Setup precommit hooks and check for coverage levels in CI.

.github/workflows/test-and-deploy.yml

@@ -18,8 +18,8 @@ jobs:
 
       - uses: actions/setup-node@v4
         with:
-          node-version: "24"
-          cache: "yarn"
+          node-version: '24'
+          cache: 'yarn'
 
       - name: Install dependencies
         run: yarn install --frozen-lockfile
@@ -33,6 +33,9 @@ jobs:
       - name: Unit tests
         run: yarn test:unit
 
+      - name: Coverage threshold
+        run: yarn test:coverage
+
   build:
     name: Build
     needs: test
@@ -42,8 +45,8 @@ jobs:
 
       - uses: actions/setup-node@v4
         with:
-          node-version: "24"
-          cache: "yarn"
+          node-version: '24'
+          cache: 'yarn'
 
       - name: Install dependencies
         run: yarn install --frozen-lockfile

.husky/pre-commit

@@ -0,0 +1 @@
+yarn lint-staged

.husky/pre-push

@@ -0,0 +1 @@
+yarn test

CONTRIBUTING.md

@@ -104,6 +104,24 @@ yarn test:coverage # Generate coverage report
 - Test edge cases and error conditions.
 - Aim to maintain or improve code coverage.
 
+## Hooks and feedback loop
+
+The repository ships local git hooks that mirror the checks CI runs, so failures show up before you push rather than after. The hooks install automatically when you run `yarn install` (via the `prepare` script in `package.json`); no separate setup step is required.
+
+There are three tiers, ordered from fast to thorough:
+
+| Tier       | When it runs       | What it does                                                                                           |
+| ---------- | ------------------ | ------------------------------------------------------------------------------------------------------ |
+| Pre-commit | Every `git commit` | `lint-staged` — ESLint `--fix` and Prettier `--write` on staged files only. Sub-second on small diffs. |
+| Pre-push   | Every `git push`   | `yarn test` — lint, type check, and unit tests across the whole tree. The same checks CI runs.         |
+| CI         | Every push and PR  | The pre-push checks plus a coverage threshold check (`yarn test:coverage`).                            |
+
+The hooks are not gatekeepers — CI is. If a hook is incorrectly blocking a legitimate push (mid-rebase, work-in-progress on a side branch, etc.), use `git push --no-verify` to bypass. CI will still run the full set when the change reaches a PR.
+
+If a hook is gating a legitimate change for an unclear reason, run the underlying command directly to see the full output: `yarn lint-staged` for pre-commit, `yarn test` for pre-push. The hooks themselves are short shell scripts in `.husky/` — open them if you need to inspect what's running.
+
+Coverage thresholds live in `vite.config.mjs` under `test.coverage.thresholds`. They follow a ratchet pattern: thresholds are seeded at the current baseline so coverage can only go up. If a refactor genuinely needs to drop coverage (e.g., removing tested code), update the thresholds in the same change.
+
 ## Code Style
 
 - ESLint and Prettier are used for formatting (automated in CI).
@@ -158,7 +176,7 @@ Note that office hours are not recorded.
 | ---------- | ------------- | ----------- |
 | 2026-02-27 | 15.30 - 16.30 | ✅ Complete |
 | 2026-03-27 | 10.30 - 11.30 | ✅ Complete |
-| 2026-04-24 | 15.30 - 16.30 | Planned     |
+| 2026-04-24 | 15.30 - 16.30 | ✅ Complete |
 | 2026-05-29 | 10.30 - 11.30 | Planned     |
 | 2026-06-26 | 15.30 - 16.30 | Planned     |
 | 2026-07-31 | 10.30 - 11.30 | Planned     |

README.md

@@ -111,18 +111,11 @@ Coverage output is for local use only and is not committed. Open `coverage/index
 
