 "torchmetrics==0.11.4",
-                "torchvision>=0.14.1",
+                "torchvision>=0.16.1",
                 "--force-reinstall",
                 "--find-links" if find_links is not None else None,
                 find_links,

installer/tag_release.sh

@@ -0,0 +1,71 @@
+#!/bin/bash
+
+set -e
+
+BCYAN="\e[1;36m"
+BYELLOW="\e[1;33m"
+BGREEN="\e[1;32m"
+BRED="\e[1;31m"
+RED="\e[31m"
+RESET="\e[0m"
+
+function does_tag_exist {
+    git rev-parse --quiet --verify "refs/tags/$1" >/dev/null
+}
+
+function git_show_ref {
+    git show-ref --dereference $1 --abbrev 7
+}
+
+function git_show {
+    git show -s --format='%h %s' $1
+}
+
+VERSION=$(
+    cd ..
+    python -c "from invokeai.version import __version__ as version; print(version)"
+)
+PATCH=""
+MAJOR_VERSION=$(echo $VERSION | sed 's/\..*$//')
+VERSION="v${VERSION}${PATCH}"
+LATEST_TAG="v${MAJOR_VERSION}-latest"
+
+if does_tag_exist $VERSION; then
+    echo -e "${BCYAN}${VERSION}${RESET} already exists:"
+    git_show_ref tags/$VERSION
+    echo
+fi
+if does_tag_exist $LATEST_TAG; then
+    echo -e "${BCYAN}${LATEST_TAG}${RESET} already exists:"
+    git_show_ref tags/$LATEST_TAG
+    echo
+fi
+
+echo -e "${BGREEN}HEAD${RESET}:"
+git_show
+echo
+
+echo -e -n "Create tags ${BCYAN}${VERSION}${RESET} and ${BCYAN}${LATEST_TAG}${RESET} @ ${BGREEN}HEAD${RESET}, ${RED}deleting existing tags on remote${RESET}? "
+read -e -p 'y/n [n]: ' input
+RESPONSE=${input:='n'}
+if [ "$RESPONSE" == 'y' ]; then
+    echo
+    echo -e "Deleting ${BCYAN}${VERSION}${RESET} tag on remote..."
+    git push --delete origin $VERSION
+
+    echo -e "Tagging ${BGREEN}HEAD${RESET} with ${BCYAN}${VERSION}${RESET} locally..."
+    if ! git tag -fa $VERSION; then
+        echo "Existing/invalid tag"
+        exit -1
+    fi
+
+    echo -e "Deleting ${BCYAN}${LATEST_TAG}${RESET} tag on remote..."
+    git push --delete origin $LATEST_TAG
+
+    echo -e "Tagging ${BGREEN}HEAD${RESET} with ${BCYAN}${LATEST_TAG}${RESET} locally..."
+    git tag -fa $LATEST_TAG
+
+    echo -e "Pushing updated tags to remote..."
+    git push origin --tags
+fi
+exit 0

invokeai/app/api/dependencies.py

@@ -2,6 +2,7 @@
 
 from logging import Logger
 
+from invokeai.app.services.shared.sqlite.sqlite_util import init_db
 from invokeai.backend.util.logging import InvokeAILogger
 from invokeai.version.invokeai_version import __version__
 
@@ -22,6 +23,7 @@ from ..services.invoker import Invoker
 from ..services.item_storage.item_storage_sqlite import SqliteItemStorage
 from ..services.latents_storage.latents_storage_disk import DiskLatentsStorage
 from ..services.latents_storage.latents_storage_forward_cache import ForwardCacheLatentsStorage
+from ..services.model_install import ModelInstallService
 from ..services.model_manager.model_manager_default import ModelManagerService
 from ..services.model_records import ModelRecordServiceSQL
 from ..services.names.names_default import SimpleNameService
@@ -29,7 +31,6 @@ from ..services.session_processor.session_processor_default import DefaultSessio
 from ..services.session_queue.session_queue_sqlite import SqliteSessionQueue
 from ..services.shared.default_graphs import create_system_graphs
 from ..services.shared.graph import GraphExecutionState, LibraryGraph
-from ..services.shared.sqlite.sqlite_database import SqliteDatabase
 from ..services.urls.urls_default import LocalUrlService
 from ..services.workflow_records.workflow_records_sqlite import SqliteWorkflowRecordsStorage
 from .events import FastAPIEventService
@@ -66,8 +67,9 @@ class ApiDependencies:
         logger.debug(f"Internet connectivity is {config.internet_available}")
 
         output_folder = config.output_path
+        image_files = DiskImageFileStorage(f"{output_folder}/images")
 
-        db = SqliteDatabase(config, logger)
+        db = init_db(config=config, logger=logger, image_files=image_files)
 
         configuration = config
         logger = logger
@@ -79,13 +81,15 @@ class ApiDependencies:
         events = FastAPIEventService(event_handler_id)
         graph_execution_manager = SqliteItemStorage[GraphExecutionState](db=db, table_name="graph_executions")
         graph_library = SqliteItemStorage[LibraryGraph](db=db, table_name="graphs")
-        image_files = DiskImageFileStorage(f"{output_folder}/images")
         image_records = SqliteImageRecordStorage(db=db)
         images = ImageService()
         invocation_cache = MemoryInvocationCache(max_cache_size=config.node_cache_size)
         latents = ForwardCacheLatentsStorage(DiskLatentsStorage(f"{output_folder}/latents"))
         model_manager = ModelManagerService(config, logger)
         model_record_service = ModelRecordServiceSQL(db=db)
+        model_install_service = ModelInstallService(
+            app_config=config, record_store=model_record_service, event_bus=events
+        )
         names = SimpleNameService()
         performance_statistics = InvocationStatsService()
         processor = DefaultInvocationProcessor()
@@ -112,6 +116,7 @@ class ApiDependencies:
             logger=logger,
             model_manager=model_manager,
             model_records=model_record_service,
+            model_install=model_install_service,
             names=names,
             performance_statistics=performance_statistics,
             processor=processor,

invokeai/app/api/routers/model_records.py

@@ -4,7 +4,7 @@
 
 from hashlib import sha1
 from random import randbytes
-from typing import List, Optional
+from typing import Any, Dict, List, Optional
 
 from fastapi import Body, Path, Query, Response
 from fastapi.routing import APIRouter
@@ -12,6 +12,7 @@ from pydantic import BaseModel, ConfigDict
 from starlette.exceptions import HTTPException
 from typing_extensions import Annotated
 
+from invokeai.app.services.model_install import ModelInstallJob, ModelSource
 from invokeai.app.services.model_records import (
     DuplicateModelException,
     InvalidModelException,
@@ -25,7 +26,7 @@ from invokeai.backend.model_manager.config import (
 
 from ..dependencies import ApiDependencies
 
-model_records_router = APIRouter(prefix="/v1/model/record", tags=["models"])
+model_records_router = APIRouter(prefix="/v1/model/record", tags=["model_manager_v2"])
 
 
 class ModelsList(BaseModel):
@@ -43,15 +44,25 @@ class ModelsList(BaseModel):
 async def list_model_records(
     base_models: Optional[List[BaseModelType]] = Query(default=None, description="Base models to include"),
     model_type: Optional[ModelType] = Query(default=None, description="The type of model to get"),
+    model_name: Optional[str] = Query(default=None, description="Exact match on the name of the model"),
+    model_format: Optional[str] = Query(
+        default=None, description="Exact match on the format of the model (e.g. 'diffusers')"
+    ),
 ) -> ModelsList:
     """Get a list of models."""
     record_store = ApiDependencies.invoker.services.model_records
     found_models: list[AnyModelConfig] = []
     if base_models:
         for base_model in base_models:
-            found_models.extend(record_store.search_by_attr(base_model=base_model, model_type=model_type))
+            found_models.extend(
+                record_store.search_by_attr(
+                    base_model=base_model, model_type=model_type, model_name=model_name, model_format=model_format
+                )
+            )
     else:
-        found_models.extend(record_store.search_by_attr(model_type=model_type))
+        found_models.extend(
+            record_store.search_by_attr(model_type=model_type, model_name=model_name, model_format=model_format)
+        )
     return ModelsList(models=found_models)
 
 
@@ -117,12 +128,17 @@ async def update_model_record(
 async def del_model_record(
     key: str = Path(description="Unique key of model to remove from model registry."),
 ) -> Response:
-    """Delete Model"""
+    """
+    Delete model record from database.
+
+    The configuration record will be removed. The corresponding weights files will be
+    deleted as well if they reside within the InvokeAI "models" directory.
+    """
     logger = ApiDependencies.invoker.services.logger
 
     try:
-        record_store = ApiDependencies.invoker.services.model_records
-        record_store.del_model(key)
+        installer = ApiDependencies.invoker.services.model_install
+        installer.delete(key)
         logger.info(f"Deleted model: {key}")
         return Response(status_code=204)
     except UnknownModelException as e:
@@ -162,3 +178,145 @@ async def add_model_record(
 
     # now fetch it out
     return record_store.get_model(config.key)
+
+
+@model_records_router.post(
+    "/import",
+    operation_id="import_model_record",
+    responses={
+        201: {"description": "The model imported successfully"},
+        415: {"description": "Unrecognized file/folder format"},
+        424: {"description": "The model appeared to import successfully, but could not be found in the model manager"},
+        409: {"description": "There is already a model corresponding to this path or repo_id"},
+    },
+    status_code=201,
+)
+async def import_model(
+    source: ModelSource,
+    config: Optional[Dict[str, Any]] = Body(
+        description="Dict of fields that override auto-probed values in the model config record, such as name, description and prediction_type ",
+        default=None,
+    ),
+) -> ModelInstallJob:
+    """Add a model using its local path, repo_id, or remote URL.
+
+    Models will be downloaded, probed, configured and installed in a
+    series of background threads. The return object has `status` attribute
+    that can be used to monitor progress.
+
+    The source object is a discriminated Union of LocalModelSource,
+    HFModelSource and URLModelSource. Set the "type" field to the
+    appropriate value:
+
+    * To install a local path using LocalModelSource, pass a source of form:
+      `{
+        "type": "local",
+        "path": "/path/to/model",
+        "inplace": false
+      }`
+       The "inplace" flag, if true, will register the model in place in its
+       current filesystem location. Otherwise, the model will be copied
+       into the InvokeAI models directory.
+
+    * To install a HuggingFace repo_id using HFModelSource, pass a source of form:
+      `{
+        "type": "hf",
+        "repo_id": "stabilityai/stable-diffusion-2.0",
+        "variant": "fp16",
+        "subfolder": "vae",
+        "access_token": "f5820a918aaf01"
+      }`
+     The `variant`, `subfolder` and `access_token` fields are optional.
+
+    * To install a remote model using an arbitrary URL, pass:
+      `{
+        "type": "url",
+        "url": "http://www.civitai.com/models/123456",
+        "access_token": "f5820a918aaf01"
+      }`
+    The `access_token` field is optonal
+
+    The model's configuration record will be probed and filled in
+    automatically.  To override the default guesses, pass "metadata"
+    with a Dict containing the attributes you wish to override.
+
+    Installation occurs in the background. Either use list_model_install_jobs()
+    to poll for completion, or listen on the event bus for the following events:
+
+      "model_install_started"
+      "model_install_completed"
+      "model_install_error"
+
+    On successful completion, the event's payload will contain the field "key"
+    containing the installed ID of the model. On an error, the event's payload
+    will contain the fields "error_type" and "error" describing the nature of the
+    error and its traceback, respectively.
+
+    """
+    logger = ApiDependencies.invoker.services.logger
+
+    try:
+        installer = ApiDependencies.invoker.services.model_install
+        result: ModelInstallJob = installer.import_model(
+            source=source,
+            config=config,
+        )
+        logger.info(f"Started installation of {source}")
+    except UnknownModelException as e:
+        logger.error(str(e))
+        raise HTTPException(status_code=424, detail=str(e))
+    except InvalidModelException as e:
+        logger.error(str(e))
+        raise HTTPException(status_code=415)
+    except ValueError as e:
+        logger.error(str(e))
+        raise HTTPException(status_code=409, detail=str(e))
+    return result
+
+
+@model_records_router.get(
+    "/import",
+    operation_id="list_model_install_jobs",
+)
+async def list_model_install_jobs() -> List[ModelInstallJob]:
+    """
+    Return list of model install jobs.
+
+    If the optional 'source' argument is provided, then the list will be filtered
+    for partial string matches against the install source.
+    """
+    jobs: List[ModelInstallJob] = ApiDependencies.invoker.services.model_install.list_jobs()
+    return jobs
+
+
+@model_records_router.patch(
+    "/import",
+    operation_id="prune_model_install_jobs",
+    responses={
+        204: {"description": "All completed and errored jobs have been pruned"},
+        400: {"description": "Bad request"},
+    },
+)
+async def prune_model_install_jobs() -> Response:
+    """
+    Prune all completed and errored jobs from the install job list.
+    """
+    ApiDependencies.invoker.services.model_install.prune_jobs()
+    return Response(status_code=204)
+
+
+@model_records_router.patch(
+    "/sync",
+    operation_id="sync_models_to_config",
+    responses={
+        204: {"description": "Model config record database resynced with files on disk"},
+        400: {"description": "Bad request"},
+    },
+)
+async def sync_models_to_config() -> Response:
+    """
+    Traverse the models and autoimport directories. Model files without a corresponding
+    record in the database are added. Orphan records without a models file are deleted.
+    """
+    ApiDependencies.invoker.services.model_install.sync_to_config()
+    return Response(status_code=204)

invokeai/app/api/sockets.py

@@ -20,6 +20,7 @@ class SocketIO:
         self.__sio.on("subscribe_queue", handler=self._handle_sub_queue)
         self.__sio.on("unsubscribe_queue", handler=self._handle_unsub_queue)
         local_handler.register(event_name=EventServiceBase.queue_event, _func=self._handle_queue_event)
+        local_handler.register(event_name=EventServiceBase.model_event, _func=self._handle_model_event)
 
     async def _handle_queue_event(self, event: Event):
         await self.__sio.emit(
@@ -28,10 +29,13 @@ class SocketIO:
             room=event[1]["data"]["queue_id"],
         )
 
-    async def _handle_sub_queue(self, sid, data, *args, **kwargs):
+    async def _handle_sub_queue(self, sid, data, *args, **kwargs) -> None:
         if "queue_id" in data:
             await self.__sio.enter_room(sid, data["queue_id"])
 
-    async def _handle_unsub_queue(self, sid, data, *args, **kwargs):
+    async def _handle_unsub_queue(self, sid, data, *args, **kwargs) -> None:
         if "queue_id" in data:
             await self.__sio.leave_room(sid, data["queue_id"])
+
+    async def _handle_model_event(self, event: Event) -> None:
+        await self.__sio.emit(event=event[1]["event"], data=event[1]["data"])

invokeai/app/api_app.py

@@ -272,6 +272,8 @@ def invoke_api() -> None:
         port=port,
         loop="asyncio",
         log_level=app_config.log_level,
+        ssl_certfile=app_config.ssl_certfile,
+        ssl_keyfile=app_config.ssl_keyfile,
     )
     server = uvicorn.Server(config)
 

invokeai/app/invocations/baseinvocation.py

@@ -39,6 +39,19 @@ class InvalidFieldError(TypeError):
     pass
 
 
+class Classification(str, Enum, metaclass=MetaEnum):
+    """
+    The classification of an Invocation.
+    - `Stable`: The invocation, including its inputs/outputs and internal logic, is stable. You may build workflows with it, having confidence that they will not break because of a change in this invocation.
+    - `Beta`: The invocation is not yet stable, but is planned to be stable in the future. Workflows built around this invocation may break, but we are committed to supporting this invocation long-term.
+    - `Prototype`: The invocation is not yet stable and may be removed from the application at any time. Workflows built around this invocation may break, and we are *not* committed to supporting this invocation.
+    """
+
+    Stable = "stable"
+    Beta = "beta"
+    Prototype = "prototype"
+
+
 class Input(str, Enum, metaclass=MetaEnum):
     """
     The type of input a field accepts.
@@ -439,6 +452,7 @@ class UIConfigBase(BaseModel):
         description='The node\'s version. Should be a valid semver string e.g. "1.0.0" or "3.8.13".',
     )
     node_pack: Optional[str] = Field(default=None, description="Whether or not this is a custom node")
+    classification: Classification = Field(default=Classification.Stable, description="The node's classification")
 
     model_config = ConfigDict(
         validate_assignment=True,
@@ -607,6 +621,7 @@ class BaseInvocation(ABC, BaseModel):
                 schema["category"] = uiconfig.category
             if uiconfig.node_pack is not None:
                 schema["node_pack"] = uiconfig.node_pack
+            schema["classification"] = uiconfig.classification
             schema["version"] = uiconfig.version
         if "required" not in schema or not isinstance(schema["required"], list):
             schema["required"] = []
@@ -782,6 +797,7 @@ def invocation(
     category: Optional[str] = None,
     version: Optional[str] = None,
     use_cache: Optional[bool] = True,
+    classification: Classification = Classification.Stable,
 ) -> Callable[[Type[TBaseInvocation]], Type[TBaseInvocation]]:
     """
     Registers an invocation.
@@ -792,6 +808,7 @@ def invocation(
     :param Optional[str] category: Adds a category to the invocation. Used to group the invocations in the UI. Defaults to None.
     :param Optional[str] version: Adds a version to the invocation. Must be a valid semver string. Defaults to None.
     :param Optional[bool] use_cache: Whether or not to use the invocation cache. Defaults to True. The user may override this in the workflow editor.
+    :param Classification classification: The classification of the invocation. Defaults to FeatureClassification.Stable. Use Beta or Prototype if the invocation is unstable.
     """
 
     def wrapper(cls: Type[TBaseInvocation]) -> Type[TBaseInvocation]:
@@ -812,6 +829,7 @@ def invocation(
         cls.UIConfig.title = title
         cls.UIConfig.tags = tags
         cls.UIConfig.category = category
+        cls.UIConfig.classification = classification
 
         # Grab the node pack's name from the module name, if it's a custom node
         is_custom_node = cls.__module__.rsplit(".", 1)[0] == "invokeai.app.invocations"

invokeai/app/invocations/image.py

@@ -13,7 +13,15 @@ from invokeai.app.shared.fields import FieldDescriptions
 from invokeai.backend.image_util.invisible_watermark import InvisibleWatermark
 from invokeai.backend.image_util.safety_checker import SafetyChecker
 
-from .baseinvocation import BaseInvocation, Input, InputField, InvocationContext, WithMetadata, invocation
+from .baseinvocation import (
+    BaseInvocation,
+    Classification,
+    Input,
+    InputField,
+    InvocationContext,
+    WithMetadata,
+    invocation,
+)
 
 
 @invocation("show_image", title="Show Image", tags=["image"], category="image", version="1.0.0")
@@ -421,6 +429,64 @@ class ImageBlurInvocation(BaseInvocation, WithMetadata):
         )
 
 
+@invocation(
+    "unsharp_mask",
+    title="Unsharp Mask",
+    tags=["image", "unsharp_mask"],
+    category="image",
+    version="1.2.0",
+    classification=Classification.Beta,
+)
+class UnsharpMaskInvocation(BaseInvocation, WithMetadata):
+    """Applies an unsharp mask filter to an image"""
+
+    image: ImageField = InputField(description="The image to use")
+    radius: float = InputField(gt=0, description="Unsharp mask radius", default=2)
+    strength: float = InputField(ge=0, description="Unsharp mask strength", default=50)
+
+    def pil_from_array(self, arr):
+        return Image.fromarray((arr * 255).astype("uint8"))
+
+    def array_from_pil(self, img):
+        return numpy.array(img) / 255
+
+    def invoke(self, context: InvocationContext) -> ImageOutput:
+        image = context.services.images.get_pil_image(self.image.image_name)
+        mode = image.mode
+
+        alpha_channel = image.getchannel("A") if mode == "RGBA" else None
+        image = image.convert("RGB")
+        image_blurred = self.array_from_pil(image.filter(ImageFilter.GaussianBlur(radius=self.radius)))
+
+        image = self.array_from_pil(image)
+        image += (image - image_blurred) * (self.strength / 100.0)
+        image = numpy.clip(image, 0, 1)
+        image = self.pil_from_array(image)
+
+        image = image.convert(mode)
+
+        # Make the image RGBA if we had a source alpha channel
+        if alpha_channel is not None:
+            image.putalpha(alpha_channel)
+
+        image_dto = context.services.images.create(
+            image=image,
+            image_origin=ResourceOrigin.INTERNAL,
+            image_category=ImageCategory.GENERAL,
+            node_id=self.id,
+            session_id=context.graph_execution_state_id,
+            is_intermediate=self.is_intermediate,
+            metadata=self.metadata,
+            workflow=context.workflow,
+        )
+
+        return ImageOutput(
+            image=ImageField(image_name=image_dto.image_name),
+            width=image.width,
+            height=image.height,
+        )
+
+
 PIL_RESAMPLING_MODES = Literal[
     "nearest",
     "box",

invokeai/app/invocations/tiles.py

@@ -1,3 +1,5 @@
+from typing import Literal
+
 import numpy as np
 from PIL import Image
 from pydantic import BaseModel
@@ -5,6 +7,8 @@ from pydantic import BaseModel
 from invokeai.app.invocations.baseinvocation import (
     BaseInvocation,
     BaseInvocationOutput,
+    Classification,
+    Input,
     InputField,
     InvocationContext,
     OutputField,
@@ -14,7 +18,13 @@ from invokeai.app.invocations.baseinvocation import (
 )
 from invokeai.app.invocations.primitives import ImageField, ImageOutput
 from invokeai.app.services.image_records.image_records_common import ImageCategory, ResourceOrigin
-from invokeai.backend.tiles.tiles import calc_tiles_with_overlap, merge_tiles_with_linear_blending
+from invokeai.backend.tiles.tiles import (
+    calc_tiles_even_split,
+    calc_tiles_min_overlap,
+    calc_tiles_with_overlap,
+    merge_tiles_with_linear_blending,
+    merge_tiles_with_seam_blending,
+)
 from invokeai.backend.tiles.utils import Tile
 
 
@@ -28,7 +38,14 @@ class CalculateImageTilesOutput(BaseInvocationOutput):
     tiles: list[Tile] = OutputField(description="The tiles coordinates that cover a particular image shape.")
 
 
-@invocation("calculate_image_tiles", title="Calculate Image Tiles", tags=["tiles"], category="tiles", version="1.0.0")
+@invocation(
+    "calculate_image_tiles",
+    title="Calculate Image Tiles",
+    tags=["tiles"],
+    category="tiles",
+    version="1.0.0",
+    classification=Classification.Beta,
+)
 class CalculateImageTilesInvocation(BaseInvocation):
     """Calculate the coordinates and overlaps of tiles that cover a target image shape."""
 
@@ -55,6 +72,79 @@ class CalculateImageTilesInvocation(BaseInvocation):
         return CalculateImageTilesOutput(tiles=tiles)
 
 
+@invocation(
+    "calculate_image_tiles_even_split",
+    title="Calculate Image Tiles Even Split",
+    tags=["tiles"],
+    category="tiles",
+    version="1.0.0",
+    classification=Classification.Beta,
+)
+class CalculateImageTilesEvenSplitInvocation(BaseInvocation):
+    """Calculate the coordinates and overlaps of tiles that cover a target image shape."""
+
+    image_width: int = InputField(ge=1, default=1024, description="The image width, in pixels, to calculate tiles for.")
+    image_height: int = InputField(
+        ge=1, default=1024, description="The image height, in pixels, to calculate tiles for."
+    )
+    num_tiles_x: int = InputField(
+        default=2,
+        ge=1,
+        description="Number of tiles to divide image into on the x axis",
+    )
+    num_tiles_y: int = InputField(
+        default=2,
+        ge=1,
+        description="Number of tiles to divide image into on the y axis",
+    )
+    overlap_fraction: float = InputField(
+        default=0.25,
+        ge=0,
+        lt=1,
+        description="Overlap between adjacent tiles as a fraction of the tile's dimensions (0-1)",
+    )
+
+    def invoke(self, context: InvocationContext) -> CalculateImageTilesOutput:
+        tiles = calc_tiles_even_split(
+            image_height=self.image_height,
+            image_width=self.image_width,
+            num_tiles_x=self.num_tiles_x,
+            num_tiles_y=self.num_tiles_y,
+            overlap_fraction=self.overlap_fraction,
+        )
+        return CalculateImageTilesOutput(tiles=tiles)
+
+
+@invocation(
+    "calculate_image_tiles_min_overlap",
+    title="Calculate Image Tiles Minimum Overlap",
+    tags=["tiles"],
+    category="tiles",
+    version="1.0.0",
+    classification=Classification.Beta,
+)
+class CalculateImageTilesMinimumOverlapInvocation(BaseInvocation):
+    """Calculate the coordinates and overlaps of tiles that cover a target image shape."""
+
+    image_width: int = InputField(ge=1, default=1024, description="The image width, in pixels, to calculate tiles for.")
+    image_height: int = InputField(
+        ge=1, default=1024, description="The image height, in pixels, to calculate tiles for."
+    )
+    tile_width: int = InputField(ge=1, default=576, description="The tile width, in pixels.")
+    tile_height: int = InputField(ge=1, default=576, description="The tile height, in pixels.")
+    min_overlap: int = InputField(default=128, ge=0, description="Minimum overlap between adjacent tiles, in pixels.")
+
+    def invoke(self, context: InvocationContext) -> CalculateImageTilesOutput:
+        tiles = calc_tiles_min_overlap(
+            image_height=self.image_height,
+            image_width=self.image_width,
+            tile_height=self.tile_height,
+            tile_width=self.tile_width,
+            min_overlap=self.min_overlap,
+        )
+        return CalculateImageTilesOutput(tiles=tiles)
+
+
 @invocation_output("tile_to_properties_output")
 class TileToPropertiesOutput(BaseInvocationOutput):
     coords_left: int = OutputField(description="Left coordinate of the tile relative to its parent image.")
@@ -76,7 +166,14 @@ class TileToPropertiesOutput(BaseInvocationOutput):
     overlap_right: int = OutputField(description="Overlap between this tile and its right neighbor.")
 
 
-@invocation("tile_to_properties", title="Tile to Properties", tags=["tiles"], category="tiles", version="1.0.0")
+@invocation(
+    "tile_to_properties",
+    title="Tile to Properties",
+    tags=["tiles"],
+    category="tiles",
+    version="1.0.0",
+    classification=Classification.Beta,
+)
 class TileToPropertiesInvocation(BaseInvocation):
     """Split a Tile into its individual properties."""
 
@@ -102,7 +199,14 @@ class PairTileImageOutput(BaseInvocationOutput):
     tile_with_image: TileWithImage = OutputField(description="A tile description with its corresponding image.")
 
 
-@invocation("pair_tile_image", title="Pair Tile with Image", tags=["tiles"], category="tiles", version="1.0.0")
+@invocation(
+    "pair_tile_image",
+    title="Pair Tile with Image",
+    tags=["tiles"],
+    category="tiles",
+    version="1.0.0",
+    classification=Classification.Beta,
+)
 class PairTileImageInvocation(BaseInvocation):
     """Pair an image with its tile properties."""
 
@@ -121,13 +225,29 @@ class PairTileImageInvocation(BaseInvocation):
         )
 
 
-@invocation("merge_tiles_to_image", title="Merge Tiles to Image", tags=["tiles"], category="tiles", version="1.1.0")
+BLEND_MODES = Literal["Linear", "Seam"]
+
+
+@invocation(
+    "merge_tiles_to_image",
+    title="Merge Tiles to Image",
+    tags=["tiles"],
+    category="tiles",
+    version="1.1.0",
+    classification=Classification.Beta,
+)
 class MergeTilesToImageInvocation(BaseInvocation, WithMetadata):
     """Merge multiple tile images into a single image."""
 
     # Inputs
     tiles_with_images: list[TileWithImage] = InputField(description="A list of tile images with tile properties.")
+    blend_mode: BLEND_MODES = InputField(
+        default="Seam",
+        description="blending type Linear or Seam",
+        input=Input.Direct,
+    )
     blend_amount: int = InputField(
+        default=32,
         ge=0,
         description="The amount to blend adjacent tiles in pixels. Must be <= the amount of overlap between adjacent tiles.",
     )
@@ -157,10 +277,18 @@ class MergeTilesToImageInvocation(BaseInvocation, WithMetadata):
         channels = tile_np_images[0].shape[-1]
         dtype = tile_np_images[0].dtype
         np_image = np.zeros(shape=(height, width, channels), dtype=dtype)
+        if self.blend_mode == "Linear":
+            merge_tiles_with_linear_blending(
+                dst_image=np_image, tiles=tiles, tile_images=tile_np_images, blend_amount=self.blend_amount
+            )
+        elif self.blend_mode == "Seam":
+            merge_tiles_with_seam_blending(
+                dst_image=np_image, tiles=tiles, tile_images=tile_np_images, blend_amount=self.blend_amount
+            )
+        else:
+            raise ValueError(f"Unsupported blend mode: '{self.blend_mode}'.")
 
-        merge_tiles_with_linear_blending(
-            dst_image=np_image, tiles=tiles, tile_images=tile_np_images, blend_amount=self.blend_amount
-        )
+        # Convert into a PIL image and save
         pil_image = Image.fromarray(np_image)
 
         image_dto = context.services.images.create(

invokeai/app/services/board_image_records/board_image_records_sqlite.py

@@ -20,63 +20,6 @@ class SqliteBoardImageRecordStorage(BoardImageRecordStorageBase):
         self._conn = db.conn
         self._cursor = self._conn.cursor()
 
-        try:
-            self._lock.acquire()
-            self._create_tables()
-            self._conn.commit()
-        finally:
-            self._lock.release()
-
-    def _create_tables(self) -> None:
-        """Creates the `board_images` junction table."""
-
-        # Create the `board_images` junction table.
-        self._cursor.execute(
-            """--sql
-            CREATE TABLE IF NOT EXISTS board_images (
-                board_id TEXT NOT NULL,
-                image_name TEXT NOT NULL,
-                created_at DATETIME NOT NULL DEFAULT(STRFTIME('%Y-%m-%d %H:%M:%f', 'NOW')),
-                -- updated via trigger
-                updated_at DATETIME NOT NULL DEFAULT(STRFTIME('%Y-%m-%d %H:%M:%f', 'NOW')),
-                -- Soft delete, currently unused
-                deleted_at DATETIME,
-                -- enforce one-to-many relationship between boards and images using PK
-                -- (we can extend this to many-to-many later)
-                PRIMARY KEY (image_name),
-                FOREIGN KEY (board_id) REFERENCES boards (board_id) ON DELETE CASCADE,
-                FOREIGN KEY (image_name) REFERENCES images (image_name) ON DELETE CASCADE
-            );
-            """
-        )
-
-        # Add index for board id
-        self._cursor.execute(
-            """--sql
-            CREATE INDEX IF NOT EXISTS idx_board_images_board_id ON board_images (board_id);
-            """
-        )
-
-        # Add index for board id, sorted by created_at
-        self._cursor.execute(
-            """--sql
-            CREATE INDEX IF NOT EXISTS idx_board_images_board_id_created_at ON board_images (board_id, created_at);
-            """
-        )
-
-        # Add trigger for `updated_at`.
-        self._cursor.execute(
-            """--sql
-            CREATE TRIGGER IF NOT EXISTS tg_board_images_updated_at
-            AFTER UPDATE
-            ON board_images FOR EACH ROW
-            BEGIN
-                UPDATE board_images SET updated_at = STRFTIME('%Y-%m-%d %H:%M:%f', 'NOW')
-                    WHERE board_id = old.board_id AND image_name = old.image_name;
-            END;
-            """
-        )
-
     def add_image_to_board(
         self,
         board_id: str,

invokeai/app/services/board_records/board_records_sqlite.py

@@ -28,52 +28,6 @@ class SqliteBoardRecordStorage(BoardRecordStorageBase):
         self._conn = db.conn
         self._cursor = self._conn.cursor()
 
-        try:
-            self._lock.acquire()
-            self._create_tables()
-            self._conn.commit()
-        finally:
-            self._lock.release()
-
-    def _create_tables(self) -> None:
-        """Creates the `boards` table and `board_images` junction table."""
-
-        # Create the `boards` table.
-        self._cursor.execute(
-            """--sql
-            CREATE TABLE IF NOT EXISTS boards (
-                board_id TEXT NOT NULL PRIMARY KEY,
-                board_name TEXT NOT NULL,
-                cover_image_name TEXT,
-                created_at DATETIME NOT NULL DEFAULT(STRFTIME('%Y-%m-%d %H:%M:%f', 'NOW')),
-                -- Updated via trigger
-                updated_at DATETIME NOT NULL DEFAULT(STRFTIME('%Y-%m-%d %H:%M:%f', 'NOW')),
-                -- Soft delete, currently unused
-                deleted_at DATETIME,
-                FOREIGN KEY (cover_image_name) REFERENCES images (image_name) ON DELETE SET NULL
-            );
-            """
-        )
-
-        self._cursor.execute(
-            """--sql
-            CREATE INDEX IF NOT EXISTS idx_boards_created_at ON boards (created_at);
-            """
-        )
-
-        # Add trigger for `updated_at`.
-        self._cursor.execute(
-            """--sql
-            CREATE TRIGGER IF NOT EXISTS tg_boards_updated_at
-            AFTER UPDATE
-            ON boards FOR EACH ROW
-            BEGIN
-                UPDATE boards SET updated_at = current_timestamp
-                    WHERE board_id = old.board_id;
-            END;
-            """
-        )
-
     def delete(self, board_id: str) -> None:
         try:
             self._lock.acquire()

invokeai/app/services/config/__init__.py

@@ -1,6 +1,5 @@
-"""
-Init file for InvokeAI configure package
-"""
+"""Init file for InvokeAI configure package."""
 
