AI on Your Lakehouse: Context Comes in Shapes, Not Queries — Zach Blumenfeld, Neo4j

An agent searching technical manuals can benefit from knowing how the documents are organized. In this Neo4j workshop, a graph is built from simulated repair manuals: documents contain ordered sections, and named links connect one section or document to another. The structure is loaded with deterministic parsing (regular expressions rather than an LLM), which works here because the manuals are consistently formatted.

The agent can use an outline to choose a section, retrieve its text and follow a link into another document, and can also run full-text search scoped to part of that hierarchy. This supplies a navigation path alongside keyword search, helping it collect connected context. In one demonstration the agent skipped the outline and took a slower search route. The presenter says no benchmarking was run on this setup and calls the course conceptual, so no accuracy or performance advantage is established.

“an ontology helps something interpret and reason about the data. A semantic layer helps something understand what the consistent agreed upon terms are so it can query the data accurately.”

Zach Blumenfeld

“you can't really search for a negative. You have to search for things that are there.”

Zach Blumenfeld

“as you start adding data to it, it's almost like a bunch of pre-joined tables and everything's already interconnected”

Zach Blumenfeld

“graphs can be very useful when you start having these more global types of questions, these estate level questions that you want to ask”

Zach Blumenfeld

“it's going to go on the internet and get the Stack Overflow questions from like four or five or six years ago and it's going to give you outdated bad cipher basically”

Zach Blumenfeld