Run the project's checks in GitHub Actions

.env.1password

@@ -0,0 +1,13 @@
+# 1Password-backed development environment. Contains REFERENCES, not secrets,
+# so this file is committed. Run the app through `op run` to expand them:
+#
+#   npm run dev:op       # = op run --env-file=.env.1password -- next dev
+#
+# Requires 1Password CLI 2.x, unlocked (desktop-app integration or `op signin`).
+# Values here are injected into the shell environment, and Next.js does NOT let
+# .env.local override an already-set shell variable (see @next/env processEnv),
+# so this reference wins over any stale key sitting in .env.local.
+TYPESAFE_API_KEY="op://Development/Jev API Key/password"
+
+# Optional: override the Jev model id (defaults to jev-latest).
+# JEV_MODEL=jev-latest

.github/workflows/ci.yml

@@ -0,0 +1,83 @@
+# Runs the checks listed in CONTRIBUTING.md on every pull request and on main.
+#
+# Node: the manifest pins neither Node nor a package manager, so CI uses the
+# version this project is developed against. Next 16 requires >= 20.9.
+name: CI
+
+on:
+  push:
+    branches: [main]
+  pull_request:
+
+permissions:
+  contents: read
+
+# A new push to the same branch supersedes an in-flight run.
+concurrency:
+  group: ${{ github.workflow }}-${{ github.ref }}
+  cancel-in-progress: true
+
+env:
+  NODE_VERSION: "24"
+
+jobs:
+  checks:
+    name: unit, types, lint, build
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v7
+
+      - uses: actions/setup-node@v7
+        with:
+          node-version: ${{ env.NODE_VERSION }}
+          cache: npm
+
+      - run: npm ci
+
+      - name: Unit tests
+        run: npm test
+
+      - name: Types
+        run: npm run typecheck
+
+      - name: Lint
+        run: npm run lint
+
+      - name: Production build
+        run: npm run build
+
+  e2e:
+    name: end-to-end (Playwright)
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v7
+
+      - uses: actions/setup-node@v7
+        with:
+          node-version: ${{ env.NODE_VERSION }}
+          cache: npm
+
+      - run: npm ci
+
+      # Keyed on the lockfile, so the cache is rebuilt when Playwright's version moves.
+      - name: Cache Playwright browsers
+        uses: actions/cache@v6
+        with:
+          path: ~/.cache/ms-playwright
+          key: playwright-${{ runner.os }}-${{ hashFiles('package-lock.json') }}
+
+      - name: Install Chromium
+        run: npx playwright install --with-deps chromium
+
+      # playwright.config.ts builds the app and serves it with the key unset, so
+      # the suite runs in demo mode and needs no credentials.
+      - name: End-to-end tests
+        run: npm run e2e
+
+      - name: Upload report
+        if: failure()
+        uses: actions/upload-artifact@v7
+        with:
+          name: playwright-report
+          path: playwright-report/
+          retention-days: 7

.gitignore

@@ -70,6 +70,7 @@ web_modules/
 .env.*
 !.env.example
 !.env.local.example
+!.env.1password
 
 # parcel-bundler cache (https://parceljs.org/)
 .cache

CONTRIBUTING.md

@@ -20,6 +20,8 @@ npm run build
 npm run e2e   # Playwright, builds and serves the app in demo mode
 ```
 
+GitHub Actions runs all of these on every pull request and on `main` (`.github/workflows/ci.yml`): one job for the unit tests, types, lint and build, and one for the end-to-end suite. The end-to-end job needs no credentials, since the suite serves the app with the key unset and routes through the simulator. A failing run uploads its Playwright report as an artifact.
+
 Automated tests should not need a real API key. For UI changes, exercise model/tool modes, option validation, threshold changes, fallback visibility, key removal, storage/export behavior, keyboard controls, themes, and narrow screens.
 
 Open a pull request explaining the intended behavior, why it preserves the execution boundary, which tests ran, and any unverified behavior. Include a before/after example or screenshot when useful. A proposed GitHub topic change in `repository-metadata.json` still needs a separate authorized repository-settings update.

README.md

@@ -38,6 +38,31 @@ cp .env.local.example .env.local
 # Edit .env.local and set TYPESAFE_API_KEY, then restart the dev server.
 ```
 
+### 1Password-backed key (recommended)
+
+Keep the key in 1Password and run the app through `op run`, so no secret ever
+lands in a file:
+
+```bash
+npm run dev:op
+```
+
+`.env.1password` is committed and holds a **reference**, not a secret:
+`TYPESAFE_API_KEY="op://Development/Jev API Key/password"`. `op run` expands it
+into the process environment at launch. Requires 1Password CLI 2.x, unlocked
+(desktop-app integration or `op signin`).
+
+Verify it took effect — `live` reports whether the SERVER has a key:
+
+```bash
+curl -s http://localhost:3000/api/route
+# {"live":true}   ({"live":false} means demo mode)
+```
+
+A shell-provided variable is not overridden by `.env.local` (`@next/env`'s
+`processEnv` only fills keys absent from the initial environment), so this path
+wins over any stale key left in `.env.local`.
+
 The browser calls this app's `/api/route` endpoint. The server forwards live evaluations to `POST https://api.typesafe.ai/v1/systemone` using `jev-latest`. Jev answers typed questions; it does not generate the code that executes a route.
 
 ### Bring your own key and privacy

package.json

@@ -2,15 +2,17 @@
   "name": "typesafe-router",
   "version": "0.1.0",
   "private": true,
-  "description": "Jev Tool & Model Router — Jev decides, your code executes.",
+  "description": "Jev Tool & Model Router \u2014 Jev decides, your code executes.",
   "scripts": {
     "dev": "next dev",
     "build": "next build",
     "start": "next start",
     "lint": "eslint",
     "test": "vitest run",
     "e2e": "playwright test",
-    "typecheck": "next typegen && tsc --noEmit"
+    "typecheck": "next typegen && tsc --noEmit",
+    "dev:op": "op run --env-file=.env.1password -- next dev",
+    "start:op": "op run --env-file=.env.1password -- next start"
   },
   "dependencies": {
     "next": "16.3.5",