rule-engine

Rule Engine

On Entry

Tell the user:

Rule Engine — I'll search for known migration fix patterns that apply to your code, and can apply them automatically (regex) or with your approval (AI-guided).

Prerequisites

• Snowflake connection active with direct SQL ability
• Session configured with a Snowflake connection and database. The configure MCP tool sets up the RULE_ENGINE schema automatically on first call.

Setup (automatic)

You do not run a separate setup step. On the first configure call that supplies both snowflake_connection and snowflake_database, the MCP server creates and seeds everything needed:

• RULE_ENGINE schema
• RULES, RULE_EVENTS, CODE_UNITS_SQL tables
• Built-in seed rules (SQL Server → Snowflake patterns)
• RULE_SEARCH and CODE_SEARCH Cortex Search services

The setup handler is idempotent: calling configure again reruns pending migrations and re-seeds missing rules without disturbing existing data.

Commands

Search

Find applicable rules for a SQL file. Sends the file content to Snowflake for regex pattern matching (REGEXP_LIKE) and Cortex semantic search.

→ Load search/SKILL.md

Apply

Apply matched rules to local SQL files. Supports single-file and batch application. Runs regex replacements mechanically, presents AI-mode fixes for review.

→ Load apply/SKILL.md

Extract Rules

Analyze code changes to extract reusable migration rules. Works on interactive fixes (before/after comparison) or git history (retroactive analysis of committed changes).

→ Load extract/SKILL.md

Propagate

Given a rule, find all code units in the project it applies to. Uses reverse search (regex + Cortex semantic) to identify candidates, then hands off to Apply for batch application.

→ Load propagate/SKILL.md

Status

Check what rules exist and recent fix history:

SELECT id, name, replacement_mode, priority, rule_sentiment, rule_applications, successes, created_from
FROM RULE_ENGINE.RULES
ORDER BY priority;