How would you design a telemetry pipeline to monitor the latency and token usage of an LLM integrated into AHEAD's Agentic Platform?
