Hey guys. Currently trying to alter the voice pipeline a bit. I want a RAG Agent, but I don't want the RAG to happen when the agent decides it needs to call a tool, but I want it to happen before each conversation turn. So this is the pipeline STT -> RAG -> LLM -> TTS The goal here is to give the FULL context to the LLM so it can generate the correct answer right away. Does anyone here have an idea how to get this working with VAPI? Appreciate any insights