Skip to content

FastEmbed embedder

Runs a local FastEmbed ONNX model with no API key. Selected via EMBEDDING_PROVIDER=fastembed.

Enable it

bash
EMBEDDING_PROVIDER=fastembed
FASTEMBED_MODEL=BAAI/bge-small-en-v1.5
EMBEDDING_DIMS=384

pip install fastembed (imported lazily only when this provider is selected).

Configuration

VariableDefaultDescription
EMBEDDING_PROVIDERopenaiSet to fastembed.
FASTEMBED_MODELBAAI/bge-small-en-v1.5FastEmbed ONNX model name.
EMBEDDING_DIMS384Vector dimension; must match the embedding model.

Notes

  • Runs locally with ONNX models and no API key.
  • fastembed pulls the ONNX runtime, so it is imported and loaded lazily only when the provider is actually selected.
  • If fastembed is not installed, selecting this provider raises an ImportError with the install hint.

The memory layer for AI agents.