Simplifies a couple things:
- Init is more straightforward
- It's clear in the migrator that the connection we are working with is related to the SqliteDatabase
- Simplify init args to path (None means use memory), logger, and verbose
- Add docstrings to SqliteDatabase (it had almost none)
- Update all usages of the class
Using default_factory to autogenerate UUIDs doesn't make sense here, and results awkward typescript types.
Remove the default factory and instead manually create a UUID for workflow id. There are only two places where this needs to happen so it's not a big change.
* chore: bump pydantic to 2.5.2
This release fixespydantic/pydantic#8175 and allows us to use `JsonValue`
* fix(ui): exclude public/en.json from prettier config
* fix(workflow_records): fix SQLite workflow insertion to ignore duplicates
* feat(backend): update workflows handling
Update workflows handling for Workflow Library.
**Updated Workflow Storage**
"Embedded Workflows" are workflows associated with images, and are now only stored in the image files. "Library Workflows" are not associated with images, and are stored only in DB.
This works out nicely. We have always saved workflows to files, but recently began saving them to the DB in addition to in image files. When that happened, we stopped reading workflows from files, so all the workflows that only existed in images were inaccessible. With this change, access to those workflows is restored, and no workflows are lost.
**Updated Workflow Handling in Nodes**
Prior to this change, workflows were embedded in images by passing the whole workflow JSON to a special workflow field on a node. In the node's `invoke()` function, the node was able to access this workflow and save it with the image. This (inaccurately) models workflows as a property of an image and is rather awkward technically.
A workflow is now a property of a batch/session queue item. It is available in the InvocationContext and therefore available to all nodes during `invoke()`.
**Database Migrations**
Added a `SQLiteMigrator` class to handle database migrations. Migrations were needed to accomodate the DB-related changes in this PR. See the code for details.
The `images`, `workflows` and `session_queue` tables required migrations for this PR, and are using the new migrator. Other tables/services are still creating tables themselves. A followup PR will adapt them to use the migrator.
**Other/Support Changes**
- Add a `has_workflow` column to `images` table to indicate that the image has an embedded workflow.
- Add handling for retrieving the workflow from an image in python. The image file must be fetched, the workflow extracted, and then sent to client, avoiding needing the browser to parse the image file. With the `has_workflow` column, the UI knows if there is a workflow to be fetched, and only fetches when the user requests to load the workflow.
- Add route to get the workflow from an image
- Add CRUD service/routes for the library workflows
- `workflow_images` table and services removed (no longer needed now that embedded workflows are not in the DB)
* feat(ui): updated workflow handling (WIP)
Clientside updates for the backend workflow changes.
Includes roughed-out workflow library UI.
* feat: revert SQLiteMigrator class
Will pursue this in a separate PR.
* feat(nodes): do not overwrite custom node module names
Use a different, simpler method to detect if a node is custom.
* feat(nodes): restore WithWorkflow as no-op class
This class is deprecated and no longer needed. Set its workflow attr value to None (meaning it is now a no-op), and issue a warning when an invocation subclasses it.
* fix(nodes): fix get_workflow from queue item dict func
* feat(backend): add WorkflowRecordListItemDTO
This is the id, name, description, created at and updated at workflow columns/attrs. Used to display lists of workflowsl
* chore(ui): typegen
* feat(ui): add workflow loading, deleting to workflow library UI
* feat(ui): workflow library pagination button styles
* wip
* feat: workflow library WIP
- Save to library
- Duplicate
- Filter/sort
- UI/queries
* feat: workflow library - system graphs - wip
* feat(backend): sync system workflows to db
* fix: merge conflicts
* feat: simplify default workflows
- Rename "system" -> "default"
- Simplify syncing logic
- Update UI to match
* feat(workflows): update default workflows
- Update TextToImage_SD15
- Add TextToImage_SDXL
- Add README
* feat(ui): refine workflow list UI
* fix(workflow_records): typo
* fix(tests): fix tests
* feat(ui): clean up workflow library hooks
* fix(db): fix mis-ordered db cleanup step
It was happening before pruning queue items - should happen afterwards, else you have to restart the app again to free disk space made available by the pruning.
* feat(ui): tweak reset workflow editor translations
* feat(ui): split out workflow redux state
The `nodes` slice is a rather complicated slice. Removing `workflow` makes it a bit more reasonable.
Also helps to flatten state out a bit.
* docs: update default workflows README
* fix: tidy up unused files, unrelated changes
* fix(backend): revert unrelated service organisational changes
* feat(backend): workflow_records.get_many arg "filter_text" -> "query"
* feat(ui): use custom hook in current image buttons
Already in use elsewhere, forgot to use it here.
* fix(ui): remove commented out property
* fix(ui): fix workflow loading
- Different handling for loading from library vs external
- Fix bug where only nodes and edges loaded
* fix(ui): fix save/save-as workflow naming
* fix(ui): fix circular dependency
* fix(db): fix bug with releasing without lock in db.clean()
* fix(db): remove extraneous lock
* chore: bump ruff
* fix(workflow_records): default `category` to `WorkflowCategory.User`
This allows old workflows to validate when reading them from the db or image files.
* hide workflow library buttons if feature is disabled
---------
Co-authored-by: Mary Hipp <maryhipp@Marys-MacBook-Air.local>
Adds logic to `DiskLatentsStorage.start()` to empty the latents folder on startup.