-from .config_base import PagingArgumentParser  # noqa F401
-from .config_default import InvokeAIAppConfig, get_invokeai_config  # noqa F401
+from .config_default import InvokeAIAppConfig, get_invokeai_config
+
+__all__ = ["InvokeAIAppConfig", "get_invokeai_config"]

invokeai/app/services/config/config_default.py

@@ -173,7 +173,7 @@ from __future__ import annotations
 
 import os
 from pathlib import Path
-from typing import ClassVar, Dict, List, Literal, Optional, Union, get_type_hints
+from typing import Any, ClassVar, Dict, List, Literal, Optional, Union, get_type_hints
 
 from omegaconf import DictConfig, OmegaConf
 from pydantic import Field, TypeAdapter
@@ -221,6 +221,9 @@ class InvokeAIAppConfig(InvokeAISettings):
     allow_credentials   : bool = Field(default=True, description="Allow CORS credentials", json_schema_extra=Categories.WebServer)
     allow_methods       : List[str] = Field(default=["*"], description="Methods allowed for CORS", json_schema_extra=Categories.WebServer)
     allow_headers       : List[str] = Field(default=["*"], description="Headers allowed for CORS", json_schema_extra=Categories.WebServer)
+    # SSL options correspond to https://www.uvicorn.org/settings/#https
+    ssl_certfile        : Optional[Path] = Field(default=None, description="SSL certificate file (for HTTPS)", json_schema_extra=Categories.WebServer)
+    ssl_keyfile         : Optional[Path] = Field(default=None, description="SSL key file", json_schema_extra=Categories.WebServer)
 
     # FEATURES
     esrgan              : bool = Field(default=True, description="Enable/disable upscaling code", json_schema_extra=Categories.Features)
@@ -334,7 +337,7 @@ class InvokeAIAppConfig(InvokeAISettings):
                 )
 
     @classmethod
-    def get_config(cls, **kwargs) -> InvokeAIAppConfig:
+    def get_config(cls, **kwargs: Dict[str, Any]) -> InvokeAIAppConfig:
         """Return a singleton InvokeAIAppConfig configuration object."""
         if (
             cls.singleton_config is None
@@ -383,17 +386,17 @@ class InvokeAIAppConfig(InvokeAISettings):
         return db_dir / DB_FILE
 
     @property
-    def model_conf_path(self) -> Optional[Path]:
+    def model_conf_path(self) -> Path:
         """Path to models configuration file."""
         return self._resolve(self.conf_path)
 
     @property
-    def legacy_conf_path(self) -> Optional[Path]:
+    def legacy_conf_path(self) -> Path:
         """Path to directory of legacy configuration files (e.g. v1-inference.yaml)."""
         return self._resolve(self.legacy_conf_dir)
 
     @property
-    def models_path(self) -> Optional[Path]:
+    def models_path(self) -> Path:
         """Path to the models directory."""
         return self._resolve(self.models_dir)
 

invokeai/app/services/events/__init__.py

@@ -0,0 +1 @@
+from .events_base import EventServiceBase  # noqa F401

invokeai/app/services/events/events_base.py

@@ -1,5 +1,6 @@
 # Copyright (c) 2022 Kyle Schouviller (https://github.com/kyle0654)
 
+
 from typing import Any, Optional
 
 from invokeai.app.services.invocation_processor.invocation_processor_common import ProgressImage
@@ -16,6 +17,7 @@ from invokeai.backend.model_management.models.base import BaseModelType, ModelTy
 
 class EventServiceBase:
     queue_event: str = "queue_event"
+    model_event: str = "model_event"
 
     """Basic event bus, to have an empty stand-in when not needed"""
 
@@ -30,6 +32,13 @@ class EventServiceBase:
             payload={"event": event_name, "data": payload},
         )
 
+    def __emit_model_event(self, event_name: str, payload: dict) -> None:
+        payload["timestamp"] = get_timestamp()
+        self.dispatch(
+            event_name=EventServiceBase.model_event,
+            payload={"event": event_name, "data": payload},
+        )
+
     # Define events here for every event in the system.
     # This will make them easier to integrate until we find a schema generator.
     def emit_generator_progress(
@@ -313,3 +322,73 @@ class EventServiceBase:
             event_name="queue_cleared",
             payload={"queue_id": queue_id},
         )
+
+    def emit_model_install_started(self, source: str) -> None:
+        """
+        Emitted when an install job is started.
+
+        :param source: Source of the model; local path, repo_id or url
+        """
+        self.__emit_model_event(
+            event_name="model_install_started",
+            payload={"source": source},
+        )
+
+    def emit_model_install_completed(self, source: str, key: str) -> None:
+        """
+        Emitted when an install job is completed successfully.
+
+        :param source: Source of the model; local path, repo_id or url
+        :param key: Model config record key
+        """
+        self.__emit_model_event(
+            event_name="model_install_completed",
+            payload={
+                "source": source,
+                "key": key,
+            },
+        )
+
+    def emit_model_install_progress(
+        self,
+        source: str,
+        current_bytes: int,
+        total_bytes: int,
+    ) -> None:
+        """
+        Emitted while the install job is in progress.
+        (Downloaded models only)
+
+        :param source: Source of the model
+        :param current_bytes: Number of bytes downloaded so far
+        :param total_bytes: Total bytes to download
+        """
+        self.__emit_model_event(
+            event_name="model_install_progress",
+            payload={
+                "source": source,
+                "current_bytes": int,
+                "total_bytes": int,
+            },
+        )
+
+    def emit_model_install_error(
+        self,
+        source: str,
+        error_type: str,
+        error: str,
+    ) -> None:
+        """
+        Emitted when an install job encounters an exception.
+
+        :param source: Source of the model
+        :param exception: The exception that raised the error
+        """
+        self.__emit_model_event(
+            event_name="model_install_error",
+            payload={
+                "source": source,
+                "error_type": error_type,
+                "error": error,
+            },
+        )

invokeai/app/services/image_records/image_records_sqlite.py

@@ -32,101 +32,6 @@ class SqliteImageRecordStorage(ImageRecordStorageBase):
         self._conn = db.conn
         self._cursor = self._conn.cursor()
 
-        try:
-            self._lock.acquire()
-            self._create_tables()
-            self._conn.commit()
-        finally:
-            self._lock.release()
-
-    def _create_tables(self) -> None:
-        """Creates the `images` table."""
-
-        # Create the `images` table.
-        self._cursor.execute(
-            """--sql
-            CREATE TABLE IF NOT EXISTS images (
-                image_name TEXT NOT NULL PRIMARY KEY,
-                -- This is an enum in python, unrestricted string here for flexibility
-                image_origin TEXT NOT NULL,
-                -- This is an enum in python, unrestricted string here for flexibility
-                image_category TEXT NOT NULL,
-                width INTEGER NOT NULL,
-                height INTEGER NOT NULL,
-                session_id TEXT,
-                node_id TEXT,
-                metadata TEXT,
-                is_intermediate BOOLEAN DEFAULT FALSE,
-                created_at DATETIME NOT NULL DEFAULT(STRFTIME('%Y-%m-%d %H:%M:%f', 'NOW')),
-                -- Updated via trigger
-                updated_at DATETIME NOT NULL DEFAULT(STRFTIME('%Y-%m-%d %H:%M:%f', 'NOW')),
-                -- Soft delete, currently unused
-                deleted_at DATETIME
-            );
-            """
-        )
-
-        self._cursor.execute("PRAGMA table_info(images)")
-        columns = [column[1] for column in self._cursor.fetchall()]
-
-        if "starred" not in columns:
-            self._cursor.execute(
-                """--sql
-                ALTER TABLE images ADD COLUMN starred BOOLEAN DEFAULT FALSE;
-                """
-            )
-
-        # Create the `images` table indices.
-        self._cursor.execute(
-            """--sql
-            CREATE UNIQUE INDEX IF NOT EXISTS idx_images_image_name ON images(image_name);
-            """
-        )
-        self._cursor.execute(
-            """--sql
-            CREATE INDEX IF NOT EXISTS idx_images_image_origin ON images(image_origin);
-            """
-        )
-        self._cursor.execute(
-            """--sql
-            CREATE INDEX IF NOT EXISTS idx_images_image_category ON images(image_category);
-            """
-        )
-        self._cursor.execute(
-            """--sql
-            CREATE INDEX IF NOT EXISTS idx_images_created_at ON images(created_at);
-            """
-        )
-
-        self._cursor.execute(
-            """--sql
-            CREATE INDEX IF NOT EXISTS idx_images_starred ON images(starred);
-            """
-        )
-
-        # Add trigger for `updated_at`.
-        self._cursor.execute(
-            """--sql
-            CREATE TRIGGER IF NOT EXISTS tg_images_updated_at
-            AFTER UPDATE
-            ON images FOR EACH ROW
-            BEGIN
-                UPDATE images SET updated_at = STRFTIME('%Y-%m-%d %H:%M:%f', 'NOW')
-                    WHERE image_name = old.image_name;
-            END;
-            """
-        )
-
-        self._cursor.execute("PRAGMA table_info(images)")
-        columns = [column[1] for column in self._cursor.fetchall()]
-        if "has_workflow" not in columns:
-            self._cursor.execute(
-                """--sql
-                ALTER TABLE images
-                ADD COLUMN has_workflow BOOLEAN DEFAULT FALSE;
-                """
-            )
-
     def get(self, image_name: str) -> ImageRecord:
         try:
             self._lock.acquire()

invokeai/app/services/invocation_services.py

@@ -21,6 +21,7 @@ if TYPE_CHECKING:
     from .invocation_stats.invocation_stats_base import InvocationStatsServiceBase
     from .item_storage.item_storage_base import ItemStorageABC
     from .latents_storage.latents_storage_base import LatentsStorageBase
+    from .model_install import ModelInstallServiceBase
     from .model_manager.model_manager_base import ModelManagerServiceBase
     from .model_records import ModelRecordServiceBase
     from .names.names_base import NameServiceBase
@@ -50,6 +51,7 @@ class InvocationServices:
     logger: "Logger"
     model_manager: "ModelManagerServiceBase"
     model_records: "ModelRecordServiceBase"
+    model_install: "ModelInstallServiceBase"
     processor: "InvocationProcessorABC"
     performance_statistics: "InvocationStatsServiceBase"
     queue: "InvocationQueueABC"
@@ -77,6 +79,7 @@ class InvocationServices:
         logger: "Logger",
         model_manager: "ModelManagerServiceBase",
         model_records: "ModelRecordServiceBase",
+        model_install: "ModelInstallServiceBase",
         processor: "InvocationProcessorABC",
         performance_statistics: "InvocationStatsServiceBase",
         queue: "InvocationQueueABC",
@@ -102,6 +105,7 @@ class InvocationServices:
         self.logger = logger
         self.model_manager = model_manager
         self.model_records = model_records
+        self.model_install = model_install
         self.processor = processor
         self.performance_statistics = performance_statistics
         self.queue = queue

invokeai/app/services/model_install/__init__.py

@@ -0,0 +1,25 @@
+"""Initialization file for model install service package."""
+
+from .model_install_base import (
+    HFModelSource,
+    InstallStatus,
+    LocalModelSource,
+    ModelInstallJob,
+    ModelInstallServiceBase,
+    ModelSource,
+    UnknownInstallJobException,
+    URLModelSource,
+)
+from .model_install_default import ModelInstallService
+
+__all__ = [
+    "ModelInstallServiceBase",
+    "ModelInstallService",
+    "InstallStatus",
+    "ModelInstallJob",
+    "UnknownInstallJobException",
+    "ModelSource",
+    "LocalModelSource",
+    "HFModelSource",
+    "URLModelSource",
+]

invokeai/app/services/model_install/model_install_base.py

@@ -0,0 +1,306 @@
+import re
+import traceback
+from abc import ABC, abstractmethod
+from enum import Enum
+from pathlib import Path
+from typing import Any, Dict, List, Literal, Optional, Union
+
+from pydantic import BaseModel, Field, field_validator
+from pydantic.networks import AnyHttpUrl
+from typing_extensions import Annotated
+
+from invokeai.app.services.config import InvokeAIAppConfig
+from invokeai.app.services.events import EventServiceBase
+from invokeai.app.services.invoker import Invoker
+from invokeai.app.services.model_records import ModelRecordServiceBase
+from invokeai.backend.model_manager import AnyModelConfig
+
+
+class InstallStatus(str, Enum):
+    """State of an install job running in the background."""
+
+    WAITING = "waiting"  # waiting to be dequeued
+    RUNNING = "running"  # being processed
+    COMPLETED = "completed"  # finished running
+    ERROR = "error"  # terminated with an error message
+
+
+class UnknownInstallJobException(Exception):
+    """Raised when the status of an unknown job is requested."""
+
+
+class StringLikeSource(BaseModel):
+    """
+    Base class for model sources, implements functions that lets the source be sorted and indexed.
+
+    These shenanigans let this stuff work:
+
+      source1 = LocalModelSource(path='C:/users/mort/foo.safetensors')
+      mydict = {source1: 'model 1'}
+      assert mydict['C:/users/mort/foo.safetensors'] == 'model 1'
+      assert mydict[LocalModelSource(path='C:/users/mort/foo.safetensors')] == 'model 1'
+
+      source2 = LocalModelSource(path=Path('C:/users/mort/foo.safetensors'))
+      assert source1 == source2
+      assert source1 == 'C:/users/mort/foo.safetensors'
+    """
+
+    def __hash__(self) -> int:
+        """Return hash of the path field, for indexing."""
+        return hash(str(self))
+
+    def __lt__(self, other: object) -> int:
+        """Return comparison of the stringified version, for sorting."""
+        return str(self) < str(other)
+
+    def __eq__(self, other: object) -> bool:
+        """Return equality on the stringified version."""
+        if isinstance(other, Path):
+            return str(self) == other.as_posix()
+        else:
+            return str(self) == str(other)
+
+
+class LocalModelSource(StringLikeSource):
+    """A local file or directory path."""
+
+    path: str | Path
+    inplace: Optional[bool] = False
+    type: Literal["local"] = "local"
+
+    # these methods allow the source to be used in a string-like way,
+    # for example as an index into a dict
+    def __str__(self) -> str:
+        """Return string version of path when string rep needed."""
+        return Path(self.path).as_posix()
+
+
+class HFModelSource(StringLikeSource):
+    """A HuggingFace repo_id, with optional variant and sub-folder."""
+
+    repo_id: str
+    variant: Optional[str] = None
+    subfolder: Optional[str | Path] = None
+    access_token: Optional[str] = None
+    type: Literal["hf"] = "hf"
+
+    @field_validator("repo_id")
+    @classmethod
+    def proper_repo_id(cls, v: str) -> str:  # noqa D102
+        if not re.match(r"^([.\w-]+/[.\w-]+)$", v):
+            raise ValueError(f"{v}: invalid repo_id format")
+        return v
+
+    def __str__(self) -> str:
+        """Return string version of repoid when string rep needed."""
+        base: str = self.repo_id
+        base += f":{self.subfolder}" if self.subfolder else ""
+        base += f" ({self.variant})" if self.variant else ""
+        return base
+
+
+class URLModelSource(StringLikeSource):
+    """A generic URL point to a checkpoint file."""
+
+    url: AnyHttpUrl
+    access_token: Optional[str] = None
+    type: Literal["generic_url"] = "generic_url"
+
+    def __str__(self) -> str:
+        """Return string version of the url when string rep needed."""
+        return str(self.url)
+
+
+ModelSource = Annotated[Union[LocalModelSource, HFModelSource, URLModelSource], Field(discriminator="type")]
+
+
+class ModelInstallJob(BaseModel):
+    """Object that tracks the current status of an install request."""
+
+    status: InstallStatus = Field(default=InstallStatus.WAITING, description="Current status of install process")
+    config_in: Dict[str, Any] = Field(
+        default_factory=dict, description="Configuration information (e.g. 'description') to apply to model."
+    )
+    config_out: Optional[AnyModelConfig] = Field(
+        default=None, description="After successful installation, this will hold the configuration object."
+    )
+    inplace: bool = Field(
+        default=False, description="Leave model in its current location; otherwise install under models directory"
+    )
+    source: ModelSource = Field(description="Source (URL, repo_id, or local path) of model")
+    local_path: Path = Field(description="Path to locally-downloaded model; may be the same as the source")
+    error_type: Optional[str] = Field(default=None, description="Class name of the exception that led to status==ERROR")
+    error: Optional[str] = Field(default=None, description="Error traceback")  # noqa #501
+
+    def set_error(self, e: Exception) -> None:
+        """Record the error and traceback from an exception."""
+        self.error_type = e.__class__.__name__
+        self.error = "".join(traceback.format_exception(e))
+        self.status = InstallStatus.ERROR
+
+
+class ModelInstallServiceBase(ABC):
+    """Abstract base class for InvokeAI model installation."""
+
+    @abstractmethod
+    def __init__(
+        self,
+        app_config: InvokeAIAppConfig,
+        record_store: ModelRecordServiceBase,
+        event_bus: Optional["EventServiceBase"] = None,
+    ):
+        """
+        Create ModelInstallService object.
+
+        :param config: Systemwide InvokeAIAppConfig.
+        :param store: Systemwide ModelConfigStore
+        :param event_bus: InvokeAI event bus for reporting events to.
+        """
+
+    def start(self, invoker: Invoker) -> None:
+        """Call at InvokeAI startup time."""
+        self.sync_to_config()
+
+    @abstractmethod
+    def stop(self) -> None:
+        """Stop the model install service. After this the objection can be safely deleted."""
+
+    @property
+    @abstractmethod
+    def app_config(self) -> InvokeAIAppConfig:
+        """Return the appConfig object associated with the installer."""
+
+    @property
+    @abstractmethod
+    def record_store(self) -> ModelRecordServiceBase:
+        """Return the ModelRecoreService object associated with the installer."""
+
+    @property
+    @abstractmethod
+    def event_bus(self) -> Optional[EventServiceBase]:
+        """Return the event service base object associated with the installer."""
+
+    @abstractmethod
+    def register_path(
+        self,
+        model_path: Union[Path, str],
+        config: Optional[Dict[str, Any]] = None,
+    ) -> str:
+        """
+        Probe and register the model at model_path.
+
+        This keeps the model in its current location.
+
+        :param model_path: Filesystem Path to the model.
+        :param config: Dict of attributes that will override autoassigned values.
+        :returns id: The string ID of the registered model.
+        """
+
+    @abstractmethod
+    def unregister(self, key: str) -> None:
+        """Remove model with indicated key from the database."""
+
+    @abstractmethod
+    def delete(self, key: str) -> None:
+        """Remove model with indicated key from the database. Delete its files only if they are within our models directory."""
+
+    @abstractmethod
+    def unconditionally_delete(self, key: str) -> None:
+        """Remove model with indicated key from the database and unconditionally delete weight files from disk."""
+
+    @abstractmethod
+    def install_path(
+        self,
+        model_path: Union[Path, str],
+        config: Optional[Dict[str, Any]] = None,
+    ) -> str:
+        """
+        Probe, register and install the model in the models directory.
+
+        This moves the model from its current location into
+        the models directory handled by InvokeAI.
+
+        :param model_path: Filesystem Path to the model.
+        :param config: Dict of attributes that will override autoassigned values.
+        :returns id: The string ID of the registered model.
+        """
+
+    @abstractmethod
+    def import_model(
+        self,
+        source: ModelSource,
+        config: Optional[Dict[str, Any]] = None,
+    ) -> ModelInstallJob:
+        """Install the indicated model.
+
+        :param source: ModelSource object
+
+        :param config: Optional dict. Any fields in this dict
+         will override corresponding autoassigned probe fields in the
+         model's config record. Use it to override
+         `name`, `description`, `base_type`, `model_type`, `format`,
+         `prediction_type`, `image_size`, and/or `ztsnr_training`.
+
+        This will download the model located at `source`,
+        probe it, and install it into the models directory.
+        This call is executed asynchronously in a separate
+        thread and will issue the following events on the event bus:
+
+             - model_install_started
+             - model_install_error
+             - model_install_completed
+
+        The `inplace` flag does not affect the behavior of downloaded
+        models, which are always moved into the `models` directory.
+
+        The call returns a ModelInstallJob object which can be
+        polled to learn the current status and/or error message.
+
+        Variants recognized by HuggingFace currently are:
+        1. onnx
+        2. openvino
+        3. fp16
+        4. None (usually returns fp32 model)
+
+        """
+
+    @abstractmethod
+    def get_job(self, source: ModelSource) -> List[ModelInstallJob]:
+        """Return the ModelInstallJob(s) corresponding to the provided source."""
+
+    @abstractmethod
+    def list_jobs(self) -> List[ModelInstallJob]:  # noqa D102
+        """
+        List active and complete install jobs.
+        """
+
+    @abstractmethod
+    def prune_jobs(self) -> None:
+        """Prune all completed and errored jobs."""
+
+    @abstractmethod
+    def wait_for_installs(self) -> List[ModelInstallJob]:
+        """
+        Wait for all pending installs to complete.
+
+        This will block until all pending installs have
+        completed, been cancelled, or errored out. It will
+        block indefinitely if one or more jobs are in the
+        paused state.
+
+        It will return the current list of jobs.
+        """
+
+    @abstractmethod
+    def scan_directory(self, scan_dir: Path, install: bool = False) -> List[str]:
+        """
+        Recursively scan directory for new models and register or install them.
+
+        :param scan_dir: Path to the directory to scan.
+        :param install: Install if True, otherwise register in place.
+        :returns list of IDs: Returns list of IDs of models registered/installed
+        """
+
+    @abstractmethod
+    def sync_to_config(self) -> None:
+        """Synchronize models on disk to those in the model record database."""

invokeai/app/services/model_install/model_install_default.py

@@ -0,0 +1,395 @@
+"""Model installation class."""
+
+import threading
+from hashlib import sha256
+from logging import Logger
+from pathlib import Path
+from queue import Queue
+from random import randbytes
+from shutil import copyfile, copytree, move, rmtree
+from typing import Any, Dict, List, Optional, Set, Union
+
+from invokeai.app.services.config import InvokeAIAppConfig
+from invokeai.app.services.events import EventServiceBase
+from invokeai.app.services.model_records import DuplicateModelException, ModelRecordServiceBase, UnknownModelException
+from invokeai.backend.model_manager.config import (
+    AnyModelConfig,
+    BaseModelType,
+    InvalidModelConfigException,
+    ModelType,
+)
+from invokeai.backend.model_manager.hash import FastModelHash
+from invokeai.backend.model_manager.probe import ModelProbe
+from invokeai.backend.model_manager.search import ModelSearch
+from invokeai.backend.util import Chdir, InvokeAILogger
+
+from .model_install_base import (
+    InstallStatus,
+    LocalModelSource,
+    ModelInstallJob,
+    ModelInstallServiceBase,
+    ModelSource,
+)
+
+# marker that the queue is done and that thread should exit
+STOP_JOB = ModelInstallJob(
+    source=LocalModelSource(path="stop"),
+    local_path=Path("/dev/null"),
+)
+
+
+class ModelInstallService(ModelInstallServiceBase):
+    """class for InvokeAI model installation."""
+
+    _app_config: InvokeAIAppConfig
+    _record_store: ModelRecordServiceBase
+    _event_bus: Optional[EventServiceBase] = None
+    _install_queue: Queue[ModelInstallJob]
+    _install_jobs: List[ModelInstallJob]
+    _logger: Logger
+    _cached_model_paths: Set[Path]
+    _models_installed: Set[str]
+
+    def __init__(
+        self,
+        app_config: InvokeAIAppConfig,
+        record_store: ModelRecordServiceBase,
+        event_bus: Optional[EventServiceBase] = None,
+    ):
+        """
+        Initialize the installer object.
+
+        :param app_config: InvokeAIAppConfig object
+        :param record_store: Previously-opened ModelRecordService database
+        :param event_bus: Optional EventService object
+        """
+        self._app_config = app_config
+        self._record_store = record_store
+        self._event_bus = event_bus
+        self._logger = InvokeAILogger.get_logger(name=self.__class__.__name__)
+        self._install_jobs = []
+        self._install_queue = Queue()
+        self._cached_model_paths = set()
+        self._models_installed = set()
+        self._start_installer_thread()
+
+    @property
+    def app_config(self) -> InvokeAIAppConfig:  # noqa D102
+        return self._app_config
+
+    @property
+    def record_store(self) -> ModelRecordServiceBase:  # noqa D102
+        return self._record_store
+
+    @property
+    def event_bus(self) -> Optional[EventServiceBase]:  # noqa D102
+        return self._event_bus
+
+    def stop(self, *args, **kwargs) -> None:
+        """Stop the install thread; after this the object can be deleted and garbage collected."""
+        self._install_queue.put(STOP_JOB)
+
+    def _start_installer_thread(self) -> None:
+        threading.Thread(target=self._install_next_item, daemon=True).start()
+
+    def _install_next_item(self) -> None:
+        done = False
+        while not done:
+            job = self._install_queue.get()
+            if job == STOP_JOB:
+                done = True
+                continue
+
+            assert job.local_path is not None
+            try:
+                self._signal_job_running(job)
+                if job.inplace:
+                    key = self.register_path(job.local_path, job.config_in)
+                else:
+                    key = self.install_path(job.local_path, job.config_in)
+                job.config_out = self.record_store.get_model(key)
+                self._signal_job_completed(job)
+
+            except (OSError, DuplicateModelException, InvalidModelConfigException) as excp:
+                self._signal_job_errored(job, excp)
+            finally:
+                self._install_queue.task_done()
+        self._logger.info("Install thread exiting")
+
+    def _signal_job_running(self, job: ModelInstallJob) -> None:
+        job.status = InstallStatus.RUNNING
+        self._logger.info(f"{job.source}: model installation started")
+        if self._event_bus:
+            self._event_bus.emit_model_install_started(str(job.source))
+
+    def _signal_job_completed(self, job: ModelInstallJob) -> None:
+        job.status = InstallStatus.COMPLETED
+        assert job.config_out
+        self._logger.info(
+            f"{job.source}: model installation completed. {job.local_path} registered key {job.config_out.key}"
+        )
+        if self._event_bus:
+            assert job.local_path is not None
+            assert job.config_out is not None
+            key = job.config_out.key
+            self._event_bus.emit_model_install_completed(str(job.source), key)
+
+    def _signal_job_errored(self, job: ModelInstallJob, excp: Exception) -> None:
+        job.set_error(excp)
+        self._logger.info(f"{job.source}: model installation encountered an exception: {job.error_type}")
+        if self._event_bus:
+            error_type = job.error_type
+            error = job.error
+            assert error_type is not None
+            assert error is not None
+            self._event_bus.emit_model_install_error(str(job.source), error_type, error)
+
+    def register_path(
+        self,
+        model_path: Union[Path, str],
+        config: Optional[Dict[str, Any]] = None,
+    ) -> str:  # noqa D102
+        model_path = Path(model_path)
+        config = config or {}
+        if config.get("source") is None:
+            config["source"] = model_path.resolve().as_posix()
+        return self._register(model_path, config)
+
+    def install_path(
+        self,
+        model_path: Union[Path, str],
+        config: Optional[Dict[str, Any]] = None,
+    ) -> str:  # noqa D102
+        model_path = Path(model_path)
+        config = config or {}
+        if config.get("source") is None:
+            config["source"] = model_path.resolve().as_posix()
+
+        info: AnyModelConfig = self._probe_model(Path(model_path), config)
+        old_hash = info.original_hash
+        dest_path = self.app_config.models_path / info.base.value / info.type.value / model_path.name
+        new_path = self._copy_model(model_path, dest_path)
+        new_hash = FastModelHash.hash(new_path)
+        assert new_hash == old_hash, f"{model_path}: Model hash changed during installation, possibly corrupted."
+
+        return self._register(
+            new_path,
+            config,
+            info,
+        )
+
+    def import_model(
+        self,
+        source: ModelSource,
+        config: Optional[Dict[str, Any]] = None,
+    ) -> ModelInstallJob:  # noqa D102
+        if not config:
+            config = {}
+
+        # Installing a local path
+        if isinstance(source, LocalModelSource) and Path(source.path).exists():  # a path that is already on disk
+            job = ModelInstallJob(
+                source=source,
+                config_in=config,
+                local_path=Path(source.path),
+            )
+            self._install_jobs.append(job)
+            self._install_queue.put(job)
+            return job
+
+        else:  # here is where we'd download a URL or repo_id. Implementation pending download queue.
+            raise UnknownModelException("File or directory not found")
+
+    def list_jobs(self) -> List[ModelInstallJob]:  # noqa D102
+        return self._install_jobs
+
+    def get_job(self, source: ModelSource) -> List[ModelInstallJob]:  # noqa D102
+        return [x for x in self._install_jobs if x.source == source]
+
+    def wait_for_installs(self) -> List[ModelInstallJob]:  # noqa D102
+        self._install_queue.join()
+        return self._install_jobs
+
+    def prune_jobs(self) -> None:
+        """Prune all completed and errored jobs."""
+        unfinished_jobs = [
+            x for x in self._install_jobs if x.status not in [InstallStatus.COMPLETED, InstallStatus.ERROR]
+        ]
+        self._install_jobs = unfinished_jobs
+
+    def sync_to_config(self) -> None:
+        """Synchronize models on disk to those in the config record store database."""
+        self._scan_models_directory()
+        if autoimport := self._app_config.autoimport_dir:
+            self._logger.info("Scanning autoimport directory for new models")
+            installed = self.scan_directory(self._app_config.root_path / autoimport)
+            self._logger.info(f"{len(installed)} new models registered")
+        self._logger.info("Model installer (re)initialized")
+
+    def scan_directory(self, scan_dir: Path, install: bool = False) -> List[str]:  # noqa D102
+        self._cached_model_paths = {Path(x.path) for x in self.record_store.all_models()}
+        callback = self._scan_install if install else self._scan_register
+        search = ModelSearch(on_model_found=callback)
+        self._models_installed: Set[str] = set()
+        search.search(scan_dir)
+        return list(self._models_installed)
+
+    def _scan_models_directory(self) -> None:
+        """
+        Scan the models directory for new and missing models.
+
+        New models will be added to the storage backend. Missing models
+        will be deleted.
+        """
+        defunct_models = set()
+        installed = set()
+
+        with Chdir(self._app_config.models_path):
+            self._logger.info("Checking for models that have been moved or deleted from disk")
+            for model_config in self.record_store.all_models():
+                path = Path(model_config.path)
+                if not path.exists():
+                    self._logger.info(f"{model_config.name}: path {path.as_posix()} no longer exists. Unregistering")
+                    defunct_models.add(model_config.key)
+            for key in defunct_models:
+                self.unregister(key)
+
+            self._logger.info(f"Scanning {self._app_config.models_path} for new and orphaned models")
+            for cur_base_model in BaseModelType:
+                for cur_model_type in ModelType:
+                    models_dir = Path(cur_base_model.value, cur_model_type.value)
+                    installed.update(self.scan_directory(models_dir))
+            self._logger.info(f"{len(installed)} new models registered; {len(defunct_models)} unregistered")
+
+    def _sync_model_path(self, key: str, ignore_hash_change: bool = False) -> AnyModelConfig:
+        """
+        Move model into the location indicated by its basetype, type and name.
+
+        Call this after updating a model's attributes in order to move
+        the model's path into the location indicated by its basetype, type and
+        name. Applies only to models whose paths are within the root `models_dir`
+        directory.
+
+        May raise an UnknownModelException.
+        """
+        model = self.record_store.get_model(key)
+        old_path = Path(model.path)
+        models_dir = self.app_config.models_path
+
+        if not old_path.is_relative_to(models_dir):
+            return model
+
+        new_path = models_dir / model.base.value / model.type.value / model.name
+        self._logger.info(f"Moving {model.name} to {new_path}.")
+        new_path = self._move_model(old_path, new_path)
+        new_hash = FastModelHash.hash(new_path)
+        model.path = new_path.relative_to(models_dir).as_posix()
+        if model.current_hash != new_hash:
+            assert (
+                ignore_hash_change
+            ), f"{model.name}: Model hash changed during installation, model is possibly corrupted"
+            model.current_hash = new_hash
+            self._logger.info(f"Model has new hash {model.current_hash}, but will continue to be identified by {key}")
+        self.record_store.update_model(key, model)
+        return model
+
+    def _scan_register(self, model: Path) -> bool:
+        if model in self._cached_model_paths:
+            return True
+        try:
+            id = self.register_path(model)
+            self._sync_model_path(id)  # possibly move it to right place in `models`
+            self._logger.info(f"Registered {model.name} with id {id}")
+            self._models_installed.add(id)
+        except DuplicateModelException:
+            pass
+        return True
+
+    def _scan_install(self, model: Path) -> bool:
+        if model in self._cached_model_paths:
+            return True
+        try:
+            id = self.install_path(model)
+            self._logger.info(f"Installed {model} with id {id}")
+            self._models_installed.add(id)
+        except DuplicateModelException:
+            pass
+        return True
+
+    def unregister(self, key: str) -> None:  # noqa D102
+        self.record_store.del_model(key)
+
+    def delete(self, key: str) -> None:  # noqa D102
+        """Unregister the model. Delete its files only if they are within our models directory."""
+        model = self.record_store.get_model(key)
+        models_dir = self.app_config.models_path
+        model_path = models_dir / model.path
+        if model_path.is_relative_to(models_dir):
+            self.unconditionally_delete(key)
+        else:
+            self.unregister(key)
+
+    def unconditionally_delete(self, key: str) -> None:  # noqa D102
+        model = self.record_store.get_model(key)
+        path = self.app_config.models_path / model.path
+        if path.is_dir():
+            rmtree(path)
+        else:
+            path.unlink()
+        self.unregister(key)
+
+    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
+
+        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_model(self, model_path: Path, config: Optional[Dict[str, Any]] = None) -> AnyModelConfig:
+        info: AnyModelConfig = ModelProbe.probe(Path(model_path))
+        if config:  # used to override probe fields
+            for key, value in config.items():
+                setattr(info, key, value)
+        return info
+
+    def _create_key(self) -> str:
+        return sha256(randbytes(100)).hexdigest()[0:32]
+
+    def _register(
+        self, model_path: Path, config: Optional[Dict[str, Any]] = None, info: Optional[AnyModelConfig] = None
+    ) -> str:
+        info = info or ModelProbe.probe(model_path, config)
+        key = self._create_key()
+
+        model_path = model_path.absolute()
+        if model_path.is_relative_to(self.app_config.models_path):
+            model_path = model_path.relative_to(self.app_config.models_path)
+
+        info.path = model_path.as_posix()
+
+        # add 'main' specific fields
+        if hasattr(info, "config"):
+            # make config relative to our root
+            legacy_conf = (self.app_config.root_dir / self.app_config.legacy_conf_dir / info.config).resolve()
+            info.config = legacy_conf.relative_to(self.app_config.root_dir).as_posix()
+        self.record_store.add_model(key, info)
+        return key

invokeai/app/services/model_records/__init__.py

@@ -6,3 +6,11 @@ from .model_records_base import (  # noqa F401
     UnknownModelException,
 )
 from .model_records_sql import ModelRecordServiceSQL  # noqa F401
+
+__all__ = [
+    "ModelRecordServiceBase",
+    "ModelRecordServiceSQL",
+    "DuplicateModelException",
+    "InvalidModelException",
+    "UnknownModelException",
+]

invokeai/app/services/model_records/model_records_base.py

@@ -7,10 +7,7 @@ from abc import ABC, abstractmethod
 from pathlib import Path
 from typing import List, Optional, Union
 
-from invokeai.backend.model_manager.config import AnyModelConfig, BaseModelType, ModelType
-
-# should match the InvokeAI version when this is first released.
-CONFIG_FILE_VERSION = "3.2.0"
+from invokeai.backend.model_manager.config import AnyModelConfig, BaseModelType, ModelFormat, ModelType
 
 
 class DuplicateModelException(Exception):
@@ -32,12 +29,6 @@ class ConfigFileVersionMismatchException(Exception):
 class ModelRecordServiceBase(ABC):
     """Abstract base class for storage and retrieval of model configs."""
 
