### Answer Based on the available information, there have been reported instances where Knowledge Base retrieval behaves differently between manual testing and other testing methods [(1) ]() . In one documented case, a user found that Knowledge Base worked during audio calls but not when testing via the "Test LLM" text interface . This suggests there can be inconsistencies in how Knowledge Base retrieval functions across different testing environments. For simulation testing specifically, the documentation describes simulation testing as a method that "evaluates AI agents in a controlled, virtual environment" and "mimics real-world scenarios without the risks of live deployment" [(2) ]() . You can create user prompts to guide interactions and evaluate results using defined metrics . To troubleshoot your issue: 1. Check that your Knowledge Base is properly linked to your agent at the agent level [(3) ]() 2. Verify the Knowledge Base has been indexed - there can be indexing delays [(4) ]() 3. Review your retrieval settings: "Chunks to retrieve" (default: 3) and "Similarity Threshold" (default: 0.6) [(3) ]() 4. Check if you're using web pages vs. uploaded files - there have been reports that web pages may not be consulted correctly while uploaded documents work as expected [(4) ]() If the issue persists after checking these settings, I'd recommend reaching out to Retell support with your specific call IDs from both the manual test (that worked) and the simulation test (that failed) so they can investigate the behavioral difference.

​ To learn more, see the sources I considered: [[1] Knowledge base retrieval not working](https://discord.com/channels/1186424376300679209/1311001039109095484) [[2] Platform Changelogs | Retell AI](https://www.retellai.com/changelog) [[3] Knowledge Base](https://docs.retellai.com/build/knowledge-base) [[4] Knowledge base retrieval is not happening](https://discord.com/channels/1186424376300679209/1352014353922002975) ​

Feel free to tag @AI Evy with additional questions. I also appreciate feedback, it helps me improve.