Adds start and stop methods to `ForwardCacheLatentsStorage`. This is required for `DiskLatentsStorage.start()` to be called, due to how this particular service breaks the direct DI pattern, wrapping the underlying storage with a cache.
Node authors may now create their own arbitrary/custom field types. Any pydantic model is supported.
Two notes:
1. Your field type's class name must be unique.
Suggest prefixing fields with something related to the node pack as a kind of namespace.
2. Custom field types function as connection-only fields.
For example, if your custom field has string attributes, you will not get a text input for that attribute when you give a node a field with your custom type.
This is the same behaviour as other complex fields that don't have custom UIs in the workflow editor - like, say, a string collection.
feat(ui): fix tooltips for custom types
We need to hold onto the original type of the field so they don't all just show up as "Unknown".
fix(ui): fix ts error with custom fields
feat(ui): custom field types connection validation
In the initial commit, a custom field's original type was added to the *field templates* only as `originalType`. Custom fields' `type` property was `"Custom"`*. This allowed for type safety throughout the UI logic.
*Actually, it was `"Unknown"`, but I changed it to custom for clarity.
Connection validation logic, however, uses the *field instance* of the node/field. Like the templates, *field instances* with custom types have their `type` set to `"Custom"`, but they didn't have an `originalType` property. As a result, all custom fields could be connected to all other custom fields.
To resolve this, we need to add `originalType` to the *field instances*, then switch the validation logic to use this instead of `type`.
This ended up needing a bit of fanagling:
- If we make `originalType` a required property on field instances, existing workflows will break during connection validation, because they won't have this property. We'd need a new layer of logic to migrate the workflows, adding the new `originalType` property.
While this layer is probably needed anyways, typing `originalType` as optional is much simpler. Workflow migration logic can come layer.
(Technically, we could remove all references to field types from the workflow files, and let the templates hold all this information. This feels like a significant change and I'm reluctant to do it now.)
- Because `originalType` is optional, anywhere we care about the type of a field, we need to use it over `type`. So there are a number of `field.originalType ?? field.type` expressions. This is a bit of a gotcha, we'll need to remember this in the future.
- We use `Array.prototype.includes()` often in the workflow editor, e.g. `COLLECTION_TYPES.includes(type)`. In these cases, the const array is of type `FieldType[]`, and `type` is is `FieldType`.
Because we now support custom types, the arg `type` is now widened from `FieldType` to `string`.
This causes a TS error. This behaviour is somewhat controversial (see https://github.com/microsoft/TypeScript/issues/14520). These expressions are now rewritten as `COLLECTION_TYPES.some((t) => t === type)` to satisfy TS. It's logically equivalent.
fix(ui): typo
feat(ui): add CustomCollection and CustomPolymorphic field types
feat(ui): add validation for CustomCollection & CustomPolymorphic types
- Update connection validation for custom types
- Use simple string parsing to determine if a field is a collection or polymorphic type.
- No longer need to keep a list of collection and polymorphic types.
- Added runtime checks in `baseinvocation.py` to ensure no fields are named in such a way that it could mess up the new parsing
chore(ui): remove errant console.log
fix(ui): rename 'nodes.currentConnectionFieldType' -> 'nodes.connectionStartFieldType'
This was confusingly named and kept tripping me up. Renamed to be consistent with the `reactflow` `ConnectionStartParams` type.
fix(ui): fix ts error
feat(nodes): add runtime check for custom field names
"Custom", "CustomCollection" and "CustomPolymorphic" are reserved field names.
chore(ui): add TODO for revising field type names
wip refactor fieldtype structured
wip refactor field types
wip refactor types
wip refactor types
fix node layout
refactor field types
chore: mypy
organisation
organisation
organisation
fix(nodes): fix field orig_required, field_kind and input statuses
feat(nodes): remove broken implementation of default_factory on InputField
Use of this could break connection validation due to the difference in node schemas required fields and invoke() required args.
Removed entirely for now. It wasn't ever actually used by the system, because all graphs always had values provided for fields where default_factory was used.
Also, pydantic is smart enough to not reuse the same object when specifying a default value - it clones the object first. So, the common pattern of `default_factory=list` is extraneous. It can just be `default=[]`.
fix(nodes): fix InputField name validation
workflow validation
validation
chore: ruff
feat(nodes): fix up baseinvocation comments
fix(ui): improve typing & logic of buildFieldInputTemplate
improved error handling in parseFieldType
fix: back compat for deprecated default_factory and UIType
feat(nodes): do not show node packs loaded log if none loaded
chore(ui): typegen
Custom nodes may be places in `$INVOKEAI_ROOT/nodes/` (configurable with `custom_nodes_dir` option).
On app startup, an `__init__.py` is copied into the custom nodes dir, which recursively loads all python files in the directory as modules (files starting with `_` are ignored). The custom nodes dir is now a python module itself.
When we `from invocations import *` to load init all invocations, we load the custom nodes dir, registering all custom nodes.
- Refactor how metadata is handled to support a user-defined metadata in graphs
- Update workflow embed handling
- Update UI to work with these changes
- Update tests to support metadata/workflow changes
This fixes a weird issue where the list images method needed to handle `None` for its `limit` and `offset` arguments, in order to get a count of all intermediates.
Upgrade pydantic and fastapi to latest.