-    @property
-    @abstractmethod
-    def version(self) -> str:
-        """Return the config file/database schema version."""
-        pass
-
     @abstractmethod
     def add_model(self, key: str, config: Union[dict, AnyModelConfig]) -> AnyModelConfig:
         """
@@ -115,6 +106,7 @@ class ModelRecordServiceBase(ABC):
         model_name: Optional[str] = None,
         base_model: Optional[BaseModelType] = None,
         model_type: Optional[ModelType] = None,
+        model_format: Optional[ModelFormat] = None,
     ) -> List[AnyModelConfig]:
         """
         Return models matching name, base and/or type.
@@ -122,6 +114,7 @@ class ModelRecordServiceBase(ABC):
         :param model_name: Filter by name of model (optional)
         :param base_model: Filter by base model (optional)
         :param model_type: Filter by type of model (optional)
+        :param model_format: Filter by model format (e.g. "diffusers") (optional)
 
         If none of the optional filters are passed, will return all
         models in the database.

invokeai/app/services/model_records/model_records_sql.py

@@ -49,12 +49,12 @@ from invokeai.backend.model_manager.config import (
     AnyModelConfig,
     BaseModelType,
     ModelConfigFactory,
+    ModelFormat,
     ModelType,
 )
 
 from ..shared.sqlite.sqlite_database import SqliteDatabase
 from .model_records_base import (
-    CONFIG_FILE_VERSION,
     DuplicateModelException,
     ModelRecordServiceBase,
     UnknownModelException,
@@ -78,85 +78,6 @@ class ModelRecordServiceSQL(ModelRecordServiceBase):
         self._db = db
         self._cursor = self._db.conn.cursor()
 
-        with self._db.lock:
-            # Enable foreign keys
-            self._db.conn.execute("PRAGMA foreign_keys = ON;")
-            self._create_tables()
-            self._db.conn.commit()
-        assert (
-            str(self.version) == CONFIG_FILE_VERSION
-        ), f"Model config version {self.version} does not match expected version {CONFIG_FILE_VERSION}"
-
-    def _create_tables(self) -> None:
-        """Create sqlite3 tables."""
-        #  model_config table breaks out the fields that are common to all config objects
-        # and puts class-specific ones in a serialized json object
-        self._cursor.execute(
-            """--sql
-            CREATE TABLE IF NOT EXISTS model_config (
-                id TEXT NOT NULL PRIMARY KEY,
-                -- The next 3 fields are enums in python, unrestricted string here
-                base TEXT NOT NULL,
-                type TEXT NOT NULL,
-                name TEXT NOT NULL,
-                path TEXT NOT NULL,
-                original_hash TEXT, -- could be null
-                -- Serialized JSON representation of the whole config object,
-                -- which will contain additional fields from subclasses
-                config TEXT NOT NULL,
-                created_at DATETIME NOT NULL DEFAULT(STRFTIME('%Y-%m-%d %H:%M:%f', 'NOW')),
-                -- Updated via trigger
-                updated_at DATETIME NOT NULL DEFAULT(STRFTIME('%Y-%m-%d %H:%M:%f', 'NOW')),
-                -- unique constraint on combo of name, base and type
-                UNIQUE(name, base, type)
-            );
-            """
-        )
-
-        #  metadata table
-        self._cursor.execute(
-            """--sql
-            CREATE TABLE IF NOT EXISTS model_manager_metadata (
-                metadata_key TEXT NOT NULL PRIMARY KEY,
-                metadata_value TEXT NOT NULL
-            );
-            """
-        )
-
-        # Add trigger for `updated_at`.
-        self._cursor.execute(
-            """--sql
-            CREATE TRIGGER IF NOT EXISTS model_config_updated_at
-            AFTER UPDATE
-            ON model_config FOR EACH ROW
-            BEGIN
-                UPDATE model_config SET updated_at = STRFTIME('%Y-%m-%d %H:%M:%f', 'NOW')
-                    WHERE id = old.id;
-            END;
-            """
-        )
-
-        # Add indexes for searchable fields
-        for stmt in [
-            "CREATE INDEX IF NOT EXISTS base_index ON model_config(base);",
-            "CREATE INDEX IF NOT EXISTS type_index ON model_config(type);",
-            "CREATE INDEX IF NOT EXISTS name_index ON model_config(name);",
-            "CREATE UNIQUE INDEX IF NOT EXISTS path_index ON model_config(path);",
-        ]:
-            self._cursor.execute(stmt)
-
-        # Add our version to the metadata table
-        self._cursor.execute(
-            """--sql
-            INSERT OR IGNORE into model_manager_metadata (
-               metadata_key,
-               metadata_value
-            )
-            VALUES (?,?);
-            """,
-            ("version", CONFIG_FILE_VERSION),
-        )
-
     def add_model(self, key: str, config: Union[dict, AnyModelConfig]) -> AnyModelConfig:
         """
         Add a model to the database.
@@ -175,21 +96,13 @@ class ModelRecordServiceSQL(ModelRecordServiceBase):
                     """--sql
                     INSERT INTO model_config (
                        id,
-                       base,
-                       type,
-                       name,
-                       path,
                        original_hash,
                        config
                       )
-                    VALUES (?,?,?,?,?,?,?);
+                    VALUES (?,?,?);
                     """,
                     (
                         key,
-                        record.base,
-                        record.type,
-                        record.name,
-                        record.path,
                         record.original_hash,
                         json_serialized,
                     ),
@@ -214,22 +127,6 @@ class ModelRecordServiceSQL(ModelRecordServiceBase):
 
         return self.get_model(key)
 
-    @property
-    def version(self) -> str:
-        """Return the version of the database schema."""
-        with self._db.lock:
-            self._cursor.execute(
-                """--sql
-                SELECT metadata_value FROM model_manager_metadata
-                WHERE metadata_key=?;
-                """,
-                ("version",),
-            )
-            rows = self._cursor.fetchone()
-            if not rows:
-                raise KeyError("Models database does not have metadata key 'version'")
-            return rows[0]
-
     def del_model(self, key: str) -> None:
         """
         Delete a model.
@@ -269,14 +166,11 @@ class ModelRecordServiceSQL(ModelRecordServiceBase):
                 self._cursor.execute(
                     """--sql
                     UPDATE model_config
-                    SET base=?,
-                        type=?,
-                        name=?,
-                        path=?,
+                    SET
                         config=?
                     WHERE id=?;
                     """,
-                    (record.base, record.type, record.name, record.path, json_serialized, key),
+                    (json_serialized, key),
                 )
                 if self._cursor.rowcount == 0:
                     raise UnknownModelException("model not found")
@@ -332,6 +226,7 @@ class ModelRecordServiceSQL(ModelRecordServiceBase):
         model_name: Optional[str] = None,
         base_model: Optional[BaseModelType] = None,
         model_type: Optional[ModelType] = None,
+        model_format: Optional[ModelFormat] = None,
     ) -> List[AnyModelConfig]:
         """
         Return models matching name, base and/or type.
@@ -339,6 +234,7 @@ class ModelRecordServiceSQL(ModelRecordServiceBase):
         :param model_name: Filter by name of model (optional)
         :param base_model: Filter by base model (optional)
         :param model_type: Filter by type of model (optional)
+        :param model_format: Filter by model format (e.g. "diffusers") (optional)
 
         If none of the optional filters are passed, will return all
         models in the database.
@@ -355,6 +251,9 @@ class ModelRecordServiceSQL(ModelRecordServiceBase):
         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 ""
         with self._db.lock:
             self._cursor.execute(
@@ -374,7 +273,7 @@ class ModelRecordServiceSQL(ModelRecordServiceBase):
             self._cursor.execute(
                 """--sql
                 SELECT config FROM model_config
-                WHERE model_path=?;
+                WHERE path=?;
                 """,
                 (str(path),),
             )

invokeai/app/services/session_queue/session_queue_sqlite.py

@@ -50,7 +50,6 @@ class SqliteSessionQueue(SessionQueueBase):
         self.__lock = db.lock
         self.__conn = db.conn
         self.__cursor = self.__conn.cursor()
-        self._create_tables()
 
     def _match_event_name(self, event: FastAPIEvent, match_in: list[str]) -> bool:
         return event[1]["event"] in match_in
@@ -98,123 +97,6 @@ class SqliteSessionQueue(SessionQueueBase):
         except SessionQueueItemNotFoundError:
             return
 
-    def _create_tables(self) -> None:
-        """Creates the session queue tables, indicies, and triggers"""
-        try:
-            self.__lock.acquire()
-            self.__cursor.execute(
-                """--sql
-                CREATE TABLE IF NOT EXISTS session_queue (
-                    item_id INTEGER PRIMARY KEY AUTOINCREMENT, -- used for ordering, cursor pagination
-                    batch_id TEXT NOT NULL, -- identifier of the batch this queue item belongs to
-                    queue_id TEXT NOT NULL, -- identifier of the queue this queue item belongs to
-                    session_id TEXT NOT NULL UNIQUE, -- duplicated data from the session column, for ease of access
-                    field_values TEXT, -- NULL if no values are associated with this queue item
-                    session TEXT NOT NULL, -- the session to be executed
-                    status TEXT NOT NULL DEFAULT 'pending', -- the status of the queue item, one of 'pending', 'in_progress', 'completed', 'failed', 'canceled'
-                    priority INTEGER NOT NULL DEFAULT 0, -- the priority, higher is more important
-                    error TEXT, -- any errors associated with this queue item
-                    created_at DATETIME NOT NULL DEFAULT(STRFTIME('%Y-%m-%d %H:%M:%f', 'NOW')),
-                    updated_at DATETIME NOT NULL DEFAULT(STRFTIME('%Y-%m-%d %H:%M:%f', 'NOW')), -- updated via trigger
-                    started_at DATETIME, -- updated via trigger
-                    completed_at DATETIME -- updated via trigger, completed items are cleaned up on application startup
-                    -- Ideally this is a FK, but graph_executions uses INSERT OR REPLACE, and REPLACE triggers the ON DELETE CASCADE...
-                    -- FOREIGN KEY (session_id) REFERENCES graph_executions (id) ON DELETE CASCADE
-                );
-                """
-            )
-
-            self.__cursor.execute(
-                """--sql
-                CREATE UNIQUE INDEX IF NOT EXISTS idx_session_queue_item_id ON session_queue(item_id);
-                """
-            )
-
-            self.__cursor.execute(
-                """--sql
-                CREATE UNIQUE INDEX IF NOT EXISTS idx_session_queue_session_id ON session_queue(session_id);
-                """
-            )
-
-            self.__cursor.execute(
-                """--sql
-                CREATE INDEX IF NOT EXISTS idx_session_queue_batch_id ON session_queue(batch_id);
-                """
-            )
-
-            self.__cursor.execute(
-                """--sql
-                CREATE INDEX IF NOT EXISTS idx_session_queue_created_priority ON session_queue(priority);
-                """
-            )
-
-            self.__cursor.execute(
-                """--sql
-                CREATE INDEX IF NOT EXISTS idx_session_queue_created_status ON session_queue(status);
-                """
-            )
-
-            self.__cursor.execute(
-                """--sql
-                CREATE TRIGGER IF NOT EXISTS tg_session_queue_completed_at
-                AFTER UPDATE OF status ON session_queue
-                FOR EACH ROW
-                WHEN
-                  NEW.status = 'completed'
-                  OR NEW.status = 'failed'
-                  OR NEW.status = 'canceled'
-                BEGIN
-                  UPDATE session_queue
-                  SET completed_at = STRFTIME('%Y-%m-%d %H:%M:%f', 'NOW')
-                  WHERE item_id = NEW.item_id;
-                END;
-                """
-            )
-
-            self.__cursor.execute(
-                """--sql
-                CREATE TRIGGER IF NOT EXISTS tg_session_queue_started_at
-                AFTER UPDATE OF status ON session_queue
-                FOR EACH ROW
-                WHEN
-                  NEW.status = 'in_progress'
-                BEGIN
-                  UPDATE session_queue
-                  SET started_at = STRFTIME('%Y-%m-%d %H:%M:%f', 'NOW')
-                  WHERE item_id = NEW.item_id;
-                END;
-                """
-            )
-
-            self.__cursor.execute(
-                """--sql
-                CREATE TRIGGER IF NOT EXISTS tg_session_queue_updated_at
-                AFTER UPDATE
-                ON session_queue FOR EACH ROW
-                BEGIN
-                    UPDATE session_queue
-                    SET updated_at = STRFTIME('%Y-%m-%d %H:%M:%f', 'NOW')
-                    WHERE item_id = old.item_id;
-                END;
-                """
-            )
-
-            self.__cursor.execute("PRAGMA table_info(session_queue)")
-            columns = [column[1] for column in self.__cursor.fetchall()]
-            if "workflow" not in columns:
-                self.__cursor.execute(
-                    """--sql
-                    ALTER TABLE session_queue ADD COLUMN workflow TEXT;
-                    """
-                )
-
-            self.__conn.commit()
-        except Exception:
-            self.__conn.rollback()
-            raise
-        finally:
-            self.__lock.release()
-
     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.

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

@@ -3,45 +3,65 @@ import threading
 from logging import Logger
 from pathlib import Path
 
-from invokeai.app.services.config import InvokeAIAppConfig
 from invokeai.app.services.shared.sqlite.sqlite_common import sqlite_memory
 
 
 class SqliteDatabase:
-    def __init__(self, config: InvokeAIAppConfig, logger: Logger):
-        self._logger = logger
-        self._config = config
+    """
+    Manages a connection to an SQLite database.
 
-        if self._config.use_memory_db:
-            self.db_path = sqlite_memory
-            logger.info("Using in-memory database")
+    :param db_path: Path to the database file. If None, an in-memory database is used.
+    :param logger: Logger to use for logging.
+    :param verbose: Whether to log SQL statements. Provides `logger.debug` as the SQLite trace callback.
+
+    This is a light wrapper around the `sqlite3` module, providing a few conveniences:
+    - The database file is written to disk if it does not exist.
+    - Foreign key constraints are enabled by default.
+    - The connection is configured to use the `sqlite3.Row` row factory.
+
+    In addition to the constructor args, the instance provides the following attributes and methods:
+    - `conn`: A `sqlite3.Connection` object. Note that the connection must never be closed if the database is in-memory.
+    - `lock`: A shared re-entrant lock, used to approximate thread safety.
+    - `clean()`: Runs the SQL `VACUUM;` command and reports on the freed space.
+    """
+
+    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
+
+        if not self.db_path:
+            logger.info("Initializing in-memory database")
         else:
-            db_path = self._config.db_path
-            db_path.parent.mkdir(parents=True, exist_ok=True)
-            self.db_path = str(db_path)
-            self._logger.info(f"Using 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(self.db_path, check_same_thread=False)
+        self.conn = sqlite3.connect(database=self.db_path or sqlite_memory, check_same_thread=False)
         self.lock = threading.RLock()
         self.conn.row_factory = sqlite3.Row
 
-        if self._config.log_sql:
-            self.conn.set_trace_callback(self._logger.debug)
+        if self.verbose:
+            self.conn.set_trace_callback(self.logger.debug)
 
         self.conn.execute("PRAGMA foreign_keys = ON;")
 
     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:
+            return
         with self.lock:
             try:
-                if self.db_path == sqlite_memory:
-                    return
                 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)")
+                    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

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

@@ -0,0 +1,32 @@
+from logging import Logger
+
+from invokeai.app.services.config.config_default import InvokeAIAppConfig
+from invokeai.app.services.image_files.image_files_base import ImageFileStorageBase
+from invokeai.app.services.shared.sqlite.sqlite_database import SqliteDatabase
+from invokeai.app.services.shared.sqlite_migrator.migrations.migration_1 import build_migration_1
+from invokeai.app.services.shared.sqlite_migrator.migrations.migration_2 import build_migration_2
+from invokeai.app.services.shared.sqlite_migrator.sqlite_migrator_impl import SqliteMigrator
+
+
+def init_db(config: InvokeAIAppConfig, logger: Logger, image_files: ImageFileStorageBase) -> SqliteDatabase:
+    """
+    Initializes the SQLite database.
+
+    :param config: The app config
+    :param logger: The logger
+    :param image_files: The image files service (used by migration 2)
+
+    This function:
+    - Instantiates a :class:`SqliteDatabase`
+    - Instantiates a :class:`SqliteMigrator` and registers all migrations
+    - Runs all migrations
+    """
+    db_path = None if config.use_memory_db else config.db_path
+    db = SqliteDatabase(db_path=db_path, logger=logger, verbose=config.log_sql)
+
+    migrator = SqliteMigrator(db=db)
+    migrator.register_migration(build_migration_1())
+    migrator.register_migration(build_migration_2(image_files=image_files, logger=logger))
+    migrator.run_migrations()
+
+    return db

invokeai/app/services/shared/sqlite_migrator/migrations/migration_1.py

@@ -0,0 +1,372 @@
+import sqlite3
+
+from invokeai.app.services.shared.sqlite_migrator.sqlite_migrator_common import Migration
+
+
+class Migration1Callback:
+    def __call__(self, cursor: sqlite3.Cursor) -> None:
+        """Migration callback for database version 1."""
+
+        self._create_board_images(cursor)
+        self._create_boards(cursor)
+        self._create_images(cursor)
+        self._create_model_config(cursor)
+        self._create_session_queue(cursor)
+        self._create_workflow_images(cursor)
+        self._create_workflows(cursor)
+
+    def _create_board_images(self, cursor: sqlite3.Cursor) -> None:
+        """Creates the `board_images` table, indices and triggers."""
+        tables = [
+            """--sql
+            CREATE TABLE IF NOT EXISTS board_images (
+                board_id TEXT NOT NULL,
+                image_name TEXT NOT NULL,
+                created_at DATETIME NOT NULL DEFAULT(STRFTIME('%Y-%m-%d %H:%M:%f', 'NOW')),
+                -- updated via trigger
+                updated_at DATETIME NOT NULL DEFAULT(STRFTIME('%Y-%m-%d %H:%M:%f', 'NOW')),
+                -- Soft delete, currently unused
+                deleted_at DATETIME,
+                -- enforce one-to-many relationship between boards and images using PK
+                -- (we can extend this to many-to-many later)
+                PRIMARY KEY (image_name),
+                FOREIGN KEY (board_id) REFERENCES boards (board_id) ON DELETE CASCADE,
+                FOREIGN KEY (image_name) REFERENCES images (image_name) ON DELETE CASCADE
+            );
+            """
+        ]
+
+        indices = [
+            "CREATE INDEX IF NOT EXISTS idx_board_images_board_id ON board_images (board_id);",
+            "CREATE INDEX IF NOT EXISTS idx_board_images_board_id_created_at ON board_images (board_id, created_at);",
+        ]
+
+        triggers = [
+            """--sql
+            CREATE TRIGGER IF NOT EXISTS tg_board_images_updated_at
+            AFTER UPDATE
+            ON board_images FOR EACH ROW
+            BEGIN
+                UPDATE board_images SET updated_at = STRFTIME('%Y-%m-%d %H:%M:%f', 'NOW')
+                    WHERE board_id = old.board_id AND image_name = old.image_name;
+            END;
+            """
+        ]
+
+        for stmt in tables + indices + triggers:
+            cursor.execute(stmt)
+
+    def _create_boards(self, cursor: sqlite3.Cursor) -> None:
+        """Creates the `boards` table, indices and triggers."""
+        tables = [
+            """--sql
+            CREATE TABLE IF NOT EXISTS boards (
+                board_id TEXT NOT NULL PRIMARY KEY,
+                board_name TEXT NOT NULL,
+                cover_image_name TEXT,
+                created_at DATETIME NOT NULL DEFAULT(STRFTIME('%Y-%m-%d %H:%M:%f', 'NOW')),
+                -- Updated via trigger
+                updated_at DATETIME NOT NULL DEFAULT(STRFTIME('%Y-%m-%d %H:%M:%f', 'NOW')),
+                -- Soft delete, currently unused
+                deleted_at DATETIME,
+                FOREIGN KEY (cover_image_name) REFERENCES images (image_name) ON DELETE SET NULL
+            );
+            """
+        ]
+
+        indices = ["CREATE INDEX IF NOT EXISTS idx_boards_created_at ON boards (created_at);"]
+
+        triggers = [
+            """--sql
+            CREATE TRIGGER IF NOT EXISTS tg_boards_updated_at
+            AFTER UPDATE
+            ON boards FOR EACH ROW
+            BEGIN
+                UPDATE boards SET updated_at = current_timestamp
+                    WHERE board_id = old.board_id;
+            END;
+            """
+        ]
+
+        for stmt in tables + indices + triggers:
+            cursor.execute(stmt)
+
+    def _create_images(self, cursor: sqlite3.Cursor) -> None:
+        """Creates the `images` table, indices and triggers. Adds the `starred` column."""
+
+        tables = [
+            """--sql
+            CREATE TABLE IF NOT EXISTS images (
+                image_name TEXT NOT NULL PRIMARY KEY,
+                -- This is an enum in python, unrestricted string here for flexibility
+                image_origin TEXT NOT NULL,
+                -- This is an enum in python, unrestricted string here for flexibility
+                image_category TEXT NOT NULL,
+                width INTEGER NOT NULL,
+                height INTEGER NOT NULL,
+                session_id TEXT,
+                node_id TEXT,
+                metadata TEXT,
+                is_intermediate BOOLEAN DEFAULT FALSE,
+                created_at DATETIME NOT NULL DEFAULT(STRFTIME('%Y-%m-%d %H:%M:%f', 'NOW')),
+                -- Updated via trigger
+                updated_at DATETIME NOT NULL DEFAULT(STRFTIME('%Y-%m-%d %H:%M:%f', 'NOW')),
+                -- Soft delete, currently unused
+                deleted_at DATETIME
+            );
+            """
+        ]
+
+        indices = [
+            "CREATE UNIQUE INDEX IF NOT EXISTS idx_images_image_name ON images(image_name);",
+            "CREATE INDEX IF NOT EXISTS idx_images_image_origin ON images(image_origin);",
+            "CREATE INDEX IF NOT EXISTS idx_images_image_category ON images(image_category);",
+            "CREATE INDEX IF NOT EXISTS idx_images_created_at ON images(created_at);",
+        ]
+
+        triggers = [
+            """--sql
+            CREATE TRIGGER IF NOT EXISTS tg_images_updated_at
+            AFTER UPDATE
+            ON images FOR EACH ROW
+            BEGIN
+                UPDATE images SET updated_at = STRFTIME('%Y-%m-%d %H:%M:%f', 'NOW')
+                    WHERE image_name = old.image_name;
+            END;
+            """
+        ]
+
+        # Add the 'starred' column to `images` if it doesn't exist
+        cursor.execute("PRAGMA table_info(images)")
+        columns = [column[1] for column in cursor.fetchall()]
+
+        if "starred" not in columns:
+            tables.append("ALTER TABLE images ADD COLUMN starred BOOLEAN DEFAULT FALSE;")
+            indices.append("CREATE INDEX IF NOT EXISTS idx_images_starred ON images(starred);")
+
+        for stmt in tables + indices + triggers:
+            cursor.execute(stmt)
+
+    def _create_model_config(self, cursor: sqlite3.Cursor) -> None:
+        """Creates the `model_config` table, `model_manager_metadata` table, indices and triggers."""
+
+        tables = [
+            """--sql
+            CREATE TABLE IF NOT EXISTS model_config (
+                id TEXT NOT NULL PRIMARY KEY,
+                -- The next 3 fields are enums in python, unrestricted string here
+                base TEXT NOT NULL,
+                type TEXT NOT NULL,
+                name TEXT NOT NULL,
+                path TEXT NOT NULL,
+                original_hash TEXT, -- could be null
+                -- Serialized JSON representation of the whole config object,
+                -- which will contain additional fields from subclasses
+                config TEXT NOT NULL,
+                created_at DATETIME NOT NULL DEFAULT(STRFTIME('%Y-%m-%d %H:%M:%f', 'NOW')),
+                -- Updated via trigger
+                updated_at DATETIME NOT NULL DEFAULT(STRFTIME('%Y-%m-%d %H:%M:%f', 'NOW')),
+                -- unique constraint on combo of name, base and type
+                UNIQUE(name, base, type)
+            );
+            """,
+            """--sql
+            CREATE TABLE IF NOT EXISTS model_manager_metadata (
+                metadata_key TEXT NOT NULL PRIMARY KEY,
+                metadata_value TEXT NOT NULL
+            );
+            """,
+        ]
+
+        # Add trigger for `updated_at`.
+        triggers = [
+            """--sql
+            CREATE TRIGGER IF NOT EXISTS model_config_updated_at
+            AFTER UPDATE
+            ON model_config FOR EACH ROW
+            BEGIN
+                UPDATE model_config SET updated_at = STRFTIME('%Y-%m-%d %H:%M:%f', 'NOW')
+                    WHERE id = old.id;
+            END;
+            """
+        ]
+
+        # Add indexes for searchable fields
+        indices = [
+            "CREATE INDEX IF NOT EXISTS base_index ON model_config(base);",
+            "CREATE INDEX IF NOT EXISTS type_index ON model_config(type);",
+            "CREATE INDEX IF NOT EXISTS name_index ON model_config(name);",
+            "CREATE UNIQUE INDEX IF NOT EXISTS path_index ON model_config(path);",
+        ]
+
+        for stmt in tables + indices + triggers:
+            cursor.execute(stmt)
+
+    def _create_session_queue(self, cursor: sqlite3.Cursor) -> None:
+        tables = [
+            """--sql
+            CREATE TABLE IF NOT EXISTS session_queue (
+                item_id INTEGER PRIMARY KEY AUTOINCREMENT, -- used for ordering, cursor pagination
+                batch_id TEXT NOT NULL, -- identifier of the batch this queue item belongs to
+                queue_id TEXT NOT NULL, -- identifier of the queue this queue item belongs to
+                session_id TEXT NOT NULL UNIQUE, -- duplicated data from the session column, for ease of access
+                field_values TEXT, -- NULL if no values are associated with this queue item
+                session TEXT NOT NULL, -- the session to be executed
+                status TEXT NOT NULL DEFAULT 'pending', -- the status of the queue item, one of 'pending', 'in_progress', 'completed', 'failed', 'canceled'
+                priority INTEGER NOT NULL DEFAULT 0, -- the priority, higher is more important
+                error TEXT, -- any errors associated with this queue item
+                created_at DATETIME NOT NULL DEFAULT(STRFTIME('%Y-%m-%d %H:%M:%f', 'NOW')),
+                updated_at DATETIME NOT NULL DEFAULT(STRFTIME('%Y-%m-%d %H:%M:%f', 'NOW')), -- updated via trigger
+                started_at DATETIME, -- updated via trigger
+                completed_at DATETIME -- updated via trigger, completed items are cleaned up on application startup
+                -- Ideally this is a FK, but graph_executions uses INSERT OR REPLACE, and REPLACE triggers the ON DELETE CASCADE...
+                -- FOREIGN KEY (session_id) REFERENCES graph_executions (id) ON DELETE CASCADE
+            );
+            """
+        ]
+
+        indices = [
+            "CREATE UNIQUE INDEX IF NOT EXISTS idx_session_queue_item_id ON session_queue(item_id);",
+            "CREATE UNIQUE INDEX IF NOT EXISTS idx_session_queue_session_id ON session_queue(session_id);",
+            "CREATE INDEX IF NOT EXISTS idx_session_queue_batch_id ON session_queue(batch_id);",
+            "CREATE INDEX IF NOT EXISTS idx_session_queue_created_priority ON session_queue(priority);",
+            "CREATE INDEX IF NOT EXISTS idx_session_queue_created_status ON session_queue(status);",
+        ]
+
+        triggers = [
+            """--sql
+            CREATE TRIGGER IF NOT EXISTS tg_session_queue_completed_at
+            AFTER UPDATE OF status ON session_queue
+            FOR EACH ROW
+            WHEN
+            NEW.status = 'completed'
+            OR NEW.status = 'failed'
+            OR NEW.status = 'canceled'
+            BEGIN
+            UPDATE session_queue
+            SET completed_at = STRFTIME('%Y-%m-%d %H:%M:%f', 'NOW')
+            WHERE item_id = NEW.item_id;
+            END;
+            """,
+            """--sql
+            CREATE TRIGGER IF NOT EXISTS tg_session_queue_started_at
+            AFTER UPDATE OF status ON session_queue
+            FOR EACH ROW
+            WHEN
+            NEW.status = 'in_progress'
+            BEGIN
+            UPDATE session_queue
+            SET started_at = STRFTIME('%Y-%m-%d %H:%M:%f', 'NOW')
+            WHERE item_id = NEW.item_id;
+            END;
+            """,
+            """--sql
+            CREATE TRIGGER IF NOT EXISTS tg_session_queue_updated_at
+            AFTER UPDATE
+            ON session_queue FOR EACH ROW
+            BEGIN
+                UPDATE session_queue
+                SET updated_at = STRFTIME('%Y-%m-%d %H:%M:%f', 'NOW')
+                WHERE item_id = old.item_id;
+            END;
+            """,
+        ]
+
+        for stmt in tables + indices + triggers:
+            cursor.execute(stmt)
+
+    def _create_workflow_images(self, cursor: sqlite3.Cursor) -> None:
+        tables = [
+            """--sql
+            CREATE TABLE IF NOT EXISTS workflow_images (
+                workflow_id TEXT NOT NULL,
+                image_name TEXT NOT NULL,
+                created_at DATETIME NOT NULL DEFAULT(STRFTIME('%Y-%m-%d %H:%M:%f', 'NOW')),
+                -- updated via trigger
+                updated_at DATETIME NOT NULL DEFAULT(STRFTIME('%Y-%m-%d %H:%M:%f', 'NOW')),
+                -- Soft delete, currently unused
+                deleted_at DATETIME,
+                -- enforce one-to-many relationship between workflows and images using PK
+                -- (we can extend this to many-to-many later)
+                PRIMARY KEY (image_name),
+                FOREIGN KEY (workflow_id) REFERENCES workflows (workflow_id) ON DELETE CASCADE,
+                FOREIGN KEY (image_name) REFERENCES images (image_name) ON DELETE CASCADE
+            );
+            """
+        ]
+
+        indices = [
+            "CREATE INDEX IF NOT EXISTS idx_workflow_images_workflow_id ON workflow_images (workflow_id);",
+            "CREATE INDEX IF NOT EXISTS idx_workflow_images_workflow_id_created_at ON workflow_images (workflow_id, created_at);",
+        ]
+
+        triggers = [
+            """--sql
+            CREATE TRIGGER IF NOT EXISTS tg_workflow_images_updated_at
+            AFTER UPDATE
+            ON workflow_images FOR EACH ROW
+            BEGIN
+                UPDATE workflow_images SET updated_at = STRFTIME('%Y-%m-%d %H:%M:%f', 'NOW')
+                    WHERE workflow_id = old.workflow_id AND image_name = old.image_name;
+            END;
+            """
+        ]
+
+        for stmt in tables + indices + triggers:
+            cursor.execute(stmt)
+
+    def _create_workflows(self, cursor: sqlite3.Cursor) -> None:
+        tables = [
+            """--sql
+            CREATE TABLE IF NOT EXISTS workflows (
+                workflow TEXT NOT NULL,
+                workflow_id TEXT GENERATED ALWAYS AS (json_extract(workflow, '$.id')) VIRTUAL NOT NULL UNIQUE, -- gets implicit index
+                created_at DATETIME NOT NULL DEFAULT(STRFTIME('%Y-%m-%d %H:%M:%f', 'NOW')),
+                updated_at DATETIME NOT NULL DEFAULT(STRFTIME('%Y-%m-%d %H:%M:%f', 'NOW')) -- updated via trigger
+            );
+            """
+        ]
+
+        triggers = [
+            """--sql
+            CREATE TRIGGER IF NOT EXISTS tg_workflows_updated_at
+            AFTER UPDATE
+            ON workflows FOR EACH ROW
+            BEGIN
+                UPDATE workflows
+                SET updated_at = STRFTIME('%Y-%m-%d %H:%M:%f', 'NOW')
+                WHERE workflow_id = old.workflow_id;
+            END;
+            """
+        ]
+
+        for stmt in tables + triggers:
+            cursor.execute(stmt)
+
+
+def build_migration_1() -> Migration:
+    """
+    Builds the migration from database version 0 (init) to 1.
+
+    This migration represents the state of the database circa InvokeAI v3.4.0, which was the last
+    version to not use migrations to manage the database.
+
+    As such, this migration does include some ALTER statements, and the SQL statements are written
+    to be idempotent.
+
+    - Create `board_images` junction table
+    - Create `boards` table
+    - Create `images` table, add `starred` column
+    - Create `model_config` table
+    - Create `session_queue` table
+    - Create `workflow_images` junction table
+    - Create `workflows` table
+    """
+
+    migration_1 = Migration(
+        from_version=0,
+        to_version=1,
+        callback=Migration1Callback(),
+    )
+
+    return migration_1

invokeai/app/services/shared/sqlite_migrator/migrations/migration_2.py

@@ -0,0 +1,198 @@
+import sqlite3
+from logging import Logger
+
+from pydantic import ValidationError
+from tqdm import tqdm
+
+from invokeai.app.services.image_files.image_files_base import ImageFileStorageBase
+from invokeai.app.services.image_files.image_files_common import ImageFileNotFoundException
+from invokeai.app.services.shared.sqlite_migrator.sqlite_migrator_common import Migration
+from invokeai.app.services.workflow_records.workflow_records_common import (
+    UnsafeWorkflowWithVersionValidator,
+)
+
+
+class Migration2Callback:
+    def __init__(self, image_files: ImageFileStorageBase, logger: Logger):
+        self._image_files = image_files
+        self._logger = logger
+
+    def __call__(self, cursor: sqlite3.Cursor):
+        self._add_images_has_workflow(cursor)
+        self._add_session_queue_workflow(cursor)
+        self._drop_old_workflow_tables(cursor)
+        self._add_workflow_library(cursor)
+        self._drop_model_manager_metadata(cursor)
+        self._recreate_model_config(cursor)
+        self._migrate_embedded_workflows(cursor)
+
+    def _add_images_has_workflow(self, cursor: sqlite3.Cursor) -> None:
+        """Add the `has_workflow` column to `images` table."""
+        cursor.execute("PRAGMA table_info(images)")
+        columns = [column[1] for column in cursor.fetchall()]
+
+        if "has_workflow" not in columns:
+            cursor.execute("ALTER TABLE images ADD COLUMN has_workflow BOOLEAN DEFAULT FALSE;")
+
+    def _add_session_queue_workflow(self, cursor: sqlite3.Cursor) -> None:
+        """Add the `workflow` column to `session_queue` table."""
+
+        cursor.execute("PRAGMA table_info(session_queue)")
+        columns = [column[1] for column in cursor.fetchall()]
+
+        if "workflow" not in columns:
+            cursor.execute("ALTER TABLE session_queue ADD COLUMN workflow TEXT;")
+
+    def _drop_old_workflow_tables(self, cursor: sqlite3.Cursor) -> None:
+        """Drops the `workflows` and `workflow_images` tables."""
+        cursor.execute("DROP TABLE IF EXISTS workflow_images;")
+        cursor.execute("DROP TABLE IF EXISTS workflows;")
+
+    def _add_workflow_library(self, cursor: sqlite3.Cursor) -> None:
+        """Adds the `workflow_library` table and drops the `workflows` and `workflow_images` tables."""
+        tables = [
+            """--sql
+            CREATE TABLE IF NOT EXISTS workflow_library (
+                workflow_id TEXT NOT NULL PRIMARY KEY,
+                workflow TEXT NOT NULL,
+                created_at DATETIME NOT NULL DEFAULT(STRFTIME('%Y-%m-%d %H:%M:%f', 'NOW')),
+                -- updated via trigger
+                updated_at DATETIME NOT NULL DEFAULT(STRFTIME('%Y-%m-%d %H:%M:%f', 'NOW')),
+                -- updated manually when retrieving workflow
+                opened_at DATETIME NOT NULL DEFAULT(STRFTIME('%Y-%m-%d %H:%M:%f', 'NOW')),
+                -- Generated columns, needed for indexing and searching
+                category TEXT GENERATED ALWAYS as (json_extract(workflow, '$.meta.category')) VIRTUAL NOT NULL,
+                name TEXT GENERATED ALWAYS as (json_extract(workflow, '$.name')) VIRTUAL NOT NULL,
+                description TEXT GENERATED ALWAYS as (json_extract(workflow, '$.description')) VIRTUAL NOT NULL
+            );
+            """,
+        ]
+
+        indices = [
+            "CREATE INDEX IF NOT EXISTS idx_workflow_library_created_at ON workflow_library(created_at);",
+            "CREATE INDEX IF NOT EXISTS idx_workflow_library_updated_at ON workflow_library(updated_at);",
+            "CREATE INDEX IF NOT EXISTS idx_workflow_library_opened_at ON workflow_library(opened_at);",
+            "CREATE INDEX IF NOT EXISTS idx_workflow_library_category ON workflow_library(category);",
+            "CREATE INDEX IF NOT EXISTS idx_workflow_library_name ON workflow_library(name);",
+            "CREATE INDEX IF NOT EXISTS idx_workflow_library_description ON workflow_library(description);",
+        ]
+
+        triggers = [
+            """--sql
+            CREATE TRIGGER IF NOT EXISTS tg_workflow_library_updated_at
+            AFTER UPDATE
+            ON workflow_library FOR EACH ROW
+            BEGIN
+                UPDATE workflow_library
+                SET updated_at = STRFTIME('%Y-%m-%d %H:%M:%f', 'NOW')
+                WHERE workflow_id = old.workflow_id;
+            END;
+            """
+        ]
+
+        for stmt in tables + indices + triggers:
+            cursor.execute(stmt)
+
+    def _drop_model_manager_metadata(self, cursor: sqlite3.Cursor) -> None:
+        """Drops the `model_manager_metadata` table."""
+        cursor.execute("DROP TABLE IF EXISTS model_manager_metadata;")
+
+    def _recreate_model_config(self, cursor: sqlite3.Cursor) -> None:
+        """
+        Drops the `model_config` table, recreating it.
+
+        In 3.4.0, this table used explicit columns but was changed to use json_extract 3.5.0.
+
+        Because this table is not used in production, we are able to simply drop it and recreate it.
+        """
+
+        cursor.execute("DROP TABLE IF EXISTS model_config;")
+
+        cursor.execute(
+            """--sql
+            CREATE TABLE IF NOT EXISTS model_config (
+                id TEXT NOT NULL PRIMARY KEY,
+                -- The next 3 fields are enums in python, unrestricted string here
+                base TEXT GENERATED ALWAYS as (json_extract(config, '$.base')) VIRTUAL NOT NULL,
+                type TEXT GENERATED ALWAYS as (json_extract(config, '$.type')) VIRTUAL NOT NULL,
+                name TEXT GENERATED ALWAYS as (json_extract(config, '$.name')) VIRTUAL NOT NULL,
+                path TEXT GENERATED ALWAYS as (json_extract(config, '$.path')) VIRTUAL NOT NULL,
+                format TEXT GENERATED ALWAYS as (json_extract(config, '$.format')) VIRTUAL NOT NULL,
+                original_hash TEXT, -- could be null
+                -- Serialized JSON representation of the whole config object,
+                -- which will contain additional fields from subclasses
+                config TEXT NOT NULL,
+                created_at DATETIME NOT NULL DEFAULT(STRFTIME('%Y-%m-%d %H:%M:%f', 'NOW')),
+                -- Updated via trigger
+                updated_at DATETIME NOT NULL DEFAULT(STRFTIME('%Y-%m-%d %H:%M:%f', 'NOW')),
+                -- unique constraint on combo of name, base and type
+                UNIQUE(name, base, type)
+            );
+            """
+        )
+
+    def _migrate_embedded_workflows(self, cursor: sqlite3.Cursor) -> None:
+        """
+        In the v3.5.0 release, InvokeAI changed how it handles embedded workflows. The `images` table in
+        the database now has a `has_workflow` column, indicating if an image has a workflow embedded.
+
+        This migrate callback checks each image for the presence of an embedded workflow, then updates its entry
+        in the database accordingly.
+        """
+        # Get all image names
+        cursor.execute("SELECT image_name FROM images")
+        image_names: list[str] = [image[0] for image in cursor.fetchall()]
+        total_image_names = len(image_names)
+
+        if not total_image_names:
+            return
+
+        self._logger.info(f"Migrating workflows for {total_image_names} images")
+
+        # Migrate the images
+        to_migrate: list[tuple[bool, str]] = []
+        pbar = tqdm(image_names)
+        for idx, image_name in enumerate(pbar):
+            pbar.set_description(f"Checking image {idx + 1}/{total_image_names} for workflow")
+            try:
+                pil_image = self._image_files.get(image_name)
+            except ImageFileNotFoundException:
+                self._logger.warning(f"Image {image_name} not found, skipping")
+                continue
+            if "invokeai_workflow" in pil_image.info:
+                try:
+                    UnsafeWorkflowWithVersionValidator.validate_json(pil_image.info.get("invokeai_workflow", ""))
+                except ValidationError:
+                    self._logger.warning(f"Image {image_name} has invalid embedded workflow, skipping")
+                    continue
+                to_migrate.append((True, image_name))
+
+        self._logger.info(f"Adding {len(to_migrate)} embedded workflows to database")
+        cursor.executemany("UPDATE images SET has_workflow = ? WHERE image_name = ?", to_migrate)
+
+
+def build_migration_2(image_files: ImageFileStorageBase, logger: Logger) -> Migration:
+    """
+    Builds the migration from database version 1 to 2.
+
+    Introduced in v3.5.0 for the new workflow library.
+
+    :param image_files: The image files service, used to check for embedded workflows
+    :param logger: The logger, used to log progress during embedded workflows handling
+
+    This migration does the following:
+    - Add `has_workflow` column to `images` table
+    - Add `workflow` column to `session_queue` table
+    - Drop `workflows` and `workflow_images` tables
+    - Add `workflow_library` table
+    - Drops the `model_manager_metadata` table
+    - Drops the `model_config` table, recreating it (at this point, there is no user data in this table)
+    - Populates the `has_workflow` column in the `images` table (requires `image_files` & `logger` dependencies)
+    """
+    migration_2 = Migration(
+        from_version=1,
+        to_version=2,
+        callback=Migration2Callback(image_files=image_files, logger=logger),
+    )
+
+    return migration_2

