Testing, Versioning & Release
How do I test a durable agent workflow that calls external AI models?
You can validate AI model-integrated durable agent workflows using Catalyst’s deterministic replay and mocking tools. First capture replayable workflow checkpoints from Catalyst’s durable execution layer, substitute external AI model calls with consistent mock outputs to isolate workflow logic, then replay the captured state to confirm step execution matches expected outcomes. This method does not validate the actual behavior of the external AI models, only the workflow’s handling of their responses.
Was this article helpful?
Your feedback helps improve Diagrid's FAQ experience.
Keep reading
More Diagrid FAQ articles
- Testing, Versioning & Release
How do I test durable agent workflows that use LLM calls?
Covers testing durable agent LLM workflows: use replayable validation, mock LLM calls to avoid real APIs in local/staging tests, verify state transitions.
- Testing, Versioning & Release
What staging environment requirements apply to durable agent workflows?
Outlines key staging considerations for deploying and validating durable agent workflows before moving them to production environments.
- Testing, Versioning & Release
How do versioning strategies work for durable agent workflows?
Explains core versioning approaches for managing updates to durable agent workflow definitions and their associated execution runs.
