Become the data-only repo: site and worker move to private nagi-bench-site

The showcase app and the arena backend now live in the private
nagi-bench-site repo and deploy to Cloudflare Pages, fetching this
repo's models/, outputs/ and cases.json at build time. CI here is a
standalone data validation (same rules as before, no build), plus an
optional repository_dispatch that triggers an instant site rebuild when
SITE_DISPATCH_TOKEN is configured — otherwise the site's 6-hour cron
picks new data up.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

.github/workflows/ci.yml

@@ -2,19 +2,39 @@ name: CI
 
 on:
   pull_request:
+  push:
+    branches: [main]
 
 permissions:
   contents: read
 
 jobs:
-  build:
+  validate:
     runs-on: ubuntu-latest
     steps:
       - uses: actions/checkout@v4
       - uses: oven-sh/setup-bun@v2
-      - name: Install dependencies
-        run: bun install
-        working-directory: site
-      - name: Type-check and build
-        run: bun run build
-        working-directory: site
+      - name: Validate contribution data
+        run: bun scripts/validate-data.ts
+
+  trigger-site-rebuild:
+    # The site lives in the private nagi-bench-site repo (Cloudflare Pages).
+    # With SITE_DISPATCH_TOKEN set the rebuild is instant; without it the
+    # site's 6-hour cron picks the new data up anyway.
+    if: github.event_name == 'push'
+    needs: validate
+    runs-on: ubuntu-latest
+    steps:
+      - name: Dispatch data-updated to nagi-bench-site
+        env:
+          TOKEN: ${{ secrets.SITE_DISPATCH_TOKEN }}
+        run: |
+          if [ -z "$TOKEN" ]; then
+            echo "SITE_DISPATCH_TOKEN not set; relying on the site's scheduled rebuild."
+            exit 0
+          fi
+          curl -sf -X POST \
+            -H "Authorization: Bearer $TOKEN" \
+            -H "Accept: application/vnd.github+json" \
+            https://api.github.com/repos/nagi-studio/nagi-bench-site/dispatches \
+            -d '{"event_type":"data-updated"}'

.github/workflows/deploy.yml

@@ -1,46 +0,0 @@
-name: Deploy site to GitHub Pages
-
-on:
-  push:
-    branches: [main]
-  workflow_dispatch:
-
-permissions:
-  contents: read
-  pages: write
-  id-token: write
-
-concurrency:
-  group: pages
-  cancel-in-progress: true
-
-jobs:
-  build:
-    runs-on: ubuntu-latest
-    steps:
-      - uses: actions/checkout@v4
-      - uses: oven-sh/setup-bun@v2
-      - name: Install dependencies
-        run: bun install --frozen-lockfile
-        working-directory: site
-      - name: Build
-        run: bun run build
-        working-directory: site
-        env:
-          # Optional arena backend (see worker/README.md); empty = local-only votes.
-          VITE_ARENA_API: ${{ vars.VITE_ARENA_API }}
-          VITE_TURNSTILE_SITE_KEY: ${{ vars.VITE_TURNSTILE_SITE_KEY }}
-      - uses: actions/configure-pages@v5
-      - uses: actions/upload-pages-artifact@v3
-        with:
-          path: site/dist
-
-  deploy:
-    needs: build
-    runs-on: ubuntu-latest
-    environment:
-      name: github-pages
-      url: ${{ steps.deployment.outputs.page_url }}
-    steps:
-      - id: deployment
-        uses: actions/deploy-pages@v4

README.en.md

@@ -49,18 +49,16 @@ Pending: Mistral Medium 3.5 · GLM-5.1 · Qwen3.7-Max (PRs welcome)
 ```
 outputs/<model-id>/<case-id>.<ext>   raw model artifacts (HTML / SVG), filename = case id
 models/<model-id>.json               combination registry: label / vendor / harness / effort / order / run notes
-site/                                showcase app (Vite + React + Tailwind v4 + GSAP, bun)
-site/src/data/cases.json             case definitions (bilingual prompts, maintainer-owned)
-.github/workflows/                   deploy.yml deploys; ci.yml validates data + builds on PRs
+cases.json                           case definitions (bilingual prompts, maintainer-owned)
+scripts/validate-data.ts             data validation (CI runs it on every PR)
 ```
 
-## Development
+> The site and vote-backend source lives in the private `nagi-bench-site` repo, deployed on Cloudflare Pages; this repo is the site's data source.
+
+## Validation
 
 ```bash
-cd site
-bun install
-bun run dev      # syncs outputs/ and serves http://localhost:5173/
-bun run build    # data validation + type-check + build (site/dist)
+bun scripts/validate-data.ts   # the same data validation CI runs
 ```
 