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

@@ -0,0 +1,164 @@
+import sqlite3
+from typing import Optional, Protocol, runtime_checkable
+
+from pydantic import BaseModel, ConfigDict, Field, model_validator
+
+
+@runtime_checkable
+class MigrateCallback(Protocol):
+    """
+    A callback that performs a migration.
+
+    Migrate callbacks are provided an open cursor to the database. They should not commit their
+    transaction; this is handled by the migrator.
+
+    If the callback needs to access additional dependencies, will be provided to the callback at runtime.
+
+    See :class:`Migration` for an example.
+    """
+
+    def __call__(self, cursor: sqlite3.Cursor) -> None:
+        ...
+
+
+class MigrationError(RuntimeError):
+    """Raised when a migration fails."""
+
+
+class MigrationVersionError(ValueError):
+    """Raised when a migration version is invalid."""
+
+
+class Migration(BaseModel):
+    """
+    Represents a migration for a SQLite database.
+
+    :param from_version: The database version on which this migration may be run
+    :param to_version: The database version that results from this migration
+    :param migrate_callback: The callback to run to perform the migration
+
+    Migration callbacks will be provided an open cursor to the database. They should not commit their
+    transaction; this is handled by the migrator.
+
+    It is suggested to use a class to define the migration callback and a builder function to create
+    the :class:`Migration`. This allows the callback to be provided with additional dependencies and
+    keeps things tidy, as all migration logic is self-contained.
+
+    Example:
+    ```py
+    # Define the migration callback class
+    class Migration1Callback:
+        # This migration needs a logger, so we define a class that accepts a logger in its constructor.
+        def __init__(self, image_files: ImageFileStorageBase) -> None:
+            self._image_files = ImageFileStorageBase
+
+        # This dunder method allows the instance of the class to be called like a function.
+        def __call__(self, cursor: sqlite3.Cursor) -> None:
+            self._add_with_banana_column(cursor)
+            self._do_something_with_images(cursor)
+
+        def _add_with_banana_column(self, cursor: sqlite3.Cursor) -> None:
+            \"""Adds the with_banana column to the sushi table.\"""
+            # Execute SQL using the cursor, taking care to *not commit* a transaction
+            cursor.execute('ALTER TABLE sushi ADD COLUMN with_banana BOOLEAN DEFAULT TRUE;')
+
+        def _do_something_with_images(self, cursor: sqlite3.Cursor) -> None:
+            \"""Does something with the image files service.\"""
+            self._image_files.get(...)
+
+    # Define the migration builder function. This function creates an instance of the migration callback
+    # class and returns a Migration.
+    def build_migration_1(image_files: ImageFileStorageBase) -> Migration:
+        \"""Builds the migration from database version 0 to 1.
+        Requires the image files service to...
+        \"""
+
+        migration_1 = Migration(
+            from_version=0,
+            to_version=1,
+            migrate_callback=Migration1Callback(image_files=image_files),
+        )
+
+        return migration_1
+
+    # Register the migration after all dependencies have been initialized
+    db = SqliteDatabase(db_path, logger)
+    migrator = SqliteMigrator(db)
+    migrator.register_migration(build_migration_1(image_files))
+    migrator.run_migrations()
+    ```
+    """
+
+    from_version: int = Field(ge=0, strict=True, description="The database version on which this migration may be run")
+    to_version: int = Field(ge=1, strict=True, description="The database version that results from this migration")
+    callback: MigrateCallback = Field(description="The callback to run to perform the migration")
+
+    @model_validator(mode="after")
+    def validate_to_version(self) -> "Migration":
+        """Validates that to_version is one greater than from_version."""
+        if self.to_version != self.from_version + 1:
+            raise MigrationVersionError("to_version must be one greater than from_version")
+        return self
+
+    def __hash__(self) -> int:
+        # Callables are not hashable, so we need to implement our own __hash__ function to use this class in a set.
+        return hash((self.from_version, self.to_version))
+
+    model_config = ConfigDict(arbitrary_types_allowed=True)
+
+
+class MigrationSet:
+    """
+    A set of Migrations. Performs validation during migration registration and provides utility methods.
+
+    Migrations should be registered with `register()`. Once all are registered, `validate_migration_chain()`
+    should be called to ensure that the migrations form a single chain of migrations from version 0 to the latest version.
+    """
+
+    def __init__(self) -> None:
+        self._migrations: set[Migration] = set()
+
+    def register(self, migration: Migration) -> None:
+        """Registers a migration."""
+        migration_from_already_registered = any(m.from_version == migration.from_version for m in self._migrations)
+        migration_to_already_registered = any(m.to_version == migration.to_version for m in self._migrations)
+        if migration_from_already_registered or migration_to_already_registered:
+            raise MigrationVersionError("Migration with from_version or to_version already registered")
+        self._migrations.add(migration)
+
+    def get(self, from_version: int) -> Optional[Migration]:
+        """Gets the migration that may be run on the given database version."""
+        # register() ensures that there is only one migration with a given from_version, so this is safe.
+        return next((m for m in self._migrations if m.from_version == from_version), None)
+
+    def validate_migration_chain(self) -> None:
+        """
+        Validates that the migrations form a single chain of migrations from version 0 to the latest version,
+        Raises a MigrationError if there is a problem.
+        """
+        if self.count == 0:
+            return
+        if self.latest_version == 0:
+            return
+        next_migration = self.get(from_version=0)
+        if next_migration is None:
+            raise MigrationError("Migration chain is fragmented")
+        touched_count = 1
+        while next_migration is not None:
+            next_migration = self.get(next_migration.to_version)
+            if next_migration is not None:
+                touched_count += 1
+        if touched_count != self.count:
+            raise MigrationError("Migration chain is fragmented")
+
+    @property
+    def count(self) -> int:
+        """The count of registered migrations."""
+        return len(self._migrations)
+
+    @property
+    def latest_version(self) -> int:
+        """Gets latest to_version among registered migrations. Returns 0 if there are no migrations registered."""
+        if self.count == 0:
+            return 0
+        return sorted(self._migrations, key=lambda m: m.to_version)[-1].to_version

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

@@ -0,0 +1,130 @@
+import sqlite3
+from pathlib import Path
+from typing import Optional
+
+from invokeai.app.services.shared.sqlite.sqlite_database import SqliteDatabase
+from invokeai.app.services.shared.sqlite_migrator.sqlite_migrator_common import Migration, MigrationError, MigrationSet
+
+
+class SqliteMigrator:
+    """
+    Manages migrations for a SQLite database.
+
+    :param db: The instance of :class:`SqliteDatabase` to migrate.
+
+    Migrations should be registered with :meth:`register_migration`.
+
+    Each migration is run in a transaction. If a migration fails, the transaction is rolled back.
+
+    Example Usage:
+    ```py
+    db = SqliteDatabase(db_path="my_db.db", logger=logger)
+    migrator = SqliteMigrator(db=db)
+    migrator.register_migration(build_migration_1())
+    migrator.register_migration(build_migration_2())
+    migrator.run_migrations()
+    ```
+    """
+
+    backup_path: Optional[Path] = None
+
+    def __init__(self, db: SqliteDatabase) -> None:
+        self._db = db
+        self._logger = db.logger
+        self._migration_set = MigrationSet()
+
+    def register_migration(self, migration: Migration) -> None:
+        """Registers a migration."""
+        self._migration_set.register(migration)
+        self._logger.debug(f"Registered migration {migration.from_version} -> {migration.to_version}")
+
+    def run_migrations(self) -> bool:
+        """Migrates the database to the latest version."""
+        with self._db.lock:
+            # This throws if there is a problem.
+            self._migration_set.validate_migration_chain()
+            cursor = self._db.conn.cursor()
+            self._create_migrations_table(cursor=cursor)
+
+            if self._migration_set.count == 0:
+                self._logger.debug("No migrations registered")
+                return False
+
+            if self._get_current_version(cursor=cursor) == self._migration_set.latest_version:
+                self._logger.debug("Database is up to date, no migrations to run")
+                return False
+
+            self._logger.info("Database update needed")
+            next_migration = self._migration_set.get(from_version=self._get_current_version(cursor))
+            while next_migration is not None:
+                self._run_migration(next_migration)
+                next_migration = self._migration_set.get(self._get_current_version(cursor))
+            self._logger.info("Database updated successfully")
+            return True
+
+    def _run_migration(self, migration: Migration) -> None:
+        """Runs a single migration."""
+        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.lock, self._db.conn as conn:
+                cursor = conn.cursor()
+                if self._get_current_version(cursor) != migration.from_version:
+                    raise MigrationError(
+                        f"Database is at version {self._get_current_version(cursor)}, expected {migration.from_version}"
+                    )
+                self._logger.debug(f"Running migration from {migration.from_version} to {migration.to_version}")
+
+                # Run the actual migration
+                migration.callback(cursor)
+
+                # Update the version
+                cursor.execute("INSERT INTO migrations (version) VALUES (?);", (migration.to_version,))
+
+                self._logger.debug(
+                    f"Successfully migrated database from {migration.from_version} to {migration.to_version}"
+                )
+        # We want to catch *any* error, mirroring the behaviour of the sqlite3 module.
+        except Exception as e:
+            # The connection context manager has already rolled back the migration, so we don't need to do anything.
+            msg = f"Error migrating database from {migration.from_version} to {migration.to_version}: {e}"
+            self._logger.error(msg)
+            raise MigrationError(msg) from e
+
+    def _create_migrations_table(self, cursor: sqlite3.Cursor) -> None:
+        """Creates the migrations table for the database, if one does not already exist."""
+        with self._db.lock:
+            try:
+                cursor.execute("SELECT name FROM sqlite_master WHERE type='table' AND name='migrations';")
+                if cursor.fetchone() is not None:
+                    return
+                cursor.execute(
+                    """--sql
+                    CREATE TABLE migrations (
+                        version INTEGER PRIMARY KEY,
+                        migrated_at DATETIME NOT NULL DEFAULT(STRFTIME('%Y-%m-%d %H:%M:%f', 'NOW'))
+                    );
+                    """
+                )
+                cursor.execute("INSERT INTO migrations (version) VALUES (0);")
+                cursor.connection.commit()
+                self._logger.debug("Created migrations table")
+            except sqlite3.Error as e:
+                msg = f"Problem creating migrations table: {e}"
+                self._logger.error(msg)
+                cursor.connection.rollback()
+                raise MigrationError(msg) from e
+
+    @classmethod
+    def _get_current_version(cls, cursor: sqlite3.Cursor) -> int:
+        """Gets the current version of the database, or 0 if the migrations table does not exist."""
+        try:
+            cursor.execute("SELECT MAX(version) FROM migrations;")
+            version: int = cursor.fetchone()[0]
+            if version is None:
+                return 0
+            return version
+        except sqlite3.OperationalError as e:
+            if "no such table" in str(e):
+                return 0
+            raise

invokeai/app/services/workflow_records/workflow_records_common.py

@@ -65,12 +65,24 @@ class WorkflowWithoutID(BaseModel):
     nodes: list[dict[str, JsonValue]] = Field(description="The nodes of the workflow.")
     edges: list[dict[str, JsonValue]] = Field(description="The edges of the workflow.")
 
-    model_config = ConfigDict(extra="forbid")
+    model_config = ConfigDict(extra="ignore")
 
 
 WorkflowWithoutIDValidator = TypeAdapter(WorkflowWithoutID)
 
 
+class UnsafeWorkflowWithVersion(BaseModel):
+    """
+    This utility model only requires a workflow to have a valid version string.
+    It is used to validate a workflow version without having to validate the entire workflow.
+    """
+
+    meta: WorkflowMeta = Field(description="The meta of the workflow.")
+
+
+UnsafeWorkflowWithVersionValidator = TypeAdapter(UnsafeWorkflowWithVersion)
+
+
 class Workflow(WorkflowWithoutID):
     id: str = Field(description="The id of the workflow.")
 

invokeai/app/services/workflow_records/workflow_records_sqlite.py

@@ -26,7 +26,6 @@ class SqliteWorkflowRecordsStorage(WorkflowRecordsStorageBase):
         self._lock = db.lock
         self._conn = db.conn
         self._cursor = self._conn.cursor()
-        self._create_tables()
 
     def start(self, invoker: Invoker) -> None:
         self._invoker = invoker
@@ -233,87 +232,3 @@ class SqliteWorkflowRecordsStorage(WorkflowRecordsStorageBase):
             raise
         finally:
             self._lock.release()
-
-    def _create_tables(self) -> None:
-        try:
-            self._lock.acquire()
-            self._cursor.execute(
-                """--sql
-                CREATE TABLE IF NOT EXISTS workflow_library (
-                    workflow_id TEXT NOT NULL PRIMARY KEY,
-                    workflow TEXT NOT NULL,
-                    created_at DATETIME NOT NULL DEFAULT(STRFTIME('%Y-%m-%d %H:%M:%f', 'NOW')),
-                    -- updated via trigger
-                    updated_at DATETIME NOT NULL DEFAULT(STRFTIME('%Y-%m-%d %H:%M:%f', 'NOW')),
-                    -- updated manually when retrieving workflow
-                    opened_at DATETIME NOT NULL DEFAULT(STRFTIME('%Y-%m-%d %H:%M:%f', 'NOW')),
-                    -- Generated columns, needed for indexing and searching
-                    category TEXT GENERATED ALWAYS as (json_extract(workflow, '$.meta.category')) VIRTUAL NOT NULL,
-                    name TEXT GENERATED ALWAYS as (json_extract(workflow, '$.name')) VIRTUAL NOT NULL,
-                    description TEXT GENERATED ALWAYS as (json_extract(workflow, '$.description')) VIRTUAL NOT NULL
-                );
-                """
-            )
-
-            self._cursor.execute(
-                """--sql
-                CREATE TRIGGER IF NOT EXISTS tg_workflow_library_updated_at
-                AFTER UPDATE
-                ON workflow_library FOR EACH ROW
-                BEGIN
-                    UPDATE workflow_library
-                    SET updated_at = STRFTIME('%Y-%m-%d %H:%M:%f', 'NOW')
-                    WHERE workflow_id = old.workflow_id;
-                END;
-                """
-            )
-
-            self._cursor.execute(
-                """--sql
-                CREATE INDEX IF NOT EXISTS idx_workflow_library_created_at ON workflow_library(created_at);
-                """
-            )
-            self._cursor.execute(
-                """--sql
-                CREATE INDEX IF NOT EXISTS idx_workflow_library_updated_at ON workflow_library(updated_at);
-                """
-            )
-            self._cursor.execute(
-                """--sql
-                CREATE INDEX IF NOT EXISTS idx_workflow_library_opened_at ON workflow_library(opened_at);
-                """
-            )
-            self._cursor.execute(
-                """--sql
-                CREATE INDEX IF NOT EXISTS idx_workflow_library_category ON workflow_library(category);
-                """
-            )
-            self._cursor.execute(
-                """--sql
-                CREATE INDEX IF NOT EXISTS idx_workflow_library_name ON workflow_library(name);
-                """
-            )
-            self._cursor.execute(
-                """--sql
-                CREATE INDEX IF NOT EXISTS idx_workflow_library_description ON workflow_library(description);
-                """
-            )
-
-            # We do not need the original `workflows` table or `workflow_images` junction table.
-            self._cursor.execute(
-                """--sql
-                DROP TABLE IF EXISTS workflow_images;
-                """
-            )
-            self._cursor.execute(
-                """--sql
-                DROP TABLE IF EXISTS workflows;
-                """
-            )
-
-            self._conn.commit()
-        except Exception:
-            self._conn.rollback()
-            raise
-        finally:
-            self._lock.release()

invokeai/backend/model_management/model_probe.py

@@ -32,6 +32,8 @@ class ModelProbeInfo(object):
     upcast_attention: bool
     format: Literal["diffusers", "checkpoint", "lycoris", "olive", "onnx"]
     image_size: int
+    name: Optional[str] = None
+    description: Optional[str] = None
 
 
 class ProbeBase(object):
@@ -113,12 +115,16 @@ class ModelProbe(object):
             base_type = probe.get_base_type()
             variant_type = probe.get_variant_type()
             prediction_type = probe.get_scheduler_prediction_type()
+            name = cls.get_model_name(model_path)
+            description = f"{base_type.value} {model_type.value} model {name}"
             format = probe.get_format()
             model_info = ModelProbeInfo(
                 model_type=model_type,
                 base_type=base_type,
                 variant_type=variant_type,
                 prediction_type=prediction_type,
+                name=name,
+                description=description,
                 upcast_attention=(
                     base_type == BaseModelType.StableDiffusion2
                     and prediction_type == SchedulerPredictionType.VPrediction
@@ -142,6 +148,13 @@ class ModelProbe(object):
 
         return model_info
 
+    @classmethod
+    def get_model_name(cls, model_path: Path) -> str:
+        if model_path.suffix in {".safetensors", ".bin", ".pt", ".ckpt"}:
+            return model_path.stem
+        else:
+            return model_path.name
+
     @classmethod
     def get_model_type_from_checkpoint(cls, model_path: Path, checkpoint: dict) -> ModelType:
         if model_path.suffix not in (".bin", ".pt", ".ckpt", ".safetensors", ".pth"):

invokeai/backend/model_manager/__init__.py

@@ -0,0 +1,29 @@
+"""Re-export frequently-used symbols from the Model Manager backend."""
+
+from .config import (
+    AnyModelConfig,
+    BaseModelType,
+    InvalidModelConfigException,
+    ModelConfigFactory,
+    ModelFormat,
+    ModelType,
+    ModelVariantType,
+    SchedulerPredictionType,
+    SubModelType,
+)
+from .probe import ModelProbe
+from .search import ModelSearch
+
+__all__ = [
+    "ModelProbe",
+    "ModelSearch",
+    "InvalidModelConfigException",
+    "ModelConfigFactory",
+    "BaseModelType",
+    "ModelType",
+    "SubModelType",
+    "ModelVariantType",
+    "ModelFormat",
+    "SchedulerPredictionType",
+    "AnyModelConfig",
+]

invokeai/backend/model_manager/config.py

@@ -23,7 +23,7 @@ from enum import Enum
 from typing import Literal, Optional, Type, Union
 
 from pydantic import BaseModel, ConfigDict, Field, TypeAdapter
-from typing_extensions import Annotated
+from typing_extensions import Annotated, Any, Dict
 
 
 class InvalidModelConfigException(Exception):
@@ -122,7 +122,7 @@ class ModelConfigBase(BaseModel):
         validate_assignment=True,
     )
 
-    def update(self, attributes: dict):
+    def update(self, attributes: Dict[str, Any]) -> None:
         """Update the object with fields in dict."""
         for key, value in attributes.items():
             setattr(self, key, value)  # may raise a validation error
@@ -195,8 +195,6 @@ class MainCheckpointConfig(_CheckpointConfig, _MainConfig):
     """Model config for main checkpoint models."""
 
     type: Literal[ModelType.Main] = ModelType.Main
-    # Note that we do not need prediction_type or upcast_attention here
-    # because they are provided in the checkpoint's own config file.
 
 
 class MainDiffusersConfig(_DiffusersConfig, _MainConfig):

invokeai/backend/model_manager/migrate_to_db.py

@@ -2,6 +2,7 @@
 """Migrate from the InvokeAI v2 models.yaml format to the v3 sqlite format."""
 
 from hashlib import sha1
+from logging import Logger
 
 from omegaconf import DictConfig, OmegaConf
 from pydantic import TypeAdapter
@@ -10,6 +11,7 @@ from invokeai.app.services.config import InvokeAIAppConfig
 from invokeai.app.services.model_records import (
     DuplicateModelException,
     ModelRecordServiceSQL,
+    UnknownModelException,
 )
 from invokeai.app.services.shared.sqlite.sqlite_database import SqliteDatabase
 from invokeai.backend.model_manager.config import (
@@ -38,24 +40,27 @@ class MigrateModelYamlToDb:
     """
 
     config: InvokeAIAppConfig
-    logger: InvokeAILogger
+    logger: Logger
 
-    def __init__(self):
+    def __init__(self) -> None:
         self.config = InvokeAIAppConfig.get_config()
         self.config.parse_args()
         self.logger = InvokeAILogger.get_logger()
 
     def get_db(self) -> ModelRecordServiceSQL:
         """Fetch the sqlite3 database for this installation."""
-        db = SqliteDatabase(self.config, self.logger)
+        db_path = None if self.config.use_memory_db else self.config.db_path
+        db = SqliteDatabase(db_path=db_path, logger=self.logger, verbose=self.config.log_sql)
         return ModelRecordServiceSQL(db)
 
     def get_yaml(self) -> DictConfig:
         """Fetch the models.yaml DictConfig for this installation."""
         yaml_path = self.config.model_conf_path
-        return OmegaConf.load(yaml_path)
+        omegaconf = OmegaConf.load(yaml_path)
+        assert isinstance(omegaconf, DictConfig)
+        return omegaconf
 
-    def migrate(self):
+    def migrate(self) -> None:
         """Do the migration from models.yaml to invokeai.db."""
         db = self.get_db()
         yaml = self.get_yaml()
@@ -69,6 +74,7 @@ class MigrateModelYamlToDb:
 
             base_type, model_type, model_name = str(model_key).split("/")
             hash = FastModelHash.hash(self.config.models_path / stanza.path)
+            assert isinstance(model_key, str)
             new_key = sha1(model_key.encode("utf-8")).hexdigest()
 
             stanza["base"] = BaseModelType(base_type)
@@ -77,12 +83,20 @@ class MigrateModelYamlToDb:
             stanza["original_hash"] = hash
             stanza["current_hash"] = hash
 
-            new_config = ModelsValidator.validate_python(stanza)
-            self.logger.info(f"Adding model {model_name} with key {model_key}")
+            new_config: AnyModelConfig = ModelsValidator.validate_python(stanza)  # type: ignore # see https://github.com/pydantic/pydantic/discussions/7094
+
             try:
-                db.add_model(new_key, new_config)
+                if original_record := db.search_by_path(stanza.path):
+                    key = original_record[0].key
+                    self.logger.info(f"Updating model {model_name} with information from models.yaml using key {key}")
+                    db.update_model(key, new_config)
+                else:
+                    self.logger.info(f"Adding model {model_name} with key {model_key}")
+                    db.add_model(new_key, new_config)
             except DuplicateModelException:
                 self.logger.warning(f"Model {model_name} is already in the database")
+            except UnknownModelException:
+                self.logger.warning(f"Model at {stanza.path} could not be found in database")
 
 
 def main():

