BotDojoDocs
Search BotDojo Docs
Search every guide

Try “stop a run,” “connect an app,” or “daily briefing.”

Browse Guides
GuidesPlatformAdd Evaluations
Guides · Platform

Add Evaluations

Attach repeatable checks that score agent responses during testing and batch runs.

1 min read

Add Evaluations

Evaluations turn an expected behavior into a repeatable check. Use them to measure requirements such as correctness, groundedness, format, tone, or response length.

  1. Open the Flow's Evaluate tab.
  2. Select Add Eval.
  3. Choose a Quick Setup template or an existing evaluator.
The evaluation template picker showing code and LLM-as-a-judge options
Screenshot previewThe evaluation template picker showing code and LLM-as-a-judge options
The evaluation template picker showing code and LLM-as-a-judge options

Choose the evaluator based on the requirement:

  • A Code Evaluation is deterministic and works well for schemas, exact values, patterns, and limits.
  • An LLM as a Judge evaluation can grade qualities that require language understanding, such as whether an answer is supported by the retrieved context.

Map the evaluator's inputs to the Flow input, output, or expected dataset fields it needs. Run a few examples and inspect the evaluation reason—not only the score—to make sure the check represents your requirement.

See Evaluations for setup, custom evaluators, and RAG evaluation patterns.