Same agent tasks, 76% fewer LLM calls – we moved semantic cache inside the graph(github.com)
2 pointsby insightits2 hours ago3 comments