AI API Latency Metrics: TTFT, First Response, and Output Speed

A request-timeline guide to upstream headers, first SSE event, first effective response, first visible text, end-to-end latency, and output speed. AI API latency is not one number. For a streaming req