User
Ollama/LM Studio don’t support passing a request to RAG from their side. You need to flip the flow manually by inserting a middle layer (a proxy) that handles RAG before the prompt ever hits the LLM.