llm-primer keeps a pool of warm sessions in the background. When you need one, it's already past the startup protocol.
Happy to answer questions. The main motivation was that I'm running LLM wikis (llm-context-base) and switching context between different agent roles constantly. The startup latency was killing the flow.
