Show HN: Fitness Functions for Dbt Projects

TFF is an open-source architectural linter and health dashboard for dbt, SQLMesh, and Google Cloud Dataform projects. Its central premise is that file-level SQL linters miss problems that emerge across a warehouse’s full transformation graph, so TFF analyzes models and their dependencies together. It can flag violations of expected layer direction - from staging through intermediate and core to marts - unauthorized dependencies between business domains, repeated transformation logic in CTEs, circular dependencies, and other model-level issues such as SELECT * or missing ownership metadata. The repository’s example audit shows these checks producing errors that can block a merge and warnings that identify opportunities to consolidate logic.

TFF is designed to run without a warehouse connection or required configuration: it infers common project conventions and can be invoked with uvx or installed as a Python package. Beyond its check command, it calculates a repository health score from 0 to 100, which teams can use as a CI threshold, and offers limited automatic correction, including synchronizing model names with filenames. A GitHub Action can compare pull-request changes against a base branch, gate merges, and add inline annotations; output formats include SARIF, JSON, and HTML. The project also documents custom rules and plugins, provides pre-commit integration, and reports an enterprise case study involving more than 2,200 dbt models. TFF is distributed under the MIT license.