Skip to main content
Connectors are the bridge between your data and OuterProduct’s training and inference pipelines. Each connector points OuterProduct at a data source, whether that’s a file on your local machine or a table in a cloud data warehouse, and returns a Dataset object you pass directly to op.reasoning.fit() or other model operations. OuterProduct supports four data sources: local files plus S3, Snowflake, and Databricks. For cloud data sources (S3, Snowflake, and Databricks), you store your credentials once in the OuterProduct Console and reference them by name in the SDK. No cloud credentials are ever passed through the SDK itself; the OuterProduct backend retrieves your data at training time using the stored credential.

File Upload

Upload a local CSV or Parquet file directly from Python. No console setup required.

Amazon S3

Reference files in your S3 bucket by stored credential. Your AWS keys never leave your account.

Snowflake

Query Snowflake tables by stored credential using a PAT.

Databricks

Read Databricks SQL warehouse or cluster tables by stored credential.

​
Choosing a connector

ConnectorSetup requiredData location
LocalDataset.upload()NoneLocal file or in-memory frame
S3ConnectorConsole + AWS IAM roleAmazon S3 bucket
SnowflakeConnectorConsole + Snowflake PATSnowflake table
DatabricksConnectorConsole + Databricks PATDatabricks SQL warehouse or cluster
Cloud connectors (S3, Snowflake, Databricks) require you to add a credential in the OuterProduct Console before using them from the SDK. Local uploads via op.LocalDataset.upload() are stateless and work immediately with no prior setup.