Skip to content

Embedding Models

Locally hosted embedding model from Hugging Face

This will download the selected model from the HF hub and make embeddings on the machine the backend is running on. 

pip install sentence_transformers


# backend/config.yaml
EmbeddingModelConfig: &EmbeddingModelConfig
  source: HuggingFaceEmbeddings
  source_config:
    model_name : 'BAAI/bge-base-en-v1.5'

Artefact Azure-hosted embedding model

# backend/config.yaml
EmbeddingModelConfig: &EmbeddingModelConfig
  source: OpenAIEmbeddings
  source_config:
    openai_api_type: azure
    openai_api_key: {{ EMBEDDING_API_KEY }}
    openai_api_base: https://poc-openai-artefact.openai.azure.com/
    deployment: embeddings
    chunk_size: 500

AWS Bedrock

You will first need to login to AWS

pip install boto3
Follow this guide to authenticate your machine

# backend/config.yaml
EmbeddingModelConfig: &EmbeddingModelConfig
  source: BedrockEmbeddings
  source_config:
    model_id: 'amazon.titan-embed-text-v1'