Skip to content

Beautiful, AI-native markdown IDE and LLM wiki

docs(gates): align repository gate guidance; scope local push and round checks to what exists (#4560)

Nick Gomez2bc8e2f216b9
diff --git a/AGENTS.md b/AGENTS.md
index 05117ffe9..8d7a2451a 100644
--- a/AGENTS.md
+++ b/AGENTS.md
@@ -26,6 +26,8 @@ pnpm run typecheck
pnpm run test
```
+For routine local Turbo runs, use an absolute `TURBO_CACHE_DIR` inside this worktree; use `TURBO_FORCE=true` only for a deliberate fresh run.
+
Run local apps:
```bash
@@ -92,7 +94,7 @@ Cadence: merging a PR with a changeset triggers a beta publish within minutes vi
## Before Finishing
-Run the smallest relevant check while iterating, then run:
+While iterating, run the tests for what you changed with `pnpm --dir packages/<pkg> run test`, plus `pnpm run lint` and `pnpm run typecheck`, widening to affected consumers for shared inputs. A red remains a blocker: diagnose it against the changed inputs, repair only an in-scope cause, and record an unrelated environment blocker with its log and owner. Then run:
```bash
pnpm run check
diff --git a/package.json b/package.json
index cb9eb16af..4fa99fde1 100644
--- a/package.json
+++ b/package.json
@@ -30,7 +30,7 @@
"lint:oxlint": "oxlint --max-warnings 0 .",
"format": "pnpm run lint:biome --write --unsafe",
"check": "pnpm run lint && pnpm exec turbo run build typecheck test",
- "check:full:parallel": "pnpm run check:drift:guards && pnpm run check:server-test-inputs && pnpm run lint && pnpm run check:doc-links && vitest run --config vitest.scripts.config.ts && turbo run build typecheck check:dts test test:dom test:integration test:conversion test:perf:regression:unit test:health:unit test:health test:e2e test:visual test:perf test:a11y --concurrency=100% --output-logs=errors-only",
+ "check:full:parallel": "pnpm run check:drift:guards && pnpm run check:server-test-inputs && pnpm run lint && pnpm run check:doc-links && vitest run --config vitest.scripts.config.ts && turbo run build typecheck check:dts test test:dom test:integration test:conversion test:perf:regression:unit test:health:unit test:health test:e2e test:perf test:a11y --concurrency=100% --output-logs=errors-only",
"tier1": "pnpm run check",
"tier2": "pnpm run tier1 && turbo run test:perf:regression:unit test:health:unit test:health test:e2e",
"tier3": "STRESS_FIDELITY=1 pnpm run tier2",
diff --git a/packages/app/playwright.visual.config.ts b/packages/app/playwright.visual.config.ts
index d3645b03d..37a3fe9cd 100644
--- a/packages/app/playwright.visual.config.ts
+++ b/packages/app/playwright.visual.config.ts
@@ -27,9 +27,8 @@ import { EXPECT_TIMEOUT_MS, resolveWorkerCount } from './playwright.config.ts';
* Off-CI worker count is derived from the same per-worker density as the main
* suite rather than left to Playwright's default, because this config consumes
* the identical per-worker fixture: each worker owns a full Vite + Hocuspocus +
- * Chromium stack, not just a browser. `check:full:parallel` runs this tier
- * concurrently with the e2e tier at `--concurrency=100%`, so an unbounded
- * default here would re-import the oversubscription the main config now avoids.
+ * Chromium stack, not just a browser. When an author invokes this suite on an
+ * otherwise unloaded machine, the cap bounds local oversubscription.
*/
const isCI = !!process.env.CI;