feat: allow inputs as dict for image/video/3d/audio inference and image upscale and Settings and Safety parameter too

[Sirsho1997]

Created based on PR of forked repo here

Fixes #244

Usage

task_uuid = str(uuid.uuid4())
deliveryMethod = "sync"

request_params = {
    "model": "....,
    "taskUUID": task_uuid,
    "positivePrompt": "..... ",
    "width": 1024,
    "height": 1024,
    "seed": 8,
    "numberResults": 1,
    "deliveryMethod": deliveryMethod,
    "inputs": {
        "referenceImages": [
            ".....",
            ".....",
        ],
    },
}

request_image = IImageInference(**request_params)

Added

• Added post_init to IImageInference, IImageCaption, IImageUpscale, IVideoInference, I3dInference, IAudioInference, IVideoBackgroundRemoval

[Copilot]

Pull request overview

This PR updates the SDK request payload types to accept dict values for nested request sections (notably inputs, settings, and safety) and coerces them into their corresponding dataclass/SerializableMixin types in __post_init__. This aligns the SDK with common “build params as dict” usage patterns and prevents runtime AttributeError during request serialization.

Changes:

• Allow inputs/settings/safety fields to be passed as dict on several inference/upscale payloads via Union[..., Dict[str, Any]].
• Add __post_init__ coercion for these fields to instantiate IInputs, ISettings, ISafety, etc.
• Extend the same coercion approach to background removal payload settings.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.