Skip to main content

SQL

superduper supports SQL databases via the ibis project. With superduper, queries may be built which conform to the ibis API, with additional support for complex data-types and vector-searches.

Installation​

pip install superduper_ibis

Inserting data​

Table data must correspond to the Schema for that table. Either create a Schema and Table or use an auto-detected Schema. Once you've got a Schema, all data inserted must conform to that Schema:

import pandas

pandas.DataFrame([
    PIL.Image.open('image.jpg'), 'some text', 4,
    PIL.Image.open('other_image.jpg'), 'some other text', 3,
])

t.insert(dataframe.to_dict(orient='records'))

Selecting data​

superduper supports selecting data via the ibis query API. For example:

db['my_table'].filter(t.rating > 3).limit(5).select(t.image).execute()

Vector-searches are supported via the like operator:

(
    db['my_table']
    .like({'text': 'something like this'}, vector_index='my-index')
    .filter(t.rating > 3)
    .limit(5)
    .select(t.image, t.id)
).execute()

Vector-searches are either first or last in a chain of operations:

(
    db['my_table']
    t.filter(t.rating > 3)
    .limit(5)
    .select(t.image, t.id)
    .like({'text': 'something like this'}, vector_index='my-index')
).execute()

Updating data​

Updates are not covered for superduper SQL integrations.

Deleting data​

db.databackend.drop_table('my-table')