Artifacts Usage Report — github/gh-aw #51535
Closed
Replies: 1 comment
|
This discussion was automatically closed because it expired on 2026-08-10T05:59:47.778Z.
|
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
Summary Table (matched sample, last ~30 days)
Methodology
GET /repos/{owner}/{repo}/actions/workflows.GET /actions/workflows/{id}/runs?created>=<30d ago>.GET /repos/{owner}/{repo}/actions/artifacts(up to page 60, ~6,000 records) and joined each artifact'sworkflow_run.idagainst the run→workflow map.CIandDependabot Updatesexceeded resolvable run counts).Insights & Recommendations
CIdominates by volume (2,371 artifacts / 44 distinct runs matched), consistent with per-job/per-matrix-leg upload patterns (test results, coverage, build outputs). This is the top candidate for consolidation — e.g., combining per-job artifacts into a single upload per run, or reducing retention days for short-lived debug artifacts.Doc Build - Deployuses the most storage despite only 10 artifacts (2.3 GB total, ~236 MB average) — likely full static-site/build bundles. Consider trimming unused build assets or reducing retention (e.g., 1–3 days) since these are typically superseded immediately by the next deploy.gh-aw-generated agentic workflows (Copilot/Codex/Claude smoke tests, daily analysis agents, etc.), each producing a fairly uniform ~1–2 MB artifact per run (logs, prompts, session state). Individually small, but in aggregate across ~30 such workflows this adds up to several hundred MB/month — a good candidate for a shared shorter retention policy (e.g., 3–7 days) rather than the GitHub default (90 days), since these are diagnostic/audit artifacts rarely needed long-term.Smoke Codex/Smoke Copilot/Smoke Claudeaverage artifact sizes (14–19 MB) are notably higher than other smoke/daily agents — worth checking if these bundle unnecessary binaries, node_modules, or verbose logs.gh-aw-generated diagnostic/session artifacts (the majority of rows belowCI), setretention-daysexplicitly (e.g., 7 days) in the workflow lock files rather than relying on the repository default, since none of these appear to be consumed beyond their originating run's lifetime.Dependabot Updates(99 runs) and several other high-volume workflows could not be joined to artifact data in this pass. A follow-up run with a dedicated, rate-limit-friendly script (e.g., usinggh api --paginatewith retry/backoff, or the artifacts REST endpoint filtered byworkflow_idif/when GitHub exposes it) is recommended for a fully authoritative report.All reactions