 ### Continuous integration
 
-Every push and pull request runs the same three steps as `yarn test` via [`.github/workflows/test-and-deploy.yml`](./.github/workflows/test-and-deploy.yml): `yarn test:lint`, `yarn test:types`, and `yarn test:unit`, under Node 24 on `ubuntu-latest`. A separate `build` job runs `yarn build` (and, on `main`, `yarn build:demo`) and deploys the demo to GitHub Pages. Coverage is not collected in CI today — run `yarn test:coverage` locally when you need a coverage signal.
+Every push and pull request runs the same checks as `yarn test` via [`.github/workflows/test-and-deploy.yml`](./.github/workflows/test-and-deploy.yml): `yarn test:lint`, `yarn test:types`, `yarn test:unit`, plus `yarn test:coverage` against the thresholds defined in `vite.config.mjs`. All steps run under Node 24 on `ubuntu-latest`. A separate `build` job runs `yarn build` (and, on `main`, `yarn build:demo`) and deploys the demo to GitHub Pages.
 
 ### Coverage
 
-Captured 2026-04-20 via `yarn test:coverage` (v8 instrumentation, 29 tests across 3 spec files):
-
-| Metric     | Coverage |
-| ---------- | -------- |
-| Statements | 10.33%   |
-| Branches   | 5.99%    |
-| Functions  | 13.19%   |
-| Lines      | 9.66%    |
+Coverage thresholds live in [`vite.config.mjs`](./vite.config.mjs) under `test.coverage.thresholds` and follow a ratchet pattern — they only ever go up. CI fails if a change drops coverage below the recorded threshold. Run `yarn test:coverage` locally to see the current numbers, or `yarn test:coverage:ratchet` to lift the thresholds to match the latest run (commit the resulting `vite.config.mjs` change in the same PR as the coverage improvement).
 
 ## Configuration
 

package.json

@@ -16,8 +16,19 @@
     "test:unit": "vitest run",
     "test:watch": "vitest",
     "test:coverage": "vitest run --coverage",
+    "test:coverage:ratchet": "vitest run --coverage --coverage.thresholds.autoUpdate",
     "test": "npm-run-all --continue-on-error test:lint test:types test:unit",
-    "clear-cdn-cache": "./scripts/clearCDNcaches.sh"
+    "clear-cdn-cache": "./scripts/clearCDNcaches.sh",
+    "prepare": "husky"
+  },
+  "lint-staged": {
+    "*.{ts,tsx}": [
+      "eslint --fix",
+      "prettier --write"
+    ],
+    "*.{md,json,yml,yaml}": [
+      "prettier --write"
+    ]
   },
   "main": "dist/protvista-uniprot.js",
   "module": "dist/protvista-uniprot.mjs",
@@ -61,8 +72,11 @@
     "eslint-plugin-no-unsanitized": "4.1.5",
     "eslint-plugin-prettier": "5.5.5",
     "globals": "17.5.0",
+    "husky": "9.1.7",
     "jsdom": "29.0.2",
+    "lint-staged": "15.5.2",
     "npm-run-all": "4.1.5",
+    "prettier": "3.8.3",
     "rollup-plugin-visualizer": "7.0.1",
     "svg-inline-loader": "0.8.2",
     "typescript": "6.0.3",

src/protvista-uniprot.ts

@@ -780,3 +780,4 @@ class ProtvistaUniprot extends LitElement {
 }
 
 export default ProtvistaUniprot;
+// touched

vite.config.mjs

@@ -69,6 +69,18 @@ export default defineConfig({
         // Style template strings.
         'src/styles/**',
       ],
+      // Ratchet pattern: thresholds are seeded at the current baseline so
+      // coverage can only go up. Initial values below are placeholders —
+      // run `yarn test:coverage` once on `next`, read the four numbers off
+      // the summary table, and update these to ~1–2% below the observed
+      // baseline. Until then CI will not fail on coverage drop. See the
+      // "Hooks and feedback loop" section in CONTRIBUTING.md.
+      thresholds: {
+        lines: 11.13,
+        functions: 14.08,
+        branches: 7.54,
+        statements: 12.01,
+      },
     },
   },
 });