Skip to main content

Add input/output validation and an error model

Stub

This How-to is a stub. We need a standard “error codes and remediation” pattern across atomic capabilities.

Goal​

You will validate inputs/outputs and use a predictable error model so failures are actionable and replayable.

When to use this​

  • You are making a capability production-grade.
  • You want deterministic failure modes and stable integration points.

Prerequisites​

  • JSON Schema for inputs and outputs
  • A shared error envelope strategy (code/message/details) for atomic execution

Steps​

  1. Validate inputs against the NodeType input schema.
  2. Validate outputs against the output schema (before completing the NodeRun).
  3. Map known failures to stable error codes and include remediation hints.

Verify​

  • Invalid inputs fail fast with actionable errors.
  • Outputs are schema-valid (or errors are explicit and durable).

Troubleshooting​

  • “Everything is 500” → add stable error codes for known failures.
  • “Outputs are inconsistent” → validate outputs and fail deterministically when invalid.
  • “Hard to debug” → ensure errors are emitted to the Event Stream and referenced in artifacts.

Cleanup / Rollback​

  • None.

Next steps​