feat(webui): Wave 2 trust & ops — health, Turnstile, rate-limit UX

.github/workflows/deploy-pages.yml

@@ -48,6 +48,7 @@ jobs:
           WORKER_URL: ${{ secrets.WORKER_URL }}
           LITELLM_URL: ${{ secrets.LITELLM_URL }}
           PROXY_GUEST_TOKEN: ${{ secrets.PROXY_GUEST_TOKEN }}
+          TURNSTILE_SITE_KEY: ${{ secrets.TURNSTILE_SITE_KEY }}
         run: |
           python3 <<'PY'
           import json
@@ -61,6 +62,7 @@ jobs:
                   endpoints.append(val)
 
           guest_token = os.environ.get("PROXY_GUEST_TOKEN") or "llm-fallbacks-public"
+          turnstile_site_key = (os.environ.get("TURNSTILE_SITE_KEY") or "").strip()
           chat_proxy_path = Path("configs/chat_proxy.json")
           if chat_proxy_path.is_file():
               chat_proxy = json.loads(chat_proxy_path.read_text(encoding="utf-8"))
@@ -80,6 +82,8 @@ jobs:
               "maxTokens": 512,
               "appVersion": os.environ.get("APP_VERSION", ""),
           }
+          if turnstile_site_key:
+              cfg["turnstileSiteKey"] = turnstile_site_key
           body = "window.LLM_FALLBACKS_CONFIG = " + json.dumps(cfg, indent=2) + ";\n"
           open("docs/config.js", "w", encoding="utf-8").write(body)
           PY
@@ -123,7 +127,7 @@ jobs:
         run: |
           cd webui && npm ci && npm run build
           cd ..
-          npx playwright test tests/e2e/pages-chat.spec.ts tests/e2e/failover-dual-endpoint.spec.ts tests/e2e/model-picker.spec.ts tests/e2e/message-actions.spec.ts
+          npx playwright test tests/e2e/pages-chat.spec.ts tests/e2e/failover-dual-endpoint.spec.ts tests/e2e/model-picker.spec.ts tests/e2e/message-actions.spec.ts tests/e2e/health-panel.spec.ts tests/e2e/turnstile-gate.spec.ts
 
       - name: Run live Pages chat e2e (real proxy)
         env:

CONCEPTS.md

@@ -34,8 +34,10 @@ Shared vocabulary for the static chat gateway and Python library.
 |------|---------|
 | **Render secondary** | LiteLLM on Render (`llm-fallbacks-gateway.onrender.com` pattern) |
 | **loadRuntimeConfig** | `readRuntimeConfig()` + `mergeChatProxyArtifact()` — single bootstrap path for `FailoverProvider` |
-| **Routing chip** | Per-reply UI showing endpoint, resolved model, and fallback hops (planned) |
+| **Routing chip** | Per-reply UI showing endpoint, resolved model, and fallback hops |
 | **Model selector** | Composer picker over `free_models.json`; separate from Failover endpoint settings |
+| **Endpoint health probe** | Client-side GET to `/health` (or LiteLLM `/health/liveliness` on Render); ok / slow / fail |
+| **Turnstile session** | Optional bot check at Worker; 1h KV pass per IP after successful siteverify |
 
 ## Learnings index
 

docs/CAVEATS.md

@@ -13,6 +13,8 @@ Honest limits for the public chat demo and the Python library.
 | **Backup server auth** | The browser sends the same guest token to every endpoint. The Worker accepts it; the Render backup may return 401 until a LiteLLM virtual key is set up. |
 | **Saved settings** | If you changed proxy URLs in the browser, old values stay in localStorage. Clear site data if the failover list looks wrong after an update. |
 | **Usage stats** | We count sessions and completions without storing message text. This is not full product analytics. |
+| **Routing headers** | The routing chip reads `x-llm-fallbacks-endpoint` and LiteLLM headers from proxy responses. After edge changes, redeploy the Worker (`Deploy Proxies` workflow) for production to expose them. |
+| **Turnstile** | Optional bot check when `TURNSTILE_SECRET` is set on the Worker and `turnstileSiteKey` is in `docs/config.js`. Skipped in local dev when secrets are absent. |
 
 ## Library and CI