How is iDBQuery different from a plain text-to-SQL tool?
A plain text-to-SQL tool just converts a sentence into a query and stops; iDBQuery generates the SQL, runs it across all your sources, validates the result, cites every number to its source row, and can investigate further on its own. iDBQuery delivers a verified answer, not just a query.
Basic text-to-SQL tools translate one sentence into one query against one database — then leave you to run it, check it, and trust it. iDBQuery is a full answering system built around that translation step.
- It runs and verifies, not just translates. You get the actual answer with the supporting data, not a query to paste and debug.
- Cited every time. Each figure traces back to the exact source row, so the answer is auditable instead of plausible-looking.
- One live model across sources. It joins across databases, spreadsheets, PDFs and APIs — most text-to-SQL tools target a single schema.
- Semantic understanding. It learns the meaning of cryptic schemas, so it picks the right tables instead of guessing from column names.
- Autonomous investigation. Its Analyst agent can break a hard question into steps, run several queries, and reason about the results — well beyond one-shot generation.
- Guardrails. Generated SQL is validated and sandboxed, so a bad translation can't return a wrong number unnoticed or touch what it shouldn't.
Text-to-SQL is one feature inside iDBQuery, not the whole product.
A text-to-SQL tool, or an answer you can defend
| Text-to-SQL | iDBQuery | |
|---|---|---|
| What comes back | A SQL statement to run yourself | The answer, with the query and the source |
| Scope | One database, one schema | Databases, spreadsheets, PDFs and APIs together |
| Understanding a cryptic schema | Depends on names being meaningful | A semantic layer learns what the columns mean |
| When it is wrong | You may not find out | The cited source row is there to check |
| Cost and simplicity | Often cheap or free, and easy to bolt on | A platform, priced as one |
When the other one is the right choice
A text-to-SQL helper is a reasonable choice when you already have an analyst who will read and run the SQL — it saves them typing, and that is a real saving. The distinction matters when the person asking cannot check the SQL: at that point a statement is not an answer, and the question of whether to trust it has simply been moved rather than solved.
Updated 2026-08-08