product
Data analysts, data engineers, hiring interviewers, and business operators who need to ask Olist ecommerce questions and inspect how a Text2SQL agent reached its answer.
Turn natural-language business questions into an inspectable analysis workflow. Success means users can understand retrieved context, review generated SQL and guardrail decisions, and trust the presented result.
Precise, composed, investigative.
Not a generic purple AI chat app, not a dense legacy BI dashboard, and not a decorative agent animation that hides the actual work.
- Make agent reasoning inspectable without overwhelming the answer.
- Put the business question and result first; keep implementation evidence close.
- Communicate safety and state with text and shape, not color alone.
- Reward expert scrutiny with useful schema, SQL, and timing detail.
- Keep the interface useful before real LLM integration exists.
Target WCAG AA contrast, visible keyboard focus, reduced-motion support, responsive layouts, and redundant labels for success, warning, and error states.