Feature idea: 3D interactive visualization for graph.html

[pedrolbacelar]

Context

I've been using graphify to index a multi-repo project (13 repos, ~4400 nodes, ~7000 edges). The current 2D vis.js HTML output works well for small graphs, but once you get past ~500 nodes the flat layout makes it hard to distinguish clusters and trace cross-community connections.

What I'm thinking

1. 3D visualization with --3d flag

Replace vis.js with 3d-force-graph (WebGL-based, MIT licensed) when a --3d flag is passed. The data format is nearly identical — nodes with id, label, color, size and edges with source, target. Benefits:

• Community clusters naturally separate in 3D space (force-directed layout gives them depth)
• Cross-community bridges become visible as arcs between clusters
• Navigation via orbit/zoom/pan feels intuitive on large graphs
• Still a self-contained HTML file, no server needed

2. Sidebar improvements

• Collapsible sidebar — on large graphs the sidebar takes too much screen space. A toggle button to hide/show it would help.
• "Hide all / Show all" communities toggle — right now you see everything. Being able to deselect all communities and then selectively enable only the ones you care about would make exploration much more targeted.

3. Community-level interaction

• Click a community in the sidebar to isolate it (hide everything else)
• Shift+click to add communities to the current selection
• Color-code nodes by community with a legend

Existing work

I've been building a separate graph viewer that currently imports GitNexus output format: code-synapse-viewer. It uses Sigma.js + ELK layout with features like:

• Explorer view (full graph with clustering)
• Context view (2-hop neighborhood around a selected symbol)
• Impact view (upstream call chain with depth-based coloring and risk assessment)

It currently only supports the GitNexus JSON format — it does not support graphify's graph.json yet. I'm planning to add a graphify parser so it can consume both formats.

Proposal

I'd like to build the --3d flag and sidebar improvements in my fork first, validate it on real graphs, and then upstream it as a PR if there's interest. The 2D vis.js output would remain the default — --3d would be opt-in.

Would this be welcome? Any preferences on the approach or the 3D library choice?

cc @safishamsi