Sentence Embeddings Lambda

Generic implementation of HuggingFace BERT sentence embeddings for AWS lambda. Uses ECR Image runtime with model and torch/transformers installed int othe image.

Features

• Currently API key authentication is always on. You will need to run the create_usage_plan.py script to create a usage plan for your API after it is deployed or you will only see 403 responses
• Future releases will include a toggle switch for API key auth

API

The API only has a single path for encoding a list of sentences into embedding vectors.

Authentication

You will need to run create_usage_plan.py or create your own api key for the api. Then you will need to include the key in the x-api-key header in every request.

POST /api/encode

Content-type: application/json

Request Schema

{
    "crid":int,
    "sentences":[str]
}

Response Schema

{
    "crid":int,
    "embeddings":[float]
}