You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
✨ benchmarks export: per-PR estimated cost from launch attribution #17790
The spoke already attributes spend to PRs: /api/launch-cost (persisted at /data/launch-cost.json, daily rollup per ref) prices each kick's tokens with pkg/tokens/pricing.go and splits it across the refs the kick named — the swarm's owner/repo#N PR ref, or the governor's repo#N issue ref.
Wire it into the export:
pkg/evidence/benchmarks: Cost{basis, usd_est, *_tokens, launches, price_table_date} on Record; Options.CostLookup hook; author records look up the PR and the issues it closed (issue_closures, both ref spellings), reviewer records the PR ref.
pkg/dashboard: implement the lookup over LaunchCostCollector.DailyRollup() (USD) + Snapshot().ByLaunch (token share). basis: tokens_x_pricing when usage landed, else unknown. Never emit fleet_model_median (the site imputes it).
Refresh the vendored testdata/result.schema.json (benchmarks schema now has cost.basis enum reported|tokens_x_pricing|fleet_model_median|unknown), golden/schema tests, docs (src/docs/benchmarks.md cost section), changelog fragment.
Privacy: only the per-PR estimate and its token share leave the spoke; no per-session usage, no billed figures.
Activity
✨ benchmarks: per-PR estimated cost in /api/benchmarks/export (hiveco…
🌱 Top up v6 with v5 latest (v5.237.2) (#18030)
✨ benchmarks: per-PR estimated cost in /api/benchmarks/export (#17791)