04 / AGENTIC DATABASE ANALYSIS / 2026

Ask the data.
See the answer.

View on GitHub ↗
Analytics dashboard on a laptopDATABASES, IN PLAIN ENGLISH

THE BRIEF

Data analysis should start with a question, not a query.

SelectStar is an agentic database analysis platform that lets a person connect a database and ask what they need in everyday language. The system determines whether the answer calls for SQL, exploration, a visualization, or a lightweight machine-learning model.

THE ORCHESTRATION

01 RouteClassify intent & select agents.
02 AnalyzeQuery, profile, chart, forecast.
03 ExplainStream clear, structured results.
Developer working with code on monitors

THE BUILD

A graph of specialists, not an unbounded loop.

SelectStar uses a custom TypeScript graph orchestrator in a Next.js-native system. Specialized agents handle SQL generation, EDA profiling, Vega-Lite charting, and pure-TypeScript ML forecasting, then a synthesis agent streams the result.

  • Next.js 16, TypeScript, Prisma, and Vega-Lite
  • SQLite and PostgreSQL connectivity, with MySQL-ready architecture
  • Read-only by default with explicit write confirmation
  • SQL, charts, tables, and model results rendered as canvas artifacts

THE DESIGN PRINCIPLE

SAFE
BY
DEFAULT

Agency without losing control.

A deterministic graph replaces the ambiguity of an open-ended agent loop. Potential database writes are intercepted for confirmation, while read analysis branches into parallel EDA, visualization, and ML work.

BACK TO

All selected work