synapse-action-development

Synapse plugin actions support both function-based and class-based patterns. You define the action logic, specify parameters, and configure result schemas to enable seamless action chaining within pipelines.

You validate parameters in Synapse plugin actions using Pydantic. This framework allows you to define robust input and output validation schemas, ensuring data integrity before the action processes the request.

Semantic types in Synapse pipelines define the specific data format of an action's output, such as model weights. They enable seamless action chaining by ensuring downstream actions receive properly typed inputs.

Function-based patterns suit simple Synapse actions, while class-based patterns handle complex logic. Both support Pydantic validation and semantic types, letting you choose based on your action's processing requirements.

Yes, you can develop a machine learning training action as a Synapse plugin. You define the train action, specify parameters like epochs and learning rate, and handle model weights as the output using semantic types.