Hugging Face Embeddings with Weaviate

New Documentation

The model provider integration pages are new and still undergoing improvements. We appreciate any feedback on this forum thread.

Weaviate's integration with Hugging Face's APIs allows you to access their models' capabilities directly from Weaviate.

Configure a Weaviate vector index to use an Hugging Face Hub embedding model, and Weaviate will generate embeddings for various operations using the specified model and your Hugging Face API key. This feature is called the vectorizer.

At import time, Weaviate generates text object embeddings and saves them into the index. For vector and hybrid search operations, Weaviate converts text queries into embeddings.

Requirements

Weaviate configuration

Your Weaviate instance must be configured with the Hugging Face vectorizer integration (text2vec-huggingface) module.

For Weaviate Cloud (WCD) users

This integration is enabled by default on Weaviate Cloud (WCD) serverless instances.

For self-hosted users

• Check the cluster metadata to verify if the module is enabled.
• Follow the how-to configure modules guide to enable the module in Weaviate.

API credentials

You must provide a valid Hugging Face API key to Weaviate for this integration. Go to Hugging Face to sign up and obtain an API key.

Provide the API key to Weaviate using one of the following methods:

• Set the HUGGINGFACE_APIKEY environment variable that is available to Weaviate.
• Provide the API key at runtime, as shown in the examples below.

Python API v4

import weaviate
from weaviate.auth import AuthApiKey
import os

# Recommended: save sensitive data as environment variables
huggingface_key = os.getenv("HUGGINGFACE_APIKEY")
headers = {
    "X-HuggingFace-Api-Key": huggingface_key,
}

client = weaviate.connect_to_wcs(
    cluster_url=weaviate_url,                       # `weaviate_url`: your Weaviate URL
    auth_credentials=AuthApiKey(weaviate_key),      # `weaviate_key`: your Weaviate API key
    headers=headers
)

# Work with Weaviate

client.close()

JS/TS API v3

import weaviate from 'weaviate-client'

const huggingFaceApiKey = process.env.HUGGINGFACE_APIKEY || '';  // Replace with your inference API key

const client = await weaviate.connectToWeaviateCloud(
  'WEAVIATE_INSTANCE_URL',  // Replace with your instance URL
  {
    authCredentials: new weaviate.ApiKey('WEAVIATE_INSTANCE_APIKEY'),
    headers: {
      'X-HuggingFace-Api-Key': huggingFaceApiKey,
    }
  }
)

// Work with Weaviate

client.close()

Configure the vectorizer

Configure a Weaviate index to use a Hugging Face embedding model by setting the vectorizer as follows:

Python API v4

from weaviate.classes.config import Configure

client.collections.create(
    "DemoCollection",
    vectorizer_config=[
        Configure.NamedVectors.text2vec_huggingface(
            name="title_vector",
            source_properties=["title"],
            model="sentence-transformers/all-MiniLM-L6-v2",
        )
    ],
    # Additional parameters not shown
)

JS/TS API v3

await client.collections.create({
  name: 'DemoCollection',
  properties: [
    {
      name: 'title',
      dataType: 'text' as const,
    },
  ],
  vectorizers: [
    weaviate.configure.vectorizer.text2VecHuggingFace({
      name: 'title_vector',
      sourceProperties: ['title'],
      model: 'sentence-transformers/all-MiniLM-L6-v2',
    }),
  ],
  // Additional parameters not shown
});

You must specify one of the available models for the vectorizer to use.

Data import

After configuring the vectorizer, import data into Weaviate. Weaviate generates embeddings for text objects using the specified model.

Python API v4

collection = client.collections.get("DemoCollection")

with collection.batch.dynamic() as batch:
    for src_obj in source_objects:
        weaviate_obj = {
            "title": src_obj["title"],
            "description": src_obj["description"],
        }

        # The model provider integration will automatically vectorize the object
        batch.add_object(
            properties=weaviate_obj,
            # vector=vector  # Optionally provide a pre-obtained vector
        )

JS/TS API v3

const collectionName = 'DemoCollection'
const myCollection = client.collections.get(collectionName)

let dataObjects = []

for (let srcObject of srcObjects) {
  dataObject.push({
    title: srcObject.title,
    description: srcObject.description,
  });
}

const response = await myCollection.data.insertMany(dataObjects);

console.log(response);

Re-use existing vectors

If you already have a compatible model vector available, you can provide it directly to Weaviate. This can be useful if you have already generated embeddings using the same model and want to use them in Weaviate, such as when migrating data from another system.

Searches

Once the vectorizer is configured, Weaviate will perform vector and hybrid search operations using the specified Hugging Face model.

Vector (near text) search

When you perform a vector search, Weaviate converts the text query into an embedding using the specified model and returns the most similar objects from the database.

