Summary
On October 17, 2026, Vercel added session affinity to AI Gateway logs, grouping individual model requests into conversations and surfacing per-session cache hit rate, cost, turn count, and the coding agent behind each thread.
What changed
Vercel AI Gateway log views now group requests into sessions and display cache hit rate, total cost, turn count, and the originating coding agent for each conversation.
Why it matters
Session-level observability makes it practical to attribute spend and latency to specific agent conversations rather than isolated calls, which teams running coding agents and long-horizon workflows need for cost control and debugging.
Evidence excerpt
Logs group requests into sessions and show cache hit rate, cost, turns, and the coding agent behind each conversation.