 ## Contributing a run
@@ -72,4 +70,4 @@ Contributions are data-only — no code changes needed:
 
 CI-enforced rules: dash-only lowercase ids matching the `outputs/` folder name; a **required bilingual provenance `note`** per run (which harness, what effort, one-shot or fixed); every declared run must have its artifact; multiple versions of the same combination x case go in an array with distinct `file` names (e.g. `pelican-cycling-2.svg`). Set `contributor` to your GitHub username — the site shows your avatar linked to your profile next to the run. Fill in `harness` and `effort` truthfully for new combinations: the case page renders Harness / Effort metadata chips from these fields and auto-matches brand icons for the model, vendor and harness (via [lobe-icons](https://github.com/lobehub/lobe-icons)) — contributors never touch icons.
 
-PRs get data validation + a build from CI; merges to `main` deploy automatically.
+PRs get data validation from CI; merges to `main` rebuild the site automatically (usually instant, at most 6 hours).

README.md

@@ -49,18 +49,16 @@ NAGI STUDIO 的 LLM 测评案例集：同一段提示词，不同「模型 × Ha
 ```
 outputs/<model-id>/<case-id>.<ext>   模型产出原文件（HTML / SVG），文件名 = 案例 id
 models/<model-id>.json               组合登记：label / vendor / harness / effort / order / 各案例运行备注
-site/                                展示站点（Vite + React + Tailwind v4 + GSAP，bun 驱动）
-site/src/data/cases.json             案例定义：双语标题与提示词（维护者维护）
-.github/workflows/                   deploy.yml 自动部署；ci.yml 对 PR 跑数据校验 + 构建
+cases.json                           案例定义：双语标题与提示词（维护者维护）
+scripts/validate-data.ts             数据校验（CI 对每个 PR 自动执行）
 ```
 
-## 本地开发 / Development
+> 站点与投票后端的源码在私有仓库 `nagi-bench-site` 中，部署于 Cloudflare Pages；本仓库是站点的数据源。
+
+## 本地校验 / Validation
 
 ```bash
-cd site
-bun install
-bun run dev      # 同步 outputs/ 并启动 http://localhost:5173/
-bun run build    # 数据校验 + 类型检查 + 产物构建（输出到 site/dist）
+bun scripts/validate-data.ts   # 与 CI 相同的数据校验
 ```
 
 ## 贡献一个模型产出 / Contributing a run
@@ -97,4 +95,4 @@ bun run build    # 数据校验 + 类型检查 + 产物构建（输出到 site/d
 - `contributor` 填你的 GitHub 用户名，站点会在产出旁展示你的头像并链接到主页；
 - 新组合的 `harness`（运行环境）与 `effort`（思考配额）请如实填写：站点会据此在测评页生成「运行环境 / 思考配额」metadata 徽章，并自动为模型、厂商、Harness 匹配品牌 icon（来自 [lobe-icons](https://github.com/lobehub/lobe-icons)），贡献者无需处理任何图标。
 
-提 PR 后 CI 会跑数据校验 + 构建；合入 `main` 自动部署。
+提 PR 后 CI 自动校验数据；合入 `main` 后站点自动重建（通常即时，最长 6 小时）。

