ci: 添加官网 GitHub Actions 生产部署（Environment production）

.github/workflows/deploy-website.yml

@@ -0,0 +1,85 @@
+name: Deploy website
+
+on:
+  push:
+    branches: [main]
+    paths:
+      - "apps/website/**"
+      - "install.sh"
+      - "install.ps1"
+      - ".github/workflows/deploy-website.yml"
+  workflow_dispatch:
+
+permissions:
+  contents: read
+
+jobs:
+  deploy:
+    name: Deploy
+    runs-on: ubuntu-latest
+    environment: production
+    permissions:
+      contents: read
+    defaults:
+      run:
+        working-directory: apps/website
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v4
+
+      - name: Setup pnpm
+        uses: pnpm/action-setup@v4
+        with:
+          version: 11.21.0
+
+      - name: Setup Node.js
+        uses: actions/setup-node@v4
+        with:
+          node-version: 22
+          cache: pnpm
+          cache-dependency-path: apps/website/pnpm-lock.yaml
+
+      - name: Install dependencies
+        run: pnpm install --frozen-lockfile
+
+      - name: Deploy to Cloudflare
+        env:
+          CLOUDFLARE_API_TOKEN: ${{ secrets.CLOUDFLARE_API_TOKEN }}
+          CLOUDFLARE_ACCOUNT_ID: ${{ vars.CLOUDFLARE_ACCOUNT_ID }}
+        run: |
+          set -euo pipefail
+          if [ -z "${CLOUDFLARE_API_TOKEN}" ]; then
+            echo "::error::Missing secret CLOUDFLARE_API_TOKEN on Environment production"
+            exit 1
+          fi
+          if [ -z "${CLOUDFLARE_ACCOUNT_ID}" ]; then
+            echo "::error::Missing variable CLOUDFLARE_ACCOUNT_ID on Environment production"
+            exit 1
+          fi
+          if ! pnpm run deploy; then
+            echo "::error::wrangler deploy failed; Worker and static assets were not published"
+            exit 1
+          fi
+
+      # wrangler 4.x takes {bucket}/{key}; --bucket is not a valid flag on r2 object put.
+      # Quote content-type so the charset parameter is not split by the shell.
+      - name: Upload install scripts to R2
+        env:
+          CLOUDFLARE_API_TOKEN: ${{ secrets.CLOUDFLARE_API_TOKEN }}
+          CLOUDFLARE_ACCOUNT_ID: ${{ vars.CLOUDFLARE_ACCOUNT_ID }}
+        run: |
+          set -euo pipefail
+          if ! npx wrangler r2 object put amber-dist/install.sh \
+            --file=../../install.sh \
+            --remote \
+            --content-type "text/plain; charset=utf-8"; then
+            echo "::error::Failed to upload install.sh to R2 bucket amber-dist"
+            exit 1
+          fi
+          if ! npx wrangler r2 object put amber-dist/install.ps1 \
+            --file=../../install.ps1 \
+            --remote \
+            --content-type "text/plain; charset=utf-8"; then
+            echo "::error::Failed to upload install.ps1 to R2 bucket amber-dist"
+            exit 1
+          fi

apps/website/README.md

@@ -19,7 +19,13 @@ The production bundle is written to `dist/`. The `prebuild` step copies repo-roo
 
 ## Cloudflare Deploy
 
-`get.amberjs.com` is a custom domain on the same Worker as `amberjs.com` (`apps/website/wrangler.toml`). There is no GitHub Actions deploy job; publishing is a local Wrangler deploy (needs a Cloudflare account with access to this Worker):
+`get.amberjs.com` is a custom domain on the same Worker as `amberjs.com` (`apps/website/wrangler.toml`).
+
+Production deploys run through GitHub Actions on the **`production`** Environment (secret `CLOUDFLARE_API_TOKEN`, variable `CLOUDFLARE_ACCOUNT_ID`). The workflow publishes on push to `main` when `apps/website/**`, `install.sh`, `install.ps1`, or the workflow file change. Tokens are not stored in the YAML.
+
+To publish manually: **Actions → Deploy website → Run workflow**.
+
+Local Wrangler is still available for dry-run and development (needs a Cloudflare account with access to this Worker):
 
 ```bash
 pnpm run deploy:dry-run
@@ -28,9 +34,9 @@ pnpm run deploy
 
 `wrangler.toml` serves `dist/` as static assets and uses `single-page-application` fallback so direct links such as `/docs/installation` work on Cloudflare.
 
-`src/worker.ts` serves `/install.sh` and `/install.ps1` from the R2 bucket `amber-dist` **when that object exists**, otherwise from the Worker static assets. Live `get.amberjs.com/install.sh` currently matches the asset-hosted copy (`content-type: application/x-sh`, `max-age=0`), so a Wrangler deploy updates the installer. If R2 later has `install.sh` / `install.ps1`, it shadows the deploy — also upload:
+`src/worker.ts` serves `/install.sh` and `/install.ps1` from the R2 bucket `amber-dist` **when that object exists**, otherwise from the Worker static assets. If R2 has `install.sh` / `install.ps1`, it shadows the Worker assets. The production workflow also uploads both scripts to `amber-dist` after deploy so `get.amberjs.com` stays in sync. Local equivalent (wrangler 4.x uses `{bucket}/{key}`):
 
 ```bash
-npx wrangler r2 object put install.sh --file=../../install.sh --bucket=amber-dist --remote --content-type text/plain
-npx wrangler r2 object put install.ps1 --file=../../install.ps1 --bucket=amber-dist --remote --content-type text/plain
+npx wrangler r2 object put amber-dist/install.sh --file=../../install.sh --remote --content-type "text/plain; charset=utf-8"
+npx wrangler r2 object put amber-dist/install.ps1 --file=../../install.ps1 --remote --content-type "text/plain; charset=utf-8"
 ```