synapse-result-schemas

You can standardize Synapse plugin action outputs by using Pydantic-based result schema classes like TrainResult, which ensure type-safe and validated return values for training and inference actions. The SDK automatically validates these structures when you specify a result_model.

Action result schemas are standardized Pydantic classes that define expected output structures for Synapse plugin actions. They ensure that returned data for training, inference, export, upload, weights, and metrics is type-safe, validated, and consistently formatted.

Yes, the Synapse SDK automatically validates custom action return values against the defined schema. When you specify a result_model in your class-based or function-based actions, the SDK catches errors early by checking the output against the Pydantic schema.

Yes, you can use result schemas with both class-based and function-based actions in Synapse. By specifying a result_model, automatic result validation is applied to ensure your action returns data in a predictable and type-safe format.

The best way to define expected output structures for Synapse export and upload actions is to use the provided Pydantic-based result schema classes. These schemas enforce type-safe returns and automatic validation, preventing integration errors and ensuring consistent data formats.