LLM Endpoint Workflow

About this model

The diagram shows a process where a user’s prompt and query are sent into a system that may search external knowledge sources to improve the final response. The search module returns relevant information depending on the size and quality of those sources, as well as a simulated time factor that can range from very fast (–10) to much slower than normal (+10).

A random gate then determines whether enhanced context is produced, and this probability increases when the knowledge sources are of higher quality. If the sources are poor, the search may fail and no enhanced context is added.

When enhanced context is available, it is combined with the original prompt and query before being sent to the Large Language Model endpoint. The LLM requires at least a prompt and query to generate a response, but the presence of enhanced context generally improves the result. By simulating this workflow, it becomes possible to observe how the quality and size of knowledge sources influence the quality, reliability, and richness of the responses returned to the user.

"LLM Endpoint Workflow" is a public Machinations diagram shared in the Community by Catalin Ichim, with 41 components.

It was created in November 2025 and last edited in March 2026.

Open the diagram in the Machinations app and press Play to watch the resources move, adjust its values to test how the system responds, or remix it to make an editable copy in My diagrams.