Files
orca/.github/workflows/unit-module-cache.yml
T
NeilandCopilot 3be8a205dd perf(ci): drop unbudgeted cache automation and duplicate draft planning
Keep producer and benchmark storage behavior unchanged from main. Do not add a daily 1.42 GiB cache generation to a near-quota cache store without retention and eviction evidence. Reuse existing draft selection artifacts instead of rebuilding the graph twice.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
2026-11-02 23:20:13 -08:00

31 lines
858 B
YAML

name: Unit transform cache producer
# Cold full unit run on main that publishes per-shard Vitest transform caches.
# Dispatch only: PR shards run on hosted runners, which cannot read this Blacksmith-store cache.
on:
workflow_dispatch:
concurrency:
group: unit-module-cache-${{ github.ref }}
cancel-in-progress: true
permissions:
contents: read
jobs:
unit_plan:
if: github.ref == 'refs/heads/main'
uses: ./.github/workflows/unit-plan.yml
test:
needs: unit_plan
uses: ./.github/workflows/unit-tests.yml
with:
# Match the PR unit matrix so the cache key's runtime and shard fields line up.
node_versions: '["24"]'
runner: blacksmith-8vcpu-ubuntu-2404
shards: ${{ needs.unit_plan.outputs.shards }}
module_cache: cold
artifact_prefix: cache-producer-unit
run_relay_integration: false