invokeai/backend/model_manager/probe.py

@@ -0,0 +1,684 @@
+import json
+import re
+from pathlib import Path
+from typing import Any, Dict, Literal, Optional, Union
+
+import safetensors.torch
+import torch
+from picklescan.scanner import scan_file_path
+
+from invokeai.backend.model_management.models.base import read_checkpoint_meta
+from invokeai.backend.model_management.models.ip_adapter import IPAdapterModelFormat
+from invokeai.backend.model_management.util import lora_token_vector_length
+from invokeai.backend.util.util import SilenceWarnings
+
+from .config import (
+    AnyModelConfig,
+    BaseModelType,
+    InvalidModelConfigException,
+    ModelConfigFactory,
+    ModelFormat,
+    ModelType,
+    ModelVariantType,
+    SchedulerPredictionType,
+)
+from .hash import FastModelHash
+
+CkptType = Dict[str, Any]
+
+LEGACY_CONFIGS: Dict[BaseModelType, Dict[ModelVariantType, Union[str, Dict[SchedulerPredictionType, str]]]] = {
+    BaseModelType.StableDiffusion1: {
+        ModelVariantType.Normal: "v1-inference.yaml",
+        ModelVariantType.Inpaint: "v1-inpainting-inference.yaml",
+    },
+    BaseModelType.StableDiffusion2: {
+        ModelVariantType.Normal: {
+            SchedulerPredictionType.Epsilon: "v2-inference.yaml",
+            SchedulerPredictionType.VPrediction: "v2-inference-v.yaml",
+        },
+        ModelVariantType.Inpaint: {
+            SchedulerPredictionType.Epsilon: "v2-inpainting-inference.yaml",
+            SchedulerPredictionType.VPrediction: "v2-inpainting-inference-v.yaml",
+        },
+    },
+    BaseModelType.StableDiffusionXL: {
+        ModelVariantType.Normal: "sd_xl_base.yaml",
+    },
+    BaseModelType.StableDiffusionXLRefiner: {
+        ModelVariantType.Normal: "sd_xl_refiner.yaml",
+    },
+}
+
+
+class ProbeBase(object):
+    """Base class for probes."""
+
+    def __init__(self, model_path: Path):
+        self.model_path = model_path
+
+    def get_base_type(self) -> BaseModelType:
+        """Get model base type."""
+        raise NotImplementedError
+
+    def get_format(self) -> ModelFormat:
+        """Get model file format."""
+        raise NotImplementedError
+
+    def get_variant_type(self) -> Optional[ModelVariantType]:
+        """Get model variant type."""
+        return None
+
+    def get_scheduler_prediction_type(self) -> Optional[SchedulerPredictionType]:
+        """Get model scheduler prediction type."""
+        return None
+
+
+class ModelProbe(object):
+    PROBES: Dict[str, Dict[ModelType, type[ProbeBase]]] = {
+        "diffusers": {},
+        "checkpoint": {},
+        "onnx": {},
+    }
+
+    CLASS2TYPE = {
+        "StableDiffusionPipeline": ModelType.Main,
+        "StableDiffusionInpaintPipeline": ModelType.Main,
+        "StableDiffusionXLPipeline": ModelType.Main,
+        "StableDiffusionXLImg2ImgPipeline": ModelType.Main,
+        "StableDiffusionXLInpaintPipeline": ModelType.Main,
+        "LatentConsistencyModelPipeline": ModelType.Main,
+        "AutoencoderKL": ModelType.Vae,
+        "AutoencoderTiny": ModelType.Vae,
+        "ControlNetModel": ModelType.ControlNet,
+        "CLIPVisionModelWithProjection": ModelType.CLIPVision,
+        "T2IAdapter": ModelType.T2IAdapter,
+    }
+
+    @classmethod
+    def register_probe(
+        cls, format: Literal["diffusers", "checkpoint", "onnx"], model_type: ModelType, probe_class: type[ProbeBase]
+    ) -> None:
+        cls.PROBES[format][model_type] = probe_class
+
+    @classmethod
+    def heuristic_probe(
+        cls,
+        model_path: Path,
+        fields: Optional[Dict[str, Any]] = None,
+    ) -> AnyModelConfig:
+        return cls.probe(model_path, fields)
+
+    @classmethod
+    def probe(
+        cls,
+        model_path: Path,
+        fields: Optional[Dict[str, Any]] = None,
+    ) -> AnyModelConfig:
+        """
+        Probe the model at model_path and return its configuration record.
+
+        :param model_path: Path to the model file (checkpoint) or directory (diffusers).
+        :param fields: An optional dictionary that can be used to override probed
+        fields. Typically used for fields that don't probe well, such as prediction_type.
+
+        Returns: The appropriate model configuration derived from ModelConfigBase.
+        """
+        if fields is None:
+            fields = {}
+
+        format_type = ModelFormat.Diffusers if model_path.is_dir() else ModelFormat.Checkpoint
+        model_info = None
+        model_type = None
+        if format_type == "diffusers":
+            model_type = cls.get_model_type_from_folder(model_path)
+        else:
+            model_type = cls.get_model_type_from_checkpoint(model_path)
+        format_type = ModelFormat.Onnx if model_type == ModelType.ONNX else format_type
+
+        probe_class = cls.PROBES[format_type].get(model_type)
+        if not probe_class:
+            raise InvalidModelConfigException(f"Unhandled combination of {format_type} and {model_type}")
+
+        hash = FastModelHash.hash(model_path)
+        probe = probe_class(model_path)
+
+        fields["path"] = model_path.as_posix()
+        fields["type"] = fields.get("type") or model_type
+        fields["base"] = fields.get("base") or probe.get_base_type()
+        fields["variant"] = fields.get("variant") or probe.get_variant_type()
+        fields["prediction_type"] = fields.get("prediction_type") or probe.get_scheduler_prediction_type()
+        fields["name"] = fields.get("name") or cls.get_model_name(model_path)
+        fields["description"] = (
+            fields.get("description") or f"{fields['base'].value} {fields['type'].value} model {fields['name']}"
+        )
+        fields["format"] = fields.get("format") or probe.get_format()
+        fields["original_hash"] = fields.get("original_hash") or hash
+        fields["current_hash"] = fields.get("current_hash") or hash
+
+        # additional fields needed for main and controlnet models
+        if fields["type"] in [ModelType.Main, ModelType.ControlNet] and fields["format"] == ModelFormat.Checkpoint:
+            fields["config"] = cls._get_checkpoint_config_path(
+                model_path,
+                model_type=fields["type"],
+                base_type=fields["base"],
+                variant_type=fields["variant"],
+                prediction_type=fields["prediction_type"],
+            ).as_posix()
+
+        # additional fields needed for main non-checkpoint models
+        elif fields["type"] == ModelType.Main and fields["format"] in [
+            ModelFormat.Onnx,
+            ModelFormat.Olive,
+            ModelFormat.Diffusers,
+        ]:
+            fields["upcast_attention"] = fields.get("upcast_attention") or (
+                fields["base"] == BaseModelType.StableDiffusion2
+                and fields["prediction_type"] == SchedulerPredictionType.VPrediction
+            )
+
+        model_info = ModelConfigFactory.make_config(fields)
+        return model_info
+
+    @classmethod
+    def get_model_name(cls, model_path: Path) -> str:
+        if model_path.suffix in {".safetensors", ".bin", ".pt", ".ckpt"}:
+            return model_path.stem
+        else:
+            return model_path.name
+
+    @classmethod
+    def get_model_type_from_checkpoint(cls, model_path: Path, checkpoint: Optional[CkptType] = None) -> ModelType:
+        if model_path.suffix not in (".bin", ".pt", ".ckpt", ".safetensors", ".pth"):
+            raise InvalidModelConfigException(f"{model_path}: unrecognized suffix")
+
+        if model_path.name == "learned_embeds.bin":
+            return ModelType.TextualInversion
+
+        ckpt = checkpoint if checkpoint else read_checkpoint_meta(model_path, scan=True)
+        ckpt = ckpt.get("state_dict", ckpt)
+
+        for key in ckpt.keys():
+            if any(key.startswith(v) for v in {"cond_stage_model.", "first_stage_model.", "model.diffusion_model."}):
+                return ModelType.Main
+            elif any(key.startswith(v) for v in {"encoder.conv_in", "decoder.conv_in"}):
+                return ModelType.Vae
+            elif any(key.startswith(v) for v in {"lora_te_", "lora_unet_"}):
+                return ModelType.Lora
+            elif any(key.endswith(v) for v in {"to_k_lora.up.weight", "to_q_lora.down.weight"}):
+                return ModelType.Lora
+            elif any(key.startswith(v) for v in {"control_model", "input_blocks"}):
+                return ModelType.ControlNet
+            elif key in {"emb_params", "string_to_param"}:
+                return ModelType.TextualInversion
+
+        else:
+            # diffusers-ti
+            if len(ckpt) < 10 and all(isinstance(v, torch.Tensor) for v in ckpt.values()):
+                return ModelType.TextualInversion
+
+        raise InvalidModelConfigException(f"Unable to determine model type for {model_path}")
+
+    @classmethod
+    def get_model_type_from_folder(cls, folder_path: Path) -> ModelType:
+        """Get the model type of a hugging-face style folder."""
+        class_name = None
+        error_hint = None
+        for suffix in ["bin", "safetensors"]:
+            if (folder_path / f"learned_embeds.{suffix}").exists():
+                return ModelType.TextualInversion
+            if (folder_path / f"pytorch_lora_weights.{suffix}").exists():
+                return ModelType.Lora
+        if (folder_path / "unet/model.onnx").exists():
+            return ModelType.ONNX
+        if (folder_path / "image_encoder.txt").exists():
+            return ModelType.IPAdapter
+
+        i = folder_path / "model_index.json"
+        c = folder_path / "config.json"
+        config_path = i if i.exists() else c if c.exists() else None
+
+        if config_path:
+            with open(config_path, "r") as file:
+                conf = json.load(file)
+            if "_class_name" in conf:
+                class_name = conf["_class_name"]
+            elif "architectures" in conf:
+                class_name = conf["architectures"][0]
+            else:
+                class_name = None
+        else:
+            error_hint = f"No model_index.json or config.json found in {folder_path}."
+
+        if class_name and (type := cls.CLASS2TYPE.get(class_name)):
+            return type
+        else:
+            error_hint = f"class {class_name} is not one of the supported classes [{', '.join(cls.CLASS2TYPE.keys())}]"
+
+        # give up
+        raise InvalidModelConfigException(
+            f"Unable to determine model type for {folder_path}" + (f"; {error_hint}" if error_hint else "")
+        )
+
+    @classmethod
+    def _get_checkpoint_config_path(
+        cls,
+        model_path: Path,
+        model_type: ModelType,
+        base_type: BaseModelType,
+        variant_type: ModelVariantType,
+        prediction_type: SchedulerPredictionType,
+    ) -> Path:
+        # look for a YAML file adjacent to the model file first
+        possible_conf = model_path.with_suffix(".yaml")
+        if possible_conf.exists():
+            return possible_conf.absolute()
+
+        if model_type == ModelType.Main:
+            config_file = LEGACY_CONFIGS[base_type][variant_type]
+            if isinstance(config_file, dict):  # need another tier for sd-2.x models
+                config_file = config_file[prediction_type]
+        elif model_type == ModelType.ControlNet:
+            config_file = (
+                "../controlnet/cldm_v15.yaml" if base_type == BaseModelType("sd-1") else "../controlnet/cldm_v21.yaml"
+            )
+        else:
+            raise InvalidModelConfigException(
+                f"{model_path}: Unrecognized combination of model_type={model_type}, base_type={base_type}"
+            )
+        assert isinstance(config_file, str)
+        return Path(config_file)
+
+    @classmethod
+    def _scan_and_load_checkpoint(cls, model_path: Path) -> CkptType:
+        with SilenceWarnings():
+            if model_path.suffix.endswith((".ckpt", ".pt", ".bin")):
+                cls._scan_model(model_path.name, model_path)
+                model = torch.load(model_path)
+                assert isinstance(model, dict)
+                return model
+            else:
+                return safetensors.torch.load_file(model_path)
+
+    @classmethod
+    def _scan_model(cls, model_name: str, checkpoint: Path) -> None:
+        """
+        Apply picklescanner to the indicated checkpoint and issue a warning
+        and option to exit if an infected file is identified.
+        """
+        # scan model
+        scan_result = scan_file_path(checkpoint)
+        if scan_result.infected_files != 0:
+            raise Exception("The model {model_name} is potentially infected by malware. Aborting import.")
+
+
+# ##################################################3
+# Checkpoint probing
+# ##################################################3
+
+
+class CheckpointProbeBase(ProbeBase):
+    def __init__(self, model_path: Path):
+        super().__init__(model_path)
+        self.checkpoint = ModelProbe._scan_and_load_checkpoint(model_path)
+
+    def get_format(self) -> ModelFormat:
+        return ModelFormat("checkpoint")
+
+    def get_variant_type(self) -> ModelVariantType:
+        model_type = ModelProbe.get_model_type_from_checkpoint(self.model_path, self.checkpoint)
+        if model_type != ModelType.Main:
+            return ModelVariantType.Normal
+        state_dict = self.checkpoint.get("state_dict") or self.checkpoint
+        in_channels = state_dict["model.diffusion_model.input_blocks.0.0.weight"].shape[1]
+        if in_channels == 9:
+            return ModelVariantType.Inpaint
+        elif in_channels == 5:
+            return ModelVariantType.Depth
+        elif in_channels == 4:
+            return ModelVariantType.Normal
+        else:
+            raise InvalidModelConfigException(
+                f"Cannot determine variant type (in_channels={in_channels}) at {self.model_path}"
+            )
+
+
+class PipelineCheckpointProbe(CheckpointProbeBase):
+    def get_base_type(self) -> BaseModelType:
+        checkpoint = self.checkpoint
+        state_dict = self.checkpoint.get("state_dict") or checkpoint
+        key_name = "model.diffusion_model.input_blocks.2.1.transformer_blocks.0.attn2.to_k.weight"
+        if key_name in state_dict and state_dict[key_name].shape[-1] == 768:
+            return BaseModelType.StableDiffusion1
+        if key_name in state_dict and state_dict[key_name].shape[-1] == 1024:
+            return BaseModelType.StableDiffusion2
+        key_name = "model.diffusion_model.input_blocks.4.1.transformer_blocks.0.attn2.to_k.weight"
+        if key_name in state_dict and state_dict[key_name].shape[-1] == 2048:
+            return BaseModelType.StableDiffusionXL
+        elif key_name in state_dict and state_dict[key_name].shape[-1] == 1280:
+            return BaseModelType.StableDiffusionXLRefiner
+        else:
+            raise InvalidModelConfigException("Cannot determine base type")
+
+    def get_scheduler_prediction_type(self) -> SchedulerPredictionType:
+        """Return model prediction type."""
+        type = self.get_base_type()
+        if type == BaseModelType.StableDiffusion2:
+            checkpoint = self.checkpoint
+            state_dict = self.checkpoint.get("state_dict") or checkpoint
+            key_name = "model.diffusion_model.input_blocks.2.1.transformer_blocks.0.attn2.to_k.weight"
+            if key_name in state_dict and state_dict[key_name].shape[-1] == 1024:
+                if "global_step" in checkpoint:
+                    if checkpoint["global_step"] == 220000:
+                        return SchedulerPredictionType.Epsilon
+                    elif checkpoint["global_step"] == 110000:
+                        return SchedulerPredictionType.VPrediction
+            return SchedulerPredictionType.VPrediction  # a guess for sd2 ckpts
+
+        elif type == BaseModelType.StableDiffusion1:
+            return SchedulerPredictionType.Epsilon  # a reasonable guess for sd1 ckpts
+        else:
+            return SchedulerPredictionType.Epsilon
+
+
+class VaeCheckpointProbe(CheckpointProbeBase):
+    def get_base_type(self) -> BaseModelType:
+        # I can't find any standalone 2.X VAEs to test with!
+        return BaseModelType.StableDiffusion1
+
+
+class LoRACheckpointProbe(CheckpointProbeBase):
+    """Class for LoRA checkpoints."""
+
+    def get_format(self) -> ModelFormat:
+        return ModelFormat("lycoris")
+
+    def get_base_type(self) -> BaseModelType:
+        checkpoint = self.checkpoint
+        token_vector_length = lora_token_vector_length(checkpoint)
+
+        if token_vector_length == 768:
+            return BaseModelType.StableDiffusion1
+        elif token_vector_length == 1024:
+            return BaseModelType.StableDiffusion2
+        elif token_vector_length == 2048:
+            return BaseModelType.StableDiffusionXL
+        else:
+            raise InvalidModelConfigException(f"Unknown LoRA type: {self.model_path}")
+
+
+class TextualInversionCheckpointProbe(CheckpointProbeBase):
+    """Class for probing embeddings."""
+
+    def get_format(self) -> ModelFormat:
+        return ModelFormat.EmbeddingFile
+
+    def get_base_type(self) -> BaseModelType:
+        checkpoint = self.checkpoint
+        if "string_to_token" in checkpoint:
+            token_dim = list(checkpoint["string_to_param"].values())[0].shape[-1]
+        elif "emb_params" in checkpoint:
+            token_dim = checkpoint["emb_params"].shape[-1]
+        elif "clip_g" in checkpoint:
+            token_dim = checkpoint["clip_g"].shape[-1]
+        else:
+            token_dim = list(checkpoint.values())[0].shape[0]
+        if token_dim == 768:
+            return BaseModelType.StableDiffusion1
+        elif token_dim == 1024:
+            return BaseModelType.StableDiffusion2
+        elif token_dim == 1280:
+            return BaseModelType.StableDiffusionXL
+        else:
+            raise InvalidModelConfigException(f"{self.model_path}: Could not determine base type")
+
+
+class ControlNetCheckpointProbe(CheckpointProbeBase):
+    """Class for probing controlnets."""
+
+    def get_base_type(self) -> BaseModelType:
+        checkpoint = self.checkpoint
+        for key_name in (
+            "control_model.input_blocks.2.1.transformer_blocks.0.attn2.to_k.weight",
+            "input_blocks.2.1.transformer_blocks.0.attn2.to_k.weight",
+        ):
+            if key_name not in checkpoint:
+                continue
+            if checkpoint[key_name].shape[-1] == 768:
+                return BaseModelType.StableDiffusion1
+            elif checkpoint[key_name].shape[-1] == 1024:
+                return BaseModelType.StableDiffusion2
+        raise InvalidModelConfigException("{self.model_path}: Unable to determine base type")
+
+
+class IPAdapterCheckpointProbe(CheckpointProbeBase):
+    def get_base_type(self) -> BaseModelType:
+        raise NotImplementedError()
+
+
+class CLIPVisionCheckpointProbe(CheckpointProbeBase):
+    def get_base_type(self) -> BaseModelType:
+        raise NotImplementedError()
+
+
+class T2IAdapterCheckpointProbe(CheckpointProbeBase):
+    def get_base_type(self) -> BaseModelType:
+        raise NotImplementedError()
+
+
+########################################################
+# classes for probing folders
+#######################################################
+class FolderProbeBase(ProbeBase):
+    def get_variant_type(self) -> ModelVariantType:
+        return ModelVariantType.Normal
+
+    def get_format(self) -> ModelFormat:
+        return ModelFormat("diffusers")
+
+
+class PipelineFolderProbe(FolderProbeBase):
+    def get_base_type(self) -> BaseModelType:
+        with open(self.model_path / "unet" / "config.json", "r") as file:
+            unet_conf = json.load(file)
+        if unet_conf["cross_attention_dim"] == 768:
+            return BaseModelType.StableDiffusion1
+        elif unet_conf["cross_attention_dim"] == 1024:
+            return BaseModelType.StableDiffusion2
+        elif unet_conf["cross_attention_dim"] == 1280:
+            return BaseModelType.StableDiffusionXLRefiner
+        elif unet_conf["cross_attention_dim"] == 2048:
+            return BaseModelType.StableDiffusionXL
+        else:
+            raise InvalidModelConfigException(f"Unknown base model for {self.model_path}")
+
+    def get_scheduler_prediction_type(self) -> SchedulerPredictionType:
+        with open(self.model_path / "scheduler" / "scheduler_config.json", "r") as file:
+            scheduler_conf = json.load(file)
+        if scheduler_conf["prediction_type"] == "v_prediction":
+            return SchedulerPredictionType.VPrediction
+        elif scheduler_conf["prediction_type"] == "epsilon":
+            return SchedulerPredictionType.Epsilon
+        else:
+            raise InvalidModelConfigException("Unknown scheduler prediction type: {scheduler_conf['prediction_type']}")
+
+    def get_variant_type(self) -> ModelVariantType:
+        # This only works for pipelines! Any kind of
+        # exception results in our returning the
+        # "normal" variant type
+        try:
+            config_file = self.model_path / "unet" / "config.json"
+            with open(config_file, "r") as file:
+                conf = json.load(file)
+
+            in_channels = conf["in_channels"]
+            if in_channels == 9:
+                return ModelVariantType.Inpaint
+            elif in_channels == 5:
+                return ModelVariantType.Depth
+            elif in_channels == 4:
+                return ModelVariantType.Normal
+        except Exception:
+            pass
+        return ModelVariantType.Normal
+
+
+class VaeFolderProbe(FolderProbeBase):
+    def get_base_type(self) -> BaseModelType:
+        if self._config_looks_like_sdxl():
+            return BaseModelType.StableDiffusionXL
+        elif self._name_looks_like_sdxl():
+            # but SD and SDXL VAE are the same shape (3-channel RGB to 4-channel float scaled down
+            # by a factor of 8), we can't necessarily tell them apart by config hyperparameters.
+            return BaseModelType.StableDiffusionXL
+        else:
+            return BaseModelType.StableDiffusion1
+
+    def _config_looks_like_sdxl(self) -> bool:
+        # config values that distinguish Stability's SD 1.x VAE from their SDXL VAE.
+        config_file = self.model_path / "config.json"
+        if not config_file.exists():
+            raise InvalidModelConfigException(f"Cannot determine base type for {self.model_path}")
+        with open(config_file, "r") as file:
+            config = json.load(file)
+        return config.get("scaling_factor", 0) == 0.13025 and config.get("sample_size") in [512, 1024]
+
+    def _name_looks_like_sdxl(self) -> bool:
+        return bool(re.search(r"xl\b", self._guess_name(), re.IGNORECASE))
+
+    def _guess_name(self) -> str:
+        name = self.model_path.name
+        if name == "vae":
+            name = self.model_path.parent.name
+        return name
+
+
+class TextualInversionFolderProbe(FolderProbeBase):
+    def get_format(self) -> ModelFormat:
+        return ModelFormat.EmbeddingFolder
+
+    def get_base_type(self) -> BaseModelType:
+        path = self.model_path / "learned_embeds.bin"
+        if not path.exists():
+            raise InvalidModelConfigException(
+                f"{self.model_path.as_posix()} does not contain expected 'learned_embeds.bin' file"
+            )
+        return TextualInversionCheckpointProbe(path).get_base_type()
+
+
+class ONNXFolderProbe(FolderProbeBase):
+    def get_format(self) -> ModelFormat:
+        return ModelFormat("onnx")
+
+    def get_base_type(self) -> BaseModelType:
+        return BaseModelType.StableDiffusion1
+
+    def get_variant_type(self) -> ModelVariantType:
+        return ModelVariantType.Normal
+
+
+class ControlNetFolderProbe(FolderProbeBase):
+    def get_base_type(self) -> BaseModelType:
+        config_file = self.model_path / "config.json"
+        if not config_file.exists():
+            raise InvalidModelConfigException(f"Cannot determine base type for {self.model_path}")
+        with open(config_file, "r") as file:
+            config = json.load(file)
+        # no obvious way to distinguish between sd2-base and sd2-768
+        dimension = config["cross_attention_dim"]
+        base_model = (
+            BaseModelType.StableDiffusion1
+            if dimension == 768
+            else (
+                BaseModelType.StableDiffusion2
+                if dimension == 1024
+                else BaseModelType.StableDiffusionXL
+                if dimension == 2048
+                else None
+            )
+        )
+        if not base_model:
+            raise InvalidModelConfigException(f"Unable to determine model base for {self.model_path}")
+        return base_model
+
+
+class LoRAFolderProbe(FolderProbeBase):
+    def get_base_type(self) -> BaseModelType:
+        model_file = None
+        for suffix in ["safetensors", "bin"]:
+            base_file = self.model_path / f"pytorch_lora_weights.{suffix}"
+            if base_file.exists():
+                model_file = base_file
+                break
+        if not model_file:
+            raise InvalidModelConfigException("Unknown LoRA format encountered")
+        return LoRACheckpointProbe(model_file).get_base_type()
+
+
+class IPAdapterFolderProbe(FolderProbeBase):
+    def get_format(self) -> IPAdapterModelFormat:
+        return IPAdapterModelFormat.InvokeAI.value
+
+    def get_base_type(self) -> BaseModelType:
+        model_file = self.model_path / "ip_adapter.bin"
+        if not model_file.exists():
+            raise InvalidModelConfigException("Unknown IP-Adapter model format.")
+
+        state_dict = torch.load(model_file, map_location="cpu")
+        cross_attention_dim = state_dict["ip_adapter"]["1.to_k_ip.weight"].shape[-1]
+        if cross_attention_dim == 768:
+            return BaseModelType.StableDiffusion1
+        elif cross_attention_dim == 1024:
+            return BaseModelType.StableDiffusion2
+        elif cross_attention_dim == 2048:
+            return BaseModelType.StableDiffusionXL
+        else:
+            raise InvalidModelConfigException(
+                f"IP-Adapter had unexpected cross-attention dimension: {cross_attention_dim}."
+            )
+
+
+class CLIPVisionFolderProbe(FolderProbeBase):
+    def get_base_type(self) -> BaseModelType:
+        return BaseModelType.Any
+
+
+class T2IAdapterFolderProbe(FolderProbeBase):
+    def get_base_type(self) -> BaseModelType:
+        config_file = self.model_path / "config.json"
+        if not config_file.exists():
+            raise InvalidModelConfigException(f"Cannot determine base type for {self.model_path}")
+        with open(config_file, "r") as file:
+            config = json.load(file)
+
+        adapter_type = config.get("adapter_type", None)
+        if adapter_type == "full_adapter_xl":
+            return BaseModelType.StableDiffusionXL
+        elif adapter_type == "full_adapter" or "light_adapter":
+            # I haven't seen any T2I adapter models for SD2, so assume that this is an SD1 adapter.
+            return BaseModelType.StableDiffusion1
+        else:
+            raise InvalidModelConfigException(
+                f"Unable to determine base model for '{self.model_path}' (adapter_type = {adapter_type})."
+            )
+
+
+############## register probe classes ######
+ModelProbe.register_probe("diffusers", ModelType.Main, PipelineFolderProbe)
+ModelProbe.register_probe("diffusers", ModelType.Vae, VaeFolderProbe)
+ModelProbe.register_probe("diffusers", ModelType.Lora, LoRAFolderProbe)
+ModelProbe.register_probe("diffusers", ModelType.TextualInversion, TextualInversionFolderProbe)
+ModelProbe.register_probe("diffusers", ModelType.ControlNet, ControlNetFolderProbe)
+ModelProbe.register_probe("diffusers", ModelType.IPAdapter, IPAdapterFolderProbe)
+ModelProbe.register_probe("diffusers", ModelType.CLIPVision, CLIPVisionFolderProbe)
+ModelProbe.register_probe("diffusers", ModelType.T2IAdapter, T2IAdapterFolderProbe)
+
+ModelProbe.register_probe("checkpoint", ModelType.Main, PipelineCheckpointProbe)
+ModelProbe.register_probe("checkpoint", ModelType.Vae, VaeCheckpointProbe)
+ModelProbe.register_probe("checkpoint", ModelType.Lora, LoRACheckpointProbe)
+ModelProbe.register_probe("checkpoint", ModelType.TextualInversion, TextualInversionCheckpointProbe)
+ModelProbe.register_probe("checkpoint", ModelType.ControlNet, ControlNetCheckpointProbe)
+ModelProbe.register_probe("checkpoint", ModelType.IPAdapter, IPAdapterCheckpointProbe)
+ModelProbe.register_probe("checkpoint", ModelType.CLIPVision, CLIPVisionCheckpointProbe)
+ModelProbe.register_probe("checkpoint", ModelType.T2IAdapter, T2IAdapterCheckpointProbe)
+
+ModelProbe.register_probe("onnx", ModelType.ONNX, ONNXFolderProbe)

invokeai/backend/model_manager/search.py

@@ -0,0 +1,190 @@
+# Copyright 2023, Lincoln D. Stein and the InvokeAI Team
+"""
+Abstract base class and implementation for recursive directory search for models.
+
+Example usage:
+```
+  from invokeai.backend.model_manager import ModelSearch, ModelProbe
+
+  def find_main_models(model: Path) -> bool:
+    info = ModelProbe.probe(model)
+    if info.model_type == 'main' and info.base_type == 'sd-1':
+        return True
+    else:
+        return False
+
+  search = ModelSearch(on_model_found=report_it)
+  found = search.search('/tmp/models')
+  print(found)   #  list of matching model paths
+  print(search.stats)  #  search stats
+```
+"""
+
+import os
+from abc import ABC, abstractmethod
+from pathlib import Path
+from typing import Callable, Optional, Set, Union
+
+from pydantic import BaseModel, Field
+
+from invokeai.backend.util.logging import InvokeAILogger
+
+default_logger = InvokeAILogger.get_logger()
+
+
+class SearchStats(BaseModel):
+    items_scanned: int = 0
+    models_found: int = 0
+    models_filtered: int = 0
+
+
+class ModelSearchBase(ABC, BaseModel):
+    """
+    Abstract directory traversal model search class
+
+    Usage:
+       search = ModelSearchBase(
+            on_search_started = search_started_callback,
+            on_search_completed = search_completed_callback,
+            on_model_found = model_found_callback,
+       )
+       models_found = search.search('/path/to/directory')
+    """
+
+    # fmt: off
+    on_search_started   : Optional[Callable[[Path], None]]      = Field(default=None, description="Called just before the search starts.")  # noqa E221
+    on_model_found      : Optional[Callable[[Path], bool]]      = Field(default=None, description="Called when a model is found.")          # noqa E221
+    on_search_completed : Optional[Callable[[Set[Path]], None]] = Field(default=None, description="Called when search is complete.")        # noqa E221
+    stats               : SearchStats                           = Field(default_factory=SearchStats, description="Summary statistics after search")  # noqa E221
+    logger              : InvokeAILogger                        = Field(default=default_logger, description="Logger instance.")     # noqa E221
+    # fmt: on
+
+    class Config:
+        arbitrary_types_allowed = True
+
+    @abstractmethod
+    def search_started(self) -> None:
+        """
+        Called before the scan starts.
+
+        Passes the root search directory to the Callable `on_search_started`.
+        """
+        pass
+
+    @abstractmethod
+    def model_found(self, model: Path) -> None:
+        """
+        Called when a model is found during search.
+
+        :param model: Model to process - could be a directory or checkpoint.
+
+        Passes the model's Path to the Callable `on_model_found`.
+        This Callable receives the path to the model and returns a boolean
+        to indicate whether the model should be returned in the search
+        results.
+        """
+        pass
+
+    @abstractmethod
+    def search_completed(self) -> None:
+        """
+        Called before the scan starts.
+
+        Passes the Set of found model Paths to the Callable `on_search_completed`.
+        """
+        pass
+
+    @abstractmethod
+    def search(self, directory: Union[Path, str]) -> Set[Path]:
+        """
+        Recursively search for models in `directory` and return a set of model paths.
+
+        If provided, the `on_search_started`, `on_model_found` and `on_search_completed`
+        Callables will be invoked during the search.
+        """
+        pass
+
+
+class ModelSearch(ModelSearchBase):
+    """
+    Implementation of ModelSearch with callbacks.
+    Usage:
+       search = ModelSearch()
+       search.model_found = lambda path : 'anime' in path.as_posix()
+       found = search.list_models(['/tmp/models1','/tmp/models2'])
+       # returns all models that have 'anime' in the path
+    """
+
+    models_found: Set[Path] = Field(default=None)
+    scanned_dirs: Set[Path] = Field(default=None)
+    pruned_paths: Set[Path] = Field(default=None)
+
+    def search_started(self) -> None:
+        self.models_found = set()
+        self.scanned_dirs = set()
+        self.pruned_paths = set()
+        if self.on_search_started:
+            self.on_search_started(self._directory)
+
+    def model_found(self, model: Path) -> None:
+        self.stats.models_found += 1
+        if not self.on_model_found or self.on_model_found(model):
+            self.stats.models_filtered += 1
+            self.models_found.add(model)
+
+    def search_completed(self) -> None:
+        if self.on_search_completed:
+            self.on_search_completed(self._models_found)
+
+    def search(self, directory: Union[Path, str]) -> Set[Path]:
+        self._directory = Path(directory)
+        self.stats = SearchStats()  # zero out
+        self.search_started()  # This will initialize _models_found to empty
+        self._walk_directory(directory)
+        self.search_completed()
+        return self.models_found
+
+    def _walk_directory(self, path: Union[Path, str]) -> None:
+        for root, dirs, files in os.walk(path, followlinks=True):
+            # don't descend into directories that start with a "."
+            # to avoid the Mac .DS_STORE issue.
+            if str(Path(root).name).startswith("."):
+                self.pruned_paths.add(Path(root))
+            if any(Path(root).is_relative_to(x) for x in self.pruned_paths):
+                continue
+
+            self.stats.items_scanned += len(dirs) + len(files)
+            for d in dirs:
+                path = Path(root) / d
+                if path.parent in self.scanned_dirs:
+                    self.scanned_dirs.add(path)
+                    continue
+                if any(
+                    (path / x).exists()
+                    for x in [
+                        "config.json",
+                        "model_index.json",
+                        "learned_embeds.bin",
+                        "pytorch_lora_weights.bin",
+                        "image_encoder.txt",
+                    ]
+                ):
+                    self.scanned_dirs.add(path)
+                    try:
+                        self.model_found(path)
+                    except KeyboardInterrupt:
+                        raise
+                    except Exception as e:
+                        self.logger.warning(str(e))
+
+            for f in files:
+                path = Path(root) / f
+                if path.parent in self.scanned_dirs:
+                    continue
+                if path.suffix in {".ckpt", ".bin", ".pth", ".safetensors", ".pt"}:
+                    try:
+                        self.model_found(path)
+                    except KeyboardInterrupt:
+                        raise
+                    except Exception as e:
+                        self.logger.warning(str(e))

invokeai/backend/stable_diffusion/diffusers_pipeline.py

@@ -242,17 +242,6 @@ class StableDiffusionGeneratorPipeline(StableDiffusionPipeline):
         control_model: ControlNetModel = None,
     ):
         super().__init__(
-            vae,
-            text_encoder,
-            tokenizer,
-            unet,
-            scheduler,
-            safety_checker,
-            feature_extractor,
-            requires_safety_checker,
-        )
-
-        self.register_modules(
             vae=vae,
             text_encoder=text_encoder,
             tokenizer=tokenizer,
@@ -260,9 +249,9 @@ class StableDiffusionGeneratorPipeline(StableDiffusionPipeline):
             scheduler=scheduler,
             safety_checker=safety_checker,
             feature_extractor=feature_extractor,
-            # FIXME: can't currently register control module
-            # control_model=control_model,
+            requires_safety_checker=requires_safety_checker,
         )
+
         self.invokeai_diffuser = InvokeAIDiffuserComponent(self.unet, self._unet_forward)
         self.control_model = control_model
         self.use_ip_adapter = False

invokeai/backend/tiles/tiles.py

@@ -3,7 +3,42 @@ from typing import Union
 
 import numpy as np
 
-from invokeai.backend.tiles.utils import TBLR, Tile, paste
+from invokeai.app.invocations.latent import LATENT_SCALE_FACTOR
+from invokeai.backend.tiles.utils import TBLR, Tile, paste, seam_blend
+
+
+def calc_overlap(tiles: list[Tile], num_tiles_x: int, num_tiles_y: int) -> list[Tile]:
+    """Calculate and update the overlap of a list of tiles.
+
+    Args:
+        tiles (list[Tile]): The list of tiles describing the locations of the respective `tile_images`.
+        num_tiles_x: the number of tiles on the x axis.
+        num_tiles_y: the number of tiles on the y axis.
+    """
+
+    def get_tile_or_none(idx_y: int, idx_x: int) -> Union[Tile, None]:
+        if idx_y < 0 or idx_y > num_tiles_y or idx_x < 0 or idx_x > num_tiles_x:
+            return None
+        return tiles[idx_y * num_tiles_x + idx_x]
+
+    for tile_idx_y in range(num_tiles_y):
+        for tile_idx_x in range(num_tiles_x):
+            cur_tile = get_tile_or_none(tile_idx_y, tile_idx_x)
+            top_neighbor_tile = get_tile_or_none(tile_idx_y - 1, tile_idx_x)
+            left_neighbor_tile = get_tile_or_none(tile_idx_y, tile_idx_x - 1)
+
+            assert cur_tile is not None
+
+            # Update cur_tile top-overlap and corresponding top-neighbor bottom-overlap.
+            if top_neighbor_tile is not None:
+                cur_tile.overlap.top = max(0, top_neighbor_tile.coords.bottom - cur_tile.coords.top)
+                top_neighbor_tile.overlap.bottom = cur_tile.overlap.top
+
+            # Update cur_tile left-overlap and corresponding left-neighbor right-overlap.
+            if left_neighbor_tile is not None:
+                cur_tile.overlap.left = max(0, left_neighbor_tile.coords.right - cur_tile.coords.left)
+                left_neighbor_tile.overlap.right = cur_tile.overlap.left
+    return tiles
 
 
 def calc_tiles_with_overlap(
@@ -63,31 +98,129 @@ def calc_tiles_with_overlap(
 
             tiles.append(tile)
 
-    def get_tile_or_none(idx_y: int, idx_x: int) -> Union[Tile, None]:
-        if idx_y < 0 or idx_y > num_tiles_y or idx_x < 0 or idx_x > num_tiles_x:
-            return None
-        return tiles[idx_y * num_tiles_x + idx_x]
+    return calc_overlap(tiles, num_tiles_x, num_tiles_y)
 
-    # Iterate over tiles again and calculate overlaps.
+
+def calc_tiles_even_split(
+    image_height: int, image_width: int, num_tiles_x: int, num_tiles_y: int, overlap_fraction: float = 0
+) -> list[Tile]:
+    """Calculate the tile coordinates for a given image shape with the number of tiles requested.
+
+    Args:
+        image_height (int): The image height in px.
+        image_width (int): The image width in px.
+        num_x_tiles (int): The number of tile to split the image into on the X-axis.
+        num_y_tiles (int): The number of tile to split the image into on the Y-axis.
+        overlap_fraction (float, optional): The target overlap as fraction of the tiles size. Defaults to 0.
+
+    Returns:
+        list[Tile]: A list of tiles that cover the image shape. Ordered from left-to-right, top-to-bottom.
+    """
+
+    # Ensure tile size is divisible by 8
+    if image_width % LATENT_SCALE_FACTOR != 0 or image_height % LATENT_SCALE_FACTOR != 0:
+        raise ValueError(f"image size (({image_width}, {image_height})) must be divisible by {LATENT_SCALE_FACTOR}")
+
+    # Calculate the overlap size based on the percentage and adjust it to be divisible by 8 (rounding up)
+    overlap_x = LATENT_SCALE_FACTOR * math.ceil(
+        int((image_width / num_tiles_x) * overlap_fraction) / LATENT_SCALE_FACTOR
+    )
+    overlap_y = LATENT_SCALE_FACTOR * math.ceil(
+        int((image_height / num_tiles_y) * overlap_fraction) / LATENT_SCALE_FACTOR
+    )
+
+    # Calculate the tile size based on the number of tiles and overlap, and ensure it's divisible by 8 (rounding down)
+    tile_size_x = LATENT_SCALE_FACTOR * math.floor(
+        ((image_width + overlap_x * (num_tiles_x - 1)) // num_tiles_x) / LATENT_SCALE_FACTOR
+    )
+    tile_size_y = LATENT_SCALE_FACTOR * math.floor(
+        ((image_height + overlap_y * (num_tiles_y - 1)) // num_tiles_y) / LATENT_SCALE_FACTOR
+    )
+
+    # tiles[y * num_tiles_x + x] is the tile for the y'th row, x'th column.
+    tiles: list[Tile] = []
+
+    # Calculate tile coordinates. (Ignore overlap values for now.)
     for tile_idx_y in range(num_tiles_y):
+        # Calculate the top and bottom of the row
+        top = tile_idx_y * (tile_size_y - overlap_y)
+        bottom = min(top + tile_size_y, image_height)
+        # For the last row adjust bottom to be the height of the image
+        if tile_idx_y == num_tiles_y - 1:
+            bottom = image_height
+
         for tile_idx_x in range(num_tiles_x):
-            cur_tile = get_tile_or_none(tile_idx_y, tile_idx_x)
-            top_neighbor_tile = get_tile_or_none(tile_idx_y - 1, tile_idx_x)
-            left_neighbor_tile = get_tile_or_none(tile_idx_y, tile_idx_x - 1)
+            # Calculate the left & right coordinate of each tile
+            left = tile_idx_x * (tile_size_x - overlap_x)
+            right = min(left + tile_size_x, image_width)
+            # For the last tile in the row adjust right to be the width of the image
+            if tile_idx_x == num_tiles_x - 1:
+                right = image_width
 
-            assert cur_tile is not None
+            tile = Tile(
+                coords=TBLR(top=top, bottom=bottom, left=left, right=right),
+                overlap=TBLR(top=0, bottom=0, left=0, right=0),
+            )
 
-            # Update cur_tile top-overlap and corresponding top-neighbor bottom-overlap.
-            if top_neighbor_tile is not None:
-                cur_tile.overlap.top = max(0, top_neighbor_tile.coords.bottom - cur_tile.coords.top)
-                top_neighbor_tile.overlap.bottom = cur_tile.overlap.top
+            tiles.append(tile)
 
-            # Update cur_tile left-overlap and corresponding left-neighbor right-overlap.
-            if left_neighbor_tile is not None:
-                cur_tile.overlap.left = max(0, left_neighbor_tile.coords.right - cur_tile.coords.left)
-                left_neighbor_tile.overlap.right = cur_tile.overlap.left
+    return calc_overlap(tiles, num_tiles_x, num_tiles_y)
 
-    return tiles
+
+def calc_tiles_min_overlap(
+    image_height: int,
+    image_width: int,
+    tile_height: int,
+    tile_width: int,
+    min_overlap: int = 0,
+) -> list[Tile]:
+    """Calculate the tile coordinates for a given image shape under a simple tiling scheme with overlaps.
+
+    Args:
+        image_height (int): The image height in px.
+        image_width (int): The image width in px.
+        tile_height (int): The tile height in px. All tiles will have this height.
+        tile_width (int): The tile width in px. All tiles will have this width.
+        min_overlap (int): The target minimum overlap between adjacent tiles. If the tiles do not evenly cover the image
+            shape, then the overlap will be spread between the tiles.
+
+    Returns:
+        list[Tile]: A list of tiles that cover the image shape. Ordered from left-to-right, top-to-bottom.
+    """
+
+    assert min_overlap < tile_height
+    assert min_overlap < tile_width
+
+    # catches the cases when the tile size is larger than the images size and adjusts the tile size
+    if image_width < tile_width:
+        tile_width = image_width
+
+    if image_height < tile_height:
+        tile_height = image_height
+
+    num_tiles_x = math.ceil((image_width - min_overlap) / (tile_width - min_overlap))
+    num_tiles_y = math.ceil((image_height - min_overlap) / (tile_height - min_overlap))
+
+    # tiles[y * num_tiles_x + x] is the tile for the y'th row, x'th column.
+    tiles: list[Tile] = []
+
+    # Calculate tile coordinates. (Ignore overlap values for now.)
+    for tile_idx_y in range(num_tiles_y):
+        top = (tile_idx_y * (image_height - tile_height)) // (num_tiles_y - 1) if num_tiles_y > 1 else 0
+        bottom = top + tile_height
+
+        for tile_idx_x in range(num_tiles_x):
+            left = (tile_idx_x * (image_width - tile_width)) // (num_tiles_x - 1) if num_tiles_x > 1 else 0
+            right = left + tile_width
+
+            tile = Tile(
+                coords=TBLR(top=top, bottom=bottom, left=left, right=right),
+                overlap=TBLR(top=0, bottom=0, left=0, right=0),
+            )
+
+            tiles.append(tile)
+
+    return calc_overlap(tiles, num_tiles_x, num_tiles_y)
 
 
 def merge_tiles_with_linear_blending(
@@ -199,3 +332,91 @@ def merge_tiles_with_linear_blending(
             ),
             mask=mask,
         )
+
+
+def merge_tiles_with_seam_blending(
+    dst_image: np.ndarray, tiles: list[Tile], tile_images: list[np.ndarray], blend_amount: int
+):
+    """Merge a set of image tiles into `dst_image` with seam blending between the tiles.
+
+    We expect every tile edge to either:
+    1) have an overlap of 0, because it is aligned with the image edge, or
+    2) have an overlap >= blend_amount.
+    If neither of these conditions are satisfied, we raise an exception.
+
+    The seam blending is centered on a seam of least energy of the overlap between adjacent tiles.
+
+    Args:
+        dst_image (np.ndarray): The destination image. Shape: (H, W, C).
+        tiles (list[Tile]): The list of tiles describing the locations of the respective `tile_images`.
+        tile_images (list[np.ndarray]): The tile images to merge into `dst_image`.
+        blend_amount (int): The amount of blending (in px) between adjacent overlapping tiles.
+    """
+    # Sort tiles and images first by left x coordinate, then by top y coordinate. During tile processing, we want to
+    # iterate over tiles left-to-right, top-to-bottom.
+    tiles_and_images = list(zip(tiles, tile_images, strict=True))
+    tiles_and_images = sorted(tiles_and_images, key=lambda x: x[0].coords.left)
+    tiles_and_images = sorted(tiles_and_images, key=lambda x: x[0].coords.top)
+
+    # Organize tiles into rows.
+    tile_and_image_rows: list[list[tuple[Tile, np.ndarray]]] = []
+    cur_tile_and_image_row: list[tuple[Tile, np.ndarray]] = []
+    first_tile_in_cur_row, _ = tiles_and_images[0]
+    for tile_and_image in tiles_and_images:
+        tile, _ = tile_and_image
+        if not (
+            tile.coords.top == first_tile_in_cur_row.coords.top
+            and tile.coords.bottom == first_tile_in_cur_row.coords.bottom
+        ):
+            # Store the previous row, and start a new one.
+            tile_and_image_rows.append(cur_tile_and_image_row)
+            cur_tile_and_image_row = []
+            first_tile_in_cur_row, _ = tile_and_image
+
+        cur_tile_and_image_row.append(tile_and_image)
+    tile_and_image_rows.append(cur_tile_and_image_row)
+
+    for tile_and_image_row in tile_and_image_rows:
+        first_tile_in_row, _ = tile_and_image_row[0]
+        row_height = first_tile_in_row.coords.bottom - first_tile_in_row.coords.top
+        row_image = np.zeros((row_height, dst_image.shape[1], dst_image.shape[2]), dtype=dst_image.dtype)
+
+        # Blend the tiles in the row horizontally.
+        for tile, tile_image in tile_and_image_row:
+            # We expect the tiles to be ordered left-to-right.
+            # For each tile:
+            # - extract the overlap regions and pass to seam_blend()
+            # - apply blended region to the row_image
+            # - apply the un-blended region to the row_image
+            tile_height, tile_width, _ = tile_image.shape
+            overlap_size = tile.overlap.left
+            # Left blending:
+            if overlap_size > 0:
+                assert overlap_size >= blend_amount
+
+                overlap_coord_right = tile.coords.left + overlap_size
+                src_overlap = row_image[:, tile.coords.left : overlap_coord_right]
+                dst_overlap = tile_image[:, :overlap_size]
+                blended_overlap = seam_blend(src_overlap, dst_overlap, blend_amount, x_seam=False)
+                row_image[:, tile.coords.left : overlap_coord_right] = blended_overlap
+                row_image[:, overlap_coord_right : tile.coords.right] = tile_image[:, overlap_size:]
+            else:
+                # no overlap just paste the tile
+                row_image[:, tile.coords.left : tile.coords.right] = tile_image
+
+        # Blend the row into the dst_image
+        # We assume that the entire row has the same vertical overlaps as the first_tile_in_row.
+        # Rows are processed in the same way as tiles (extract overlap, blend, apply)
+        row_overlap_size = first_tile_in_row.overlap.top
+        if row_overlap_size > 0:
+            assert row_overlap_size >= blend_amount
+
+            overlap_coords_bottom = first_tile_in_row.coords.top + row_overlap_size
+            src_overlap = dst_image[first_tile_in_row.coords.top : overlap_coords_bottom, :]
+            dst_overlap = row_image[:row_overlap_size, :]
+            blended_overlap = seam_blend(src_overlap, dst_overlap, blend_amount, x_seam=True)
+            dst_image[first_tile_in_row.coords.top : overlap_coords_bottom, :] = blended_overlap
+            dst_image[overlap_coords_bottom : first_tile_in_row.coords.bottom, :] = row_image[row_overlap_size:, :]
+        else:
+            # no overlap just paste the row
+            dst_image[first_tile_in_row.coords.top : first_tile_in_row.coords.bottom, :] = row_image

invokeai/backend/tiles/utils.py

@@ -1,5 +1,7 @@
+import math
 from typing import Optional
 
+import cv2
 import numpy as np
 from pydantic import BaseModel, Field
 
@@ -31,10 +33,10 @@ def paste(dst_image: np.ndarray, src_image: np.ndarray, box: TBLR, mask: Optiona
     """Paste a source image into a destination image.
 