- pydantic~=2.4.2
- fastapi~=103.2
- fastapi-events~=0.9.1
**Big Changes**
There are a number of logic changes needed to support pydantic v2. Most changes are very simple, like using the new methods to serialized and deserialize models, but there are a few more complex changes.
**Invocations**
The biggest change relates to invocation creation, instantiation and validation.
Because pydantic v2 moves all validation logic into the rust pydantic-core, we may no longer directly stick our fingers into the validation pie.
Previously, we (ab)used models and fields to allow invocation fields to be optional at instantiation, but required when `invoke()` is called. We directly manipulated the fields and invocation models when calling `invoke()`.
With pydantic v2, this is much more involved. Changes to the python wrapper do not propagate down to the rust validation logic - you have to rebuild the model. This causes problem with concurrent access to the invocation classes and is not a free operation.
This logic has been totally refactored and we do not need to change the model any more. The details are in `baseinvocation.py`, in the `InputField` function and `BaseInvocation.invoke_internal()` method.
In the end, this implementation is cleaner.
**Invocation Fields**
In pydantic v2, you can no longer directly add or remove fields from a model.
Previously, we did this to add the `type` field to invocations.
**Invocation Decorators**
With pydantic v2, we instead use the imperative `create_model()` API to create a new model with the additional field. This is done in `baseinvocation.py` in the `invocation()` wrapper.
A similar technique is used for `invocation_output()`.
**Minor Changes**
There are a number of minor changes around the pydantic v2 models API.
**Protected `model_` Namespace**
All models' pydantic-provided methods and attributes are prefixed with `model_` and this is considered a protected namespace. This causes some conflict, because "model" means something to us, and we have a ton of pydantic models with attributes starting with "model_".
Forunately, there are no direct conflicts. However, in any pydantic model where we define an attribute or method that starts with "model_", we must tell set the protected namespaces to an empty tuple.
```py
class IPAdapterModelField(BaseModel):
model_name: str = Field(description="Name of the IP-Adapter model")
base_model: BaseModelType = Field(description="Base model")
model_config = ConfigDict(protected_namespaces=())
```
**Model Serialization**
Pydantic models no longer have `Model.dict()` or `Model.json()`.
Instead, we use `Model.model_dump()` or `Model.model_dump_json()`.
**Model Deserialization**
Pydantic models no longer have `Model.parse_obj()` or `Model.parse_raw()`, and there are no `parse_raw_as()` or `parse_obj_as()` functions.
Instead, you need to create a `TypeAdapter` object to parse python objects or JSON into a model.
```py
adapter_graph = TypeAdapter(Graph)
deserialized_graph_from_json = adapter_graph.validate_json(graph_json)
deserialized_graph_from_dict = adapter_graph.validate_python(graph_dict)
```
**Field Customisation**
Pydantic `Field`s no longer accept arbitrary args.
Now, you must put all additional arbitrary args in a `json_schema_extra` arg on the field.
**Schema Customisation**
FastAPI and pydantic schema generation now follows the OpenAPI version 3.1 spec.
This necessitates two changes:
- Our schema customization logic has been revised
- Schema parsing to build node templates has been revised
The specific aren't important, but this does present additional surface area for bugs.
**Performance Improvements**
Pydantic v2 is a full rewrite with a rust backend. This offers a substantial performance improvement (pydantic claims 5x to 50x depending on the task). We'll notice this the most during serialization and deserialization of sessions/graphs, which happens very very often - a couple times per node.
I haven't done any benchmarks, but anecdotally, graph execution is much faster. Also, very larges graphs - like with massive iterators - are much, much faster.
Refactor services folder/module structure.
**Motivation**
While working on our services I've repeatedly encountered circular imports and a general lack of clarity regarding where to put things. The structure introduced goes a long way towards resolving those issues, setting us up for a clean structure going forward.
**Services**
Services are now in their own folder with a few files:
- `services/{service_name}/__init__.py`: init as needed, mostly empty now
- `services/{service_name}/{service_name}_base.py`: the base class for the service
- `services/{service_name}/{service_name}_{impl_type}.py`: the default concrete implementation of the service - typically one of `sqlite`, `default`, or `memory`
- `services/{service_name}/{service_name}_common.py`: any common items - models, exceptions, utilities, etc
Though it's a bit verbose to have the service name both as the folder name and the prefix for files, I found it is _extremely_ confusing to have all of the base classes just be named `base.py`. So, at the cost of some verbosity when importing things, I've included the service name in the filename.
There are some minor logic changes. For example, in `InvocationProcessor`, instead of assigning the model manager service to a variable to be used later in the file, the service is used directly via the `Invoker`.
**Shared**
Things that are used across disparate services are in `services/shared/`:
- `default_graphs.py`: previously in `services/`
- `graphs.py`: previously in `services/`
- `paginatation`: generic pagination models used in a few services
- `sqlite`: the `SqliteDatabase` class, other sqlite-specific things
**Service Dependencies**
Services that depend on other services now access those services via the `Invoker` object. This object is provided to the service as a kwarg to its `start()` method.
Until now, most services did not utilize this feature, and several services required their dependencies to be initialized and passed in on init.
Additionally, _all_ services are now registered as invocation services - including the low-level services. This obviates issues with inter-dependent services we would otherwise experience as we add workflow storage.
