Rotorsoft / act-root / 25640979309 / 1

100%

master: 100%

push

github

fix: unify workspace bench config + repair CI bench summary + npm keywords

Bench config unification (monorepo-standard, single root config):

- Add root `vitest.bench.config.ts` extending root `vite.config.ts`,
  with `test.include` for scenarios and `benchmark.include` for
  microbenches. One process, no `pnpm -r` fan-out, no aggregated-output
  prefix mess.
- Delete per-lib `libs/{act,act-pg}/vitest.bench.config.ts` and the
  per-lib `bench:micro`/`bench:scenarios` scripts.
- Rename bench files so shape is encoded in the filename:
  - `*.micro.bench.ts` — Shape A (vitest `bench()` blocks)
  - `*.scenario.bench.ts` — Shape C (vitest `it()` + assertions)
  - `libs/act-pg/scripts/*.bench.ts` → `*.ts` (Shape B is plain tsx
    scripts, not vitest).
- Add `reporters: ["verbose"]` to the unified config so scenario
  benches' `console.table` blocks actually reach stdout.

CI step-summary parser (`.github/workflows/ci-cd.yml`):

- Drop the awk prefix-strip acrobatics (no longer needed without
  `pnpm -r --aggregate-output`).
- Key on structural `=== .+ ===` headers instead of hardcoded `ACT-`
  ticket prefix.

Bench correctness:

- Repair anonymous handlers in `libs/act/bench/batch-projection.scenario.bench.ts`
  and `libs/act-pg/bench/drain-skip.micro.bench.ts` that violated the
  named-function constraint added in 7b6406aa5 (these benches were
  dead code in the old per-lib include lists).

npm metadata:

- Per-lib `keywords` arrays — descriptive, accurate to each lib's
  actual purpose (no event-sourcing tag on patch/pino/sse).

Recovery side-effects (semantic-release-monorepo paths-filter):

- act-patch republish should restore the missing `dist/` in the
  npm tarball (current 1.2.0 ships only LICENSE/README/package.json).
- act-sqlite, act-pino, act-sse, act-diagram get the patch releases
  that ACT-102's CI failure (Coveralls 526) prevented.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

1511 of 1511 branches covered (100.0%)

Branch coverage included in aggregate %.

2666 of 2666 relevant lines covered (100.0%)

372.03 hits per line