scripts/validate-data.ts

@@ -0,0 +1,110 @@
+// Validates the contribution data: models/<model-id>.json registrations and
+// their artifacts under outputs/. Runs in CI for every PR and push, so a
+// malformed or undocumented contribution fails before it can ship.
+import { existsSync, readFileSync, readdirSync } from 'node:fs'
+import { dirname, join } from 'node:path'
+import { fileURLToPath } from 'node:url'
+
+const root = join(dirname(fileURLToPath(import.meta.url)), '..')
+const modelsDir = join(root, 'models')
+const outputsDir = join(root, 'outputs')
+
+const cases = JSON.parse(readFileSync(join(root, 'cases.json'), 'utf8')) as Array<{
+  id: string
+  kind: string
+}>
+const caseKind = new Map(cases.map((c) => [c.id, c.kind]))
+
+const ID_RE = /^[a-z0-9]+(-[a-z0-9]+)*$/
+const errors: string[] = []
+
+const isBilingual = (v: unknown): boolean => {
+  if (!v || typeof v !== 'object') return false
+  const { zh, en } = v as { zh?: unknown; en?: unknown }
+  return typeof zh === 'string' && zh.trim() !== '' && typeof en === 'string' && en.trim() !== ''
+}
+
+const modelIds = new Set<string>()
+for (const f of readdirSync(modelsDir).filter((f) => f.endsWith('.json'))) {
+  const id = f.replace(/\.json$/, '')
+  modelIds.add(id)
+  if (!ID_RE.test(id)) {
+    errors.push(`models/${f}: id "${id}" must be lowercase letters/digits/dashes only (it doubles as the outputs/ folder name)`)
+  }
+
+  type RunJson = { note?: unknown; file?: unknown; contributor?: unknown }
+  let def: { label?: unknown; vendor?: unknown; order?: unknown; runs?: Record<string, RunJson | RunJson[]> }
+  try {
+    def = JSON.parse(readFileSync(join(modelsDir, f), 'utf8'))
+  } catch (e) {
+    errors.push(`models/${f}: invalid JSON (${e})`)
+    continue
+  }
+
+  if (typeof def.label !== 'string' || !def.label.trim()) errors.push(`models/${f}: "label" is required`)
+  if (typeof def.vendor !== 'string' || !def.vendor.trim()) errors.push(`models/${f}: "vendor" is required`)
+  if (def.order !== undefined && typeof def.order !== 'number') errors.push(`models/${f}: "order" must be a number`)
+
+  for (const [caseId, runRaw] of Object.entries(def.runs ?? {})) {
+    if (!caseKind.has(caseId)) {
+      errors.push(`models/${f}: unknown case "${caseId}" (known: ${[...caseKind.keys()].join(', ')})`)
+      continue
+    }
+    const variants = Array.isArray(runRaw) ? runRaw : [runRaw]
+    if (!variants.length) errors.push(`models/${f}: runs.${caseId} is an empty array`)
+    const seenFiles = new Set<string>()
+    variants.forEach((run, i) => {
+      const where = variants.length > 1 ? `runs.${caseId}[${i}]` : `runs.${caseId}`
+      if (!isBilingual(run?.note)) {
+        errors.push(`models/${f}: ${where} needs a bilingual "note" ({ "zh", "en" }) documenting provenance (tool, effort level, one-shot or fixed)`)
+      }
+      if (run?.file !== undefined && typeof run.file !== 'string') {
+        errors.push(`models/${f}: ${where}.file must be a string`)
+      }
+      if (run?.contributor !== undefined && (typeof run.contributor !== 'string' || !run.contributor.trim())) {
+        errors.push(`models/${f}: ${where}.contributor must be a non-empty GitHub username`)
+      }
+      const file = (typeof run?.file === 'string' && run.file) || `${caseId}.${caseKind.get(caseId)}`
+      if (seenFiles.has(file)) {
+        errors.push(`models/${f}: ${where} resolves to the same artifact "${file}" as another variant — extra variants must set a distinct "file"`)
+      }
+      seenFiles.add(file)
+      if (!existsSync(join(outputsDir, id, file))) {
+        errors.push(`models/${f}: ${where} declared but artifact missing: outputs/${id}/${file}`)
+      }
+    })
+  }
+}
+
+for (const entry of readdirSync(outputsDir, { withFileTypes: true })) {
+  if (entry.isDirectory() && !modelIds.has(entry.name)) {
+    errors.push(`outputs/${entry.name}/ exists but models/${entry.name}.json is missing`)
+  }
+}
+
+// Every file under outputs/<id>/ must be declared by a run — loose artifacts
+// would otherwise ship without provenance.
+for (const id of modelIds) {
+  const dir = join(outputsDir, id)
+  if (!existsSync(dir)) continue
+  const def = JSON.parse(readFileSync(join(modelsDir, `${id}.json`), 'utf8')) as {
+    runs?: Record<string, { file?: string } | Array<{ file?: string }>>
+  }
+  const declared = new Set<string>()
+  for (const [caseId, runRaw] of Object.entries(def.runs ?? {})) {
+    for (const run of Array.isArray(runRaw) ? runRaw : [runRaw]) {
+      declared.add(run?.file ?? `${caseId}.${caseKind.get(caseId)}`)
+    }
+  }
+  for (const file of readdirSync(dir)) {
+    if (!declared.has(file)) {
+      errors.push(`outputs/${id}/${file} is not declared by any run in models/${id}.json`)
+    }
+  }
+}
+
+if (errors.length) {
+  console.error(`Data validation failed (${errors.length}):\n- ${errors.join('\n- ')}`)
+  process.exit(1)
+}
+console.log(`Data OK: ${modelIds.size} models, ${caseKind.size} cases validated`)