I'm already sort of doing this by using an LLM router which caches identical prompts and their outputs. But a way to reduce compute and cost overhead for this project would be to utilize cached results. Could even aggregate them on a server if needed, but local caching should be fine. This probably only applies directly to PKGBUILDs that don't change like devel versions based on git branch.
I'm already sort of doing this by using an LLM router which caches identical prompts and their outputs. But a way to reduce compute and cost overhead for this project would be to utilize cached results. Could even aggregate them on a server if needed, but local caching should be fine. This probably only applies directly to PKGBUILDs that don't change like devel versions based on git branch.