Brooklyn Nicholson
20fcef5207
The existing `stream` scenario measures streaming into an empty thread, which is exactly the case where per-delta transcript work does not show up. This adds `--historyTurns`, which mounts a settled transcript and lets it drain BEFORE the recorders start, so the measurement window contains streaming work only — and `stream-history`, a report-only preset that turns it on. Report-only (tier: manual) rather than gated: how much history a host can mount varies, so the absolute number is not comparable across machines. It is meant for same-machine before/after runs. Co-authored-by: Jakub Wolniewicz <frizikk@users.noreply.github.com>
19 lines
721 B
JavaScript
19 lines
721 B
JavaScript
// Streaming into an ALREADY-LONG transcript. Same measurement as `stream`, but
|
|
// the history is mounted and allowed to settle before the recorders start, so
|
|
// what it captures is the per-delta cost that scales with transcript length —
|
|
// the regression reported in #69120.
|
|
//
|
|
// Report-only (tier: manual): the number depends on how much history the host
|
|
// can mount, so it is not gated against the committed baseline.
|
|
|
|
import stream from './stream.mjs'
|
|
|
|
export default {
|
|
name: 'stream-history',
|
|
tier: 'manual',
|
|
description: 'Streaming cost with a long settled transcript already mounted.',
|
|
run(cdp, opts = {}) {
|
|
return stream.run(cdp, { ...opts, historyTurns: Number(opts.historyTurns ?? 200) })
|
|
}
|
|
}
|