> ## Documentation Index
> Fetch the complete documentation index at: https://netter.ai/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# Promote a database

> Turn a raw project database into a typed ontology entity. Reversible via demote.

A project **database** is just a Parquet table written by a `data_sink` step. Promoting it turns the table into an **ontology entity** with a primary key, named properties, and a place in the company-wide object graph.

## Before you begin

* You must be **admin** or **owner** on the company.
* The database must already have at least one successful run so the schema is known.

## Promote

<Steps>
  <Step title="Open the database">
    From **Ontology → Databases**, pick the source database. Tables are grouped by origin: from projects, from integrations, from files.
  </Step>

  <Step title="Pick the primary key">
    Click **Promote**. For step-backed databases the platform uses `_row_key` automatically. For file-backed databases (CSV uploads), you choose which column identifies a row.
  </Step>

  <Step title="Name the entity">
    Set the entity's display name (singular, business-y—"Customer", not "customers\_table"). Properties are derived from the table's columns and can be renamed afterward.
  </Step>
</Steps>

Once promoted, the entity is visible to dashboards, apps, and the natural-language query.

## Demote

Demoting drops the entity layer but keeps the underlying database intact. Use it when you promoted by mistake, or when the entity has been replaced by a different source.

<Warning>
  Demoting deletes all property and relationship metadata for the entity. Row overlays are also removed. The source database is untouched.
</Warning>

## Clear overlays

If you've made manual edits via [row overlays](/docs/ontology/edit-rows) and want to reset to whatever the pipeline produces, use **Clear overlays**. The entity stays promoted; only the overlay rows are dropped.