Supabase Memory icon

Supabase Memory

AI Agent memory node — save, load, search & delete structured data in any Supabase table with custom field mapping and SQL-like filters.

Actions6

Search Memory

AI-generated

Overview

The node performs an advanced search operation on a specified Supabase table, allowing users to retrieve structured data with multiple filters, sorting, pagination, and column selection. It is useful for querying large datasets with complex conditions, such as searching agent memory records or any structured data stored in Supabase tables.

Use Case Examples

  1. Search for memory records in a table with specific session_id and date range filters.
  2. Retrieve a sorted list of user activity logs with pagination and selected columns.
  3. Perform a full-text search on a content column with case-insensitive matching.

Properties

Name Meaning
Table Name The Supabase table to operate on, selected either from a list or by manual name entry.
Columns to Return Comma-separated list of columns to return from the table. Supports PostgREST syntax and '*' for all columns.
Filters SQL-like filters to apply to the search query. Multiple filters are combined with AND. Supports various operators like equals, not equals, greater than, LIKE, IN, JSON contains, and full-text search.
Order By Sort the returned results by specified columns and directions (ascending or descending).
Limit Maximum number of records to return, with a range from 1 to 10,000.
Offset Number of records to skip for pagination purposes.

Output

JSON

  • json
    • _totalCount - The total count of records matching the search filters, useful for pagination.

Dependencies

  • Supabase API credentials including Supabase URL and service role key for authentication.

Troubleshooting

  • Ensure the Supabase credentials are correctly configured and have sufficient permissions to query the specified table.
  • Filters must be correctly formatted; invalid JSON or unsupported operators can cause errors.
  • The 'Order By' columns must exist in the table schema to avoid query errors.
  • Large result sets should be paginated using 'Limit' and 'Offset' to prevent timeouts or memory issues.
  • If the RPC function 'get_tables_list' is not available, the table list may not auto-load; use manual table name entry instead.

Discussion