**Database Access**
Previously, we were passing in a separate sqlite connection and corresponding lock as args to services in their init. A good amount of posturing was done in each service that uses the db.
These objects, along with the sqlite startup and cleanup logic, is now abstracted into a simple `SqliteDatabase` class. This creates the shared connection and lock objects, enables foreign keys, and provides a `clean()` method to do startup db maintenance.
This is not a service as it's only used by sqlite services.
The UI will always re-fetch queue and batch status on receiving this event, so we may as well jsut include that data in the event and save the extra network roundtrips.
When the processor has an error and it has a queue item, mark that item failed.
This addresses processor errors resulting in `in_progress` queue items, which create a soft lock of the processor, requiring the user to cancel the `in_progress` item before anything else processes.
Makes graph validation logic more rigorous, validating graphs when they are created as part of a session or batch.
`validate_self()` method added to `Graph` model. It does all the validation that `is_valid()` did, plus a few extras:
- unique `node.id` values across graph
- node ids match their key in `Graph.nodes`
- recursively validate subgraphs
- validate all edges
- validate graph is acyclical
The new method is required because `is_valid()` just returned a boolean. That behaviour is retained, but `validate_self()` now raises appropriate exceptions for validation errors. This are then surfaced to the client.
The function is named `validate_self()` because pydantic reserves `validate()`.
There are two main places where graphs are created - in batches and in sessions.
Field validators are added to each of these for their `graph` fields, which call the new validation logic.
**Closes #4744**
In this issue, a batch is enqueued with an invalid graph. The output field is typed as optional while the input field is required. The field types themselves are not relevant - this change addresses the case where an invalid graph was created.
The mismatched types problem is not noticed until we attempt to invoke the graph, because the graph was never *fully* validated. An error is raised during the call to `graph_execution_state.next()` in `invoker.py`. This function prepares the edges and validates them, raising an exception due to the mismatched types.
This exception is caught by the session processor, but it doesn't handle this situation well - the graph is not marked as having an error and the queue item status is never changed. The queue item is therefore forever `in_progress`, so no new queue items are popped - the app won't do anything until the queue item is canceled manually.
This commit addresses this by preventing invalid graphs from being created in the first place, addressing a substantial number of fail cases.
The compress_level setting of PIL.Image.save(), used for PNG encoding. All settings are lossless. 0 = fastest, largest filesize, 9 = slowest, smallest filesize
Closes#4786
* fix(nodes): do not disable invocation cache delete methods
When the runtime disabled flag is on, do not skip the delete methods. This could lead to a hit on a missing resource.
Do skip them when the cache size is 0, because the user cannot change this (must restart app to change it).
* fix(nodes): do not use double-underscores in cache service
* Thread lock for cache
* Making cache LRU
* Bug fixes
* bugfix
* Switching to one Lock and OrderedDict cache
* Removing unused imports
* Move lock cache instance
* Addressing PR comments
---------
Co-authored-by: psychedelicious <4822129+psychedelicious@users.noreply.github.com>
Co-authored-by: Martin Kristiansen <martin@modyfi.io>
When the runtime disabled flag is on, do not skip the delete methods. This could lead to a hit on a missing resource.
Do skip them when the cache size is 0, because the user cannot change this (must restart app to change it).
- New routes to clear, enable, disable and get the status of the cache
- Status includes hits, misses, size, max size, enabled
- Add client cache queries and mutations, abstracted into hooks
- Add invocation cache status area (next to queue status) w/ buttons
Add `batch_id` to outbound events. This necessitates adding it to both `InvocationContext` and `InvocationQueueItem`. This allows the canvas to receive images.
When the user enqueues a batch on the canvas, it is expected that all images from that batch are directed to the canvas.
The simplest, most flexible solution is to add the `batch_id` to the invocation context-y stuff. Then everything knows what batch it came from, and we can have the canvas pick up images associated with its list of canvas `batch_id`s.
This change enhances the invocation cache logic to delete cache entries when the resources to which they refer are deleted.
For example, a cached output may refer to "some_image.png". If that image is deleted, and this particular cache entry is later retrieved by a node, that node's successors will receive references to the now non-existent "some_image.png". When they attempt to use that image, they will fail.
To resolve this, we need to invalidate the cache when the resources to which it refers are deleted. Two options:
- Invalidate the whole cache on every image/latents/etc delete
- Selectively invalidate cache entries when their resources are deleted
Node outputs can be any shape, with any number of resource references in arbitrarily nested pydantic models. Traversing that structure to identify resources is not trivial.
But invalidating the whole cache is a bit heavy-handed. It would be nice to be more selective.
Simple solution:
- Invocation outputs' resource references are always string identifiers - like the image's or latents' name
- Invocation outputs can be stringified, which includes said identifiers
- When the invocation is cached, we store the stringified output alongside the "live" output classes
- When a resource is deleted, pass its identifier to the cache service, which can then invalidate any cache entries that refer to it
The images and latents storage services have been outfitted with `on_deleted()` callbacks, and the cache service registers itself to handle those events. This logic was copied from `ItemStorageABC`.
`on_changed()` callback are also added to the images and latents services, though these are not currently used. Just following the existing pattern.
