Configuration and administration
Field hiding and masking
Hide fields that are not needed for analysis and mask sensitive values in query results.
Guide · Module: data · Users · Deployment: Not specified · Confirmed product versions: Not specified · Source: Existing documentation
Field hiding excludes columns that are unnecessary for analysis. Field masking preserves a column's use in calculations while replacing sensitive values in results.
| Intended effect | What to use | Example |
|---|---|---|
| The column is irrelevant and should not distract AI | Hide switch | Internal auto-increment ID, creator ID |
| The column is needed for statistics, but its actual values should not be visible | Masking dropdown | Phone number, national ID number, bank card number |
| The column is sensitive and not used for analysis at all | Both | An archived national ID column never used in analysis |
Before you start
| Item | Requirement |
|---|---|
| Permissions | Project owner or administrator |
| Prerequisite | Data source is available |
| Entry point | Main navigation on the left → “More” → “Data sources” → open a source card → select a table in the left tree |
Steps
Hide a column
The rightmost column in “Field structure” is Hide, with a switch in each row. Turning it on grays out the entire row, indicating that the field no longer participates in AI analysis.
The explanation says AI ignores the column when answering after it is hidden. The switch takes effect immediately; there is no additional save button.
Restore a column
Turn off the Hide switch. The field's semantic vectors are retained, so it becomes searchable immediately when visible again without waiting for a rebuild.
Set a masking type
Click the field row to expand its detail panel and find the masking dropdown, whose explanation says values are masked according to the selected type. Choose one of six options:
Disabled, Person name, Phone number, National ID number, Bank card number, or Email.
Your selection saves immediately without confirmation. Choose Disabled to turn masking off.
Verify masking
After the change, ask a question that returns the column and check that values in the result table have been replaced. Masking occurs before query results reach the model, so the model sees replacement values.
Key constraints
| Item | Value or default | Where to change it |
|---|---|---|
| Hiding scope | AI skips the column when generating queries, answering questions, and retrieving field semantics | Hide switch in Field structure |
| Data deletion when hidden | Nothing is deleted; the column and values remain in the table | — |
| Restoring visibility | Turn the switch off; retained vectors make the field searchable immediately | Same switch |
| Masking types | Disabled, Person name, Phone number, National ID number, Bank card number, Email | Masking dropdown in the field detail panel |
| Where masking happens | Values are replaced before query results are serialized for the model; original values do not enter the model | — |
| Hiding versus masking | Hiding excludes the entire column; masking keeps the column in use but replaces values | — |
| Access-control boundary | These switches affect only what AI sees; project roles and data access policies determine who can access which data | Data access control |
Troubleshooting
| Symptom | How to identify it | What to do |
|---|---|---|
| Switch has no effect | Inline message says field visibility could not be updated | Refresh and retry; contact an administrator if it persists |
| AI still mentions a field | Its row is not gray and Hide is off | Turn the switch on |
| Phone numbers remain unmasked | Masking dropdown is still Disabled | Choose Phone number |
| Cannot change the masking type | Inline message says masking type could not be updated | Refresh and retry; contact an administrator if it persists |
| A metric no longer works after hiding | It depends on the hidden column | Hiding changes AI visibility, but the metric needs that field; use masking instead |
| Original values are needed but unavailable | The column has a masking type and conversation results contain replacements | Masking deliberately keeps original values from the model; query the database directly if you need them |
Next steps
- Explain the business meaning of retained fields: Table and field descriptions
- Match abbreviations to actual values: AI indexes
- Manage files, training sets, and metadata imports/exports: Manage data sources, files, and training sets
- Verify with a real question: Conversations