How node-editor agent nodes map to the six real roles
The node editor's palette includes an Agents group, and it's important to understand how the agent nodes you drop onto the canvas map to the actual specialist roles that run behind the scenes.
The six real roles
There are exactly six specialist sub-agent roles a flow (or Camelia) can use: research, reasoning, design, document, patent, and manufacturing. Each has a fixed purpose:
- research — a concise markdown brief (uses live web access).
- reasoning — step-by-step structured analysis with trade-offs (uses live web access).
- design — hardware/product design notes covering dimensions, materials, components, and cost.
- document — a clean markdown document.
- patent — structured claims and a summary, flagging what needs human or legal review.
- manufacturing — a practical, AI-suggested sourcing/BOM markdown plan.
How agent nodes compile
When you configure an Agent node, the role dropdown offers exactly these six roles. At compile time, each Agent node becomes a role sub-agent step that runs that role — the same runRole execution Camelia uses when she delegates.
About cosmetic labels
The palette also shows some friendly-looking labels that aren't real, separate agents — for example tiles named "PCB Engineer" or "Validation." These are cosmetic: they map onto the real roles under the hood (for instance onto design or reasoning) via an internal role map. There is no distinct "PCB Engineer," "Validation," or similarly named agent in the system — only the six roles above.
Practical takeaway
When you build a flow, think in terms of the six real roles. Pick the role whose output matches what you need at that step, wire its dependencies so it receives the right upstream context, and trust that whatever label the palette shows, the work is done by one of these six specialists.