ExFastembed

ExFastembed is an Elixir wrapper around the fastembed-rs crate. It provides a simple interface to load text embedding models and reranker models, generate embeddings for a list of texts, and rerank documents based on a query.

Installation

Add ex_fastembed to your list of dependencies in mix.exs:

def deps do
  [
    {:ex_fastembed, github: "elchemista/ex_fastembed", branch: "master"}
  ]
end

Requirements:

• Elixir 1.18 or later
• Rust 1.91 or later
• clang (e.g., sudo apt install clang)

Compile the dependency:

export CC=/usr/bin/clang CXX=/usr/bin/clang++ RUSTLER_PRECOMPILATION_EXAMPLE_BUILD=1 && mix compile

Usage

Loading an Embedding Model

Before generating embeddings, you must load one of the supported text embedding models. For example, to load the BAAI/bge-small-en-v1.5 model:

iex> ExFastembed.load("BAAI/bge-small-en-v1.5")
{:ok, 384}

Calling load/1 again replaces the active embedding model.

Generating Embeddings

After loading the embedding model, you can generate embeddings for a list of texts:

iex> ExFastembed.embed_text(["Hello, world!", "Elixir is awesome"])
{:ok, [[...], [...]]}

An empty input returns {:ok, []} without running inference.

Loading a Reranker Model

To load a reranker model, use the load_reranker/1 function. For example, to load jinaai/jina-reranker-v1-turbo-en:

iex> ExFastembed.load_reranker("jinaai/jina-reranker-v1-turbo-en")
{:ok, true}

Calling load_reranker/1 again replaces the active reranker model.

Reranking Documents

Once the reranker model is loaded, you can rerank documents based on a query:

iex> ExFastembed.rerank("search query", ["Document 1", "Document 2"], true)
{:ok, [{0, 0.95, "Document 1"}, {1, 0.90, "Document 2"}]}

An empty document list returns {:ok, []} without running inference.

Supported Models

The runtime source is ExFastembed.embed_models/0 and ExFastembed.reranker_models/0. These lists are generated from the bundled fastembed-rs 5.17.3 metadata, with legacy ExFastembed aliases retained. Quantized variants whose repository name is ambiguous are exposed by their enum-style variant names.

Embedding Models

• "Alibaba-NLP/gte-base-en-v1.5"
• "Alibaba-NLP/gte-large-en-v1.5"
• "AllMiniLML12V2Q"
• "AllMiniLML6V2Q"
• "BAAI/bge-base-en-v1.5"
• "BAAI/bge-large-en-v1.5"
• "BAAI/bge-large-zh-v1.5"
• "BAAI/bge-m3"
• "BAAI/bge-small-en-v1.5"
• "BAAI/bge-small-zh-v1.5"
• "BGEBaseENV15Q"
• "BGELargeENV15Q"
• "BGESmallENV15Q"
• "EmbeddingGemma300MQ"
• "EmbeddingGemma300MQ4"
• "GTEBaseENV15Q"
• "GTELargeENV15Q"
• "intfloat/multilingual-e5-base"
• "intfloat/multilingual-e5-large"
• "intfloat/multilingual-e5-small"
• "jinaai/jina-embeddings-v2-base-code"
• "jinaai/jina-embeddings-v2-base-en"
• "lightonai/ModernBERT-embed-large"
• "lightonai/modernbert-embed-large"
• "mixedbread-ai/mxbai-embed-large-v1"
• "MxbaiEmbedLargeV1Q"
• "nomic-ai/nomic-embed-text-v1"
• "nomic-ai/nomic-embed-text-v1.5"
• "NomicEmbedTextV15Q"
• "onnx-community/embeddinggemma-300m-ONNX"
• "ParaphraseMLMiniLML12V2Q"
• "Qdrant/all-MiniLM-L6-v2-onnx"
• "Qdrant/bge-base-en-v1.5-onnx-Q"
• "Qdrant/bge-large-en-v1.5-onnx-Q"
• "Qdrant/bge-small-en-v1.5-onnx-Q"
• "Qdrant/clip-ViT-B-32-text"
• "Qdrant/multilingual-e5-large-onnx"
• "Qdrant/paraphrase-multilingual-MiniLM-L12-v2-onnx-Q"
• "sentence-transformers/all-MiniLM-L12-v2"
• "sentence-transformers/all-MiniLM-L6-v2"
• "sentence-transformers/all-mpnet-base-v2"
• "sentence-transformers/paraphrase-MiniLM-L12-v2"
• "sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2"
• "sentence-transformers/paraphrase-multilingual-mpnet-base-v2"
• "snowflake/snowflake-arctic-embed-l"
• "Snowflake/snowflake-arctic-embed-m"
• "snowflake/snowflake-arctic-embed-m"
• "snowflake/snowflake-arctic-embed-m-long"
• "snowflake/snowflake-arctic-embed-s"
• "snowflake/snowflake-arctic-embed-xs"
• "SnowflakeArcticEmbedLQ"
• "SnowflakeArcticEmbedMLongQ"
• "SnowflakeArcticEmbedMQ"
• "SnowflakeArcticEmbedSQ"
• "SnowflakeArcticEmbedXSQ"
• "Xenova/all-MiniLM-L12-v2"
• "Xenova/all-MiniLM-L6-v2"
• "Xenova/all-mpnet-base-v2"
• "Xenova/bge-base-en-v1.5"
• "Xenova/bge-large-en-v1.5"
• "Xenova/bge-large-zh-v1.5"
• "Xenova/bge-small-en-v1.5"
• "Xenova/bge-small-zh-v1.5"
• "Xenova/paraphrase-multilingual-MiniLM-L12-v2"
• "Xenova/paraphrase-multilingual-mpnet-base-v2"

Reranker Models

• "BAAI/bge-reranker-base"
• "BAAI/bge-reranker-v2-m3"
• "jinaai/jina-reranker-v1-turbo-en"
• "jinaai/jina-reranker-v2-base-multiligual"
• "jinaai/jina-reranker-v2-base-multilingual"
• "rozgo/bge-reranker-v2-m3"

License

ExFastembed is released under the Apache License 2.0. See LICENSE for details.

Acknowledgments

ExFastembed is a wrapper around fastembed-rs, a fast Rust-based text embedding library.