Skip to main content
Version: Main branch

Connect

The standard way to connect to Superduper is via the superduper decorator:

Development mode​

In development mode, you may provide the URI/ connection details of your data deployment directly

db = superduper('<database-uri>')

For example if you are running a (not secure) MongoDB deployment locally, and you want to connect to the "documents" database, you might write:

from superduper import superduper
db = superduper('mongodb://localhost:27017/documents')

Complete connection guide​

For a semi-exhaustive list of possible connections see here.

Fine grained configuration​

superduper chooses default artifact_store (file blob storage) and metadata_store (AI metadata) values for your connection. These defaults may be overridden directly:

db = superduper(
    '<database-uri>',
    artifact_store='<artifact-store-uri>,
    metadata_store='<metadata-store-uri>,
)

Cluster mode​

In cluster mode, the connection string needs to be provided in a configuration file or via environment variable so that all services can connect correctly:

Add these lines to your configuration:

data_backend: mongodb://localhost:27018/documents

Read more about configuration here.

After doing this, you may connect directly with the superduper decorator:

db = superduper()

Fine grained configuration​

For more granular configuration add these lines:

data_backend: <database-uri>,
artifact_store: <artifact-store-uri>
metadata_store: <metadata-store-uri>

Next steps​

db is now your connection to your data, models, and model meta-data. Now that you have established this connection you are ready to build, deploy and manage AI with Superduper.