ai tool calling vs rag

Feeding a model your whole repository is the expensive way to get a worse answer. Build it small, fast, specialised tools instead. Here is what that looks like in practice.