Small cache-regression check for the pnpm + Playwright CI lane

[eivindsjursen-lab]

Hi — your main CI lane already has two separate cache surfaces: pnpm and Playwright binaries.

I tried a fork-side dry run of that workflow shape with a small warn-only cache check. In a controlled bad-case, the pnpm cache lane degraded while the Playwright cache stayed healthy, which is exactly the kind of partial cache drift that can be easy to miss while the workflow still looks green.

If useful, I can paste the minimal warn-only snippet I used for a pnpm + Playwright lane. No backend, easy rollback.

[aniruddhaadak80]

Hi @eivindsjursen-lab!

Vite is usually blazingly fast, but issues like this often crop up around dependency pre-bundling or rollup configurations.

1. Pre-bundling: Try deleting your node_modules/.vite cache and running npm run dev -- --force to make Vite rebuild the dependencies.
2. Externalizing Modules: If you are seeing issues where built-in node modules (like "stream" or "path") are breaking, it is because Vite is primarily a frontend build tool. You might need to add polyfills via vite-plugin-node-polyfills or explicitly externalize it in build.rollupOptions.external.

Let me know if clearing the Vite cache or tweaking the Rollup config gets you past this hurdle.

Hope this points you in the right direction! Let me know how it goes. Happy coding!

[eivindsjursen-lab]

Thanks - to clarify: this is about CI cache behavior, not Vite runtime/build config.

I’m referring to the GitHub Actions cache shape in your CI: the pnpm cache lane plus the separate Playwright browser cache.

I ran a fork-side dry run of that workflow shape with a small warn-only cache check. In a controlled bad-case, only the pnpm cache degraded while the Playwright cache stayed warm. The gate surfaced:

• WARN_HIT_RATE_DROP
• WARN_RESTORE_REGRESSION

Minimal workflow used:
https://github.com/eivindsjursen-lab/vite-cache-alpha/blob/main/.github/workflows/cache-alpha.yml

Gate step:

- name: Cache Health Gate
  id: gate
  uses: eivindsjursen-lab/gates-suite-public-beta/packages/cache-health-gate@cache-health-gate/v1
  with:
    mode: warn
    no_baseline_behavior: warn
    baseline_runs: "8"
    baseline_event_filter: push
    debug: "true"

So the point is not “there’s a Vite cache bug,” but that this CI shape is a good fit for catching partial cache drift before the workflow obviously slows down.

If useful, I can paste a tighter version mapped directly to your current ci.yml so it only targets the existing pnpm/Playwright cache surfaces?