* fix(config): fix typing issues in `config/`
`config/invokeai_config.py`:
- use `Optional` for things that are optional
- fix typing of `ram_cache_size()` and `vram_cache_size()`
- remove unused and incorrectly typed method `autoconvert_path`
- fix types and logic for `parse_args()`, in which `InvokeAIAppConfig.initconf` *must* be a `DictConfig`, but function would allow it to be set as a `ListConfig`, which presumably would cause issues elsewhere
`config/base.py`:
- use `cls` for first arg of class methods
- use `Optional` for things that are optional
- fix minor type issue related to setting of `env_prefix`
- remove unused `add_subparser()` method, which calls `add_parser()` on an `ArgumentParser` (method only available on the `_SubParsersAction` object, which is returned from ArgumentParser.add_subparsers()`)
* feat: queued generation and batches
Due to a very messy branch with broad addition of `isort` on `main` alongside it, some git surgery was needed to get an agreeable git history. This commit represents all of the work on queued generation. See PR for notes.
* chore: flake8, isort, black
* fix(nodes): fix incorrect service stop() method
* fix(nodes): improve names of a few variables
* fix(tests): fix up tests after changes to batches/queue
* feat(tests): add unit tests for session queue helper functions
* feat(ui): dynamic prompts is always enabled
* feat(queue): add queue_status_changed event
* feat(ui): wip queue graphs
* feat(nodes): move cleanup til after invoker startup
* feat(nodes): add cancel_by_batch_ids
* feat(ui): wip batch graphs & UI
* fix(nodes): remove `Batch.batch_id` from required
* fix(ui): cleanup and use fixedCacheKey for all mutations
* fix(ui): remove orphaned nodes from canvas graphs
* fix(nodes): fix cancel_by_batch_ids result count
* fix(ui): only show cancel batch tooltip when batches were canceled
* chore: isort
* fix(api): return `[""]` when dynamic prompts generates no prompts
Just a simple fallback so we always have a prompt.
* feat(ui): dynamicPrompts.combinatorial is always on
There seems to be little purpose in using the combinatorial generation for dynamic prompts. I've disabled it by hiding it from the UI and defaulting combinatorial to true. If we want to enable it again in the future it's straightforward to do so.
* feat: add queue_id & support logic
* feat(ui): fix upscale button
It prepends the upscale operation to queue
* feat(nodes): return queue item when enqueuing a single graph
This facilitates one-off graph async workflows in the client.
* feat(ui): move controlnet autoprocess to queue
* fix(ui): fix non-serializable DOMRect in redux state
* feat(ui): QueueTable performance tweaks
* feat(ui): update queue list
Queue items expand to show the full queue item. Just as JSON for now.
* wip threaded session_processor
* feat(nodes,ui): fully migrate queue to session_processor
* feat(nodes,ui): add processor events
* feat(ui): ui tweaks
* feat(nodes,ui): consolidate events, reduce network requests
* feat(ui): cleanup & abstract queue hooks
* feat(nodes): optimize batch permutation
Use a generator to do only as much work as is needed.
Previously, though we only ended up creating exactly as many queue items as was needed, there was still some intermediary work that calculated *all* permutations. When that number was very high, the system had a very hard time and used a lot of memory.
The logic has been refactored to use a generator. Additionally, the batch validators are optimized to return early and use less memory.
* feat(ui): add seed behaviour parameter
This dynamic prompts parameter allows the seed to be randomized per prompt or per iteration:
- Per iteration: Use the same seed for all prompts in a single dynamic prompt expansion
- Per prompt: Use a different seed for every single prompt
"Per iteration" is appropriate for exploring a the latents space with a stable starting noise, while "Per prompt" provides more variation.
* fix(ui): remove extraneous random seed nodes from linear graphs
* fix(ui): fix controlnet autoprocess not working when queue is running
* feat(queue): add timestamps to queue status updates
Also show execution time in queue list
* feat(queue): change all execution-related events to use the `queue_id` as the room, also include `queue_item_id` in InvocationQueueItem
This allows for much simpler handling of queue items.
* feat(api): deprecate sessions router
* chore(backend): tidy logging in `dependencies.py`
* fix(backend): respect `use_memory_db`
* feat(backend): add `config.log_sql` (enables sql trace logging)
* feat: add invocation cache
Supersedes #4574
The invocation cache provides simple node memoization functionality. Nodes that use the cache are memoized and not re-executed if their inputs haven't changed. Instead, the stored output is returned.
## Results
This feature provides anywhere some significant to massive performance improvement.
The improvement is most marked on large batches of generations where you only change a couple things (e.g. different seed or prompt for each iteration) and low-VRAM systems, where skipping an extraneous model load is a big deal.
## Overview
A new `invocation_cache` service is added to handle the caching. There's not much to it.
All nodes now inherit a boolean `use_cache` field from `BaseInvocation`. This is a node field and not a class attribute, because specific instances of nodes may want to opt in or out of caching.
The recently-added `invoke_internal()` method on `BaseInvocation` is used as an entrypoint for the cache logic.
To create a cache key, the invocation is first serialized using pydantic's provided `json()` method, skipping the unique `id` field. Then python's very fast builtin `hash()` is used to create an integer key. All implementations of `InvocationCacheBase` must provide a class method `create_key()` which accepts an invocation and outputs a string or integer key.
## In-Memory Implementation
An in-memory implementation is provided. In this implementation, the node outputs are stored in memory as python classes. The in-memory cache does not persist application restarts.
Max node cache size is added as `node_cache_size` under the `Generation` config category.