     Args:
-        dst_image (torch.Tensor): The destination image to paste into. Shape: (H, W, C).
-        src_image (torch.Tensor): The source image to paste. Shape: (H, W, C). H and W must be compatible with 'box'.
+        dst_image (np.array): The destination image to paste into. Shape: (H, W, C).
+        src_image (np.array): The source image to paste. Shape: (H, W, C). H and W must be compatible with 'box'.
         box (TBLR): Box defining the region in the 'dst_image' where 'src_image' will be pasted.
-        mask (Optional[torch.Tensor]): A mask that defines the blending between 'src_image' and 'dst_image'.
+        mask (Optional[np.array]): A mask that defines the blending between 'src_image' and 'dst_image'.
             Range: [0.0, 1.0], Shape: (H, W). The output is calculate per-pixel according to
             `src * mask + dst * (1 - mask)`.
     """
@@ -45,3 +47,106 @@ def paste(dst_image: np.ndarray, src_image: np.ndarray, box: TBLR, mask: Optiona
         mask = np.expand_dims(mask, -1)
         dst_image_box = dst_image[box.top : box.bottom, box.left : box.right]
         dst_image[box.top : box.bottom, box.left : box.right] = src_image * mask + dst_image_box * (1.0 - mask)
+
+
+def seam_blend(ia1: np.ndarray, ia2: np.ndarray, blend_amount: int, x_seam: bool) -> np.ndarray:
+    """Blend two overlapping tile sections using a seams to find a path.
+
+    It is assumed that input images will be RGB np arrays and are the same size.
+
+    Args:
+        ia1 (np.array): Image array 1 Shape: (H, W, C).
+        ia2 (np.array): Image array 2 Shape: (H, W, C).
+        x_seam (bool): If the images should be blended on the x axis or not.
+        blend_amount (int): The size of the blur to use on the seam. Half of this value will be used to avoid the edges of the image.
+    """
+    assert ia1.shape == ia2.shape
+    assert ia2.size == ia2.size
+
+    def shift(arr, num, fill_value=255.0):
+        result = np.full_like(arr, fill_value)
+        if num > 0:
+            result[num:] = arr[:-num]
+        elif num < 0:
+            result[:num] = arr[-num:]
+        else:
+            result[:] = arr
+        return result
+
+    # Assume RGB and convert to grey
+    # Could offer other options for the luminance conversion
+    # BT.709 [0.2126, 0.7152, 0.0722], BT.2020 [0.2627, 0.6780, 0.0593])
+    # it might not have a huge impact due to the blur that is applied over the seam
+    iag1 = np.dot(ia1, [0.2989, 0.5870, 0.1140])  # BT.601 perceived brightness
+    iag2 = np.dot(ia2, [0.2989, 0.5870, 0.1140])
+
+    # Calc Difference between the images
+    ia = iag2 - iag1
+
+    # If the seam is on the X-axis rotate the array so we can treat it like a vertical seam
+    if x_seam:
+        ia = np.rot90(ia, 1)
+
+    # Calc max and min X & Y limits
+    # gutter is used to avoid the blur hitting the edge of the image
+    gutter = math.ceil(blend_amount / 2) if blend_amount > 0 else 0
+    max_y, max_x = ia.shape
+    max_x -= gutter
+    min_x = gutter
+
+    # Calc the energy in the difference
+    # Could offer different energy calculations e.g. Sobel or Scharr
+    energy = np.abs(np.gradient(ia, axis=0)) + np.abs(np.gradient(ia, axis=1))
+
+    # Find the starting position of the seam
+    res = np.copy(energy)
+    for y in range(1, max_y):
+        row = res[y, :]
+        rowl = shift(row, -1)
+        rowr = shift(row, 1)
+        res[y, :] = res[y - 1, :] + np.min([row, rowl, rowr], axis=0)
+
+    # create an array max_y long
+    lowest_energy_line = np.empty([max_y], dtype="uint16")
+    lowest_energy_line[max_y - 1] = np.argmin(res[max_y - 1, min_x : max_x - 1])
+
+    # Calc the path of the seam
+    # could offer options for larger search than just 1 pixel by adjusting lpos and rpos
+    for ypos in range(max_y - 2, -1, -1):
+        lowest_pos = lowest_energy_line[ypos + 1]
+        lpos = lowest_pos - 1
+        rpos = lowest_pos + 1
+        lpos = np.clip(lpos, min_x, max_x - 1)
+        rpos = np.clip(rpos, min_x, max_x - 1)
+        lowest_energy_line[ypos] = np.argmin(energy[ypos, lpos : rpos + 1]) + lpos
+
+    # Draw the mask
+    mask = np.zeros_like(ia)
+    for ypos in range(0, max_y):
+        to_fill = lowest_energy_line[ypos]
+        mask[ypos, :to_fill] = 1
+
+    # If the seam is on the X-axis rotate the array back
+    if x_seam:
+        mask = np.rot90(mask, 3)
+
+    # blur the seam mask if required
+    if blend_amount > 0:
+        mask = cv2.blur(mask, (blend_amount, blend_amount))
+
+    # for visual debugging
+    # from PIL import Image
+    # m_image = Image.fromarray((mask * 255.0).astype("uint8"))
+
+    # copy ia2 over ia1 while applying the seam mask
+    mask = np.expand_dims(mask, -1)
+    blended_image = ia1 * mask + ia2 * (1.0 - mask)
+
+    # for visual debugging
+    # i1 = Image.fromarray(ia1.astype("uint8"))
+    # i2 = Image.fromarray(ia2.astype("uint8"))
+    # b_image = Image.fromarray(blended_image.astype("uint8"))
+    # print(f"{ia1.shape}, {ia2.shape}, {mask.shape}, {blended_image.shape}")
+    # print(f"{i1.size}, {i2.size}, {m_image.size}, {b_image.size}")
+
+    return blended_image

invokeai/backend/util/__init__.py

@@ -11,4 +11,7 @@ from .devices import (  # noqa: F401
     normalize_device,
     torch_dtype,
 )
+from .logging import InvokeAILogger
 from .util import Chdir, ask_user, download_with_resume, instantiate_from_config, url_attachment_name  # noqa: F401
+
+__all__ = ["Chdir", "InvokeAILogger", "choose_precision", "choose_torch_device"]

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

@@ -950,9 +950,9 @@
         "problemSettingTitle": "Problem Setting Title",
         "reloadNodeTemplates": "Reload Node Templates",
         "removeLinearView": "Remove from Linear View",
-        "resetWorkflow": "Reset Workflow Editor",
-        "resetWorkflowDesc": "Are you sure you want to reset the Workflow Editor?",
-        "resetWorkflowDesc2": "Resetting the Workflow Editor will clear all nodes, edges and workflow details. Saved workflows will not be affected.",
+        "newWorkflow": "New Workflow",
+        "newWorkflowDesc": "Create a new workflow?",
+        "newWorkflowDesc2": "Your current workflow has unsaved changes.",
         "scheduler": "Scheduler",
         "schedulerDescription": "TODO",
         "sDXLMainModelField": "SDXL Model",
@@ -1032,7 +1032,9 @@
         "workflowValidation": "Workflow Validation Error",
         "workflowVersion": "Version",
         "zoomInNodes": "Zoom In",
-        "zoomOutNodes": "Zoom Out"
+        "zoomOutNodes": "Zoom Out",
+        "betaDesc": "This invocation is in beta. Until it is stable, it may have breaking changes during app updates. We plan to support this invocation long-term.",
+        "prototypeDesc": "This invocation is a prototype. It may have breaking changes during app updates and may be removed at any time."
     },
     "parameters": {
         "aspectRatio": "Aspect Ratio",
@@ -1632,10 +1634,10 @@
         "userWorkflows": "My Workflows",
         "defaultWorkflows": "Default Workflows",
         "openWorkflow": "Open Workflow",
-        "uploadWorkflow": "Upload Workflow",
+        "uploadWorkflow": "Load from File",
         "deleteWorkflow": "Delete Workflow",
         "unnamedWorkflow": "Unnamed Workflow",
-        "downloadWorkflow": "Download Workflow",
+        "downloadWorkflow": "Save to File",
         "saveWorkflow": "Save Workflow",
         "saveWorkflowAs": "Save Workflow As",
         "savingWorkflow": "Saving Workflow...",
@@ -1650,7 +1652,7 @@
         "searchWorkflows": "Search Workflows",
         "clearWorkflowSearchFilter": "Clear Workflow Search Filter",
         "workflowName": "Workflow Name",
-        "workflowEditorReset": "Workflow Editor Reset",
+        "newWorkflowCreated": "New Workflow Created",
         "workflowEditorMenu": "Workflow Editor Menu",
         "workflowIsOpen": "Workflow is Open"
     },

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

@@ -104,7 +104,16 @@
         "copyError": "$t(gallery.copy) Errore",
         "input": "Ingresso",
         "notInstalled": "Non $t(common.installed)",
-        "unknownError": "Errore sconosciuto"
+        "unknownError": "Errore sconosciuto",
+        "updated": "Aggiornato",
+        "save": "Salva",
+        "created": "Creato",
+        "prevPage": "Pagina precedente",
+        "delete": "Elimina",
+        "orderBy": "Ordinato per",
+        "nextPage": "Pagina successiva",
+        "saveAs": "Salva come",
+        "unsaved": "Non salvato"
     },
     "gallery": {
         "generations": "Generazioni",
@@ -763,7 +772,10 @@
         "setIPAdapterImage": "Imposta come immagine per l'Adattatore IP",
         "problemSavingMaskDesc": "Impossibile salvare la maschera",
         "setAsCanvasInitialImage": "Imposta come immagine iniziale della tela",
-        "invalidUpload": "Caricamento non valido"
+        "invalidUpload": "Caricamento non valido",
+        "problemDeletingWorkflow": "Problema durante l'eliminazione del flusso di lavoro",
+        "workflowDeleted": "Flusso di lavoro eliminato",
+        "problemRetrievingWorkflow": "Problema nel recupero del flusso di lavoro"
     },
     "tooltip": {
         "feature": {
@@ -886,11 +898,11 @@
         "zoomInNodes": "Ingrandire",
         "fitViewportNodes": "Adatta vista",
         "showGraphNodes": "Mostra sovrapposizione grafico",
-        "resetWorkflowDesc2": "Reimpostare il flusso di lavoro cancellerà tutti i nodi, i bordi e i dettagli del flusso di lavoro.",
+        "resetWorkflowDesc2": "Il ripristino dell'editor del flusso di lavoro cancellerà tutti i nodi, le connessioni e i dettagli del flusso di lavoro. I flussi di lavoro salvati non saranno interessati.",
         "reloadNodeTemplates": "Ricarica i modelli di nodo",
         "loadWorkflow": "Importa flusso di lavoro JSON",
-        "resetWorkflow": "Reimposta flusso di lavoro",
-        "resetWorkflowDesc": "Sei sicuro di voler reimpostare questo flusso di lavoro?",
+        "resetWorkflow": "Reimposta l'editor del flusso di lavoro",
+        "resetWorkflowDesc": "Sei sicuro di voler reimpostare l'editor del flusso di lavoro?",
         "downloadWorkflow": "Esporta flusso di lavoro JSON",
         "scheduler": "Campionatore",
         "addNode": "Aggiungi nodo",
@@ -1080,25 +1092,27 @@
         "collectionOrScalarFieldType": "{{name}} Raccolta|Scalare",
         "nodeVersion": "Versione Nodo",
         "inputFieldTypeParseError": "Impossibile analizzare il tipo di campo di input {{node}}.{{field}} ({{message}})",
-        "unsupportedArrayItemType": "tipo di elemento dell'array non supportato \"{{type}}\"",
+        "unsupportedArrayItemType": "Tipo di elemento dell'array non supportato \"{{type}}\"",
         "targetNodeFieldDoesNotExist": "Connessione non valida: il campo di destinazione/input {{node}}.{{field}} non esiste",
         "unsupportedMismatchedUnion": "tipo CollectionOrScalar non corrispondente con tipi di base {{firstType}} e {{secondType}}",
         "allNodesUpdated": "Tutti i nodi sono aggiornati",
         "sourceNodeDoesNotExist": "Connessione non valida: il nodo di origine/output {{node}} non esiste",
-        "unableToExtractEnumOptions": "impossibile estrarre le opzioni enum",
-        "unableToParseFieldType": "impossibile analizzare il tipo di campo",
+        "unableToExtractEnumOptions": "Impossibile estrarre le opzioni enum",
+        "unableToParseFieldType": "Impossibile analizzare il tipo di campo",
         "unrecognizedWorkflowVersion": "Versione dello schema del flusso di lavoro non riconosciuta {{version}}",
         "outputFieldTypeParseError": "Impossibile analizzare il tipo di campo di output {{node}}.{{field}} ({{message}})",
         "sourceNodeFieldDoesNotExist": "Connessione non valida: il campo di origine/output {{node}}.{{field}} non esiste",
         "unableToGetWorkflowVersion": "Impossibile ottenere la versione dello schema del flusso di lavoro",
         "nodePack": "Pacchetto di nodi",
-        "unableToExtractSchemaNameFromRef": "impossibile estrarre il nome dello schema dal riferimento",
+        "unableToExtractSchemaNameFromRef": "Impossibile estrarre il nome dello schema dal riferimento",
         "unknownOutput": "Output sconosciuto: {{name}}",
         "unknownNodeType": "Tipo di nodo sconosciuto",
         "targetNodeDoesNotExist": "Connessione non valida: il nodo di destinazione/input {{node}} non esiste",
         "unknownFieldType": "$t(nodes.unknownField) tipo: {{type}}",
         "deletedInvalidEdge": "Eliminata connessione non valida {{source}} -> {{target}}",
-        "unknownInput": "Input sconosciuto: {{name}}"
+        "unknownInput": "Input sconosciuto: {{name}}",
+        "prototypeDesc": "Questa invocazione è un prototipo. Potrebbe subire modifiche sostanziali durante gli aggiornamenti dell'app e potrebbe essere rimossa in qualsiasi momento.",
+        "betaDesc": "Questa invocazione è in versione beta. Fino a quando non sarà stabile, potrebbe subire modifiche importanti durante gli aggiornamenti dell'app. Abbiamo intenzione di supportare questa invocazione a lungo termine."
     },
     "boards": {
         "autoAddBoard": "Aggiungi automaticamente bacheca",
@@ -1594,5 +1608,34 @@
         "hrf": "Correzione Alta Risoluzione",
         "hrfStrength": "Forza della Correzione Alta Risoluzione",
         "strengthTooltip": "Valori più bassi comportano meno dettagli, il che può ridurre potenziali artefatti."
+    },
+    "workflows": {
+        "saveWorkflowAs": "Salva flusso di lavoro come",
+        "workflowEditorMenu": "Menu dell'editor del flusso di lavoro",
+        "noSystemWorkflows": "Nessun flusso di lavoro del sistema",
+        "workflowName": "Nome del flusso di lavoro",
+        "noUserWorkflows": "Nessun flusso di lavoro utente",
+        "defaultWorkflows": "Flussi di lavoro predefiniti",
+        "saveWorkflow": "Salva flusso di lavoro",
+        "openWorkflow": "Apri flusso di lavoro",
+        "clearWorkflowSearchFilter": "Cancella il filtro di ricerca del flusso di lavoro",
+        "workflowEditorReset": "Reimpostazione dell'editor del flusso di lavoro",
+        "workflowLibrary": "Libreria",
+        "noRecentWorkflows": "Nessun flusso di lavoro recente",
+        "workflowSaved": "Flusso di lavoro salvato",
+        "workflowIsOpen": "Il flusso di lavoro è aperto",
+        "unnamedWorkflow": "Flusso di lavoro senza nome",
+        "savingWorkflow": "Salvataggio del flusso di lavoro...",
+        "problemLoading": "Problema durante il caricamento dei flussi di lavoro",
+        "loading": "Caricamento dei flussi di lavoro",
+        "searchWorkflows": "Cerca flussi di lavoro",
+        "problemSavingWorkflow": "Problema durante il salvataggio del flusso di lavoro",
+        "deleteWorkflow": "Elimina flusso di lavoro",
+        "workflows": "Flussi di lavoro",
+        "noDescription": "Nessuna descrizione",
+        "userWorkflows": "I miei flussi di lavoro"
+    },
+    "app": {
+        "storeNotInitialized": "Il negozio non è inizializzato"
     }
 }

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

@@ -109,7 +109,18 @@
         "somethingWentWrong": "出了点问题",
         "copyError": "$t(gallery.copy) 错误",
         "input": "输入",
-        "notInstalled": "非 $t(common.installed)"
+        "notInstalled": "非 $t(common.installed)",
+        "delete": "删除",
+        "updated": "已上传",
+        "save": "保存",
+        "created": "已创建",
+        "prevPage": "上一页",
+        "unknownError": "未知错误",
+        "direction": "指向",
+        "orderBy": "排序方式：",
+        "nextPage": "下一页",
+        "saveAs": "保存为",
+        "unsaved": "未保存"
     },
     "gallery": {
         "generations": "生成的图像",
@@ -145,7 +156,11 @@
         "image": "图像",
         "drop": "弃用",
         "dropOrUpload": "$t(gallery.drop) 或上传",
-        "dropToUpload": "$t(gallery.drop) 以上传"
+        "dropToUpload": "$t(gallery.drop) 以上传",
+        "problemDeletingImagesDesc": "有一张或多张图像无法被删除",
+        "problemDeletingImages": "删除图像时出现问题",
+        "unstarImage": "取消收藏图像",
+        "starImage": "收藏图像"
     },
     "hotkeys": {
         "keyboardShortcuts": "键盘快捷键",
@@ -723,7 +738,7 @@
         "nodesUnrecognizedTypes": "无法加载。节点图有无法识别的节点类型",
         "nodesNotValidJSON": "无效的 JSON",
         "nodesNotValidGraph": "无效的 InvokeAi 节点图",
-        "nodesLoadedFailed": "节点图加载失败",
+        "nodesLoadedFailed": "节点加载失败",
         "modelAddedSimple": "已添加模型",
         "modelAdded": "已添加模型: {{modelName}}",
         "imageSavingFailed": "图像保存失败",
@@ -759,7 +774,10 @@
         "problemImportingMask": "导入遮罩时出现问题",
         "baseModelChangedCleared_other": "基础模型已更改, 已清除或禁用 {{count}} 个不兼容的子模型",
         "setAsCanvasInitialImage": "设为画布初始图像",
-        "invalidUpload": "无效的上传"
+        "invalidUpload": "无效的上传",
+        "problemDeletingWorkflow": "删除工作流时出现问题",
+        "workflowDeleted": "已删除工作流",
+        "problemRetrievingWorkflow": "检索工作流时发生问题"
     },
     "unifiedCanvas": {
         "layer": "图层",
@@ -874,11 +892,11 @@
     },
     "nodes": {
         "zoomInNodes": "放大",
-        "resetWorkflowDesc": "是否确定要清空节点图?",
-        "resetWorkflow": "清空节点图",
-        "loadWorkflow": "读取节点图",
+        "resetWorkflowDesc": "是否确定要重置工作流编辑器?",
+        "resetWorkflow": "重置工作流编辑器",
+        "loadWorkflow": "加载工作流",
         "zoomOutNodes": "缩小",
-        "resetWorkflowDesc2": "重置节点图将清除所有节点、边际和节点图详情.",
+        "resetWorkflowDesc2": "重置工作流编辑器将清除所有节点、边际和节点图详情。不影响已保存的工作流。",
         "reloadNodeTemplates": "重载节点模板",
         "hideGraphNodes": "隐藏节点图信息",
         "fitViewportNodes": "自适应视图",
@@ -887,7 +905,7 @@
         "showLegendNodes": "显示字段类型图例",
         "hideLegendNodes": "隐藏字段类型图例",
         "showGraphNodes": "显示节点图信息",
-        "downloadWorkflow": "下载节点图 JSON",
+        "downloadWorkflow": "下载工作流 JSON",
         "workflowDescription": "简述",
         "versionUnknown": " 未知版本",
         "noNodeSelected": "无选中的节点",
@@ -1102,7 +1120,9 @@
         "collectionOrScalarFieldType": "{{name}} 合集 | 标量",
         "nodeVersion": "节点版本",
         "deletedInvalidEdge": "已删除无效的边缘 {{source}} -> {{target}}",
-        "unknownInput": "未知输入：{{name}}"
+        "unknownInput": "未知输入：{{name}}",
+        "prototypeDesc": "此调用是一个原型 (prototype)。它可能会在本项目更新期间发生破坏性更改，并且随时可能被删除。",
+        "betaDesc": "此调用尚处于测试阶段。在稳定之前，它可能会在项目更新期间发生破坏性更改。本项目计划长期支持这种调用。"
     },
     "controlnet": {
         "resize": "直接缩放",
@@ -1606,5 +1626,36 @@
         "hrf": "高分辨率修复",
         "hrfStrength": "高分辨率修复强度",
         "strengthTooltip": "值越低细节越少，但可以减少部分潜在的伪影。"
+    },
+    "workflows": {
+        "saveWorkflowAs": "保存工作流为",
+        "workflowEditorMenu": "工作流编辑器菜单",
+        "noSystemWorkflows": "无系统工作流",
+        "workflowName": "工作流名称",
+        "noUserWorkflows": "无用户工作流",
+        "defaultWorkflows": "默认工作流",
+        "saveWorkflow": "保存工作流",
+        "openWorkflow": "打开工作流",
+        "clearWorkflowSearchFilter": "清除工作流检索过滤器",
+        "workflowEditorReset": "工作流编辑器重置",
+        "workflowLibrary": "工作流库",
+        "downloadWorkflow": "下载工作流",
+        "noRecentWorkflows": "无最近工作流",
+        "workflowSaved": "已保存工作流",
+        "workflowIsOpen": "工作流已打开",
+        "unnamedWorkflow": "未命名的工作流",
+        "savingWorkflow": "保存工作流中...",
+        "problemLoading": "加载工作流时出现问题",
+        "loading": "加载工作流中",
+        "searchWorkflows": "检索工作流",
+        "problemSavingWorkflow": "保存工作流时出现问题",
+        "deleteWorkflow": "删除工作流",
+        "workflows": "工作流",
+        "noDescription": "无描述",
+        "uploadWorkflow": "上传工作流",
+        "userWorkflows": "我的工作流"
+    },
+    "app": {
+        "storeNotInitialized": "商店尚未初始化"
     }
 }

invokeai/frontend/web/src/features/gallery/components/ImageGrid/GalleryImage.tsx

@@ -3,6 +3,7 @@ import { useStore } from '@nanostores/react';
 import { $customStarUI } from 'app/store/nanostores/customStarUI';
 import { useAppDispatch, useAppSelector } from 'app/store/storeHooks';
 import IAIDndImage from 'common/components/IAIDndImage';
+import IAIDndImageIcon from 'common/components/IAIDndImageIcon';
 import IAIFillSkeleton from 'common/components/IAIFillSkeleton';
 import { imagesToDeleteSelected } from 'features/deleteImageModal/store/slice';
 import {
@@ -10,7 +11,9 @@ import {
   ImageDraggableData,
   TypesafeDraggableData,
 } from 'features/dnd/types';
+import { VirtuosoGalleryContext } from 'features/gallery/components/ImageGrid/types';
 import { useMultiselect } from 'features/gallery/hooks/useMultiselect';
+import { useScrollToVisible } from 'features/gallery/hooks/useScrollToVisible';
 import { MouseEvent, memo, useCallback, useMemo, useState } from 'react';
 import { useTranslation } from 'react-i18next';
 import { FaTrash } from 'react-icons/fa';
@@ -20,15 +23,16 @@ import {
   useStarImagesMutation,
   useUnstarImagesMutation,
 } from 'services/api/endpoints/images';
-import IAIDndImageIcon from 'common/components/IAIDndImageIcon';
 
 interface HoverableImageProps {
   imageName: string;
+  index: number;
+  virtuosoContext: VirtuosoGalleryContext;
 }
 
 const GalleryImage = (props: HoverableImageProps) => {
   const dispatch = useAppDispatch();
-  const { imageName } = props;
+  const { imageName, virtuosoContext } = props;
   const { currentData: imageDTO } = useGetImageDTOQuery(imageName);
   const shift = useAppSelector((state) => state.hotkeys.shift);
   const { t } = useTranslation();
@@ -38,6 +42,13 @@ const GalleryImage = (props: HoverableImageProps) => {
 
   const customStarUi = useStore($customStarUI);
 
+  const imageContainerRef = useScrollToVisible(
+    isSelected,
+    props.index,
+    selectionCount,
+    virtuosoContext
+  );
+
   const handleDelete = useCallback(
     (e: MouseEvent<HTMLButtonElement>) => {
       e.stopPropagation();
@@ -122,6 +133,7 @@ const GalleryImage = (props: HoverableImageProps) => {
       data-testid={`image-${imageDTO.image_name}`}
     >
       <Flex
+        ref={imageContainerRef}
         userSelect="none"
         sx={{
           position: 'relative',

invokeai/frontend/web/src/features/gallery/components/ImageGrid/GalleryImageGrid.tsx

@@ -1,7 +1,10 @@
 import { Box, Flex } from '@chakra-ui/react';
+import { EntityId } from '@reduxjs/toolkit';
 import { useAppSelector } from 'app/store/storeHooks';
 import IAIButton from 'common/components/IAIButton';
 import { IAINoContentFallback } from 'common/components/IAIImageFallback';
+import { VirtuosoGalleryContext } from 'features/gallery/components/ImageGrid/types';
+import { $useNextPrevImageState } from 'features/gallery/hooks/useNextPrevImage';
 import { selectListImagesBaseQueryArgs } from 'features/gallery/store/gallerySelectors';
 import { IMAGE_LIMIT } from 'features/gallery/store/types';
 import {
@@ -11,7 +14,12 @@ import {
 import { memo, useCallback, useEffect, useMemo, useRef, useState } from 'react';
 import { useTranslation } from 'react-i18next';
 import { FaExclamationCircle, FaImage } from 'react-icons/fa';
-import { VirtuosoGrid } from 'react-virtuoso';
+import {
+  ItemContent,
+  ListRange,
+  VirtuosoGrid,
+  VirtuosoGridHandle,
+} from 'react-virtuoso';
 import {
   useLazyListImagesQuery,
   useListImagesQuery,
@@ -20,7 +28,6 @@ import { useBoardTotal } from 'services/api/hooks/useBoardTotal';
 import GalleryImage from './GalleryImage';
 import ImageGridItemContainer from './ImageGridItemContainer';
 import ImageGridListContainer from './ImageGridListContainer';
-import { EntityId } from '@reduxjs/toolkit';
 
 const overlayScrollbarsConfig: UseOverlayScrollbarsParams = {
   defer: true,
@@ -48,6 +55,10 @@ const GalleryImageGrid = () => {
   const { currentViewTotal } = useBoardTotal(selectedBoardId);
   const queryArgs = useAppSelector(selectListImagesBaseQueryArgs);
 
+  const virtuosoRangeRef = useRef<ListRange | null>(null);
+
+  const virtuosoRef = useRef<VirtuosoGridHandle>(null);
+
   const { currentData, isFetching, isSuccess, isError } =
     useListImagesQuery(queryArgs);
 
@@ -72,12 +83,26 @@ const GalleryImageGrid = () => {
     });
   }, [areMoreAvailable, listImages, queryArgs, currentData?.ids.length]);
 
-  const itemContentFunc = useCallback(
-    (index: number, imageName: EntityId) => (
-      <GalleryImage key={imageName} imageName={imageName as string} />
-    ),
-    []
-  );
+  const virtuosoContext = useMemo<VirtuosoGalleryContext>(() => {
+    return {
+      virtuosoRef,
+      rootRef,
+      virtuosoRangeRef,
+    };
+  }, []);
+
+  const itemContentFunc: ItemContent<EntityId, VirtuosoGalleryContext> =
+    useCallback(
+      (index, imageName, virtuosoContext) => (
+        <GalleryImage
+          key={imageName}
+          index={index}
+          imageName={imageName as string}
+          virtuosoContext={virtuosoContext}
+        />
+      ),
+      []
+    );
 
   useEffect(() => {
     // Initialize the gallery's custom scrollbar
@@ -93,6 +118,15 @@ const GalleryImageGrid = () => {
     return () => osInstance()?.destroy();
   }, [scroller, initialize, osInstance]);
 
+  const onRangeChanged = useCallback((range: ListRange) => {
+    virtuosoRangeRef.current = range;
+  }, []);
+
+  useEffect(() => {
+    $useNextPrevImageState.setKey('virtuosoRef', virtuosoRef);
+    $useNextPrevImageState.setKey('virtuosoRangeRef', virtuosoRangeRef);
+  }, []);
+
   if (!currentData) {
     return (
       <Flex
@@ -140,6 +174,10 @@ const GalleryImageGrid = () => {
             }}
             scrollerRef={setScroller}
             itemContent={itemContentFunc}
+            ref={virtuosoRef}
+            rangeChanged={onRangeChanged}
+            context={virtuosoContext}
+            overscan={10}
           />
         </Box>
         <IAIButton

invokeai/frontend/web/src/features/gallery/components/ImageGrid/types.ts

@@ -0,0 +1,8 @@
+import { RefObject } from 'react';
+import { ListRange, VirtuosoGridHandle } from 'react-virtuoso';
+
+export type VirtuosoGalleryContext = {
+  virtuosoRef: RefObject<VirtuosoGridHandle>;
+  rootRef: RefObject<HTMLDivElement>;
+  virtuosoRangeRef: RefObject<ListRange>;
+};

invokeai/frontend/web/src/features/gallery/components/ImageMetadataViewer/ImageMetadataWorkflowTabContent.tsx

@@ -1,7 +1,7 @@
 import { IAINoContentFallback } from 'common/components/IAIImageFallback';
 import { memo } from 'react';
 import { useTranslation } from 'react-i18next';
-import { useGetImageWorkflowQuery } from 'services/api/endpoints/images';
+import { useDebouncedImageWorkflow } from 'services/api/hooks/useDebouncedImageWorkflow';
 import { ImageDTO } from 'services/api/types';
 import DataViewer from './DataViewer';
 
@@ -11,7 +11,7 @@ type Props = {
 
 const ImageMetadataWorkflowTabContent = ({ image }: Props) => {
   const { t } = useTranslation();
-  const { currentData: workflow } = useGetImageWorkflowQuery(image.image_name);
+  const { workflow } = useDebouncedImageWorkflow(image);
 
   if (!workflow) {
     return <IAINoContentFallback label={t('nodes.noWorkflow')} />;

invokeai/frontend/web/src/features/gallery/hooks/useNextPrevImage.ts

@@ -4,8 +4,11 @@ import { useAppDispatch, useAppSelector } from 'app/store/storeHooks';
 import { selectListImagesBaseQueryArgs } from 'features/gallery/store/gallerySelectors';
 import { imageSelected } from 'features/gallery/store/gallerySlice';
 import { IMAGE_LIMIT } from 'features/gallery/store/types';
+import { getScrollToIndexAlign } from 'features/gallery/util/getScrollToIndexAlign';
 import { clamp } from 'lodash-es';
-import { useCallback } from 'react';
+import { map } from 'nanostores';
+import { RefObject, useCallback } from 'react';
+import { ListRange, VirtuosoGridHandle } from 'react-virtuoso';
 import { boardsApi } from 'services/api/endpoints/boards';
 import {
   imagesApi,
@@ -14,6 +17,16 @@ import {
 import { ListImagesArgs } from 'services/api/types';
 import { imagesAdapter } from 'services/api/util';
 
+export type UseNextPrevImageState = {
+  virtuosoRef: RefObject<VirtuosoGridHandle> | undefined;
+  virtuosoRangeRef: RefObject<ListRange> | undefined;
+};
+
+export const $useNextPrevImageState = map<UseNextPrevImageState>({
+  virtuosoRef: undefined,
+  virtuosoRangeRef: undefined,
+});
+
 export const nextPrevImageButtonsSelector = createMemoizedSelector(
   [stateSelector, selectListImagesBaseQueryArgs],
   (state, baseQueryArgs) => {
@@ -78,6 +91,8 @@ export const nextPrevImageButtonsSelector = createMemoizedSelector(
       isFetching: status === 'pending',
       nextImage,
       prevImage,
+      nextImageIndex,
+      prevImageIndex,
       queryArgs,
     };
   }
@@ -88,7 +103,9 @@ export const useNextPrevImage = () => {
 
   const {
     nextImage,
+    nextImageIndex,
     prevImage,
+    prevImageIndex,
     areMoreImagesAvailable,
     isFetching,
     queryArgs,
@@ -98,11 +115,43 @@ export const useNextPrevImage = () => {
 
   const handlePrevImage = useCallback(() => {
     prevImage && dispatch(imageSelected(prevImage));
-  }, [dispatch, prevImage]);
+    const range = $useNextPrevImageState.get().virtuosoRangeRef?.current;
+    const virtuoso = $useNextPrevImageState.get().virtuosoRef?.current;
+    if (!range || !virtuoso) {
+      return;
+    }
+
+    if (
+      prevImageIndex !== undefined &&
+      (prevImageIndex < range.startIndex || prevImageIndex > range.endIndex)
+    ) {
+      virtuoso.scrollToIndex({
+        index: prevImageIndex,
+        behavior: 'smooth',
+        align: getScrollToIndexAlign(prevImageIndex, range),
+      });
+    }
+  }, [dispatch, prevImage, prevImageIndex]);
 
   const handleNextImage = useCallback(() => {
     nextImage && dispatch(imageSelected(nextImage));
-  }, [dispatch, nextImage]);
+    const range = $useNextPrevImageState.get().virtuosoRangeRef?.current;
+    const virtuoso = $useNextPrevImageState.get().virtuosoRef?.current;
+    if (!range || !virtuoso) {
+      return;
+    }
+
+    if (
+      nextImageIndex !== undefined &&
+      (nextImageIndex < range.startIndex || nextImageIndex > range.endIndex)
+    ) {
+      virtuoso.scrollToIndex({
+        index: nextImageIndex,
+        behavior: 'smooth',
+        align: getScrollToIndexAlign(nextImageIndex, range),
+      });
+    }
+  }, [dispatch, nextImage, nextImageIndex]);
 
   const [listImages] = useLazyListImagesQuery();
 

invokeai/frontend/web/src/features/gallery/hooks/useScrollToVisible.ts

@@ -0,0 +1,46 @@
+import { VirtuosoGalleryContext } from 'features/gallery/components/ImageGrid/types';
+import { getScrollToIndexAlign } from 'features/gallery/util/getScrollToIndexAlign';
+import { useEffect, useRef } from 'react';
+
+export const useScrollToVisible = (
+  isSelected: boolean,
+  index: number,
+  selectionCount: number,
+  virtuosoContext: VirtuosoGalleryContext
+) => {
+  const imageContainerRef = useRef<HTMLDivElement>(null);
+
+  useEffect(() => {
+    if (
+      !isSelected ||
+      selectionCount !== 1 ||
+      !virtuosoContext.rootRef.current ||
+      !virtuosoContext.virtuosoRef.current ||
+      !virtuosoContext.virtuosoRangeRef.current ||
+      !imageContainerRef.current
+    ) {
+      return;
+    }
+
+    const itemRect = imageContainerRef.current.getBoundingClientRect();
+    const rootRect = virtuosoContext.rootRef.current.getBoundingClientRect();
+    const itemIsVisible =
+      itemRect.top >= rootRect.top &&
+      itemRect.bottom <= rootRect.bottom &&
+      itemRect.left >= rootRect.left &&
+      itemRect.right <= rootRect.right;
+
+    if (!itemIsVisible) {
+      virtuosoContext.virtuosoRef.current.scrollToIndex({
+        index,
+        behavior: 'smooth',
+        align: getScrollToIndexAlign(
+          index,
+          virtuosoContext.virtuosoRangeRef.current
+        ),
+      });
+    }
+  }, [isSelected, index, selectionCount, virtuosoContext]);
+
+  return imageContainerRef;
+};

invokeai/frontend/web/src/features/gallery/util/getScrollToIndexAlign.ts

@@ -0,0 +1,17 @@
+import { ListRange } from 'react-virtuoso';
+
+/**
+ * Gets the alignment for react-virtuoso's scrollToIndex function.
+ * @param index The index of the item.
+ * @param range The range of items currently visible.
+ * @returns
+ */
+export const getScrollToIndexAlign = (
+  index: number,
+  range: ListRange
+): 'start' | 'end' => {
+  if (index > (range.endIndex - range.startIndex) / 2 + range.startIndex) {
+    return 'end';
+  }
+  return 'start';
+};

invokeai/frontend/web/src/features/nodes/components/flow/nodes/Invocation/InvocationNodeClassificationIcon.tsx

