FastEmbed from Qdrant is a lightweight, fast, Python library built for embedding generation.
  • Quantized model weights
  • ONNX Runtime, no PyTorch dependency
  • CPU-first design
  • Data-parallelism for encoding of large datasets.

​
Dependencies

To use FastEmbed with LangChain, install the fastembed Python package. 
%pip install --upgrade --quiet  fastembed

​
Imports

from langchain_community.embeddings.fastembed import FastEmbedEmbeddings

​
Instantiating FastEmbed

​
Parameters

  • model_name: str (default: “BAAI/bge-small-en-v1.5”)
    Name of the FastEmbedding model to use. You can find the list of supported models here.
  • max_length: int (default: 512)
    The maximum number of tokens. Unknown behavior for values > 512.
  • cache_dir: Optional[str] (default: None)
    The path to the cache directory. Defaults to local_cache in the parent directory.
  • threads: Optional[int] (default: None)
    The number of threads a single onnxruntime session can use.
  • doc_embed_type: Literal["default", "passage"] (default: “default”)
    “default”: Uses FastEmbed’s default embedding method.
    “passage”: Prefixes the text with “passage” before embedding.
  • batch_size: int (default: 256)
    Batch size for encoding. Higher values will use more memory, but be faster.
  • parallel: Optional[int] (default: None)
    If >1, data-parallel encoding will be used, recommended for offline encoding of large datasets. If 0, use all available cores. If None, don’t use data-parallel processing, use default onnxruntime threading instead. 
embeddings = FastEmbedEmbeddings()

​
Usage

​
Generating document embeddings

document_embeddings = embeddings.embed_documents(
    ["This is a document", "This is some other document"]
)

​
Generating query embeddings

query_embeddings = embeddings.embed_query("This is a query")