Skip to content
Redgold.AI Blog

Notes from production

Field notes from running coding agents, an LLM gateway and a data engine in production.

  1. · 9 min read

    How We Keep the Prompt Cache Warm While Rewriting Every Request

    Our gateway adds context to every agent request so the agent knows who it is working for, which repositories and apps exist, and what the project’s rules are. Here is why that context matters, and how we add it without paying full price for the prompt each time.