DB Neuron Graph: your database as a 3D brain
Database records as a 3D neuron-like brain — every node a record, every thread a relation, every pulse a synapse firing. ~500 sample records, force-directed in WebGL.
We asked a simple question: what does a database actually look like if you could see every relationship at once? Tables hide connections. So we drew ~500 records as a 3D network and let physics arrange them — and they organised themselves into something that looks unmistakably like a brain.
What you're looking at
Every white node is a record. Every thread is a relationship — a foreign key, a tag, a membership. The sample models four tables: clients link to projects, projects link to invoices, and projects share a pool of tags. A force-directed layout treats each link as a spring and each node as a charged particle, then runs the simulation in three dimensions until it settles. Densely-connected records pull together into clusters; loosely-connected ones drift to the edges. The spider-web shape is emergent, never drawn.
The neuron part
Hover any node and its relationships light up — neighbours stay bright, everything else dims, and the connecting synapses fire a pulse of light. Left alone, the graph keeps firing at random, the way a resting brain never goes quiet. It's the same data a table would show as rows; here the structure is the story.
How it's built
One HTML file, no build step. 3d-force-graph (which wraps Three.js and d3-force-3d) does the heavy lifting on the GPU via WebGL. ~500 nodes and ~500 edges run smoothly on a laptop; for much larger graphs this is exactly the workload where WebGPU's compute path starts to matter. Swap the sample-data builder for a fetch('/graph.json') and it visualises any {nodes, links} set.
Where it's useful
- Schema maps. Point it at a SQL database — tables as nodes, foreign keys as edges — and the shape of the schema appears instantly.
- CRM / content graphs. Clients ↔ projects ↔ invoices, or posts ↔ tags ↔ authors. Hubs and orphans jump out.
- Explaining data to non-engineers. Nobody reads a foreign-key diagram for fun. People will happily spin a brain.
The data here is a public sample — fictional account names, generated locally, no backend. Want this pointed at your data? Tell us the shape and we'll wire it up.