What problem does it solve?
It solves slow and unnecessarily expensive dbt runs on StarRocks caused by poor incremental partition filtering, inefficient rebuild patterns, and stale optimizer statistics.
Core Features & Use Cases
- Partition-aware incremental models that reprocess only a bounded window of partitions instead of scanning or rebuilding whole tables, suitable for large fact tables with late-arriving data.
- Performance-critical StarRocks tuning in dbt SQL using join/order hints, materialization choices, and model selection so complex DAGs run faster with fewer rebuilds.
- Post-hook and pre-hook optimizations like targeted ANALYZE TABLE for better CBO plans and pre-creating partitions for safe INSERT OVERWRITE incremental strategies.
Use case: you have a daily gold model that currently rebuilds too much history and runs for hours; this skill guides you to switch to partition-replacing increments, analyze only touched partitions, and run only modified downstream models.
Quick Start
Ask the agent to generate a revised incremental dbt model for StarRocks that uses partition-aware filters with a configurable late-arrival lookback and adds a post-hook to run ANALYZE TABLE for only the current partitions.