From question to dashboard.
Ask in plain English — Datarelix runs a validated, read-only query and shows its work.
Ask in plain English
-
Every answer comes from a query that actually ran — with the SQL on screen.
-
Each statement passes a read-only validator; row limits are automatic.
-
@ mentions pin tables into context; follow-up questions carry it forward.
-
Failed queries revise and retry automatically, within strict limits.
SELECT c.name, SUM(o.total) AS paid_order_value_usd
FROM customers c
JOIN orders o ON c.id = o.customer_id
WHERE o.status = 'paid'
AND o.created_at >= DATE_TRUNC('quarter', CURRENT_DATE)
AND o.created_at < DATE_TRUNC('quarter', CURRENT_DATE) + INTERVAL '3 months'
GROUP BY c.id, c.name
ORDER BY paid_order_value_usd DESC, c.id
LIMIT 10; | Customer | USD |
|---|---|
| Enterprise Plus | 352,940 |
| Acme Corp | 284,110 |
| Globex | 219,803 |
| Initech | 201,442 |
Enterprise Plus: 352,940 USD; Acme Corp: 284,110 USD; Globex: 219,803 USD; Initech: 201,442 USD; Northstar: 169,000 USD; River: 148,000 USD; Oak: 127,000 USD; Cedar: 106,000 USD; Pine: 91,000 USD; Maple: 78,000 USD
Connect what you already run
-
On Azure SQL, Databricks, and Kusto, queries can run with each user’s own permissions.
-
Parquet, Delta, CSV, JSON, and Excel files in cloud storage query like tables.
-
Schema discovery maps your tables, with AI-written descriptions you can edit.
-
Ask every source you connect the same way, in plain English.
- PostgreSQL
- MySQL
- Azure SQL
- Databricks SQL
- Kusto
- BigQuery
- Amazon Athena
- Elasticsearch
- Snowflake
- Files & Object Storage
- Parquet
- Delta
- CSV
- JSON
- Excel
Chart it. Pin it. Refresh it.
-
Line to heatmap — the chart fits the result, restyleable without code.
-
“Add this to my Sales dashboard” pins the chart straight from the conversation.
-
A refresh re-runs the stored query — one card or all of them.
-
Results export as CSV; charts as PNG or SVG.
Revenue by customer
Bar · 10 rows
Revenue by month
Line · 12 rows
Guardrails you can check
-
Read-only by design: every generated query is validated before it runs.
-
Analysis code runs isolated, with no internet access.
-
Credentials are encrypted and never enter the AI model.
-
Every run is a step-by-step trace you can reopen.
Read the full security posture
Go deeper: how the read-only architecture is enforced and how credentials stay out of the AI model.
- Plan validated against a strict schema
- Query passed read-only validation
- Executed — 10 rows returned
tables · customers, orders
Platform
- Fully managed — nothing to operate
- SSO or email sign-in