Module embedder 

Embedding generation via fastembed (ONNX-backed, no server required).

The embedder wraps a fastembed TextEmbedding model and serializes/ deserializes raw f32 vectors for storage in SQLite BLOBs.

Structs

Embedder

Wraps the fastembed model and provides encode/decode helpers.

Constants

DEFAULT_MODEL

Default embedding model — nomic-embed-text-v1.5 gives 768 dimensions and good code+text mixed-domain performance with Matryoshka support.

Functions

cosine_similarity

Cosine similarity between two equal-length vectors. Returns 0.0 if either vector has zero magnitude.

decode_vector

Decode a little-endian byte blob back to f32 values.

dims_for_model

Return the known output dimensionality for a model without loading it.

encode_vector

Convert a slice of f32 to a little-endian byte blob for SQLite storage.