Ask ChatPlotDB.
Every question planned. Every query shown. Every answer verified.
AAPL revenue came to $391B in FY24, up 2.0% year over year.
iPhone and Services drove most of it, together about three quarters of the total.
- revenue_by_segment: sql_macro, FY24 revenue grouped by segment
- yoy_growth: sql_macro, this year against last, per segment
- segment_share: python, each segment as a share of the total
- revenue_flow: pipeline, segment → income, for the chart
Ask your data anything,get an answer in seconds
Connect a database, ask in plain English, and ChatPlotDB writes the SQL, runs it, and hands back clean tables and charts your whole team can read.
Your weekly revenue review is four repeated steps. I drafted each as its own tool and ran it against portfolio_db:
No syntax to learn,and no analyst to wait for
These are questions, not queries. Each one is answered from your own tables, with the SQL attached so anyone who wants to check the working can.
“How many accounts churned last month, and which plan were they on?”
A table, grouped by plan, with the SQL that produced it.
“Plot weekly signups since January.”
A line chart you can save straight to a dashboard.
“Compare this quarter's revenue with the same quarter last year.”
Both periods side by side, with the delta worked out.
“Which customers in this CSV aren't in the database yet?”
A hybrid answer: the uploaded file and your tables in one query.
“What counts as an active account here?”
The definition from your config, not a guess from column names.
“Alert me in Slack if refunds go above 2% this week.”
A saved card with a threshold on it, watching on a schedule.
Asking questions
No. You ask in plain English and ChatPlotDB writes the SQL for you: the way you'd ask a colleague who happens to know the schema.
The SQL is still shown, because the people who do read it are usually the ones being asked to trust the number.
Yes. The generated query is shown alongside the answer, so you can read exactly what ran.
Questions answered from an uploaded CSV or Excel file skip the step: there's no database there to protect.
From your config: the semantic layer that records which tables matter, how they join, and how your metrics are defined, so "active account" resolves to the same thing every time.
Config Studio writes the first draft of it by reading your schema, and you review it in plain language rather than YAML.
No. The agent only ever issues read queries, and that restriction is enforced in the product rather than left to the prompt.
Pair it with a scoped database role anyway: grant SELECT on what should be queryable and nothing else.