@@ -0,0 +1,68 @@
+import { Icon, Tooltip } from '@chakra-ui/react';
+import { memo } from 'react';
+import { useTranslation } from 'react-i18next';
+import { FaFlask } from 'react-icons/fa';
+import { useNodeClassification } from 'features/nodes/hooks/useNodeClassification';
+import { Classification } from 'features/nodes/types/common';
+import { FaHammer } from 'react-icons/fa6';
+
+interface Props {
+  nodeId: string;
+}
+
+const InvocationNodeClassificationIcon = ({ nodeId }: Props) => {
+  const classification = useNodeClassification(nodeId);
+
+  if (!classification || classification === 'stable') {
+    return null;
+  }
+
+  return (
+    <Tooltip
+      label={<ClassificationTooltipContent classification={classification} />}
+      placement="top"
+      shouldWrapChildren
+    >
+      <Icon
+        as={getIcon(classification)}
+        sx={{
+          display: 'block',
+          boxSize: 4,
+          color: 'base.400',
+        }}
+      />
+    </Tooltip>
+  );
+};
+
+export default memo(InvocationNodeClassificationIcon);
+
+const ClassificationTooltipContent = memo(
+  ({ classification }: { classification: Classification }) => {
+    const { t } = useTranslation();
+
+    if (classification === 'beta') {
+      return t('nodes.betaDesc');
+    }
+
+    if (classification === 'prototype') {
+      return t('nodes.prototypeDesc');
+    }
+
+    return null;
+  }
+);
+
+ClassificationTooltipContent.displayName = 'ClassificationTooltipContent';
+
+const getIcon = (classification: Classification) => {
+  if (classification === 'beta') {
+    return FaHammer;
+  }
+
+  if (classification === 'prototype') {
+    return FaFlask;
+  }
+
+  return undefined;
+};

invokeai/frontend/web/src/features/nodes/components/flow/nodes/Invocation/InvocationNodeHeader.tsx

@@ -5,6 +5,7 @@ import NodeTitle from 'features/nodes/components/flow/nodes/common/NodeTitle';
 import InvocationNodeCollapsedHandles from './InvocationNodeCollapsedHandles';
 import InvocationNodeInfoIcon from './InvocationNodeInfoIcon';
 import InvocationNodeStatusIndicator from './InvocationNodeStatusIndicator';
+import InvocationNodeClassificationIcon from 'features/nodes/components/flow/nodes/Invocation/InvocationNodeClassificationIcon';
 
 type Props = {
   nodeId: string;
@@ -31,6 +32,7 @@ const InvocationNodeHeader = ({ nodeId, isOpen }: Props) => {
       }}
     >
       <NodeCollapseButton nodeId={nodeId} isOpen={isOpen} />
+      <InvocationNodeClassificationIcon nodeId={nodeId} />
       <NodeTitle nodeId={nodeId} />
       <Flex alignItems="center">
         <InvocationNodeStatusIndicator nodeId={nodeId} />

invokeai/frontend/web/src/features/nodes/hooks/useNodeClassification.ts

@@ -0,0 +1,23 @@
+import { createMemoizedSelector } from 'app/store/createMemoizedSelector';
+import { stateSelector } from 'app/store/store';
+import { useAppSelector } from 'app/store/storeHooks';
+import { isInvocationNode } from 'features/nodes/types/invocation';
+import { useMemo } from 'react';
+
+export const useNodeClassification = (nodeId: string) => {
+  const selector = useMemo(
+    () =>
+      createMemoizedSelector(stateSelector, ({ nodes }) => {
+        const node = nodes.nodes.find((node) => node.id === nodeId);
+        if (!isInvocationNode(node)) {
+          return false;
+        }
+        const nodeTemplate = nodes.nodeTemplates[node?.data.type ?? ''];
+        return nodeTemplate?.classification;
+      }),
+    [nodeId]
+  );
+
+  const title = useAppSelector(selector);
+  return title;
+};

invokeai/frontend/web/src/features/nodes/types/common.ts

@@ -19,6 +19,9 @@ export const zColorField = z.object({
 });
 export type ColorField = z.infer<typeof zColorField>;
 
+export const zClassification = z.enum(['stable', 'beta', 'prototype']);
+export type Classification = z.infer<typeof zClassification>;
+
 export const zSchedulerField = z.enum([
   'euler',
   'deis',

invokeai/frontend/web/src/features/nodes/types/invocation.ts

@@ -1,6 +1,6 @@
 import { Edge, Node } from 'reactflow';
 import { z } from 'zod';
-import { zProgressImage } from './common';
+import { zClassification, zProgressImage } from './common';
 import {
   zFieldInputInstance,
   zFieldInputTemplate,
@@ -21,6 +21,7 @@ export const zInvocationTemplate = z.object({
   version: zSemVer,
   useCache: z.boolean(),
   nodePack: z.string().min(1).nullish(),
+  classification: zClassification,
 });
 export type InvocationTemplate = z.infer<typeof zInvocationTemplate>;
 // #endregion

invokeai/frontend/web/src/features/nodes/util/schema/parseSchema.ts

@@ -83,6 +83,7 @@ export const parseSchema = (
     const description = schema.description ?? '';
     const version = schema.version;
     const nodePack = schema.node_pack;
+    const classification = schema.classification;
 
     const inputs = reduce(
       schema.properties,
@@ -245,6 +246,7 @@ export const parseSchema = (
       outputs,
       useCache,
       nodePack,
+      classification,
     };
 
     Object.assign(invocationsAccumulator, { [type]: invocation });

invokeai/frontend/web/src/features/nodes/util/workflow/migrations.ts

@@ -64,7 +64,10 @@ const migrateV1toV2 = (workflowToMigrate: WorkflowV1): WorkflowV2 => {
       const nodePack = invocationTemplate
         ? invocationTemplate.nodePack
         : t('common.unknown');
+
       (node.data as unknown as InvocationNodeData).nodePack = nodePack;
+      // Fallback to 1.0.0 if not specified - this matches the behavior of the backend
+      node.data.version ||= '1.0.0';
     }
   });
   // Bump version

invokeai/frontend/web/src/features/workflowLibrary/components/WorkflowLibraryMenu/NewWorkflowMenuItem.tsx

@@ -11,44 +11,48 @@ import {
   Text,
   useDisclosure,
 } from '@chakra-ui/react';
-import { useAppDispatch } from 'app/store/storeHooks';
+import { useAppDispatch, useAppSelector } from 'app/store/storeHooks';
 import { nodeEditorReset } from 'features/nodes/store/nodesSlice';
 import { addToast } from 'features/system/store/systemSlice';
 import { makeToast } from 'features/system/util/makeToast';
 import { memo, useCallback, useRef } from 'react';
 import { useTranslation } from 'react-i18next';
-import { FaTrash } from 'react-icons/fa';
+import { FaCircleNodes } from 'react-icons/fa6';
 
-const ResetWorkflowEditorMenuItem = () => {
+const NewWorkflowMenuItem = () => {
   const { t } = useTranslation();
   const dispatch = useAppDispatch();
   const { isOpen, onOpen, onClose } = useDisclosure();
   const cancelRef = useRef<HTMLButtonElement | null>(null);
+  const isTouched = useAppSelector((state) => state.workflow.isTouched);
 
-  const handleConfirmClear = useCallback(() => {
+  const handleNewWorkflow = useCallback(() => {
     dispatch(nodeEditorReset());
 
     dispatch(
       addToast(
         makeToast({
-          title: t('workflows.workflowEditorReset'),
+          title: t('workflows.newWorkflowCreated'),
           status: 'success',
         })
       )
     );
 
     onClose();
-  }, [dispatch, t, onClose]);
+  }, [dispatch, onClose, t]);
+
+  const onClick = useCallback(() => {
+    if (!isTouched) {
+      handleNewWorkflow();
+      return;
+    }
+    onOpen();
+  }, [handleNewWorkflow, isTouched, onOpen]);
 
   return (
     <>
-      <MenuItem
-        as="button"
-        icon={<FaTrash />}
-        sx={{ color: 'error.600', _dark: { color: 'error.300' } }}
-        onClick={onOpen}
-      >
-        {t('nodes.resetWorkflow')}
+      <MenuItem as="button" icon={<FaCircleNodes />} onClick={onClick}>
+        {t('nodes.newWorkflow')}
       </MenuItem>
 
       <AlertDialog
@@ -61,13 +65,13 @@ const ResetWorkflowEditorMenuItem = () => {
 
         <AlertDialogContent>
           <AlertDialogHeader fontSize="lg" fontWeight="bold">
-            {t('nodes.resetWorkflow')}
+            {t('nodes.newWorkflow')}
           </AlertDialogHeader>
 
           <AlertDialogBody py={4}>
             <Flex flexDir="column" gap={2}>
-              <Text>{t('nodes.resetWorkflowDesc')}</Text>
-              <Text variant="subtext">{t('nodes.resetWorkflowDesc2')}</Text>
+              <Text>{t('nodes.newWorkflowDesc')}</Text>
+              <Text variant="subtext">{t('nodes.newWorkflowDesc2')}</Text>
             </Flex>
           </AlertDialogBody>
 
@@ -75,7 +79,7 @@ const ResetWorkflowEditorMenuItem = () => {
             <Button ref={cancelRef} onClick={onClose}>
               {t('common.cancel')}
             </Button>
-            <Button colorScheme="error" ml={3} onClick={handleConfirmClear}>
+            <Button colorScheme="error" ml={3} onClick={handleNewWorkflow}>
               {t('common.accept')}
             </Button>
           </AlertDialogFooter>
@@ -85,4 +89,4 @@ const ResetWorkflowEditorMenuItem = () => {
   );
 };
 
-export default memo(ResetWorkflowEditorMenuItem);
+export default memo(NewWorkflowMenuItem);

invokeai/frontend/web/src/features/workflowLibrary/components/WorkflowLibraryMenu/WorkflowLibraryMenu.tsx

@@ -9,7 +9,7 @@ import IAIIconButton from 'common/components/IAIIconButton';
 import { useGlobalMenuCloseTrigger } from 'common/hooks/useGlobalMenuCloseTrigger';
 import { useFeatureStatus } from 'features/system/hooks/useFeatureStatus';
 import DownloadWorkflowMenuItem from 'features/workflowLibrary/components/WorkflowLibraryMenu/DownloadWorkflowMenuItem';
-import ResetWorkflowEditorMenuItem from 'features/workflowLibrary/components/WorkflowLibraryMenu/ResetWorkflowEditorMenuItem';
+import NewWorkflowMenuItem from 'features/workflowLibrary/components/WorkflowLibraryMenu/NewWorkflowMenuItem';
 import SaveWorkflowAsMenuItem from 'features/workflowLibrary/components/WorkflowLibraryMenu/SaveWorkflowAsMenuItem';
 import SaveWorkflowMenuItem from 'features/workflowLibrary/components/WorkflowLibraryMenu/SaveWorkflowMenuItem';
 import SettingsMenuItem from 'features/workflowLibrary/components/WorkflowLibraryMenu/SettingsMenuItem';
@@ -39,7 +39,7 @@ const WorkflowLibraryMenu = () => {
         {isWorkflowLibraryEnabled && <SaveWorkflowAsMenuItem />}
         <DownloadWorkflowMenuItem />
         <UploadWorkflowMenuItem />
-        <ResetWorkflowEditorMenuItem />
+        <NewWorkflowMenuItem />
         <MenuDivider />
         <SettingsMenuItem />
       </MenuList>

invokeai/frontend/web/src/services/api/hooks/useDebouncedImageWorkflow.ts

@@ -0,0 +1,22 @@
+import { skipToken } from '@reduxjs/toolkit/query';
+import { useAppSelector } from 'app/store/storeHooks';
+import { useGetImageWorkflowQuery } from 'services/api/endpoints/images';
+import { ImageDTO } from 'services/api/types';
+import { useDebounce } from 'use-debounce';
+
+export const useDebouncedImageWorkflow = (imageDTO?: ImageDTO | null) => {
+  const workflowFetchDebounce = useAppSelector(
+    (state) => state.config.workflowFetchDebounce ?? 300
+  );
+
+  const [debouncedImageName] = useDebounce(
+    imageDTO?.has_workflow ? imageDTO.image_name : null,
+    workflowFetchDebounce
+  );
+
+  const { data: workflow, isLoading } = useGetImageWorkflowQuery(
+    debouncedImageName ?? skipToken
+  );
+
+  return { workflow, isLoading };
+};

invokeai/frontend/web/src/services/api/hooks/useDebouncedMetadata.ts

@@ -1,17 +1,14 @@
 import { skipToken } from '@reduxjs/toolkit/query';
-import { useDebounce } from 'use-debounce';
-import { useGetImageMetadataQuery } from 'services/api/endpoints/images';
 import { useAppSelector } from 'app/store/storeHooks';
+import { useGetImageMetadataQuery } from 'services/api/endpoints/images';
+import { useDebounce } from 'use-debounce';
 
 export const useDebouncedMetadata = (imageName?: string | null) => {
   const metadataFetchDebounce = useAppSelector(
-    (state) => state.config.metadataFetchDebounce
+    (state) => state.config.metadataFetchDebounce ?? 300
   );
 
-  const [debouncedImageName] = useDebounce(
-    imageName,
-    metadataFetchDebounce ?? 0
-  );
+  const [debouncedImageName] = useDebounce(imageName, metadataFetchDebounce);
 
   const { data: metadata, isLoading } = useGetImageMetadataQuery(
     debouncedImageName ?? skipToken

mkdocs.yml

@@ -101,6 +101,8 @@ plugins:
 extra_javascript:
   - https://unpkg.com/tablesort@5.3.0/dist/tablesort.min.js
   - javascripts/tablesort.js
+  - https://widget.kapa.ai/kapa-widget.bundle.js
+  - javascript/init_kapa_widget.js
 
 extra:
   analytics:
@@ -164,6 +166,7 @@ nav:
         - Overview: 'contributing/contribution_guides/development.md'
         - New Contributors: 'contributing/contribution_guides/newContributorChecklist.md'
         - InvokeAI Architecture: 'contributing/ARCHITECTURE.md'
+        - Model Manager v2: 'contributing/MODEL_MANAGER.md'
         - Frontend Documentation: 'contributing/contribution_guides/contributingToFrontend.md'
         - Local Development: 'contributing/LOCAL_DEVELOPMENT.md'
         - Adding Tests: 'contributing/TESTS.md'

pyproject.toml

@@ -32,7 +32,7 @@ classifiers = [
   'Topic :: Scientific/Engineering :: Image Processing',
 ]
 dependencies = [
-  "accelerate~=0.24.0",
+  "accelerate~=0.25.0",
   "albumentations",
   "basicsr",
   "click",
@@ -41,15 +41,15 @@ dependencies = [
   "controlnet-aux>=0.0.6",
   "timm==0.6.13",               # needed to override timm latest in controlnet_aux, see  https://github.com/isl-org/ZoeDepth/issues/26
   "datasets",
-  "diffusers[torch]~=0.23.0",
+  "diffusers[torch]~=0.24.0",
   "dnspython~=2.4.0",
   "dynamicprompts",
   "easing-functions",
   "einops",
   "facexlib",
-  "fastapi~=0.104.1",
+  "fastapi~=0.105.0",
   "fastapi-events~=0.9.1",
-  "huggingface-hub~=0.16.4",
+  "huggingface-hub~=0.19.4",
   "imohash",
   "invisible-watermark~=0.2.0", # needed to install SDXL base and refiner using their repo_ids
   "matplotlib",                 # needed for plotting of Penner easing functions
@@ -80,11 +80,11 @@ dependencies = [
   "semver~=3.0.1",
   "send2trash",
   "test-tube~=0.7.5",
-  "torch==2.1.0",
-  "torchvision==0.16.0",
+  "torch==2.1.1",
+  "torchvision==0.16.1",
   "torchmetrics~=0.11.0",
   "torchsde~=0.2.5",
-  "transformers~=4.35.0",
+  "transformers~=4.36.0",
   "uvicorn[standard]~=0.21.1",
   "windows-curses; sys_platform=='win32'",
 ]
@@ -107,7 +107,7 @@ dependencies = [
   "pytest-datadir",
 ]
 "xformers" = [
-  "xformers==0.0.22post7; sys_platform!='darwin'",
+  "xformers==0.0.23; sys_platform!='darwin'",
   "triton; sys_platform=='linux'",
 ]
 "onnx" = ["onnxruntime"]
@@ -221,6 +221,8 @@ exclude = [
 # global mypy config
 [tool.mypy]
 ignore_missing_imports = true # ignores missing types in third-party libraries
+strict = true
+exclude = ["tests/*"]
 
 # overrides for specific modules
 [[tool.mypy.overrides]]

scripts/probe-model.py

@@ -1,9 +1,11 @@
 #!/bin/env python
 
+"""Little command-line utility for probing a model on disk."""
+
 import argparse
 from pathlib import Path
 
-from invokeai.backend.model_management.model_probe import ModelProbe
+from invokeai.backend.model_manager import InvalidModelConfigException, ModelProbe
 
 parser = argparse.ArgumentParser(description="Probe model type")
 parser.add_argument(
@@ -14,5 +16,8 @@ parser.add_argument(
 args = parser.parse_args()
 
 for path in args.model_path:
-    info = ModelProbe().probe(path)
-    print(f"{path}: {info}")
+    try:
+        info = ModelProbe.probe(path)
+        print(f"{path}:{info.model_dump_json(indent=4)}")
+    except InvalidModelConfigException as exc:
+        print(exc)

tests/aa_nodes/test_graph_execution_state.py

@@ -28,8 +28,8 @@ from invokeai.app.services.shared.graph import (
     IterateInvocation,
     LibraryGraph,
 )
-from invokeai.app.services.shared.sqlite.sqlite_database import SqliteDatabase
 from invokeai.backend.util.logging import InvokeAILogger
+from tests.fixtures.sqlite_database import create_mock_sqlite_database
 
 from .test_invoker import create_edge
 
@@ -49,7 +49,8 @@ def simple_graph():
 @pytest.fixture
 def mock_services() -> InvocationServices:
     configuration = InvokeAIAppConfig(use_memory_db=True, node_cache_size=0)
-    db = SqliteDatabase(configuration, InvokeAILogger.get_logger())
+    logger = InvokeAILogger.get_logger()
+    db = create_mock_sqlite_database(configuration, logger)
     # NOTE: none of these are actually called by the test invocations
     graph_execution_manager = SqliteItemStorage[GraphExecutionState](db=db, table_name="graph_executions")
     return InvocationServices(
@@ -69,6 +70,7 @@ def mock_services() -> InvocationServices:
         logger=logging,  # type: ignore
         model_manager=None,  # type: ignore
         model_records=None,  # type: ignore
+        model_install=None,  # type: ignore
         names=None,  # type: ignore
         performance_statistics=InvocationStatsService(),
         processor=DefaultInvocationProcessor(),

tests/aa_nodes/test_invoker.py

@@ -4,6 +4,7 @@ import pytest
 
 from invokeai.app.services.config.config_default import InvokeAIAppConfig
 from invokeai.backend.util.logging import InvokeAILogger
+from tests.fixtures.sqlite_database import create_mock_sqlite_database
 
 # This import must happen before other invoke imports or test in other files(!!) break
 from .test_nodes import (  # isort: split
@@ -24,7 +25,6 @@ from invokeai.app.services.invoker import Invoker
 from invokeai.app.services.item_storage.item_storage_sqlite import SqliteItemStorage
 from invokeai.app.services.session_queue.session_queue_common import DEFAULT_QUEUE_ID
 from invokeai.app.services.shared.graph import Graph, GraphExecutionState, GraphInvocation, LibraryGraph
-from invokeai.app.services.shared.sqlite.sqlite_database import SqliteDatabase
 
 
 @pytest.fixture
@@ -52,8 +52,9 @@ def graph_with_subgraph():
 # the test invocations.
 @pytest.fixture
 def mock_services() -> InvocationServices:
-    db = SqliteDatabase(InvokeAIAppConfig(use_memory_db=True), InvokeAILogger.get_logger())
     configuration = InvokeAIAppConfig(use_memory_db=True, node_cache_size=0)
+    logger = InvokeAILogger.get_logger()
+    db = create_mock_sqlite_database(configuration, logger)
 
     # NOTE: none of these are actually called by the test invocations
     graph_execution_manager = SqliteItemStorage[GraphExecutionState](db=db, table_name="graph_executions")
@@ -74,6 +75,7 @@ def mock_services() -> InvocationServices:
         logger=logging,  # type: ignore
         model_manager=None,  # type: ignore
         model_records=None,  # type: ignore
+        model_install=None,  # type: ignore
         names=None,  # type: ignore
         performance_statistics=InvocationStatsService(),
         processor=DefaultInvocationProcessor(),

tests/aa_nodes/test_session_queue.py

@@ -12,7 +12,7 @@ from invokeai.app.services.session_queue.session_queue_common import (
     prepare_values_to_insert,
 )
 from invokeai.app.services.shared.graph import Graph, GraphExecutionState, GraphInvocation
-from tests.nodes.test_nodes import PromptTestInvocation
+from tests.aa_nodes.test_nodes import PromptTestInvocation
 
 
 @pytest.fixture

tests/aa_nodes/test_sqlite.py

@@ -15,8 +15,11 @@ class TestModel(BaseModel):
 
 @pytest.fixture
 def db() -> SqliteItemStorage[TestModel]:
-    sqlite_db = SqliteDatabase(InvokeAIAppConfig(use_memory_db=True), InvokeAILogger.get_logger())
-    sqlite_item_storage = SqliteItemStorage[TestModel](db=sqlite_db, table_name="test", id_field="id")
+    config = InvokeAIAppConfig(use_memory_db=True)
+    logger = InvokeAILogger.get_logger()
+    db_path = None if config.use_memory_db else config.db_path
+    db = SqliteDatabase(db_path=db_path, logger=logger, verbose=config.log_sql)
+    sqlite_item_storage = SqliteItemStorage[TestModel](db=db, table_name="test", id_field="id")
     return sqlite_item_storage
 
 

tests/app/services/model_install/test_model_install.py

@@ -0,0 +1,198 @@
+"""
+Test the model installer
+"""
+
+from pathlib import Path
+from typing import Any, Dict, List
+
+import pytest
+from pydantic import BaseModel, ValidationError
+
+from invokeai.app.services.config import InvokeAIAppConfig
+from invokeai.app.services.events.events_base import EventServiceBase
+from invokeai.app.services.model_install import (
+    InstallStatus,
+    LocalModelSource,
+    ModelInstallJob,
+    ModelInstallService,
+    ModelInstallServiceBase,
+)
+from invokeai.app.services.model_records import ModelRecordServiceBase, ModelRecordServiceSQL, UnknownModelException
+from invokeai.backend.model_manager.config import BaseModelType, ModelType
+from invokeai.backend.util.logging import InvokeAILogger
+from tests.fixtures.sqlite_database import create_mock_sqlite_database
+
+
+@pytest.fixture
+def test_file(datadir: Path) -> Path:
+    return datadir / "test_embedding.safetensors"
+
+
+@pytest.fixture
+def app_config(datadir: Path) -> InvokeAIAppConfig:
+    return InvokeAIAppConfig(
+        root=datadir / "root",
+        models_dir=datadir / "root/models",
+    )
+
+
+@pytest.fixture
+def store(
+    app_config: InvokeAIAppConfig,
+) -> ModelRecordServiceBase:
+    logger = InvokeAILogger.get_logger(config=app_config)
+    db = create_mock_sqlite_database(app_config, logger)
+    store: ModelRecordServiceBase = ModelRecordServiceSQL(db)
+    return store
+
+
+@pytest.fixture
+def installer(app_config: InvokeAIAppConfig, store: ModelRecordServiceBase) -> ModelInstallServiceBase:
+    return ModelInstallService(
+        app_config=app_config,
+        record_store=store,
+        event_bus=DummyEventService(),
+    )
+
+
+class DummyEvent(BaseModel):
+    """Dummy Event to use with Dummy Event service."""
+
+    event_name: str
+    payload: Dict[str, Any]
+
+
+class DummyEventService(EventServiceBase):
+    """Dummy event service for testing."""
+
+    events: List[DummyEvent]
+
+    def __init__(self) -> None:
+        super().__init__()
+        self.events = []
+
+    def dispatch(self, event_name: str, payload: Any) -> None:
+        """Dispatch an event by appending it to self.events."""
+        self.events.append(DummyEvent(event_name=payload["event"], payload=payload["data"]))
+
+
+def test_registration(installer: ModelInstallServiceBase, test_file: Path) -> None:
+    store = installer.record_store
+    matches = store.search_by_attr(model_name="test_embedding")
+    assert len(matches) == 0
+    key = installer.register_path(test_file)
+    assert key is not None
+    assert len(key) == 32
+
+
+def test_registration_meta(installer: ModelInstallServiceBase, test_file: Path) -> None:
+    store = installer.record_store
+    key = installer.register_path(test_file)
+    model_record = store.get_model(key)
+    assert model_record is not None
+    assert model_record.name == "test_embedding"
+    assert model_record.type == ModelType.TextualInversion
+    assert Path(model_record.path) == test_file
+    assert model_record.base == BaseModelType("sd-1")
+    assert model_record.description is not None
+    assert model_record.source is not None
+    assert Path(model_record.source) == test_file
+
+
+def test_registration_meta_override_fail(installer: ModelInstallServiceBase, test_file: Path) -> None:
+    key = None
+    with pytest.raises(ValidationError):
+        key = installer.register_path(test_file, {"name": "banana_sushi", "type": ModelType("lora")})
+    assert key is None
+
+
+def test_registration_meta_override_succeed(installer: ModelInstallServiceBase, test_file: Path) -> None:
+    store = installer.record_store
+    key = installer.register_path(
+        test_file, {"name": "banana_sushi", "source": "fake/repo_id", "current_hash": "New Hash"}
+    )
+    model_record = store.get_model(key)
+    assert model_record.name == "banana_sushi"
+    assert model_record.source == "fake/repo_id"
+    assert model_record.current_hash == "New Hash"
+
+
+def test_install(installer: ModelInstallServiceBase, test_file: Path, app_config: InvokeAIAppConfig) -> None:
+    store = installer.record_store
+    key = installer.install_path(test_file)
+    model_record = store.get_model(key)
+    assert model_record.path == "sd-1/embedding/test_embedding.safetensors"
+    assert model_record.source == test_file.as_posix()
+
+
+def test_background_install(installer: ModelInstallServiceBase, test_file: Path, app_config: InvokeAIAppConfig) -> None:
+    """Note: may want to break this down into several smaller unit tests."""
+    path = test_file
+    description = "Test of metadata assignment"
+    source = LocalModelSource(path=path, inplace=False)
+    job = installer.import_model(source, config={"description": description})
+    assert job is not None
+    assert isinstance(job, ModelInstallJob)
+
+    # See if job is registered properly
+    assert job in installer.get_job(source)
+
+    # test that the job object tracked installation correctly
+    jobs = installer.wait_for_installs()
+    assert len(jobs) > 0
+    my_job = [x for x in jobs if x.source == source]
+    assert len(my_job) == 1
+    assert my_job[0].status == InstallStatus.COMPLETED
+
+    # test that the expected events were issued
+    bus = installer.event_bus
+    assert bus is not None  # sigh - ruff is a stickler for type checking
+    assert isinstance(bus, DummyEventService)
+    assert len(bus.events) == 2
+    event_names = [x.event_name for x in bus.events]
+    assert "model_install_started" in event_names
+    assert "model_install_completed" in event_names
+    assert Path(bus.events[0].payload["source"]) == source
+    assert Path(bus.events[1].payload["source"]) == source
+    key = bus.events[1].payload["key"]
+    assert key is not None
+
+    # see if the thing actually got installed at the expected location
+    model_record = installer.record_store.get_model(key)
+    assert model_record is not None
+    assert model_record.path == "sd-1/embedding/test_embedding.safetensors"
+    assert Path(app_config.models_dir / model_record.path).exists()
+
+    # see if metadata was properly passed through
+    assert model_record.description == description
+
+    # see if prune works properly
+    installer.prune_jobs()
+    assert not installer.get_job(source)
+
+
+def test_delete_install(installer: ModelInstallServiceBase, test_file: Path, app_config: InvokeAIAppConfig):
+    store = installer.record_store
+    key = installer.install_path(test_file)
+    model_record = store.get_model(key)
+    assert Path(app_config.models_dir / model_record.path).exists()
+    assert test_file.exists()  # original should still be there after installation
+    installer.delete(key)
+    assert not Path(
+        app_config.models_dir / model_record.path
+    ).exists()  # after deletion, installed copy should not exist
+    assert test_file.exists()  # but original should still be there
+    with pytest.raises(UnknownModelException):
+        store.get_model(key)
+
+
+def test_delete_register(installer: ModelInstallServiceBase, test_file: Path, app_config: InvokeAIAppConfig):
+    store = installer.record_store
+    key = installer.register_path(test_file)
+    model_record = store.get_model(key)
+    assert Path(app_config.models_dir / model_record.path).exists()
+    assert test_file.exists()  # original should still be there after installation
+    installer.delete(key)
+    assert Path(app_config.models_dir / model_record.path).exists()
+    with pytest.raises(UnknownModelException):
+        store.get_model(key)

tests/app/services/model_install/test_model_install/README

@@ -0,0 +1 @@
+This directory is used by pytest-datadir.

tests/app/services/model_install/test_model_install/root/configs/stable-diffusion/v1-inference.yaml

@@ -0,0 +1,79 @@
+model:
+  base_learning_rate: 1.0e-04
+  target: invokeai.backend.models.diffusion.ddpm.LatentDiffusion
+  params:
+    linear_start: 0.00085
+    linear_end: 0.0120
+    num_timesteps_cond: 1
+    log_every_t: 200
+    timesteps: 1000
+    first_stage_key: "jpg"
+    cond_stage_key: "txt"
+    image_size: 64
+    channels: 4
+    cond_stage_trainable: false   # Note: different from the one we trained before
+    conditioning_key: crossattn
+    monitor: val/loss_simple_ema
+    scale_factor: 0.18215
+    use_ema: False
+
+    scheduler_config: # 10000 warmup steps
+      target: invokeai.backend.stable_diffusion.lr_scheduler.LambdaLinearScheduler
+      params:
+        warm_up_steps: [ 10000 ]
+        cycle_lengths: [ 10000000000000 ] # incredibly large number to prevent corner cases
+        f_start: [ 1.e-6 ]
+        f_max: [ 1. ]
+        f_min: [ 1. ]
+
+    personalization_config:
+      target: invokeai.backend.stable_diffusion.embedding_manager.EmbeddingManager
+      params:
+        placeholder_strings: ["*"]
+        initializer_words: ['sculpture']
+        per_image_tokens: false
+        num_vectors_per_token: 1
+        progressive_words: False
+
+    unet_config:
+      target: invokeai.backend.stable_diffusion.diffusionmodules.openaimodel.UNetModel
+      params:
+        image_size: 32 # unused
+        in_channels: 4
+        out_channels: 4
+        model_channels: 320
+        attention_resolutions: [ 4, 2, 1 ]
+        num_res_blocks: 2
+        channel_mult: [ 1, 2, 4, 4 ]
+        num_heads: 8
+        use_spatial_transformer: True
+        transformer_depth: 1
+        context_dim: 768
+        use_checkpoint: True
+        legacy: False
+
+    first_stage_config:
+      target: invokeai.backend.stable_diffusion.autoencoder.AutoencoderKL
+      params:
+        embed_dim: 4
+        monitor: val/rec_loss
+        ddconfig:
+          double_z: true
+          z_channels: 4
+          resolution: 256
+          in_channels: 3
+          out_ch: 3
+          ch: 128
+          ch_mult:
+          - 1
+          - 2
+          - 4
+          - 4
+          num_res_blocks: 2
+          attn_resolutions: []
+          dropout: 0.0
+        lossconfig:
+          target: torch.nn.Identity
+
+    cond_stage_config:
+      target: invokeai.backend.stable_diffusion.encoders.modules.WeightedFrozenCLIPEmbedder

tests/app/services/model_install/test_model_install/root/models/placeholder

@@ -0,0 +1 @@
+Dummy file to establish git path.

tests/app/services/model_records/test_model_records_sql.py

@@ -3,6 +3,7 @@ Test the refactored model config classes.
 """
 
 from hashlib import sha256
+from typing import Any
 
 import pytest
 
@@ -13,7 +14,6 @@ from invokeai.app.services.model_records import (
     ModelRecordServiceSQL,
     UnknownModelException,
 )
-from invokeai.app.services.shared.sqlite.sqlite_database import SqliteDatabase
 from invokeai.backend.model_manager.config import (
     BaseModelType,
     MainCheckpointConfig,
@@ -23,13 +23,16 @@ from invokeai.backend.model_manager.config import (
     VaeDiffusersConfig,
 )
 from invokeai.backend.util.logging import InvokeAILogger
+from tests.fixtures.sqlite_database import create_mock_sqlite_database
 
 
 @pytest.fixture
-def store(datadir) -> ModelRecordServiceBase:
+def store(
+    datadir: Any,
+) -> ModelRecordServiceBase:
     config = InvokeAIAppConfig(root=datadir)
     logger = InvokeAILogger.get_logger(config=config)
-    db = SqliteDatabase(config, logger)
+    db = create_mock_sqlite_database(config, logger)
     return ModelRecordServiceSQL(db)
 
 

tests/backend/tiles/test_tiles.py

@@ -1,7 +1,12 @@
 import numpy as np
 import pytest
 
-from invokeai.backend.tiles.tiles import calc_tiles_with_overlap, merge_tiles_with_linear_blending
+from invokeai.backend.tiles.tiles import (
+    calc_tiles_even_split,
+    calc_tiles_min_overlap,
+    calc_tiles_with_overlap,
+    merge_tiles_with_linear_blending,
+)
 from invokeai.backend.tiles.utils import TBLR, Tile
 
