Skip to content

fix(ci): scope the /benchmark suite pre-check to BENCHMARKS_DISPATCH_TOKEN #6588

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
pranavraja99 merged 2 commits into from
Jul 23, 2026
Merged

fix(ci): scope the /benchmark suite pre-check to BENCHMARKS_DISPATCH_TOKEN #6588

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
fe7100c
fix(ci): use BENCHMARKS_DISPATCH_TOKEN for the pre-dispatch suite-exi…
pranavraja99 Jul 23, 2026
1d334bd
chore: retrigger checks with skip-regression-check label applied
pranavraja99 Jul 23, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
24 changes: 18 additions & 6 deletions .github/workflows/nearai-bench.yml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -113,10 +113,21 @@ jobs:
PR: ${{ github.event.issue.number }}
REPO: ${{ github.repository }}
GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
# nearai/benchmarks is private, so reading its suites/ contents
# below (an ironclaw-repo-scoped GITHUB_TOKEN can't see another
# private repo) needs the same PAT the `bench` job uses to
# dispatch — it must also carry Contents: Read-only on
# nearai/benchmarks (in addition to Actions: Read and write) for
# this specific lookup. Scoped to just those two `gh api` calls
# below via a per-command GH_TOKEN override — everything else in
# this step (gh pr view/comment on THIS repo) still uses the
# ambient GITHUB_TOKEN above, since this PAT has no access to
# nearai/ironclaw at all.
BENCH_GH_TOKEN: ${{ secrets.BENCHMARKS_DISPATCH_TOKEN }}
# Ref on nearai/benchmarks used to validate the requested suite
# exists. Must match the ref the `bench` job actually runs against
# (its `uses:` below) so the pre-dispatch check gives the same
# answer as the dispatched workflow. That `uses:` is `@main`, so a
# exists. Must match the ref the `bench` job actually dispatches
# against so the pre-dispatch check gives the same answer as the
# dispatched workflow. That dispatch targets `--ref main`, so a
# pinned SHA here drifts stale and rejects suites that already exist
# on main (e.g. pinchbench26, officeqa, terminal-bench-2). Track
# `main` so the two stay in lock-step.
Expand Down Expand Up @@ -157,9 +168,10 @@ jobs:
# Pre-dispatch suite check: refuse unknown suite names before
# we acknowledge with 🚀 or spend any LLM credit. Cheap GitHub
# API lookup against the pinned benchmarks SHA so the user gets
# the same answer the dispatched workflow would.
if ! gh api "repos/nearai/benchmarks/contents/suites/${suite}.toml?ref=${BENCH_PIN}" --silent >/dev/null 2>&1; then
available=$(gh api "repos/nearai/benchmarks/contents/suites?ref=${BENCH_PIN}" \
# the same answer the dispatched workflow would. Uses
# BENCH_GH_TOKEN (not the step's ambient GH_TOKEN) — see env block.
if ! GH_TOKEN="$BENCH_GH_TOKEN" gh api "repos/nearai/benchmarks/contents/suites/${suite}.toml?ref=${BENCH_PIN}" --silent >/dev/null 2>&1; then
available=$(GH_TOKEN="$BENCH_GH_TOKEN" gh api "repos/nearai/benchmarks/contents/suites?ref=${BENCH_PIN}" \
--jq '[.[] | select(.name | endswith(".toml")) | .name | rtrimstr(".toml")] | sort | join(", ")')
gh pr comment "$PR" --repo "$REPO" --body "Unknown suite \`${suite}\`. Available: ${available}"
exit 0
Expand Down
Loading