Core Concepts

To truly master pConAI, it's important to understand the fundamental concepts that power our agentic space-planning engine. Unlike traditional CAD software where you manually place every vertex, pConAI operates on higher-level architectural logic.

1. The Scene Graph

Every object, wall, and light source you request is stored in a structured Scene Graph. This means pConAI doesn't just "draw" a chair; it understands that the chair belongs to a specific room, sits on the floor, and relates to the desk next to it.

• When you ask to "Move all chairs to the left", the agent queries the Scene Graph to identify all objects classified as chairs and translates their coordinates.

2. Intent Routing

When you type a prompt like "Build a glass partition and make the frame black", the system's Intent Router breaks this down into multiple sub-tasks:

1. Architecture Tool: Create a partition wall entity.
2. Material Tool: Target the frame property of the new partition and apply a black anodized material.

3. Parametric Entities

Most items inserted via the catalog are parametric. This means they are not static 3D meshes. A desk might have dynamic properties for width, height, and leg_style.

• You can say: "Change the desk width to 180cm". The agent will automatically adjust the parametric boundaries without stretching or distorting the geometry.

4. Agentic Verification

pConAI uses an internal feedback loop. If you ask it to place a 3-meter sofa in a 2-meter alcove, the agent will detect the collision mathematically before rendering it. It will either adjust the position or warn you that the item does not fit.

Understanding these concepts will help you write better prompts and unlock the true potential of the system.