Module local_onnx 

Local ONNX Runtime provider (local/onnx).

Single backend, multiple tasks. The provider declares ModelTask::Raw and ModelTask::Rerank in ProviderCapabilities and dispatches inside ModelProvider::load — matching the pattern used by cohere.rs, mistralrs.rs, etc.

Task implementations live in private submodules:

• raw — arbitrary ONNX tensor execution via RawTensorModel.
• rerank — cross-encoder reranking via RerankerModel.
• embedding — dense text embeddings via EmbeddingModel.

Structs

LocalOnnxProvider