Dbt power tools AI based Documentation

A powerful CLI tool that generates LLM-powered documentation for dbt models and columns — written directly into your schema.yml, ready to appear in dbt Docs.

It supports both schema-only documentation and data-aware documentation by profiling your warehouse data and feeding summary statistics to the LLM for enhanced accuracy.

🚀 Features

✅ LLM‑generated dbt documentation

• Generates rich model & column descriptions

• Writes directly into schema.yml

• Fully dbt‑docs compatible

✅ Customizable Jinja prompt templates

Located in <project>/prompts/
You can rewrite the tone, style, or structure.

✅ dbt-aware model selection

Supports:

• --select, --exclude, --tags

• Glob patterns (stg_*, marts.*)

• Parent/child expansion (+model_name)

✅ Optional data profiling (--use_data Y)

When enabled:

1. Reads warehouse credentials from profiles.yml

2. Runs the model’s compiled SQL

3. Profiles real data (Postgres & Redshift supported today)

4. Computes:

   • Missing %

   • Unique %

   • Min / Max

   • Mean / Std

   • Example values

5. LLM uses this context to produce far higher‑quality column docs

6. A Markdown stats table is appended to each column description

7. These appear in dbt Docs → Documentation UI