OnnxAdapterFactory 0.1.2 API

All Implemented Interfaces:

ai.zerogpu.android.sdk.adapter.AdapterFactory
```
public final class OnnxAdapterFactory
 implements AdapterFactory
                    
```
The ONNX engine, backed by Microsoft ONNX Runtime (inference) + the HuggingFace tokenizer (reads the standard tokenizer.json). It consumes the same model artifacts as the browser SDK (Transformers.js) and the cloud fallback — model.onnx + tokenizer.json + config.json — so a model never needs an Android-specific re-export.
Behaviour is data: ModelConfig.inferenceKind selects a small, bounded post-processor in OnnxAdapter; labels come from config.json. Adding a model of a supported kind is a catalog row, not SDK code.

Nested Class Summary

Nested Classes
Modifier and Type	Class	Description
`public class`	`OnnxAdapterFactory.Companion`

Field Summary

Fields
Modifier and Type	Field	Description
`private final List<Capability>`	`capabilities`

Constructor Summary

Constructors
Constructor	Description
`OnnxAdapterFactory()`

Method Summary

Modifier and Type	Method	Description
`List<Capability>`	`getCapabilities()`	The kinds this engine implements.
`InferenceAdapter`	`load(ModelConfig config, Map<String, File> files)`

Methods inherited from class java.lang.Object
clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

- Constructor Detail
  - OnnxAdapterFactory
```
OnnxAdapterFactory()
```
- Method Detail
  - getCapabilities
```
 List<Capability> getCapabilities()
```
    The kinds this engine implements. The only thing the registry routes on.
  - load
```
 InferenceAdapter load(ModelConfig config, Map<String, File> files)
```