Pure-Java FFM bindings to Windows D3D12, DirectML and DXGI, plus DirectML kernel primitives (Linear, LayerNorm, Attention, GELU, ...). Supports Windows in-box DirectML with feature-level fallbacks; some fast paths use newer DirectML feature levels when available.

BERT-class encoder pipelines (MiniLM, E5, cross-encoder reranker) on DirectML with CPU fallback, embeddings, and reranking with per-layer coalesced GPU submissions.

Neutral, swing-free Java 21 local generation runtime for Windows DirectML/WARP. Public API in com.aresstack.windirectml.inference.api (GenerationRuntime, GenerationModelHandle, catalog-driven family adapters for Qwen, SmolLM2, Gemma 3, Phi-3 and ...