Trouble modelling letter-connection paths for a small word-puzzle experiment

I’m playing around with a simple word-puzzle idea—something along the lines of a little letter-link challenge, and I wanted to see how well it could be represented in neo4j. The setup is straightforward, a few letters placed around a virtual “box,” and the player forms words by connecting them without reusing nodes.

My issue is that my current graph model keeps producing unintended paths. Sometimes Cypher returns valid words I didn’t expect, and other times it misses obvious ones. I’ve tried modelling each letter as a node with directional relationships, but the traversal rules feel too loose or too strict depending on the constraints I add.

Has anyone here tackled a similar problem involving constrained pathfinding or letter-sequence modelling? I’d love advice on the best way to represent adjacency and prevent node reuse without making the queries overly complex.

1 Like

what is your model and query?