A 2.5-million-row warehouse learns to answer in plain English
A read-only AI analyst over a PostgreSQL star schema that answers business questions in plain English and shows its SQL.
- Client
- Solinovo build
- Industry
- Retail analytics
- Country
- Pakistan
- Timeline
- V1 complete
The challenge
The situation
A data warehouse only answers questions for the people who can write SQL against it. Everyone else joins a queue for a report, and by the time the report arrives the question has usually moved on.
How we approached it
The thinking
The failure mode of natural-language querying is confident nonsense, so we built for verifiability first. A semantic layer maps business vocabulary to the actual tables, and the generated SQL is always shown in full. An analyst can check the query before trusting the number. Access is read-only by construction, so the worst case is a wrong answer, never a damaged warehouse.
What we built
The engineering
A read-only analyst sitting over a PostgreSQL Kimball star schema, with a semantic layer between the question and the query. Results are charted automatically, and the SQL behind every answer is available on the same screen.
- Plain-English querying over a star schema
- Semantic layer mapping business terms to tables
- Generated SQL shown in full for verification
- Automated charting of results
- Read-only access across a 2.5-million-row warehouse
Technology
- PostgreSQL
- Kimball star schema
- MCP
How it helped
The difference
The reporting queue stops being the bottleneck between a question and an answer. Analysts keep their audit trail, because nothing is returned that can't be checked.
Measured result
Anyone can ask a 2.5-million-row warehouse a business question in plain English.
Next step
Got a problem shaped like this one?
Tell us the situation in plain language. We'll tell you how we'd approach it, what we'd build first, and where the risk sits.
Direct
contact@solinovo.comWe reply within one business day. If we're not the right fit, we'll say so.