Skip to content

Capabilities ​

@anvia/fastembed implements local dense and sparse Anvia embedding contracts.

CapabilitySupport
Dense document/query vectorsFastEmbedEmbeddingModel
Sparse passage vectorsFastEmbedSparseEmbeddingModel.embedTexts()
Sparse query vectorsFastEmbedSparseEmbeddingModel.embedQuery()
Hybrid retrievalCombine dense and sparse with a capable vector store
Remote API callsNone after model assets are available locally
Browser runtimeNot the intended target

Dense embeddings ​

The default model is fast-bge-small-en-v1.5. The adapter accepts FastEmbed’s non-custom model enum values, batches 256 inputs by default, and returns one Anvia Embedding per input.

Runtime batches may contain plain arrays or typed arrays. The adapter validates the batch/vector shapes and final output count.

Sparse embeddings ​

The default sparse model is prithivida/Splade_PP_en_v1. Passage embedding accepts batches; query embedding accepts one query. Returned parallel indices and values arrays must be numeric and have equal lengths.

Sparse output is useful only with a store and search configuration that understands sparse or hybrid vectors. Qdrant can combine named dense and sparse vectors with reciprocal-rank fusion.

What the package does not own ​

FastEmbed does not split documents, select vector-store metadata, schedule ingestion, create a collection, or authorize retrieval. It also does not expose completion or media models. Those concerns remain with Core, the selected store adapter, and application code.

Failure behavior ​

Initialization rejects when the native runtime or model load fails. Embedding rejects invalid batch containers, malformed vectors, mismatched sparse arrays, and a final count different from the number of inputs. Empty input returns an empty array without invoking the runtime.

Built for Anvia.