Module embedding 

Embedding abstraction for semantic similarity.

Supports two embedding modes:

• Sparse (TF-IDF): Zero-dependency, works for any language.
• Dense (f32): Produced by ONNX or API-based models (OpenAI, etc.).

Dense vectors are used by the HNSW index for fast ANN search. Sparse vectors serve as a fallback when no embedding model is available.

Structs

TfIdfEmbeddingProvider

TF-IDF based embedding provider (zero dependencies).

Enums

EmbeddingVector

An embedding vector for semantic similarity comparison.

Traits

EmbeddingProvider

Provider for generating text embeddings.