Knowledge base
It learns your data before it answers.
Choose the tables and press Build. Every table you pick is read, profiled and described, then checked, and questions only open once it's complete.
orders
CheckedOne row per customer order, across every sales channel.
| Column | Description | Profile |
|---|---|---|
| order_id | Unique order number | unique · 0% empty |
| amount | Order value in USD | $4 – $2,180 |
| region | Sales region | North, East, South, West |
| created_at | When the order was placed | Jan 2024 – Oct 2026 |
From tables to answers.
Five steps, once per source.
- 1
Connect a source
Point atomiq at your database or warehouse. Nothing is copied out.
- 2
Choose the tables
Only the ones your team actually asks about.
- 3
Build
Every table is read, profiled and described: columns, values and how tables relate.
- 4
Check
The knowledge is checked before anyone can ask. Questions open once it passes.
- 5
Ask
Every answer now uses your tables, your rules and your definitions.
What the knowledge base holds.
Everything atomiq needs to answer the way your team would.
- Table and column descriptions
- Plain-language descriptions of what each table and column holds.
- Relationships
- How your tables join, so answers combine them correctly.
- Business rules
- The rules your team works by, applied to every answer.
- Metric definitions
- One definition of revenue, churn or margin that everyone shares.
- Issues
- Problems the check found in the data, listed so you can fix them.
- Import and export
- Move knowledge between projects.
Watch a knowledge base build.
Connect a source, choose the tables, press Build. Every table is read, profiled, described and checked.
Better with atomhub.
atomiq runs on its own in your cloud. Run it on atomhub, the data operating system, and atomhub takes care of ingestion, storage, governance and the agents underneath — so atomiq can focus on the answers.
Connects to the data you already have
- PostgreSQL
- MySQL
- MariaDB
- SQL Server
- Oracle
- ClickHouse
- Snowflake
- BigQuery
- Redshift
- Databricks
- Athena
- Trino
AI-powered observability for every log
Ask your data what happened, and why
- Data plane
- Execution runtime
- Governance kernel
- Observability & lineage
- Agentic operations
Agents that run it
- AWS
- Google Cloud
- Azure
- On-premises
Questions
Do I have to write a semantic layer first?
No. atomiq builds the knowledge base from your tables. You can add business rules and metric definitions on top whenever you want.
What happens if the check finds problems?
It lists them as issues — duplicate keys, empty columns, values that don't line up — so you can fix them before questions open.
Can I move a knowledge base between projects?
Yes. Knowledge can be exported and imported.
Build a knowledge base on your data.
Pick a few tables. We'll build it, check it, and answer your first question live.