Implementation:Explodinggradients Ragas HuggingFaceEmbeddings Provider

Knowledge Sources	Explodinggradients_Ragas
Domains	Embeddings, HuggingFace
Last Updated	2026-02-10 00:00 GMT

Overview

Modern HuggingFace embeddings provider supporting both local sentence-transformer models and API-based embedding services.

Description

HuggingFaceEmbeddings extends BaseRagasEmbedding to provide text embeddings using HuggingFace models. It supports local inference via sentence-transformers and remote inference via the HuggingFace Inference API. Batch processing and caching are built in.

Usage

Use this provider for local or API-based HuggingFace embeddings in evaluation pipelines.

Code Reference

Source Location

• Repository: Explodinggradients_Ragas
• File: src/ragas/embeddings/huggingface_provider.py
• Lines: 11-206

Signature

class HuggingFaceEmbeddings(BaseRagasEmbedding):
    PROVIDER_NAME = "huggingface"

    def __init__(
        self,
        model: str,
        use_api: bool = False,
        api_key: Optional[str] = None,
        device: Optional[str] = None,
        normalize_embeddings: bool = True,
        batch_size: int = 32,
        cache: Optional[CacheInterface] = None,
        **model_kwargs: Any,
    ) -> None:
        ...

Import

from ragas.embeddings.huggingface_provider import HuggingFaceEmbeddings

I/O Contract

Inputs

Name	Type	Required	Description
model	str	Yes	HuggingFace model name or path
use_api	bool	No	Use HuggingFace Inference API (default False)
device	Optional[str]	No	Device for local inference (e.g., "cuda")
batch_size	int	No	Batch size for embedding (default 32)

Outputs

Name	Type	Description
embed_text returns	List[float]	Single text embedding vector
embed_texts returns	List[List[float]]	Batch embedding vectors

Usage Examples

from ragas.embeddings.huggingface_provider import HuggingFaceEmbeddings

# Local model
embeddings = HuggingFaceEmbeddings(model="BAAI/bge-small-en-v1.5")
vector = embeddings.embed_text("What is AI?")

# API-based
embeddings_api = HuggingFaceEmbeddings(
    model="BAAI/bge-small-en-v1.5",
    use_api=True,
    api_key="hf_xxx",
)