The query below returns the n most similar objects from the database, set by limit.

Python API v4

collection = client.collections.get("DemoCollection")

response = collection.query.near_text(
    query="A holiday film",  # The model provider integration will automatically vectorize the query
    limit=2
)

for obj in response.objects:
    print(obj.properties["title"])

JS/TS API v3

const collectionName = 'DemoCollection'
const myCollection = client.collections.get(collectionName)

let result;

result = await myCollection.query.nearText(
  'A holiday film',  // The model provider integration will automatically vectorize the query
  {
    limit: 2,
  }
)

console.log(JSON.stringify(result.objects, null, 2));

Hybrid search

What is a hybrid search?

A hybrid search performs a vector search and a keyword (BM25) search, before combining the results to return the best matching objects from the database.

When you perform a hybrid search, Weaviate converts the text query into an embedding using the specified model and returns the best scoring objects from the database.

The query below returns the n best scoring objects from the database, set by limit.

Python API v4

collection = client.collections.get("DemoCollection")

response = collection.query.hybrid(
    query="A holiday film",  # The model provider integration will automatically vectorize the query
    limit=2
)

for obj in response.objects:
    print(obj.properties["title"])

JS/TS API v3

const collectionName = 'DemoCollection'
const myCollection = client.collections.get(collectionName)

result = await myCollection.query.hybrid(
  'A holiday film',  // The model provider integration will automatically vectorize the query
  {
    limit: 2,
  }
)

console.log(JSON.stringify(result.objects, null, 2));

References

Vectorizer parameters

The following examples show how to configure Hugging Face-specific options.

Model selection parameters

Only select one of the following parameters to specify the model:

• model,
• passageModel and queryModel, or
• endpointURL

Differences between model, passageModel/queryModel and endpointURL

The passageModel and queryModel parameters are used together to specify a DPR passage and query model.

The endpointURL parameter is used to specify a custom Hugging Face Inference Endpoint. This parameter overrides the model, passageModel, and queryModel parameters.

Other Parameters

• options.waitForModel: If the model is not ready, wait for it rather than returning a 503 error.
• options.useGPU: Use a GPU for inference if your account plan supports it.
• options.useCache: Use a cached result if available. (For non-deterministic models to prevent the caching mechanism from being used.)

Python API v4

from weaviate.classes.config import Configure

client.collections.create(
    "DemoCollection",
    vectorizer_config=[
        Configure.NamedVectors.text2vec_huggingface(
            name="title_vector",
            source_properties=["title"],
            # NOTE: Use only one of (`model`), (`passage_model` and `query_model`), or (`endpoint_url`)
            model="sentence-transformers/all-MiniLM-L6-v2",
            # passage_model="sentence-transformers/facebook-dpr-ctx_encoder-single-nq-base",    # Required if using `query_model`
            # query_model="sentence-transformers/facebook-dpr-question_encoder-single-nq-base", # Required if using `passage_model`
            # endpoint_url="<custom_huggingface_url>",
            #
            # wait_for_model=True,
            # use_cache=True,
            # use_gpu=True,
        )
    ],
    # Additional parameters not shown
)

JS/TS API v3

await client.collections.create({
  name: 'DemoCollection',
  properties: [
    {
      name: 'title',
      dataType: 'text' as const,
    },
  ],
  vectorizers: [
    weaviate.configure.vectorizer.text2VecHuggingFace({
      name: 'title_vector',
      sourceProperties: ['title'],
      model: 'sentence-transformers/all-MiniLM-L6-v2',
      // passage_model: 'sentence-transformers/facebook-dpr-ctx_encoder-single-nq-base',    // Required if using `query_model`
      // query_model: 'sentence-transformers/facebook-dpr-question_encoder-single-nq-base', // Required if using `passage_model`
      // endpoint_url: '<custom_huggingface_url>',
      //
      // wait_for_model: true,
      // use_cache: true,
      // use_gpu: true,
    }),
  ],
  // Additional parameters not shown
});

Available models

You can use any Hugging Face embedding model with text2vec-huggingface, including public and private Hugging Face models. Sentence similarity models generally work best.

Further resources

• Hugging Face integrations (locally hosted)

Code examples

Once the integrations are configured at the collection, the data management and search operations in Weaviate work identically to any other collection. See the following model-agnostic examples:

• The how-to: manage data guides show how to perform data operations (i.e. create, update, delete).
• The how-to: search guides show how to perform search operations (i.e. vector, keyword, hybrid) as well as retrieval augmented generation.

External resources

• Hugging Face Inference API documentation
• Hugging Face Model Hub

If you have any questions or feedback, let us know in the user forum.

• Technical questions
• Documentation feedback