stamp-transactions

What problem does it solve?

This Skill eliminates manual and inconsistent categorization of transaction records by automatically assigning provider names and category identifiers to unclassified transactions, reducing reconciliation time and user intervention.

Core Features & Use Cases

• Four-stage matching engine (PATTERN, LEARNED, EXACT, FUZZY) that evaluates description patterns, learned hashes, exact substrings, and fuzzy token overlap to propose matches with confidence scores.
• Dry-run preview mode to inspect proposed updates without writing to the database and automatic confirmation for matches above confidence thresholds.
• Batch processing for a Supabase user's transactions with mappings from global providers, user overrides, and category inference fallbacks for institution-specific categories.
• Use case: Clean up a bank feed for a user by stamping hundreds of unclassified transactions so downstream analytics and budgets use consistent provider and category data.

Quick Start

Run the stamp-transactions command for a user to preview or apply auto-categorization by running tsx bin/optimal.ts stamp-transactions --user-id USER_UUID and include --dry-run to preview results without updating the database.