Improve Codex Live architecture benchmark

Compare production-equivalent direct and app-server paths, include annotated UI work, expose first-usable latency, and keep semantic quality gates honest.\n\nAI-assisted: OpenAI Codex.
This commit is contained in:
Paul Bakaus
2026-07-12 20:46:53 -07:00
parent 395953ae1d
commit 27bbf90ede
5 changed files with 152 additions and 20 deletions
+2
View File
@@ -89,6 +89,8 @@ export function summarizeArchitectureRuns(runs) {
return {
runs: runs.length,
passed: completed.filter((run) => run.passed).length,
medianFirstUsableMs: percentile(completed.map((run) => run.firstUsableMs).filter(Number.isFinite), 0.5),
p95FirstUsableMs: percentile(completed.map((run) => run.firstUsableMs).filter(Number.isFinite), 0.95),
medianTotalMs: percentile(completed.map((run) => run.totalMs), 0.5),
p95TotalMs: percentile(completed.map((run) => run.totalMs), 0.95),
medianStartupMs: percentile(completed.map((run) => run.startupMs).filter(Number.isFinite), 0.5),