In standard Genie Space chat, space managers can use the Monitoring tab to review prompts/conversations. However, in Agent Mode, we see the warning:
โGenie agent responses may contain results obtained using other usersโ credentials and are hidden from space managers.โ
A few questions:
- Why is Agent Mode treated differently from standard Genie chat if both are ultimately querying the same underlying tables/data assets?
- Is there currently any admin/workspace setting, governance control, or future roadmap item that would allow space managers to have fuller visibility into Agent Mode conversations/results for testing and governance purposes?
- Do benchmarks, thumbs up/down feedback, and saved benchmark questions improve/tune both standard mode and Agent Mode behavior equally, or are they handled separately?
Would appreciate any clarification from anyone who has implemented governance/testing processes around Genie Agent Mode. Thank you!!
