Task/ccl 8779 txt files for llms fixed

.github/workflows/llms-txt.yml

@@ -0,0 +1,43 @@
+# Builds the site exactly like production (npm run build = generate-llms + docusaurus build)
+# and verifies that the LLM/MCP text files (llms.txt, llms-full.txt, /docs/**.txt)
+# are present and in sync with the HTML pages.
+#
+# Production deploys are done by Cloudflare Pages from `master`; this workflow does not
+# deploy anything. It only guards against regressions before they reach master.
+name: llms.txt parity
+
+on:
+  pull_request:
+  push:
+    branches: [master]
+
+permissions:
+  contents: read
+
+jobs:
+  build-and-verify:
+    runs-on: ubuntu-latest
+    timeout-minutes: 20
+    env:
+      # Placeholders only. The Algolia theme refuses to build with empty values.
+      # Real search keys live in the Cloudflare Pages environment, never in this repo.
+      ALGOLIA_APP_ID: placeholder
+      ALGOLIA_API_KEY: placeholder
+      ALGOLIA_INDEX_NAME: placeholder
+    steps:
+      - uses: actions/checkout@v4
+        with:
+          fetch-depth: 0 # showLastUpdateTime reads git history
+
+      - uses: actions/setup-node@v4
+        with:
+          node-version: 20
+          cache: npm
+
+      - run: npm ci
+
+      - name: Build site (same command as production)
+        run: npm run build
+
+      - name: Verify LLM text files match the built pages
+        run: npm run verify-llms

README.md

@@ -99,6 +99,40 @@ yarn serve
 
 ---
 
+### 🚀 Deployment
+
+The live site is built and published by **Cloudflare Pages** (project `capmonstercloud-docs-v2`) directly from this GitHub repository. Every push to `master` triggers a production build. There is no deploy step in this repo — merging to `master` **is** the deploy.
+
+The Cloudflare build runs `npm run build`, which is:
+
+1. `node scripts/generate-llms.mjs` — generates the plain-text documentation for AI agents (see below);
+2. `docusaurus build` — builds the site into `build/`.
+
+Search (Algolia) credentials are provided as environment variables in the Cloudflare Pages project settings. **Never commit them to this repository** — it is public.
+
+The `Dockerfile` is a legacy artifact from a previous hosting setup and is not used by the current deployment.
+
+### 🤖 Text version for AI agents (llms.txt / MCP)
+
+Alongside the HTML pages the build publishes plain-text copies of the documentation, used by the [CapMonster Cloud MCP server](https://github.com/CapMonsterCloud/capmonster-mcp-captcha-solver) and other AI agents:
+
+- `https://docs.capmonster.cloud/llms.txt` — index of all pages with short descriptions;
+- `https://docs.capmonster.cloud/llms-full.txt` — the whole documentation in one file;
+- `https://docs.capmonster.cloud/docs/<page>.txt` — one file per page, e.g. `/docs/captchas/recaptcha-v3-task/` → `/docs/captchas/recaptcha-v3-task.txt`.
+
+They are generated from the English MDX sources on every build by `scripts/generate-llms.mjs` (output is git-ignored), so they are always in sync with the deployed pages. Content that lives in React components rather than MDX (`<McpNotice/>`, `<PriceBlock/>`, `<BlogLink/>`) is rendered to text by that script — if you add a new content-bearing component, teach the script about it too.
+
+To check locally:
+
+```bash
+npm run build          # generates txt files + builds the site
+npm run verify-llms    # checks that txt files exist and match the built pages
+```
+
+The same check runs in GitHub Actions (`.github/workflows/llms-txt.yml`) on every pull request.
+
+---
+
 ## 📄 License
 
 [MIT](./LICENSE) © [CapMonster Cloud](https://capmonster.cloud/en/?utm_source=github&utm_medium=referral&utm_campaign=docs_repo_read)

package.json

@@ -6,6 +6,7 @@
     "docusaurus": "docusaurus",
     "start": "docusaurus start --host 0.0.0.0",
     "generate-llms": "node scripts/generate-llms.mjs",
+    "verify-llms": "node scripts/verify-llms.mjs",
     "build": "node scripts/generate-llms.mjs && docusaurus build",
     "swizzle": "docusaurus swizzle",
     "deploy": "docusaurus deploy",