It defaults to 512 - this number is up for discussion, but given that these are relatively lightweight pydantic models, I think it's safe to up this even higher.
Note that the cache isn't storing the big stuff - tensors and images are store on disk, and outputs include only references to them.
## Node Definition
The default for all nodes is to use the cache. The `@invocation` decorator now accepts an optional `use_cache: bool` argument to override the default of `True`.
Non-deterministic nodes, however, should set this to `False`. Currently, all random-stuff nodes, including `dynamic_prompt`, are set to `False`.
The field name `use_cache` is now effectively a reserved field name and possibly a breaking change if any community nodes use this as a field name. In hindsight, all our reserved field names should have been prefixed with underscores or something.
## One Gotcha
Leaf nodes probably want to opt out of the cache, because if they are not cached, their outputs are not saved again.
If you run the same graph multiple times, you only end up with a single image output, because the image storage side-effects are in the `invoke()` method, which is bypassed if we have a cache hit.
## Linear UI
The linear graphs _almost_ just work, but due to the gotcha, we need to be careful about the final image-outputting node. To resolve this, a `SaveImageInvocation` node is added and used in the linear graphs.
This node is similar to `ImagePrimitive`, except it saves a copy of its input image, and has `use_cache` set to `False` by default.
This is now the leaf node in all linear graphs, and is the only node in those graphs with `use_cache == False` _and_ the only node with `is_intermedate == False`.
## Workflow Editor
All nodes now have a footer with a new `Use Cache [ ]` checkbox. It defaults to the value set by the invocation in its python definition, but can be changed by the user.
The workflow/node validation logic has been updated to migrate old workflows to use the new default values for `use_cache`. Users may still want to review the settings that have been chosen. In the event of catastrophic failure when running this migration, the default value of `True` is applied, as this is correct for most nodes.
Users should consider saving their workflows after loading them in and having them updated.
## Future Enhancements - Callback
A future enhancement would be to provide a callback to the `use_cache` flag that would be run as the node is executed to determine, based on its own internal state, if the cache should be used or not.
This would be useful for `DynamicPromptInvocation`, where the deterministic behaviour is determined by the `combinatorial: bool` field.
## Future Enhancements - Persisted Cache
Similar to how the latents storage is backed by disk, the invocation cache could be persisted to the database or disk. We'd need to be very careful about deserializing outputs, but it's perhaps worth exploring in the future.
* fix(ui): fix queue list item width
* feat(nodes): do not send the whole node on every generator progress
* feat(ui): strip out old logic related to sessions
Things like `isProcessing` are no longer relevant with queue. Removed them all & updated everything be appropriate for queue. May be a few little quirks I've missed...
* feat(ui): fix up param collapse labels
* feat(ui): click queue count to go to queue tab
* tidy(queue): update comment, query format
* feat(ui): fix progress bar when canceling
* fix(ui): fix circular dependency
* feat(nodes): bail on node caching logic if `node_cache_size == 0`
* feat(nodes): handle KeyError on node cache pop
* feat(nodes): bypass cache codepath if caches is disabled
more better no do thing
* fix(ui): reset api cache on connect/disconnect
* feat(ui): prevent enqueue when no prompts generated
* feat(ui): add queue controls to workflow editor
* feat(ui): update floating buttons & other incidental UI tweaks
* fix(ui): fix missing/incorrect translation keys
* fix(tests): add config service to mock invocation services
invoking needs access to `node_cache_size` to occur
* optionally remove pause/resume buttons from queue UI
* option to disable prepending
* chore(ui): remove unused file
* feat(queue): remove `order_id` entirely, `item_id` is now an autoinc pk
---------
Co-authored-by: Mary Hipp <maryhipp@Marys-MacBook-Air.local>
Allow denying and explicitly allowing nodes. When a not-allowed node is used, a pydantic `ValidationError` will be raised.
- When collecting all invocations, check against the allowlist and denylist first. When pydantic constructs any unions related to nodes, the denied nodes will be omitted
- Add `allow_nodes` and `deny_nodes` to `InvokeAIAppConfig`. These are `Union[list[str], None]`, and may be populated with the `type` of invocations.
- When `allow_nodes` is `None`, allow all nodes, else if it is `list[str]`, only allow nodes in the list
- When `deny_nodes` is `None`, deny no nodes, else if it is `list[str]`, deny nodes in the list
- `deny_nodes` overrides `allow_nodes`
All invocation metadata (type, title, tags and category) are now defined in decorators.
The decorators add the `type: Literal["invocation_type"]: "invocation_type"` field to the invocation.
Category is a new invocation metadata, but it is not used by the frontend just yet.
- `@invocation()` decorator for invocations
```py
@invocation(
"sdxl_compel_prompt",
title="SDXL Prompt",
tags=["sdxl", "compel", "prompt"],
category="conditioning",
)
class SDXLCompelPromptInvocation(BaseInvocation, SDXLPromptInvocationBase):
...
```
- `@invocation_output()` decorator for invocation outputs
```py
@invocation_output("clip_skip_output")
class ClipSkipInvocationOutput(BaseInvocationOutput):
...
