Why Permissions Matter

When building apps with database access, you need to control what data the AI can see and modify. Permissions let you:

Protect sensitive information
Limit AI actions to specific data
Create secure multi-user apps

Permission Types

Row-Level Permissions

Row-level permissions control which records the AI can access. Think of rows as individual items in your data.

Example: In a customer database, you might want the AI to only see customers from a specific region:

{
  "filter": {
    "region": "North America"
  }
}

With this filter, the AI only sees customers where region equals "North America".

Column-Level Permissions

Column-level permissions control which fields the AI can access. Think of columns as the properties of each record.

Example: In a customer database, you might hide sensitive financial data:

Visible columns:

name
email
region
last_contact

Hidden columns:

credit_card
annual_revenue
internal_notes

Setting Up Permissions

Step 1: Enable Database

Open your app settings
Navigate to the Database tab
Enable database access

Step 2: Create Collections

Collections are like tables in your database:

Click Add Collection
Name your collection (e.g., "customers")
Define the schema (fields and types)

Step 3: Configure Row Filters

Select a collection
Click Row Permissions
Define filter conditions

Filter operators:

equals: Exact match
contains: Partial text match
greaterThan: Numeric comparison
in: Match any value in a list

Step 4: Configure Column Access

Select a collection
Click Column Permissions
Check/uncheck columns the AI can access

Common Patterns

User-Specific Data

Limit the AI to the current user's data:

{
  "filter": {
    "user_id": "{{current_user_id}}"
  }
}

Read-Only Access

Allow reading but prevent modifications:

{
  "read": true,
  "write": false,
  "delete": false
}

Hiding Sensitive Columns

Hide PII and financial data:

{
  "hidden_columns": [
    "ssn",
    "credit_card",
    "bank_account",
    "salary"
  ]
}

Testing Permissions

Always test your permissions:

Create test data with various permission scenarios
Ask the AI to retrieve filtered and unfiltered data
Verify hidden columns aren't revealed
Test write/delete operations if enabled

Best Practices

Start restrictive: Give minimal access, then expand as needed
Document your filters: Keep notes on why each filter exists
Regular audits: Review permissions periodically
Test edge cases: What happens with empty filters? Null values?

Understanding Permissions