Tests failing

Failures in the Test Center help detect issues in the flow before putting the agent into production. Use this guide to interpret results.

Initial checks

1. The agent has a published version.
2. Success criteria are clear and measurable.
3. The test suite covers different intents and exceptions.
4. The test environment has access to the API and voice providers.

How to analyze

1. Open the run in Test Center → History.
2. Review the full transcript and the evaluated criteria.
3. Check whether the agent halted, ignored instructions, or deviated from the topic.
4. Compare with the flow version used in the run.

Recommended adjustments

• Reword criteria that are too vague.
• Add test cases for exceptions and edge cases.
• Generate additional cases automatically with AI for extra coverage.
• Confirm you published the latest flow before running the suite.
• Verify the Knowledge Base used in the test is indexed.