22-25 April 2026 
Video unavailable
SQLBits 2026
Untangling dbt Dependencies with Knowledge Graphs and AI
Learn how dbt-kg turns dbt metadata into a queryable knowledge graph, enabling AI-powered analysis of dependencies, circular refs, and model relationships to help you understand and optimize large dbt projects.
As analytics teams scale their dbt projects, what starts as a clean set of staging models often evolves into hundreds of interconnected models with complex, hard-to-navigate dependencies. Traditional documentation falls short when trying to answer questions like "Which source tables are most widely reused across the entire dependency chain?" or "How do I break this monolithic project into smaller, orchestratable jobs without creating circular dependencies?"
In this session, we'll introduce dbt-kg, an open-source project that transforms your dbt metadata into a queryable knowledge graph using graph databases like FalkorDB and Neo4j. By leveraging the naturally graph-shaped structure of dbt projects, we enable AI-powered natural language queries that can analyze complex project-wide patterns and dependencies.
I wrote about this approach in detail here: https://www.linkedin.com/posts/egorseno_dbt-dataengineering-graphdatabase-activity-7391107790909456384-iu_s?utm_source=share&utm_medium=member_desktop&rcm=ACoAAC1CwQsBx7nr1t6yomC0WgSTBfXKQAYQvv8
Attendees will learn:
How to upload dbt manifest and catalog files into a graph database
Real-world examples of querying dbt dependencies using natural language chat powered by LLMs
Practical use cases including identifying circular dependencies, tracing indirect model relationships, and creating intelligent job grouping strategies for orchestration
Live demo of the open-source tool analyzing a sample dbt project
This approach is particularly valuable for education data teams and other organizations where analytics infrastructure grows organically and data engineers need efficient ways to understand and refactor increasingly sophisticated dbt projects. Whether you're managing a single warehouse or a multi-organization data platform, this session will show you how to make your dbt metadata more intelligent, conversational, and actionable.
Target Audience: Analytics engineers, data engineers, and dbt practitioners looking to better understand and optimize their dbt projects at scale.
https://www.youtube.com/watch?v=FZra9fM7-ks&t=2124s
In this session, we'll introduce dbt-kg, an open-source project that transforms your dbt metadata into a queryable knowledge graph using graph databases like FalkorDB and Neo4j. By leveraging the naturally graph-shaped structure of dbt projects, we enable AI-powered natural language queries that can analyze complex project-wide patterns and dependencies.
I wrote about this approach in detail here: https://www.linkedin.com/posts/egorseno_dbt-dataengineering-graphdatabase-activity-7391107790909456384-iu_s?utm_source=share&utm_medium=member_desktop&rcm=ACoAAC1CwQsBx7nr1t6yomC0WgSTBfXKQAYQvv8
Attendees will learn:
How to upload dbt manifest and catalog files into a graph database
Real-world examples of querying dbt dependencies using natural language chat powered by LLMs
Practical use cases including identifying circular dependencies, tracing indirect model relationships, and creating intelligent job grouping strategies for orchestration
Live demo of the open-source tool analyzing a sample dbt project
This approach is particularly valuable for education data teams and other organizations where analytics infrastructure grows organically and data engineers need efficient ways to understand and refactor increasingly sophisticated dbt projects. Whether you're managing a single warehouse or a multi-organization data platform, this session will show you how to make your dbt metadata more intelligent, conversational, and actionable.
Target Audience: Analytics engineers, data engineers, and dbt practitioners looking to better understand and optimize their dbt projects at scale.
https://www.youtube.com/watch?v=FZra9fM7-ks&t=2124s