Conversation that executes
Natural-language intent becomes queries and executable code without hiding the path to the answer.
This product does not have an official name yet. The title used here describes what it does.
Product in development
AI data analysis
Ask the database. Get an answer grounded in measured data.
In developmentAvailable evidence
What can already be verified.
- SQL + Python
- execution in one flowQueries and calculations remain visible next to the answer.
- Notebook
- inspectable historyCells and results preserve the path taken by the analysis.
- In progress
- impact not measured yetThe case does not present a product hypothesis as a production result.
Context
What needed to change.
A simple question about a real database can require schema discovery, queries, calculations, and several attempts. In conventional tools, that work is split across chat, editor, terminal, and memory — while the final answer may not show where it came from.
The product brings conversation, database, notebook, and reporting into the same flow. The agent investigates tables, writes and runs code in an isolated environment, and supports each conclusion with observable results, while a durable runtime preserves analysis state.
The delivery
A system designed end to end.
Traceable notebook
Cells, results, and artifacts remain available for inspection, continuation, and reproduction.
State that survives
Sessions, events, and executions are designed to remain consistent even when a connection drops.
Evaluation with real cases
A dedicated console checks the end-to-end delivery against real questions and datasets, not only isolated tests.
Technology
The foundation behind the experience.
- 01React
- 02TypeScript
- 03FastAPI
- 04Python
- 05PostgreSQL
- 06Vite
- 07Playwright