 ####################################
@@ -14,7 +19,10 @@ def test_calc_tiles_with_overlap_single_tile():
     tiles = calc_tiles_with_overlap(image_height=512, image_width=1024, tile_height=512, tile_width=1024, overlap=64)
 
     expected_tiles = [
-        Tile(coords=TBLR(top=0, bottom=512, left=0, right=1024), overlap=TBLR(top=0, bottom=0, left=0, right=0))
+        Tile(
+            coords=TBLR(top=0, bottom=512, left=0, right=1024),
+            overlap=TBLR(top=0, bottom=0, left=0, right=0),
+        )
     ]
 
     assert tiles == expected_tiles
@@ -27,13 +35,31 @@ def test_calc_tiles_with_overlap_evenly_divisible():
 
     expected_tiles = [
         # Row 0
-        Tile(coords=TBLR(top=0, bottom=320, left=0, right=576), overlap=TBLR(top=0, bottom=64, left=0, right=64)),
-        Tile(coords=TBLR(top=0, bottom=320, left=512, right=1088), overlap=TBLR(top=0, bottom=64, left=64, right=64)),
-        Tile(coords=TBLR(top=0, bottom=320, left=1024, right=1600), overlap=TBLR(top=0, bottom=64, left=64, right=0)),
+        Tile(
+            coords=TBLR(top=0, bottom=320, left=0, right=576),
+            overlap=TBLR(top=0, bottom=64, left=0, right=64),
+        ),
+        Tile(
+            coords=TBLR(top=0, bottom=320, left=512, right=1088),
+            overlap=TBLR(top=0, bottom=64, left=64, right=64),
+        ),
+        Tile(
+            coords=TBLR(top=0, bottom=320, left=1024, right=1600),
+            overlap=TBLR(top=0, bottom=64, left=64, right=0),
+        ),
         # Row 1
-        Tile(coords=TBLR(top=256, bottom=576, left=0, right=576), overlap=TBLR(top=64, bottom=0, left=0, right=64)),
-        Tile(coords=TBLR(top=256, bottom=576, left=512, right=1088), overlap=TBLR(top=64, bottom=0, left=64, right=64)),
-        Tile(coords=TBLR(top=256, bottom=576, left=1024, right=1600), overlap=TBLR(top=64, bottom=0, left=64, right=0)),
+        Tile(
+            coords=TBLR(top=256, bottom=576, left=0, right=576),
+            overlap=TBLR(top=64, bottom=0, left=0, right=64),
+        ),
+        Tile(
+            coords=TBLR(top=256, bottom=576, left=512, right=1088),
+            overlap=TBLR(top=64, bottom=0, left=64, right=64),
+        ),
+        Tile(
+            coords=TBLR(top=256, bottom=576, left=1024, right=1600),
+            overlap=TBLR(top=64, bottom=0, left=64, right=0),
+        ),
     ]
 
     assert tiles == expected_tiles
@@ -46,16 +72,30 @@ def test_calc_tiles_with_overlap_not_evenly_divisible():
 
     expected_tiles = [
         # Row 0
-        Tile(coords=TBLR(top=0, bottom=256, left=0, right=512), overlap=TBLR(top=0, bottom=112, left=0, right=64)),
-        Tile(coords=TBLR(top=0, bottom=256, left=448, right=960), overlap=TBLR(top=0, bottom=112, left=64, right=272)),
-        Tile(coords=TBLR(top=0, bottom=256, left=688, right=1200), overlap=TBLR(top=0, bottom=112, left=272, right=0)),
-        # Row 1
-        Tile(coords=TBLR(top=144, bottom=400, left=0, right=512), overlap=TBLR(top=112, bottom=0, left=0, right=64)),
         Tile(
-            coords=TBLR(top=144, bottom=400, left=448, right=960), overlap=TBLR(top=112, bottom=0, left=64, right=272)
+            coords=TBLR(top=0, bottom=256, left=0, right=512),
+            overlap=TBLR(top=0, bottom=112, left=0, right=64),
         ),
         Tile(
-            coords=TBLR(top=144, bottom=400, left=688, right=1200), overlap=TBLR(top=112, bottom=0, left=272, right=0)
+            coords=TBLR(top=0, bottom=256, left=448, right=960),
+            overlap=TBLR(top=0, bottom=112, left=64, right=272),
+        ),
+        Tile(
+            coords=TBLR(top=0, bottom=256, left=688, right=1200),
+            overlap=TBLR(top=0, bottom=112, left=272, right=0),
+        ),
+        # Row 1
+        Tile(
+            coords=TBLR(top=144, bottom=400, left=0, right=512),
+            overlap=TBLR(top=112, bottom=0, left=0, right=64),
+        ),
+        Tile(
+            coords=TBLR(top=144, bottom=400, left=448, right=960),
+            overlap=TBLR(top=112, bottom=0, left=64, right=272),
+        ),
+        Tile(
+            coords=TBLR(top=144, bottom=400, left=688, right=1200),
+            overlap=TBLR(top=112, bottom=0, left=272, right=0),
         ),
     ]
 
@@ -75,7 +115,12 @@ def test_calc_tiles_with_overlap_not_evenly_divisible():
     ],
 )
 def test_calc_tiles_with_overlap_input_validation(
-    image_height: int, image_width: int, tile_height: int, tile_width: int, overlap: int, raises: bool
+    image_height: int,
+    image_width: int,
+    tile_height: int,
+    tile_width: int,
+    overlap: int,
+    raises: bool,
 ):
     """Test that calc_tiles_with_overlap() raises an exception if the inputs are invalid."""
     if raises:
@@ -85,6 +130,328 @@ def test_calc_tiles_with_overlap_input_validation(
         calc_tiles_with_overlap(image_height, image_width, tile_height, tile_width, overlap)
 
 
+####################################
+# Test calc_tiles_min_overlap(...)
+####################################
+
+
+def test_calc_tiles_min_overlap_single_tile():
+    """Test calc_tiles_min_overlap() behavior when a single tile covers the image."""
+    tiles = calc_tiles_min_overlap(
+        image_height=512,
+        image_width=1024,
+        tile_height=512,
+        tile_width=1024,
+        min_overlap=64,
+    )
+
+    expected_tiles = [
+        Tile(
+            coords=TBLR(top=0, bottom=512, left=0, right=1024),
+            overlap=TBLR(top=0, bottom=0, left=0, right=0),
+        )
+    ]
+
+    assert tiles == expected_tiles
+
+
+def test_calc_tiles_min_overlap_evenly_divisible():
+    """Test calc_tiles_min_overlap() behavior when the image is evenly covered by multiple tiles."""
+    # Parameters mimic roughly the same output as the original tile generations of the same test name
+    tiles = calc_tiles_min_overlap(
+        image_height=576,
+        image_width=1600,
+        tile_height=320,
+        tile_width=576,
+        min_overlap=64,
+    )
+
+    expected_tiles = [
+        # Row 0
+        Tile(
+            coords=TBLR(top=0, bottom=320, left=0, right=576),
+            overlap=TBLR(top=0, bottom=64, left=0, right=64),
+        ),
+        Tile(
+            coords=TBLR(top=0, bottom=320, left=512, right=1088),
+            overlap=TBLR(top=0, bottom=64, left=64, right=64),
+        ),
+        Tile(
+            coords=TBLR(top=0, bottom=320, left=1024, right=1600),
+            overlap=TBLR(top=0, bottom=64, left=64, right=0),
+        ),
+        # Row 1
+        Tile(
+            coords=TBLR(top=256, bottom=576, left=0, right=576),
+            overlap=TBLR(top=64, bottom=0, left=0, right=64),
+        ),
+        Tile(
+            coords=TBLR(top=256, bottom=576, left=512, right=1088),
+            overlap=TBLR(top=64, bottom=0, left=64, right=64),
+        ),
+        Tile(
+            coords=TBLR(top=256, bottom=576, left=1024, right=1600),
+            overlap=TBLR(top=64, bottom=0, left=64, right=0),
+        ),
+    ]
+
+    assert tiles == expected_tiles
+
+
+def test_calc_tiles_min_overlap_not_evenly_divisible():
+    """Test calc_tiles_min_overlap() behavior when the image requires 'uneven' overlaps to achieve proper coverage."""
+    # Parameters mimic roughly the same output as the original tile generations of the same test name
+    tiles = calc_tiles_min_overlap(
+        image_height=400,
+        image_width=1200,
+        tile_height=256,
+        tile_width=512,
+        min_overlap=64,
+    )
+
+    expected_tiles = [
+        # Row 0
+        Tile(
+            coords=TBLR(top=0, bottom=256, left=0, right=512),
+            overlap=TBLR(top=0, bottom=112, left=0, right=168),
+        ),
+        Tile(
+            coords=TBLR(top=0, bottom=256, left=344, right=856),
+            overlap=TBLR(top=0, bottom=112, left=168, right=168),
+        ),
+        Tile(
+            coords=TBLR(top=0, bottom=256, left=688, right=1200),
+            overlap=TBLR(top=0, bottom=112, left=168, right=0),
+        ),
+        # Row 1
+        Tile(
+            coords=TBLR(top=144, bottom=400, left=0, right=512),
+            overlap=TBLR(top=112, bottom=0, left=0, right=168),
+        ),
+        Tile(
+            coords=TBLR(top=144, bottom=400, left=344, right=856),
+            overlap=TBLR(top=112, bottom=0, left=168, right=168),
+        ),
+        Tile(
+            coords=TBLR(top=144, bottom=400, left=688, right=1200),
+            overlap=TBLR(top=112, bottom=0, left=168, right=0),
+        ),
+    ]
+
+    assert tiles == expected_tiles
+
+
+def test_calc_tiles_min_overlap_tile_bigger_than_image():
+    """Test calc_tiles_min_overlap() behavior when the tile is nigger than the image"""
+    # Parameters mimic roughly the same output as the original tile generations of the same test name
+    tiles = calc_tiles_min_overlap(
+        image_height=1024,
+        image_width=1024,
+        tile_height=1408,
+        tile_width=1408,
+        min_overlap=128,
+    )
+
+    expected_tiles = [
+        # single tile
+        Tile(
+            coords=TBLR(top=0, bottom=1024, left=0, right=1024),
+            overlap=TBLR(top=0, bottom=0, left=0, right=0),
+        ),
+    ]
+
+    assert tiles == expected_tiles
+
+
+@pytest.mark.parametrize(
+    [
+        "image_height",
+        "image_width",
+        "tile_height",
+        "tile_width",
+        "min_overlap",
+        "raises",
+    ],
+    [
+        (128, 128, 128, 128, 127, False),  # OK
+        (128, 128, 128, 128, 0, False),  # OK
+        (128, 128, 64, 64, 0, False),  # OK
+        (128, 128, 129, 128, 0, False),  # tile_height exceeds image_height defaults to 1 tile.
+        (128, 128, 128, 129, 0, False),  # tile_width exceeds image_width defaults to 1 tile.
+        (128, 128, 64, 128, 64, True),  # overlap equals tile_height.
+        (128, 128, 128, 64, 64, True),  # overlap equals tile_width.
+    ],
+)
+def test_calc_tiles_min_overlap_input_validation(
+    image_height: int,
+    image_width: int,
+    tile_height: int,
+    tile_width: int,
+    min_overlap: int,
+    raises: bool,
+):
+    """Test that calc_tiles_min_overlap() raises an exception if the inputs are invalid."""
+    if raises:
+        with pytest.raises(AssertionError):
+            calc_tiles_min_overlap(image_height, image_width, tile_height, tile_width, min_overlap)
+    else:
+        calc_tiles_min_overlap(image_height, image_width, tile_height, tile_width, min_overlap)
+
+
+####################################
+# Test calc_tiles_even_split(...)
+####################################
+
+
+def test_calc_tiles_even_split_single_tile():
+    """Test calc_tiles_even_split() behavior when a single tile covers the image."""
+    tiles = calc_tiles_even_split(
+        image_height=512, image_width=1024, num_tiles_x=1, num_tiles_y=1, overlap_fraction=0.25
+    )
+
+    expected_tiles = [
+        Tile(
+            coords=TBLR(top=0, bottom=512, left=0, right=1024),
+            overlap=TBLR(top=0, bottom=0, left=0, right=0),
+        )
+    ]
+
+    assert tiles == expected_tiles
+
+
+def test_calc_tiles_even_split_evenly_divisible():
+    """Test calc_tiles_even_split() behavior when the image is evenly covered by multiple tiles."""
+    # Parameters mimic roughly the same output as the original tile generations of the same test name
+    tiles = calc_tiles_even_split(
+        image_height=576, image_width=1600, num_tiles_x=3, num_tiles_y=2, overlap_fraction=0.25
+    )
+
+    expected_tiles = [
+        # Row 0
+        Tile(
+            coords=TBLR(top=0, bottom=320, left=0, right=624),
+            overlap=TBLR(top=0, bottom=72, left=0, right=136),
+        ),
+        Tile(
+            coords=TBLR(top=0, bottom=320, left=488, right=1112),
+            overlap=TBLR(top=0, bottom=72, left=136, right=136),
+        ),
+        Tile(
+            coords=TBLR(top=0, bottom=320, left=976, right=1600),
+            overlap=TBLR(top=0, bottom=72, left=136, right=0),
+        ),
+        # Row 1
+        Tile(
+            coords=TBLR(top=248, bottom=576, left=0, right=624),
+            overlap=TBLR(top=72, bottom=0, left=0, right=136),
+        ),
+        Tile(
+            coords=TBLR(top=248, bottom=576, left=488, right=1112),
+            overlap=TBLR(top=72, bottom=0, left=136, right=136),
+        ),
+        Tile(
+            coords=TBLR(top=248, bottom=576, left=976, right=1600),
+            overlap=TBLR(top=72, bottom=0, left=136, right=0),
+        ),
+    ]
+    assert tiles == expected_tiles
+
+
+def test_calc_tiles_even_split_not_evenly_divisible():
+    """Test calc_tiles_even_split() behavior when the image requires 'uneven' overlaps to achieve proper coverage."""
+    # Parameters mimic roughly the same output as the original tile generations of the same test name
+    tiles = calc_tiles_even_split(
+        image_height=400, image_width=1200, num_tiles_x=3, num_tiles_y=2, overlap_fraction=0.25
+    )
+
+    expected_tiles = [
+        # Row 0
+        Tile(
+            coords=TBLR(top=0, bottom=224, left=0, right=464),
+            overlap=TBLR(top=0, bottom=56, left=0, right=104),
+        ),
+        Tile(
+            coords=TBLR(top=0, bottom=224, left=360, right=824),
+            overlap=TBLR(top=0, bottom=56, left=104, right=104),
+        ),
+        Tile(
+            coords=TBLR(top=0, bottom=224, left=720, right=1200),
+            overlap=TBLR(top=0, bottom=56, left=104, right=0),
+        ),
+        # Row 1
+        Tile(
+            coords=TBLR(top=168, bottom=400, left=0, right=464),
+            overlap=TBLR(top=56, bottom=0, left=0, right=104),
+        ),
+        Tile(
+            coords=TBLR(top=168, bottom=400, left=360, right=824),
+            overlap=TBLR(top=56, bottom=0, left=104, right=104),
+        ),
+        Tile(
+            coords=TBLR(top=168, bottom=400, left=720, right=1200),
+            overlap=TBLR(top=56, bottom=0, left=104, right=0),
+        ),
+    ]
+
+    assert tiles == expected_tiles
+
+
+def test_calc_tiles_even_split_difficult_size():
+    """Test calc_tiles_even_split() behavior when the image is a difficult size to spilt evenly and keep div8."""
+    # Parameters are a difficult size for other tile gen routines to calculate
+    tiles = calc_tiles_even_split(
+        image_height=1000, image_width=1000, num_tiles_x=2, num_tiles_y=2, overlap_fraction=0.25
+    )
+
+    expected_tiles = [
+        # Row 0
+        Tile(
+            coords=TBLR(top=0, bottom=560, left=0, right=560),
+            overlap=TBLR(top=0, bottom=128, left=0, right=128),
+        ),
+        Tile(
+            coords=TBLR(top=0, bottom=560, left=432, right=1000),
+            overlap=TBLR(top=0, bottom=128, left=128, right=0),
+        ),
+        # Row 1
+        Tile(
+            coords=TBLR(top=432, bottom=1000, left=0, right=560),
+            overlap=TBLR(top=128, bottom=0, left=0, right=128),
+        ),
+        Tile(
+            coords=TBLR(top=432, bottom=1000, left=432, right=1000),
+            overlap=TBLR(top=128, bottom=0, left=128, right=0),
+        ),
+    ]
+
+    assert tiles == expected_tiles
+
+
+@pytest.mark.parametrize(
+    ["image_height", "image_width", "num_tiles_x", "num_tiles_y", "overlap_fraction", "raises"],
+    [
+        (128, 128, 1, 1, 0.25, False),  # OK
+        (128, 128, 1, 1, 0, False),  # OK
+        (128, 128, 2, 1, 0, False),  # OK
+        (127, 127, 1, 1, 0, True),  # image size must be dividable by 8
+    ],
+)
+def test_calc_tiles_even_split_input_validation(
+    image_height: int,
+    image_width: int,
+    num_tiles_x: int,
+    num_tiles_y: int,
+    overlap_fraction: float,
+    raises: bool,
+):
+    """Test that calc_tiles_even_split() raises an exception if the inputs are invalid."""
+    if raises:
+        with pytest.raises(ValueError):
+            calc_tiles_even_split(image_height, image_width, num_tiles_x, num_tiles_y, overlap_fraction)
+    else:
+        calc_tiles_even_split(image_height, image_width, num_tiles_x, num_tiles_y, overlap_fraction)
+
+
 #############################################
 # Test merge_tiles_with_linear_blending(...)
 #############################################
@@ -95,8 +462,14 @@ def test_merge_tiles_with_linear_blending_horizontal(blend_amount: int):
     """Test merge_tiles_with_linear_blending(...) behavior when merging horizontally."""
     # Initialize 2 tiles side-by-side.
     tiles = [
-        Tile(coords=TBLR(top=0, bottom=512, left=0, right=512), overlap=TBLR(top=0, bottom=0, left=0, right=64)),
-        Tile(coords=TBLR(top=0, bottom=512, left=448, right=960), overlap=TBLR(top=0, bottom=0, left=64, right=0)),
+        Tile(
+            coords=TBLR(top=0, bottom=512, left=0, right=512),
+            overlap=TBLR(top=0, bottom=0, left=0, right=64),
+        ),
+        Tile(
+            coords=TBLR(top=0, bottom=512, left=448, right=960),
+            overlap=TBLR(top=0, bottom=0, left=64, right=0),
+        ),
     ]
 
     dst_image = np.zeros((512, 960, 3), dtype=np.uint8)
@@ -116,7 +489,10 @@ def test_merge_tiles_with_linear_blending_horizontal(blend_amount: int):
     expected_output[:, 480 + (blend_amount // 2) :, :] = 128
 
     merge_tiles_with_linear_blending(
-        dst_image=dst_image, tiles=tiles, tile_images=tile_images, blend_amount=blend_amount
+        dst_image=dst_image,
+        tiles=tiles,
+        tile_images=tile_images,
+        blend_amount=blend_amount,
     )
 
     np.testing.assert_array_equal(dst_image, expected_output, strict=True)
@@ -127,8 +503,14 @@ def test_merge_tiles_with_linear_blending_vertical(blend_amount: int):
     """Test merge_tiles_with_linear_blending(...) behavior when merging vertically."""
     # Initialize 2 tiles stacked vertically.
     tiles = [
-        Tile(coords=TBLR(top=0, bottom=512, left=0, right=512), overlap=TBLR(top=0, bottom=64, left=0, right=0)),
-        Tile(coords=TBLR(top=448, bottom=960, left=0, right=512), overlap=TBLR(top=64, bottom=0, left=0, right=0)),
+        Tile(
+            coords=TBLR(top=0, bottom=512, left=0, right=512),
+            overlap=TBLR(top=0, bottom=64, left=0, right=0),
+        ),
+        Tile(
+            coords=TBLR(top=448, bottom=960, left=0, right=512),
+            overlap=TBLR(top=64, bottom=0, left=0, right=0),
+        ),
     ]
 
     dst_image = np.zeros((960, 512, 3), dtype=np.uint8)
@@ -148,7 +530,10 @@ def test_merge_tiles_with_linear_blending_vertical(blend_amount: int):
     expected_output[480 + (blend_amount // 2) :, :, :] = 128
 
     merge_tiles_with_linear_blending(
-        dst_image=dst_image, tiles=tiles, tile_images=tile_images, blend_amount=blend_amount
+        dst_image=dst_image,
+        tiles=tiles,
+        tile_images=tile_images,
+        blend_amount=blend_amount,
     )
 
     np.testing.assert_array_equal(dst_image, expected_output, strict=True)
@@ -160,8 +545,14 @@ def test_merge_tiles_with_linear_blending_blend_amount_exceeds_vertical_overlap(
     """
     # Initialize 2 tiles stacked vertically.
     tiles = [
-        Tile(coords=TBLR(top=0, bottom=512, left=0, right=512), overlap=TBLR(top=0, bottom=64, left=0, right=0)),
-        Tile(coords=TBLR(top=448, bottom=960, left=0, right=512), overlap=TBLR(top=64, bottom=0, left=0, right=0)),
+        Tile(
+            coords=TBLR(top=0, bottom=512, left=0, right=512),
+            overlap=TBLR(top=0, bottom=64, left=0, right=0),
+        ),
+        Tile(
+            coords=TBLR(top=448, bottom=960, left=0, right=512),
+            overlap=TBLR(top=64, bottom=0, left=0, right=0),
+        ),
     ]
 
     dst_image = np.zeros((960, 512, 3), dtype=np.uint8)
@@ -180,8 +571,14 @@ def test_merge_tiles_with_linear_blending_blend_amount_exceeds_horizontal_overla
     """
     # Initialize 2 tiles side-by-side.
     tiles = [
-        Tile(coords=TBLR(top=0, bottom=512, left=0, right=512), overlap=TBLR(top=0, bottom=0, left=0, right=64)),
-        Tile(coords=TBLR(top=0, bottom=512, left=448, right=960), overlap=TBLR(top=0, bottom=0, left=64, right=0)),
+        Tile(
+            coords=TBLR(top=0, bottom=512, left=0, right=512),
+            overlap=TBLR(top=0, bottom=0, left=0, right=64),
+        ),
+        Tile(
+            coords=TBLR(top=0, bottom=512, left=448, right=960),
+            overlap=TBLR(top=0, bottom=0, left=64, right=0),
+        ),
     ]
 
     dst_image = np.zeros((512, 960, 3), dtype=np.uint8)
@@ -198,7 +595,12 @@ def test_merge_tiles_with_linear_blending_tiles_overflow_dst_image():
     """Test that merge_tiles_with_linear_blending(...) raises an exception if any of the tiles overflows the
     dst_image.
     """
-    tiles = [Tile(coords=TBLR(top=0, bottom=512, left=0, right=512), overlap=TBLR(top=0, bottom=0, left=0, right=0))]
+    tiles = [
+        Tile(
+            coords=TBLR(top=0, bottom=512, left=0, right=512),
+            overlap=TBLR(top=0, bottom=0, left=0, right=0),
+        )
+    ]
 
     dst_image = np.zeros((256, 512, 3), dtype=np.uint8)
 
@@ -213,7 +615,12 @@ def test_merge_tiles_with_linear_blending_mismatched_list_lengths():
     """Test that merge_tiles_with_linear_blending(...) raises an exception if the lengths of 'tiles' and 'tile_images'
     do not match.
     """
-    tiles = [Tile(coords=TBLR(top=0, bottom=512, left=0, right=512), overlap=TBLR(top=0, bottom=0, left=0, right=0))]
+    tiles = [
+        Tile(
+            coords=TBLR(top=0, bottom=512, left=0, right=512),
+            overlap=TBLR(top=0, bottom=0, left=0, right=0),
+        )
+    ]
 
     dst_image = np.zeros((256, 512, 3), dtype=np.uint8)
 

tests/fixtures/sqlite_database.py

@@ -0,0 +1,13 @@
+from logging import Logger
+from unittest import mock
+
+from invokeai.app.services.config.config_default import InvokeAIAppConfig
+from invokeai.app.services.image_files.image_files_base import ImageFileStorageBase
+from invokeai.app.services.shared.sqlite.sqlite_database import SqliteDatabase
+from invokeai.app.services.shared.sqlite.sqlite_util import init_db
+
+
+def create_mock_sqlite_database(config: InvokeAIAppConfig, logger: Logger) -> SqliteDatabase:
+    image_files = mock.Mock(spec=ImageFileStorageBase)
+    db = init_db(config=config, logger=logger, image_files=image_files)
+    return db

tests/test_sqlite_migrator.py

@@ -0,0 +1,272 @@
+import sqlite3
+from contextlib import closing
+from logging import Logger
+from pathlib import Path
+from tempfile import TemporaryDirectory
+
+import pytest
+from pydantic import ValidationError
+
+from invokeai.app.services.shared.sqlite.sqlite_database import SqliteDatabase
+from invokeai.app.services.shared.sqlite_migrator.sqlite_migrator_common import (
+    MigrateCallback,
+    Migration,
+    MigrationError,
+    MigrationSet,
+    MigrationVersionError,
+)
+from invokeai.app.services.shared.sqlite_migrator.sqlite_migrator_impl import (
+    SqliteMigrator,
+)
+
+
+@pytest.fixture
+def logger() -> Logger:
+    return Logger("test_sqlite_migrator")
+
+
+@pytest.fixture
+def memory_db_conn() -> sqlite3.Connection:
+    return sqlite3.connect(":memory:")
+
+
+@pytest.fixture
+def memory_db_cursor(memory_db_conn: sqlite3.Connection) -> sqlite3.Cursor:
+    return memory_db_conn.cursor()
+
+
+@pytest.fixture
+def migrator(logger: Logger) -> SqliteMigrator:
+    db = SqliteDatabase(db_path=None, logger=logger, verbose=False)
+    return SqliteMigrator(db=db)
+
+
+@pytest.fixture
+def no_op_migrate_callback() -> MigrateCallback:
+    def no_op_migrate(cursor: sqlite3.Cursor, **kwargs) -> None:
+        pass
+
+    return no_op_migrate
+
+
+@pytest.fixture
+def migration_no_op(no_op_migrate_callback: MigrateCallback) -> Migration:
+    return Migration(from_version=0, to_version=1, callback=no_op_migrate_callback)
+
+
+@pytest.fixture
+def migrate_callback_create_table_of_name() -> MigrateCallback:
+    def migrate(cursor: sqlite3.Cursor, **kwargs) -> None:
+        table_name = kwargs["table_name"]
+        cursor.execute(f"CREATE TABLE {table_name} (id INTEGER PRIMARY KEY);")
+
+    return migrate
+
+
+@pytest.fixture
+def migrate_callback_create_test_table() -> MigrateCallback:
+    def migrate(cursor: sqlite3.Cursor, **kwargs) -> None:
+        cursor.execute("CREATE TABLE test (id INTEGER PRIMARY KEY);")
+
+    return migrate
+
+
+@pytest.fixture
+def migration_create_test_table(migrate_callback_create_test_table: MigrateCallback) -> Migration:
+    return Migration(from_version=0, to_version=1, callback=migrate_callback_create_test_table)
+
+
+@pytest.fixture
+def failing_migration() -> Migration:
+    def failing_migration(cursor: sqlite3.Cursor, **kwargs) -> None:
+        raise Exception("Bad migration")
+
+    return Migration(from_version=0, to_version=1, callback=failing_migration)
+
+
+@pytest.fixture
+def failing_migrate_callback() -> MigrateCallback:
+    def failing_migrate(cursor: sqlite3.Cursor, **kwargs) -> None:
+        raise Exception("Bad migration")
+
+    return failing_migrate
+
+
+def create_migrate(i: int) -> MigrateCallback:
+    def migrate(cursor: sqlite3.Cursor, **kwargs) -> None:
+        cursor.execute(f"CREATE TABLE test{i} (id INTEGER PRIMARY KEY);")
+
+    return migrate
+
+
+def test_migration_to_version_is_one_gt_from_version(no_op_migrate_callback: MigrateCallback) -> None:
+    with pytest.raises(ValidationError, match="to_version must be one greater than from_version"):
+        Migration(from_version=0, to_version=2, callback=no_op_migrate_callback)
+    # not raising is sufficient
+    Migration(from_version=1, to_version=2, callback=no_op_migrate_callback)
+
+
+def test_migration_hash(no_op_migrate_callback: MigrateCallback) -> None:
+    migration = Migration(from_version=0, to_version=1, callback=no_op_migrate_callback)
+    assert hash(migration) == hash((0, 1))
+
+
+def test_migration_set_add_migration(migrator: SqliteMigrator, migration_no_op: Migration) -> None:
+    migration = migration_no_op
+    migrator._migration_set.register(migration)
+    assert migration in migrator._migration_set._migrations
+
+
+def test_migration_set_may_not_register_dupes(
+    migrator: SqliteMigrator, no_op_migrate_callback: MigrateCallback
+) -> None:
+    migrate_0_to_1_a = Migration(from_version=0, to_version=1, callback=no_op_migrate_callback)
+    migrate_0_to_1_b = Migration(from_version=0, to_version=1, callback=no_op_migrate_callback)
+    migrator._migration_set.register(migrate_0_to_1_a)
+    with pytest.raises(MigrationVersionError, match=r"Migration with from_version or to_version already registered"):
+        migrator._migration_set.register(migrate_0_to_1_b)
+    migrate_1_to_2_a = Migration(from_version=1, to_version=2, callback=no_op_migrate_callback)
+    migrate_1_to_2_b = Migration(from_version=1, to_version=2, callback=no_op_migrate_callback)
+    migrator._migration_set.register(migrate_1_to_2_a)
+    with pytest.raises(MigrationVersionError, match=r"Migration with from_version or to_version already registered"):
+        migrator._migration_set.register(migrate_1_to_2_b)
+
+
+def test_migration_set_gets_migration(migration_no_op: Migration) -> None:
+    migration_set = MigrationSet()
+    migration_set.register(migration_no_op)
+    assert migration_set.get(0) == migration_no_op
+    assert migration_set.get(1) is None
+
+
+def test_migration_set_validates_migration_chain(no_op_migrate_callback: MigrateCallback) -> None:
+    migration_set = MigrationSet()
+    migration_set.register(Migration(from_version=1, to_version=2, callback=no_op_migrate_callback))
+    with pytest.raises(MigrationError, match="Migration chain is fragmented"):
+        # no migration from 0 to 1
+        migration_set.validate_migration_chain()
+    migration_set.register(Migration(from_version=0, to_version=1, callback=no_op_migrate_callback))
+    migration_set.validate_migration_chain()
+    migration_set.register(Migration(from_version=2, to_version=3, callback=no_op_migrate_callback))
+    migration_set.validate_migration_chain()
+    migration_set.register(Migration(from_version=4, to_version=5, callback=no_op_migrate_callback))
+    with pytest.raises(MigrationError, match="Migration chain is fragmented"):
+        # no migration from 3 to 4
+        migration_set.validate_migration_chain()
+
+
+def test_migration_set_counts_migrations(no_op_migrate_callback: MigrateCallback) -> None:
+    migration_set = MigrationSet()
+    assert migration_set.count == 0
+    migration_set.register(Migration(from_version=0, to_version=1, callback=no_op_migrate_callback))
+    assert migration_set.count == 1
+    migration_set.register(Migration(from_version=1, to_version=2, callback=no_op_migrate_callback))
+    assert migration_set.count == 2
+
+
+def test_migration_set_gets_latest_version(no_op_migrate_callback: MigrateCallback) -> None:
+    migration_set = MigrationSet()
+    assert migration_set.latest_version == 0
+    migration_set.register(Migration(from_version=1, to_version=2, callback=no_op_migrate_callback))
+    assert migration_set.latest_version == 2
+    migration_set.register(Migration(from_version=0, to_version=1, callback=no_op_migrate_callback))
+    assert migration_set.latest_version == 2
+
+
+def test_migration_runs(memory_db_cursor: sqlite3.Cursor, migrate_callback_create_test_table: MigrateCallback) -> None:
+    migration = Migration(
+        from_version=0,
+        to_version=1,
+        callback=migrate_callback_create_test_table,
+    )
+    migration.callback(memory_db_cursor)
+    memory_db_cursor.execute("SELECT name FROM sqlite_master WHERE type='table' AND name='test';")
+    assert memory_db_cursor.fetchone() is not None
+
+
+def test_migrator_registers_migration(migrator: SqliteMigrator, migration_no_op: Migration) -> None:
+    migration = migration_no_op
+    migrator.register_migration(migration)
+    assert migration in migrator._migration_set._migrations
+
+
+def test_migrator_creates_migrations_table(migrator: SqliteMigrator) -> None:
+    cursor = migrator._db.conn.cursor()
+    migrator._create_migrations_table(cursor)
+    cursor.execute("SELECT * FROM sqlite_master WHERE type='table' AND name='migrations';")
+    assert cursor.fetchone() is not None
+
+
+def test_migrator_migration_sets_version(migrator: SqliteMigrator, migration_no_op: Migration) -> None:
+    cursor = migrator._db.conn.cursor()
+    migrator._create_migrations_table(cursor)
+    migrator.register_migration(migration_no_op)
+    migrator.run_migrations()
+    cursor.execute("SELECT MAX(version) FROM migrations;")
+    assert cursor.fetchone()[0] == 1
+
+
+def test_migrator_gets_current_version(migrator: SqliteMigrator, migration_no_op: Migration) -> None:
+    cursor = migrator._db.conn.cursor()
+    assert migrator._get_current_version(cursor) == 0
+    migrator._create_migrations_table(cursor)
+    assert migrator._get_current_version(cursor) == 0
+    migrator.register_migration(migration_no_op)
+    migrator.run_migrations()
+    assert migrator._get_current_version(cursor) == 1
+
+
+def test_migrator_runs_single_migration(migrator: SqliteMigrator, migration_create_test_table: Migration) -> None:
+    cursor = migrator._db.conn.cursor()
+    migrator._create_migrations_table(cursor)
+    migrator._run_migration(migration_create_test_table)
+    assert migrator._get_current_version(cursor) == 1
+    cursor.execute("SELECT name FROM sqlite_master WHERE type='table' AND name='test';")
+    assert cursor.fetchone() is not None
+
+
+def test_migrator_runs_all_migrations_in_memory(migrator: SqliteMigrator) -> None:
+    cursor = migrator._db.conn.cursor()
+    migrations = [Migration(from_version=i, to_version=i + 1, callback=create_migrate(i)) for i in range(0, 3)]
+    for migration in migrations:
+        migrator.register_migration(migration)
+    migrator.run_migrations()
+    assert migrator._get_current_version(cursor) == 3
+
+
+def test_migrator_runs_all_migrations_file(logger: Logger) -> None:
+    with TemporaryDirectory() as tempdir:
+        original_db_path = Path(tempdir) / "invokeai.db"
+        db = SqliteDatabase(db_path=original_db_path, logger=logger, verbose=False)
+        migrator = SqliteMigrator(db=db)
+        migrations = [Migration(from_version=i, to_version=i + 1, callback=create_migrate(i)) for i in range(0, 3)]
+        for migration in migrations:
+            migrator.register_migration(migration)
+        migrator.run_migrations()
+        with closing(sqlite3.connect(original_db_path)) as original_db_conn:
+            original_db_cursor = original_db_conn.cursor()
+            assert SqliteMigrator._get_current_version(original_db_cursor) == 3
+        # Must manually close else we get an error on Windows
+        db.conn.close()
+
+
+def test_migrator_makes_no_changes_on_failed_migration(
+    migrator: SqliteMigrator, migration_no_op: Migration, failing_migrate_callback: MigrateCallback
+) -> None:
+    cursor = migrator._db.conn.cursor()
+    migrator.register_migration(migration_no_op)
+    migrator.run_migrations()
+    assert migrator._get_current_version(cursor) == 1
+    migrator.register_migration(Migration(from_version=1, to_version=2, callback=failing_migrate_callback))
+    with pytest.raises(MigrationError, match="Bad migration"):
+        migrator.run_migrations()
+    assert migrator._get_current_version(cursor) == 1
+
+
+def test_idempotent_migrations(migrator: SqliteMigrator, migration_create_test_table: Migration) -> None:
+    cursor = migrator._db.conn.cursor()
+    migrator.register_migration(migration_create_test_table)
+    migrator.run_migrations()
+    # not throwing is sufficient
+    migrator.run_migrations()
+    assert migrator._get_current_version(cursor) == 1