Real semantic embeddings via @xenova/transformers. The model (all-MiniLM-L6-v2, 384-dim) downloads once and is cached by the browser — later sessions run offline. The library is lazy-loaded (code-split) so it's never in the core bundle if you don't use it. The vector store itself runs on CPU fallback (WASM-SIMD) if WebGPU is unavailable.
ready.