```
- update invocation docs
- add category to decorator
- regen frontend types
- move docstrings to ABC
- `start_time: int` -> `start_time: float`
- remove class attribute assignments in `StatsContext`
- add `update_mem_stats()` to ABC
- add class attributes to ABC, because they are referenced in instances of the class. if they should not be on the ABC, then maybe there needs to be some restructuring
When retrieving a graph, it is parsed through pydantic. It is possible that this graph is invalid, and an error is thrown.
Handle this by deleting the failed graph from the stats if this occurs.
- move docstrings to ABC
- `start_time: int` -> `start_time: float`
- remove class attribute assignments in `StatsContext`
- add `update_mem_stats()` to ABC
- add class attributes to ABC, because they are referenced in instances of the class. if they should not be on the ABC, then maybe there needs to be some restructuring
When retrieving a graph, it is parsed through pydantic. It is possible that this graph is invalid, and an error is thrown.
Handle this by deleting the failed graph from the stats if this occurs.
Refine concept of "parameter" nodes to "primitives":
- integer
- float
- string
- boolean
- image
- latents
- conditioning
- color
Each primitive has:
- A field definition, if it is not already python primitive value. The field is how this primitive value is passed between nodes. Collections are lists of the field in node definitions. ex: `ImageField` & `list[ImageField]`
- A single output class. ex: `ImageOutput`
- A collection output class. ex: `ImageCollectionOutput`
- A node, which functions to load or pass on the primitive value. ex: `ImageInvocation` (in this case, `ImageInvocation` replaces `LoadImage`)
Plus a number of related changes:
- Reorganize these into `primitives.py`
- Update all nodes and logic to use primitives
- Consolidate "prompt" outputs into "string" & "mask" into "image" (there's no reason for these to be different, the function identically)
- Update default graphs & tests
- Regen frontend types & minor frontend tidy related to changes
multi-select actions include:
- drag to board to move all to that board
- right click to add all to board or delete all
backend changes:
- add routes for changing board for list of image names, deleting list of images
- change image-specific routes to `images/i/{image_name}` to not clobber other routes (like `images/upload`, `images/delete`)
- subclass pydantic `BaseModel` as `BaseModelExcludeNull`, which excludes null values when calling `dict()` on the model. this fixes inconsistent types related to JSON parsing null values into `null` instead of `undefined`
- remove `board_id` from `remove_image_from_board`
frontend changes:
- multi-selection stuff uses `ImageDTO[]` as payloads, for dnd and other mutations. this gives us access to image `board_id`s when hitting routes, and enables efficient cache updates.
- consolidate change board and delete image modals to handle single and multiples
- board totals are now re-fetched on mutation and not kept in sync manually - was way too tedious to do this
- fixed warning about nested `<p>` elements
- closes#4088 , need to handle case when `autoAddBoardId` is `"none"`
- add option to show gallery image delete button on every gallery image
frontend refactors/organisation:
- make typegen script js instead of ts
- enable `noUncheckedIndexedAccess` to help avoid bugs when indexing into arrays, many small changes needed to satisfy TS after this
- move all image-related endpoints into `endpoints/images.ts`, its a big file now, but this fixes a number of circular dependency issues that were otherwise felt impossible to resolve
- Create abstract base class InvocationStatsServiceBase
- Store InvocationStatsService in the InvocationServices object
- Collect and report stats on simultaneous graph execution
independently for each graph id
- Track VRAM usage for each node
- Handle cancellations and other exceptions gracefully
When a queue item is popped for processing, we need to retrieve its session from the DB. Pydantic serializes the graph at this stage.
It's possible for a graph to have been made invalid during the graph preparation stage (e.g. an ancestor node executes, and its output is not valid for its successor node's input field).
When this occurs, the session in the DB will fail validation, but we don't have a chance to find out until it is retrieved and parsed by pydantic.
This logic was previously not wrapped in any exception handling.
Just after retrieving a session, we retrieve the specific invocation to execute from the session. It's possible that this could also have some sort of error, though it should be impossible for it to be a pydantic validation error (that would have been caught during session validation). There was also no exception handling here.
When either of these processes fail, the processor gets soft-locked because the processor's cleanup logic is never run. (I didn't dig deeper into exactly what cleanup is not happening, because the fix is to just handle the exceptions.)
This PR adds exception handling to both the session retrieval and node retrieval and events for each: `session_retrieval_error` and `invocation_retrieval_error`.
These events are caught and displayed in the UI as toasts, along with the type of the python exception (e.g. `Validation Error`). The events are also logged to the browser console.
* feat(ui): enhance clear intermediates feature
- retrieve the # of intermediates using a new query (just uses list images endpoint w/ limit of 0)
- display the count in the UI
- add types for clearIntermediates mutation
- minor styling and verbiage changes
* feat(ui): remove unused settings option for guides
* feat(ui): use solid badge variant
consistent with the rest of the usage of badges
* feat(ui): update board ctx menu, add board auto-add
- add context menu to system boards - only open is select board. did this so that you dont think its broken when you click it
- add auto-add board. you can right click a user board to enable it for auto-add, or use the gallery settings popover to select it. the invoke button has a tooltip on a short delay to remind you that you have auto-add enabled
- made useBoardName hook, provide it a board id and it gets your the board name
- removed `boardIdToAdTo` state & logic, updated workflows to auto-switch and auto-add on image generation
* fix(ui): clear controlnet when clearing intermediates
* feat: Make Add Board icon a button
* feat(db, api): clear intermediates now clears all of them
* feat(ui): make reset webui text subtext style
* feat(ui): board name change submits on blur
---------
Co-authored-by: blessedcoolant <54517381+blessedcoolant@users.noreply.github.com>
* feat(ui): migrate listImages to RTK query using createEntityAdapter
- see comments in `endpoints/images.ts` for explanation of the caching
- so far, only manually updating `all` images when new image is generated. no other manual cache updates are implemented, but will be needed.
