voxora-engine

Canonical adapter contract for voxora engine crates (ASR-specific).

Every voxora-<engine> crate (voxora-whisper, voxora-qwen3asr, future voxora-parakeet, voxora-voxtral, …) builds on top of this crate so consumers see a uniform shape: an [EngineAdapter] trait that exposes family, info, backend, and the wrapped [voxora_traits::AsrEngine].

The adapter wraps voxora_traits::AsrEngine — there is no generic Model trait at this layer. The next planned engine family (parakeet, voxtral, …) lands as a new variant on [EngineFamily] in 0.2.0; today the enum covers Whisper and Qwen3Asr and is #[non_exhaustive].

Re-exports: [AnyEngine], [BackendDescriptor], [BackendKind], [EngineFamily], [EngineInfo], [InvalidEngineFamily], and the testing::MockAdapter helper for downstream test suites.

Hardware backends

[BackendKind] is the canonical enum shared by every engine adapter (Cpu / Cuda / Metal / Vulkan); [BackendDescriptor] is what an adapter returns from EngineAdapter::backend. The cross-engine hardware / GPU support matrix — which engine ships which backend, CUDA compute capability per backend, the voxora-bridge forwarding rules — is documented in docs/GPU_SUPPORT.md. The runtime picker ("CUDA → Metal → CPU at process start") lives in voxora-backend behind its opt-in candle Cargo feature.

Examples

• adapter_dispatch — wrap a [MockAdapter] behind [AnyEngine], dispatch on family(), and call the borrowed [voxora_traits::AsrEngine] synchronously. The smallest possible program that proves the adapter contract works without touching Hugging Face or loading a real model. Run with cargo run --example adapter_dispatch -p voxora-engine.