Alex has just uploaded the intermediate formats on the datajs CodePlex Wiki. These describe the shape of results provided by reading OData as well as the expected format to send data back to the server.

Like I mentioned before, these values do not have any classes or prototypes associated with them - you can simply recognize them by their shape, access them through fields, and create them with literals or in whichever way you prefer.

Another thing that we did when designing these was to keep them as close as possible to the on-the-wire representation for OData JSON payloads. This way, if you are using a tool like Fiddler to look at the data going back and forth, you will have a very good idea about how datajs is going to represent this in memory.