GloriousFlywheel Template Consumer Mode Decision 2026-04-18

GloriousFlywheel Template Consumer Mode Decision 2026-04-18

Snapshot date: 2026-04-18

Purpose

Resolve the remaining ambiguity around recent template consumers.

The key question is not “do these repos use ci-templates?” The key question is:

  • what mode are they actually using
  • should that mode count toward runner enrollment

Current Evidence

tinyland-inc/tinyvectors

Both current workflow entrypoints on main are explicit:

  • ci.yml
    • runner_mode: hosted
    • workspace_mode: isolated
    • publish_mode: same_runner
  • publish.yml
    • runner_mode: hosted
    • workspace_mode: isolated
    • publish_mode: same_runner

This means tinyvectors is no longer ambiguous. It is an explicit hosted consumer of js-bazel-package.yml.

Hosted npm-publish.yml Family

The reusable workflow tinyland-inc/ci-templates/.github/workflows/npm-publish.yml still hardcodes GitHub-hosted execution:

  • build-and-test
    • runs-on: ubuntu-latest
  • publish-gpr
    • runs-on: ubuntu-latest
  • publish-npm
    • runs-on: ubuntu-latest

Recent repos consuming that hosted workflow include:

  • tinyland-inc/vite-plugin-a11y
  • tinyland-inc/tinyland-rate-limit
  • tinyland-inc/tinyland-physics
  • tinyland-inc/tinyland-color-utils
  • tinyland-inc/tinyland-a11y-engine

Those repos are therefore explicit hosted template consumers too.

Repo-Owned Template Consumers

The current clean repo-owned template consumers on main remain:

  • Jesssullivan/scheduling-kit
  • Jesssullivan/acuity-middleware

Those repos pass:

  • runner_labels_json: ${{ vars.PRIMARY_LINUX_RUNNER_LABELS_JSON }}

through js-bazel-package.yml.

Decision

Use three template-consumer buckets in reporting:

1. Hosted Template Consumers

Count here:

  • tinyland-inc/tinyvectors
  • tinyland-inc/vite-plugin-a11y
  • tinyland-inc/tinyland-rate-limit
  • tinyland-inc/tinyland-physics
  • tinyland-inc/tinyland-color-utils
  • tinyland-inc/tinyland-a11y-engine

These repos should not count toward runner enrollment.

2. Repo-Owned Template Consumers

Count here:

  • Jesssullivan/scheduling-kit
  • Jesssullivan/acuity-middleware

These repos do count toward runner enrollment.

3. Shared Template Consumers

Count here only when a repo explicitly uses a shared ARC mode in workflow code.

Current recent examples on the scanned surface:

  • none proven through template call sites

What Changes Operationally

tinyvectors

This repo no longer needs a “make runner intent explicit” task.

That task is already done.

The real question is now:

  • should tinyvectors remain a hosted template consumer
  • or should it be explicitly promoted to shared or repo_owned

That is a future migration choice, not a current truthing task.

npm-publish.yml Family

This family should be treated as hosted-by-current-policy unless and until the workflow itself changes.

That means:

  • no more overcounting those repos as runner-adjacent enrollments
  • no more treating raw ci-templates consumption as evidence of runner use

Recommendation

Immediately update GloriousFlywheel reporting and queue language so that:

  • tinyvectors is classified as an explicit hosted template consumer
  • the npm-publish.yml family is classified as hosted template consumers
  • only js-bazel-package.yml repos using explicit shared or repo_owned inputs count toward runner adoption

Exit Condition

This note can retire once:

  • template mode reporting is standard everywhere in GloriousFlywheel docs
  • tinyvectors either remains intentionally hosted or is explicitly promoted
  • the npm-publish.yml family is either clearly hosted-by-policy or migrated onto a new self-hosted-capable contract

GloriousFlywheel