The hardest part of joining a project is building a mental model of how it fits together. Vibgrate Graph shortens that ramp by letting newcomers explore structure directly instead of reading every file. This guide is for both the new developer and the lead setting them up.
Set up the graph
Have the new developer build the graph from the repository root:
vg build
After the first build, all exploration works offline — no account needed for graph queries.
Ask the codebase questions
When a newcomer does not yet know the right names, natural language is the fastest entry:
vg ask
vg "…"
Replace the quoted text with a real question, such as where requests enter the system or how a feature is implemented. The result is a focused context block that explains the relevant code — and it can be pasted into an AI assistant for a guided explanation.
Explore from an entry point
Once they have a starting node, walk outward to see how behavior flows:
vg tree
This shows what the entry point calls, level by level, giving a map of the feature without manual file hopping. Bound it with --depth to keep it readable.
Zoom into specific pieces
For any node that looks important, vg show explains what it is, what it calls, and what calls it:
vg show
Seeing the callers immediately reveals which functions are central hubs worth learning first.
A suggested onboarding path
- Build the graph.
- Ask a few natural-language questions about the area you will work in.
- Pick an entry point and walk its call tree.
- Use vg show on the hubs you keep encountering.
- Before your first change, run vg impact to learn the blast radius.
Going further with AI assistance
For a continuous experience, Vibgrate AI Context (via vg serve) serves the code map and version-correct library docs to an AI assistant offline, so a newcomer can ask their assistant accurate questions about the real codebase.
Related
See the articles on asking questions with vg ask, exploring the call tree with vg tree, and impact analysis before a change.