Skip to main content

Community support

In order to specify the action of models on the data, we provide an interface to pythonic ecosystem query APIs. In particular, we provide wrappers to these projects to create database queries:

ibis also allows users to use raw SQL in their workflows.

Queries in these two-worlds can be built by importing the table/collection class from each data backend. With pymongo, one can write:

query = db['products'].find({'brand': 'Nike'}, {'_id': 1}).limit(10)

In ibis, one would write:

query = db['products'].filter(products.brand == 'Nike').select('id').limit(10)

Hybrid API​

On top of the native features of pymongo and ibis, superduper builds several novel features:

  • Additional ways to query the database with the outputs of machine learning models
    • Query model-outputs directly
    • Vector-search
  • Ways to encode and query more sophisticated data-types using the Document-Encoder pattern.