checked once at startup · 1.13.5 · read-only · cloud
VIEW
💬 Chat
🔍 Full Analysis
HISTORY
Success rate by country
Users lost — last 30 days
Worst protocol this week
🔒 Read-only session · version-locked to 1.13.5.
Results are always explained locally — no data row reaches a cloud API.
💬 Chat
Ask anything — it routes itself to the database, the web, or plain conversation.
Routes per message · curated checks for in-house metrics, generated SQL for everything else
🔍 Full Analysis
Read-only · locked to app_version 1.13.5 · curated SQL anchors every number
Runs DQ statements → failure taxonomy → researched fixes and saves two colour-coded reports (Comparison + Action Plan) to reports/.
✔ Two reports created in reports/: ⇩ Comparison_Report_1.13.5.ipynb⇩ Action_Plan_1.13.5.ipynb
What you're watching. The agent runs the analyst's own curated SQL (it never invents the anchor queries),
adds generative diagnostics clearly labelled as leads, researches a fix per failure type through a
self-hosted search, fact-checks each one, and renders two reports. Numbers are synthetic.
Live agent activity
LIVE
Idle — press “Run full analysis” to start.
Illustrative UI mockup · synthetic values · no production data shown · database accessed read-only