Photo credit I’ve been working with a Goodreads book dataset for a while now, and my next goal was to generate vector embeddings for 2.3 million book descriptions using Neo4j’s new Cypher AI procedures and a local Ollama model. I figured it’d be pretty straightforward - set up the config, write a query, let it run. What could go wrong? Turns out…​quite a bit. What I expected to be a few query tweaks turned into a multi-step adventure through configuration quirks, batching strategies, query optimization, and a crash course in how tokens actually work.

Photo credit When text-to-query approaches (specifically, text2cypher) first entered the scene, I was a bit uncertain how it was useful, especially when existing models were hit-or-miss on result accuracy. It would be hard to justify the benefits over a human expert in the domain and query language. However, as technologies have evolved over the last couple of years, I’ve started to see how a text-to-query approach adds flexibility to rigid applications that could previously only answer a set of pre-defined questions with limited parameters.