dbt-buddy 0.1.0

AI-based documentation for dbt-models

dbt-buddy

Autogenerated documentation for dbt-models using YandexGPT LLM

dbt-buddy is a python package with CLI that allows automatically create YAML-based documenation for existing dbt-model. Built-in method fill_yaml_with_column_description() will add columns description in Russian language using LLM-model YandexGPT.

Installation

1. Install uv if you haven't already:

curl -LsSf https://astral.sh/uv/install.sh | sh

2. Clone the repository and navigate to the project directory:

git clone <repository-url>
cd dbt-buddy

3. Install dependencies:

uv sync

Prerequisites

1. You need to create .env file in the dbt-project working directory and add the following secret:
   • API_KEY=<secret key> - required to access YandexGPT API.
   • CATALOG_ID=<catalog identifier> - Yandex Cloud catalog ID is a part of API request.

Available commands

1. document - generates YAML-based documentation with AI-proposed columms description.

document

You can create documentation by simply running the command:

$ buddy document --model <dbt-model name>

The result will be a text string in the console, formatted in a documentation format acceptable for dbt.

CLI Options

1. -m <model_name>, --model <model_name>(required). The name of existing dbt-model.
2. --project-dir. The path to directory with dbt_project.yml. Default is the current working directory.
3. --profiles-dir. The path to directory with profiles.yml. Default is the current working directory.
4. -s, --save. If specified, the generated documentation is saved in a YAML-file in the same directory and with the same name as the specified model.
5. -e, --examples. If specified, YandexGPT will try to add column's possible accepted values (especially relevant if the SQL-query explicitly specifies values with the CASE statement).
6. -v, --verbose. If specified, the response from the YandexGPT API will be displayed in the console.

You can get the full list of existing options by running the command:

$ buddy document --help