FastEmbed fromQdrant 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 thefastembed Python package.

%pip install--upgrade--quiet  fastembed

Imports​

from langchain_community.embeddings.fastembedimport 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 modelshere.

• 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 tolocal_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.If0, use all available cores.IfNone, 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")

Related​

• Embedding modelconceptual guide
• Embedding modelhow-to guides