Seven Components, in Order

This reading lists the components of a web tool that answers questions with a language model, in the order the work passes through them.

The Seven Components

A person opens the tool's page, types a question and reads the answer on the same page. Between the question and the answer, the work passes through seven components, in this order. The interface is the page in the browser, where a person types a question and reads the answer. The request is the message that the browser sends to the server, and it carries the person's question. The server receives the request, adds the system prompt and sends the work on to the model. The system prompt is the set of instructions that the tool's builder writes ahead of time, and it joins the request just before the model. The model has two parts, its parameters and its runtime: the parameters are a large file of numbers, and the runtime is the program that uses them to predict the answer. The hardware is the memory that holds the parameters and the GPU that does the calculations. The response is the answer, which travels back from the model through the server to the interface.

What a Specification Is

A specification is a written description of each component: what it is, what it receives, what it does and what it sends on. This is the idea behind spec-driven development, where the specification is the part of the work that lasts and the code is built from it. Using the seven names above, in the same order, lets every specification be read the same way.

Sources: MDN Web Docs, Client-server overview (2026); Mistral AI, Prompting (n.d.); Hugging Face, Optimizing LLMs for Speed and Memory (n.d.); ggml-org, llama.cpp (n.d.); MuggsAI request path, class site source code (2026).