Exposing the dbt manifest through LLMs to make data discoverable

See how M1 used dbt metadata and LLMs to build an AI-powered catalog that makes data searchable and intuitive.

Using dbt metadata and LLMs, we built a system that helps users find and understand data assets beyond traditional search. Our solution creates a rich catalog focusing on business problems each model solves, enabling users to search assets and get guidance through an AI interface.

Speaker:

Matthew Kramer, Senior Data Engineer at M1

Daniella Vecchionacce, Analytics Engineer at M1

Coalesce 2025: Exposing the dbt manifest through LLMs to make data discoverable | dbt Labs