- fixed some weirdness with loading state components (like the spinners in gallery)
- added `useThumbnailFallback` for `IAIDndImage`, this displays the tiny webp thumbnail while the full-size images load
- comment out some old thunk related stuff in gallerySlice, which is no longer needed
* feat(ui): add manual cache updates for board changes (wip)
- update RTK Query caches when adding/removing single image to/from board
- work more on migrating all image-related operations to RTK Query
* update AddImagesToBoardContext so that it works when user uses context menu + modal
* handle case where no image is selected
* get assets working for main list and boards - dnd only
* feat(ui): migrate image uploads to RTK Query
- minor refactor of `ImageUploader` and `useImageUploadButton` hooks, simplify some logic
- style filesystem upload overlay to match existing UI
- replace all old `imageUploaded` thunks with `uploadImage` RTK Query calls, update associated logic including canvas related uploads
- simplify `PostUploadAction`s that only need to display user input
* feat(ui): remove `receivedPageOfImages` thunks
* feat(ui): remove `receivedImageUrls` thunk
* feat(ui): finish removing all images thunks
stuff now broken:
- image usage
- delete board images
- on first load, no image selected
* feat(ui): simplify `updateImage` cache manipulation
- we don't actually ever change categories, so we can remove a lot of logic
* feat(ui): simplify canvas autosave
- instead of using a network request to set the canvas generation as not intermediate, we can just do that in the graph
* feat(ui): simplify & handle edge cases in cache updates
* feat(db, api): support `board_id='none'` for `get_many` images queries
This allows us to get all images that are not on a board.
* chore(ui): regen types
* feat(ui): add `All Assets`, `No Board` boards
Restructure boards:
- `all images` is all images
- `all assets` is all assets
- `no board` is all images/assets without a board set
- user boards may have images and assets
Update caching logic
- much simpler without every board having sub-views of images and assets
- update drag and drop operations for all possible interactions
* chore(ui): regen types
* feat(ui): move download to top of context menu
* feat(ui): improve drop overlay styles
* fix(ui): fix image not selected on first load
- listen for first load of all images board, then select the first image
* feat(ui): refactor board deletion
api changes:
- add route to list all image names for a board. this is required to handle board + image deletion. we need to know every image in the board to determine the image usage across the app. this is fetched only when the delete board and images modal is opened so it's as efficient as it can be.
- update the delete board route to respond with a list of deleted `board_images` and `images`, as image names. this is needed to perform accurate clientside state & cache updates after deleting.
db changes:
- remove unused `board_images` service method to get paginated images dtos for a board. this is now done thru the list images endpoint & images service. needs a small logic change on `images.delete_images_on_board`
ui changes:
- simplify the delete board modal - no context, just minor prop drilling. this is feasible for boards only because the components that need to trigger and manipulate the modal are very close together in the tree
- add cache updates for `deleteBoard` & `deleteBoardAndImages` mutations
- the only thing we cannot do directly is on `deleteBoardAndImages`, update the `No Board` board. we'd need to insert image dtos that we may not have loaded. instead, i am just invalidating the tags for that `listImages` cache. so when you `deleteBoardAndImages`, the `No Board` will re-fetch the initial image limit. i think this is more efficient than e.g. fetching all image dtos to insert then inserting them.
- handle image usage for `deleteBoardAndImages`
- update all (i think/hope) the little bits and pieces in the UI to accomodate these changes
* fix(ui): fix board selection logic
* feat(ui): add delete board modal loading state
* fix(ui): use thumbnails for board cover images
* fix(ui): fix race condition with board selection
when selecting a board that doesn't have any images loaded, we need to wait until the images haveloaded before selecting the first image.
this logic is debounced to ~1000ms.
* feat(ui): name 'No Board' correctly, change icon
* fix(ui): do not cache listAllImageNames query
if we cache it, we can end up with stale image usage during deletion.
we could of course manually update the cache as we are doing elsewhere. but because this is a relatively infrequent network request, i'd like to trade increased cache mgmt complexity here for increased resource usage.
* feat(ui): reduce drag preview opacity, remove border
* fix(ui): fix incorrect queryArg used in `deleteImage` and `updateImage` cache updates
* fix(ui): fix doubled open in new tab
* fix(ui): fix new generations not getting added to 'No Board'
* fix(ui): fix board id not changing on new image when autosave enabled
* fix(ui): context menu when selection is 0
need to revise how context menu is triggered later, when we approach multi select
* fix(ui): fix deleting does not update counts for all images and all assets
* fix(ui): fix all assets board name in boards list collapse button
* fix(ui): ensure we never go under 0 for total board count
* fix(ui): fix text overflow on board names
---------
Co-authored-by: Mary Hipp <maryhipp@Marys-MacBook-Air.local>
* new route to clear intermediates
* UI to clear intermediates from settings modal
* cleanup
* PR feedback
---------
Co-authored-by: Mary Hipp <maryhipp@Marys-MacBook-Air.local>
