Once the sampling ray is defined it is also necessary to specify what the ray should see in the scene, either polygons, volumetrics, and so on. If nothing is connected to these inputs, the current sample position and/or ray direction is used. The inputs are connected to the Origin and Direction, and represent the ray that is fired for sampling, and are evaluated per sample. These can be a specific defined vector value or be representative of shading values within the scene, such as a position and normal sample from a surface using a shader input connection, see Shader Inputs.
To create a ray, you need to have both a position and a direction. For example, the Raycast node makes it possible to make a virtual camera for a scene, what it sees, the results of the ray evaluations, would be displayed on another surface elsewhere in the scene, determined by the output connections. The Raycast node adds the ability to fire rays based on user defined criteria, gather information from where the ray hits or doesn't hit, and then use this information to drive other channel values.
