Add coverage reporting and publish the real numbers

.github/workflows/ci.yml

@@ -51,7 +51,10 @@ jobs:
         # Tests that require a live local Ollama daemon skip themselves
         # automatically (pytest.mark.skipif) — no CI-specific filtering
         # needed.
-        run: uv run pytest tests -q
+        # --cov-fail-under guards the 92% the README advertises. The gate is
+        # 91, not 92: the headline figure is coverage.py rounding ~91.9%
+        # (1545/1681), and it drifts to ~91.7% between runs.
+        run: uv run pytest tests -q --cov --cov-report=term --cov-fail-under=91
 
   frontend:
     runs-on: ubuntu-latest
@@ -78,7 +81,7 @@ jobs:
 
       - name: Unit test
         working-directory: apps/web
-        run: npm run test
+        run: npm run test:coverage
 
       - name: Install Playwright browsers
         working-directory: apps/web

.gitignore

@@ -11,6 +11,7 @@ apps/api/.pytest_cache/
 apps/api/.mypy_cache/
 apps/api/.ruff_cache/
 apps/api/*.egg-info/
+apps/api/.coverage
 
 # Node
 apps/web/node_modules/

Makefile

@@ -1,5 +1,5 @@
 .PHONY: setup up down migrate ingest ingest-fixtures db-reset \
-        test test-backend test-frontend test-e2e lint format typecheck \
+        test test-backend test-frontend test-e2e coverage lint format typecheck \
         eval-seed eval-run eval demo
 
 setup:
@@ -44,6 +44,10 @@ test-frontend:
 test-e2e:
 	cd apps/web && npm run test:e2e
 
+coverage:
+	cd apps/api && uv run pytest tests -q --cov --cov-report=term-missing
+	cd apps/web && npm run test:coverage
+
 demo:
 	./scripts/demo.sh
 

README.md

@@ -6,6 +6,7 @@
   ### A local-first AI business analyst that investigates *why* a metric changed — not just that it did
 
   [![CI](https://github.com/Bruce1508/RootLens/actions/workflows/ci.yml/badge.svg)](https://github.com/Bruce1508/RootLens/actions/workflows/ci.yml)
+  [![Backend coverage](https://img.shields.io/badge/backend%20coverage-92%25-2ea44f?style=flat-square)](#status)
   [![License: MIT](https://img.shields.io/badge/license-MIT-black?style=flat-square)](LICENSE)
   [![Status](https://img.shields.io/badge/milestones-0--6%20complete-2ea44f?style=flat-square)](#status)
   [![Backend](https://img.shields.io/badge/backend-FastAPI%20%2F%20Python%203.13-3776AB?style=flat-square&logo=python&logoColor=white)](apps/api)
@@ -137,8 +138,21 @@ engine makes — including, since Milestone 6, the engine's own reads.
 The dashboard's investigate/history flow (PRD acceptance criterion #4)
 is wired end-to-end, not just reachable via the API.
 
-What this *doesn't* mean: RootLens is not under active maintenance,
-does not have CI/CD or a hosted deployment, and its [known
+**Tests** — 141 backend tests at **92% coverage** over `app/`, 22
+frontend unit tests, and 4 Playwright end-to-end tests. Reproduce both
+numbers with `make coverage`.
+
+Frontend *unit* coverage is 12.4%, and that figure is a deliberate
+consequence of where the tests live rather than a gap: `lib/` and the
+presentational components are unit-tested in jsdom, while the four page
+components are covered end-to-end by Playwright, which the unit
+coverage run cannot see. Vitest is configured to measure the whole
+`app/`, `components/`, and `lib/` tree — not only the files a test
+happens to import — so that number stays honest as the app grows.
+
+What this *doesn't* mean: RootLens is not under active maintenance, and
+while CI runs on every push, there is no continuous deployment and no
+hosted demo — evaluating it means running it locally. Its [known
 limitations](#known-limitations) are recorded deliberately rather than
 smoothed over — read them before assuming a given behavior is a bug.
 
@@ -241,6 +255,7 @@ observed here.
 | `make eval-seed` / `make eval-run` / `make eval` | Seed and run the incident benchmark |
 | `make test` | Run backend (pytest) and frontend (Vitest) tests |
 | `make test-e2e` | Run frontend end-to-end tests (Playwright, mocked API) |
+| `make coverage` | Print backend and frontend coverage reports |
 | `make lint` / `make format` / `make typecheck` | Quality gates for both apps |
 
 <br>

apps/api/pyproject.toml

@@ -21,13 +21,17 @@ dev = [
     "mypy>=2.3.1",
     "pytest>=9.1.1",
     "pytest-asyncio>=1.4.0",
+    "pytest-cov>=7.0",
     "ruff>=0.16.5",
 ]
 
 [tool.pytest.ini_options]
 testpaths = ["tests"]
 asyncio_mode = "auto"
 
+[tool.coverage.run]
+source = ["app"]
+
 [tool.ruff]
 line-length = 100
 target-version = "py313"

apps/web/package.json

@@ -8,6 +8,7 @@
     "start": "next start",
     "lint": "eslint",
     "test": "vitest run",
+    "test:coverage": "vitest run --coverage",
     "test:e2e": "playwright test",
     "format": "prettier --write .",
     "typecheck": "tsc --noEmit"
@@ -27,6 +28,7 @@
     "@types/react": "^19",
     "@types/react-dom": "^19",
     "@vitejs/plugin-react": "^6.1.1",
+    "@vitest/coverage-v8": "^4.1.11",
     "eslint": "^9",
     "eslint-config-next": "15.5.25",
     "jsdom": "^29.1.1",

apps/web/vitest.config.ts

@@ -16,5 +16,15 @@ export default defineConfig({
     // e2e/ holds @playwright/test specs (a different runner, different
     // `test()` signature) — excluded so vitest doesn't try to collect them.
     exclude: ["**/node_modules/**", "**/e2e/**"],
+    coverage: {
+      // Without an explicit include, v8 only reports on files a test
+      // imported — which reads as high coverage of a very small app.
+      // Name the real source tree so untested files count against us.
+      include: ["app/**/*.{ts,tsx}", "components/**/*.tsx", "lib/**/*.ts"],
+      exclude: ["**/__tests__/**"],
+      // Text only: the default reporters also write a coverage/ HTML
+      // bundle whose vendored JS then trips `npm run lint`.
+      reporter: ["text"],
+    },
   },
 });