Filters
The Filters tab controls which parts of your data source are included in the semantic layer. Precise filtering is critical — it determines what the AI knows about and can query. Including too many irrelevant schemas, tables, or columns reduces query accuracy and increases extraction time.
Note: The Filters tab displays the live schema fetched from the connected data source. The connection must be valid for filtering to work.
Schema Filter
Determines which database schemas are visible to the semantic layer. Use Include or Exclude — not both simultaneously.
| Option | Behavior |
|---|---|
| Include Schemas | Only the selected schemas are passed to the semantic layer. All others are ignored. |
| Exclude Schemas | All schemas are included except the selected ones. |
Include and Exclude are mutually exclusive. Selecting a schema in Include clears the Exclude list, and vice versa.
Recommendation: Use Include to explicitly whitelist only the schemas relevant to your model. This avoids unintended exposure of system schemas or unrelated data.
Schema filtering is only available for datasource types that expose a schema hierarchy (e.g., SQL databases). Lake format and file-based sources do not have a schema level.
Table Selection
Controls which tables within the filtered schemas are included. Select individual tables by checking them in the list.
| Option | Behavior |
|---|---|
| Include Tables | Only the checked tables are included. All others are ignored. |
| Exclude Tables | All tables are included except the checked ones. |
Include and Exclude are mutually exclusive. Checking a table in Include clears the Exclude list, and vice versa.
Column Selections
Per-table column filtering. Use this to expose only the columns that are meaningful for analysis.
For each column selection entry:
- Select a table from the dropdown (based on the current schema and table filters).
- Choose Include Columns (whitelist) or Exclude Columns (blacklist) for that table.
Column selections are additive — you can define selections for multiple tables within the same data source.
Include and Exclude are mutually exclusive per table.
Column Row-Level Filters
Adds row-level data restrictions equivalent to a WHERE clause at the semantic layer level.
Each filter has four parts:
| Part | Description |
|---|---|
| Table | The table to filter |
| Column | The column to filter on |
| Operator | The comparison operator: EQUALS, NOT_EQUALS, IN, CONTAINS, etc. |
| Value | The filter value. For IN operators, provide a comma-separated list. |
Use cases:
- Multi-tenant isolation:
tenant_id EQUALS acme - Exclude test data:
environment NOT_EQUALS test - Restrict to active records:
status EQUALS active
Saving Filters
Click Save to apply the filter configuration. Filters become active in the next semantic extraction — changes do not take effect until you re-run extraction from the Operations tab.