Data Connector

A data connector implements the interface between a source and a destination. It may package API requests, pagination, authentication, incremental state, normalization, and loading. The supported source objects and the destination behavior are part of its contract; the product name alone does not establish them.

For an orders API, test initial loading, repeated reads, changed orders, removed items, hard deletes, and expired credentials. A successful run with zero rows can mean there was nothing new. It can also coexist with an upstream system that stopped sending, so check expected deliveries and source reconciliation separately.

A vendor-maintained connector assigns some maintenance to a provider. A self-managed platform or library leaves deployment and more operations with the team. Custom extensions need explicit owners even inside a managed service. Compare supported versions, incident handling, source limits, and migration costs alongside the service bill.

The connector’s flattened tables still reflect source definitions. Consumers need stable analytical grain, business rules, and history. Before accepting a connector, trace a representative update and a failed load through recovery into the destination.

See Managed Ingestion for the worked examples. Reference: Airbyte sync modes.