SentenceTransformersSparseTextEmbedder

Embed text strings, such as queries, using sparse embedding models from Sentence Transformers. The model runs locally, so no external API calls are made during embedding. Use this component in query pipelines to embed user queries for sparse retrieval.

Embedding Models in Query Pipelines and Indexes

The embedding model you use to embed documents in your indexing pipeline must be the same as the embedding model you use to embed the query in your query pipeline.

This means the embedders for your indexing and query pipelines must match. For example, if you use CohereDocumentEmbedder to embed your documents, you should use CohereTextEmbedder with the same model to embed your queries.

Key Features

• Downloads and runs sparse Sentence Transformers models locally — no external API required.
• Based on SPLADE (SParse Lexical AnD Expansion), combining the benefits of learned sparse representations with efficient sparse retrieval.
• Outputs a SparseEmbedding where each non-zero value represents the importance weight of a term in the vocabulary.
• Compatible with private Hugging Face models via API token authentication.
• Works with sparse embedding retrievers such as QdrantSparseEmbeddingRetriever.

Configuration

1. Drag the SentenceTransformersSparseTextEmbedder component onto the canvas from the Component Library.
2. Click the component to open the configuration panel.
3. On the General tab:
   1. Enter the model name. Specify the path to a local model or the ID of the model on Hugging Face.
4. Go to the Advanced tab to configure device, token, prefix, suffix, trust_remote_code, model_kwargs, tokenizer_kwargs, config_kwargs, backend, and revision.

Connections

SentenceTransformersSparseTextEmbedder accepts a text string as input. It outputs sparse_embedding — a sparse vector representation of the input text.

Typically, you connect the pipeline Input component to the text input and send sparse_embedding to a sparse embedding retriever.

To use private models from Hugging Face, connect the platform to Hugging Face first. For details, see Use Hugging Face Models.

Usage Example

components:
  sparse_text_embedder:
    type: haystack.components.embedders.sentence_transformers_sparse_text_embedder.SentenceTransformersSparseTextEmbedder
    init_parameters:
      model: prithivida/Splade_PP_en_v2  # SPLADE model for sparse embeddings
      prefix: ""
      suffix: ""

  sparse_retriever:
    type: haystack_integrations.components.retrievers.qdrant.retriever.QdrantSparseEmbeddingRetriever
    init_parameters:
      document_store:
        type: haystack_integrations.document_stores.qdrant.document_store.QdrantDocumentStore
        init_parameters:
          location: ${QDRANT_HOST}
          api_key: ${QDRANT_API_KEY}
          index: default
          use_sparse_embeddings: true
          return_embedding: false
      top_k: 10
      scale_score: false

  prompt_builder:
    type: haystack.components.builders.prompt_builder.PromptBuilder
    init_parameters:
      template: |-
        You are a helpful assistant.
        Answer the question based on the provided documents.
        If the documents don't contain enough information, say so.
        
        Documents:
        {% for document in documents %}
        Document[{{ loop.index }}]:
        {{ document.content }}
        {% endfor %}
        
        Question: {{ question }}
        Answer:

  llm:
    type: haystack.components.generators.openai.OpenAIGenerator
    init_parameters:
      api_key: {"type": "env_var", "env_vars": ["OPENAI_API_KEY"], "strict": false}
      model: gpt-4o
      generation_kwargs:
        max_tokens: 500
        temperature: 0.0

  answer_builder:
    type: haystack.components.builders.answer_builder.AnswerBuilder
    init_parameters: {}

connections:
- sender: sparse_text_embedder.sparse_embedding
  receiver: sparse_retriever.query_sparse_embedding
- sender: sparse_retriever.documents
  receiver: prompt_builder.documents
- sender: sparse_retriever.documents
  receiver: answer_builder.documents
- sender: prompt_builder.prompt
  receiver: llm.prompt
- sender: llm.replies
  receiver: answer_builder.replies

max_runs_per_component: 100

inputs:
  query:
  - sparse_text_embedder.text
  - prompt_builder.question
  - answer_builder.query
  filters:
  - sparse_retriever.filters

outputs:
  documents: sparse_retriever.documents
  answers: answer_builder.answers

Parameters

Inputs

Parameter	Type	Default	Description
text	str		Text to embed.

Outputs

Parameter	Type	Default	Description
sparse_embedding	SparseEmbedding		The sparse embedding of the input text.

Init Parameters

These are the parameters you can configure in Pipeline Builder:

Parameter	Type	Default	Description
model	str	prithivida/Splade_PP_en_v2	The model to use for calculating sparse embeddings. Specify the path to a local model or the ID of the model on Hugging Face. For available models, check Hugging Face.
device	Optional[ComponentDevice]	None	Overrides the default device used to load the model.
token	Optional[Secret]		An API token to use private models from Hugging Face.
prefix	str	""	A string to add at the beginning of each text to be embedded. Some models may benefit from adding a prefix or suffix to the text before embedding. For example, the prithivida/Splade_PP_en_v2 model may benefit from adding a prefix of "query: " to the text before embedding.
suffix	str	""	A string to add at the end of each text to embed. Some models may benefit from adding a prefix or suffix to the text before embedding.
trust_remote_code	bool	False	If True, permits custom models and scripts.
local_files_only	bool	False	If True, only looks at local files without downloading from Hugging Face Hub.
model_kwargs	Optional[Dict[str, Any]]	None	Additional keyword arguments for AutoModelForSequenceClassification.from_pretrained when loading the model. Refer to specific model documentation for available kwargs.
tokenizer_kwargs	Optional[Dict[str, Any]]	None	Additional keyword arguments for AutoTokenizer.from_pretrained when loading the tokenizer. Refer to specific model documentation for available kwargs.
config_kwargs	Optional[Dict[str, Any]]	None	Additional keyword arguments for AutoConfig.from_pretrained when loading the model configuration.
backend	Literal["torch", "onnx", "openvino"]	torch	The backend to use for the Sentence Transformers model. Choose from torch, onnx, or openvino. Refer to the Sentence Transformers documentation for more information on acceleration and quantization options.
revision	Optional[str]	None	The specific model version to use. It can be a branch name, a tag name, or a commit ID for a stored model on Hugging Face.

Run Method Parameters

These are the parameters you can configure for the component's run() method. This means you can pass these parameters at query time through the API, in Playground, or when running a job. For details, see Modify Pipeline Parameters at Query Time.

Parameter	Type	Default	Description
text	str		Text to embed.