Knowledge Graph
Rook is backed by a knowledge graph that remembers. It’s why Rook can resolve your intent to the right command or component so quickly — and why it gets better over time.
What’s in it
Section titled “What’s in it”- 197 Rhino commands — with 543 logged observations of what works and what fails
- 945 Grasshopper components — mapped to 1,533 intents
When you describe what you want, Rook consults this graph to find the right pieces and how they fit together.
It learns from corrections
Section titled “It learns from corrections”The graph improves from failure, not routine success. When a mistake gets fixed, correction detection records the learning — so the next agent, and the one after, won’t repeat it.
Tiered retrieval
Section titled “Tiered retrieval”Rook pulls only as much knowledge as a task needs, keeping responses fast and focused:
| Tier | Cost | When |
|---|---|---|
| quick | ~20 tokens | fast check |
| context | ~50 tokens | routine use |
| errors | ~30 tokens | on failure |
| raw | 500+ tokens | forensic / deep dive |
Pulling the right tier instead of everything is roughly a 97% token reduction versus reading raw patterns.
Why it matters to you
Section titled “Why it matters to you”You get the benefit of accumulated expertise — the patterns experienced users rely on — without having to learn or maintain any of it yourself.
Related
Section titled “Related”- The Design Cascade — its Consolidate phase feeds the graph
- Multi-Agent — how the graph keeps cheap agents competent