spencerheywood.com

get_live_lab_stats

Fetch live stats for Spencer's self-hosted AI lab straight from his LLM gateway: decode throughput in tokens/sec (mean aggregate across concurrent streams, peak single-stream, 24-hour average), lifetime token and request totals, build cost vs hosted-API spend avoided, wall power draw, energy used, GPU utilization, the local-vs-api-overflow routing split, input-context lengths (median/p95 per request), and the top serving models. Use this whenever asked about the lab's performance or any of those numbers — never quote them from memory.

Input schema

No input properties (empty object).

{
  "type": "object",
  "properties": {},
  "additionalProperties": false
}