10-Layer Architecture
Ten layers of a production LLM platform in five tiers, each with the failure it absorbs and its share of a per-query budget derived from visible inputs; which layers ship first; and one gateway decision written with the number that reopens it.
Last updated
After this section you can
- Name the layer a failure belongs to, and the control that absorbs it
- Price a request per query and find the line that moves the bill
- Write a gateway decision with the number that reopens it
10-Layer Architecture
Ten layers in five tiers: the failure each one absorbs, its share of a per-query budget, and the order to ship them in.
A platform is ten layers, and each one exists because of a failure you have not had yet. At staff level you name the layer you are in, put a number on it, and say what you defer and the number that brings it back.
Read left to right, then down. Teal is a layer, blue the request, amber the model and gateway decisions. The dashed rail watches every layer. Pans sideways on a phone.