fix: classify HTML provider error pages correctly (#67642) (thanks @stainlu)

* fix(agents): classify Cloudflare/CDN HTML error pages as transport failures Fixes #67517 When a provider endpoint returns an HTML error page (e.g. Cloudflare 502/503/520-524), the pattern-based message classifiers would scan the HTML body and misinterpret embedded text like "Rate limit exceeded" as a structured rate_limit API error. This caused incorrect failover behavior (profile rotation instead of clean retry/fallback) and left the TUI stuck. Two fixes: 1. classifyFailoverSignal now short-circuits on HTML responses before running pattern matchers, returning "timeout" (transport failure) so retry/fallback handles them correctly. 2. classifyProviderRuntimeFailureKind now detects HTML errors at any status (not just 403), returning "upstream_html" for non-403 statuses with a clear user-facing message about CDN/gateway errors. Adds regression tests covering Cloudflare 502/503 HTML with embedded rate-limit text, 403 HTML (still classified as auth), and JSON rate-limit responses (still classified correctly). * fix: preserve auth and proxy HTML classification * fix: classify HTML provider error pages correctly (#67642) (thanks @stainlu) --------- Co-authored-by: Ayaan Zaidi <hi@obviy.us>
2026-05-06 08:30:42 +00:00 · 2026-04-16 20:49:53 +08:00
parent 55f05df77e
commit e588e904a7
4 changed files with 110 additions and 12 deletions
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -35,6 +35,7 @@ Docs: https://docs.openclaw.ai
 - Agents/tools: resolve non-workspace host tilde paths against the OS home directory and keep edit recovery aligned with that same path target, so `~/...` host edit/write operations stop failing or reading back the wrong file when `OPENCLAW_HOME` differs. (#62804) Thanks @stainlu.
 - Speech/TTS: auto-enable the bundled Microsoft and ElevenLabs speech providers, and route generic TTS directive tokens through the explicit or active provider first so overrides like `[[tts:speed=1.2]]` stop silently landing on the wrong provider. (#62846) Thanks @stainlu.
 - OpenAI Codex/models: normalize stale native transport metadata in both runtime resolution and discovery/listing so legacy `openai-codex` rows with missing `api` or `https://chatgpt.com/backend-api/v1` self-heal to the canonical Codex transport instead of routing requests through broken HTML/Cloudflare paths, combining the original fixes proposed in #66969 (saamuelng601-pixel) and #67159 (hclsys). (#67635)
+- Agents/failover: treat HTML provider error pages as upstream transport failures for CDN-style 5xx responses without misclassifying embedded body text as API rate limits, while still preserving auth remediation for HTML 401/403 pages and proxy remediation for HTML 407 pages. (#67642) Thanks @stainlu.

 ## 2026.4.15-beta.1

--- a/src/agents/pi-embedded-helpers.formatassistanterrortext.test.ts
+++ b/src/agents/pi-embedded-helpers.formatassistanterrortext.test.ts
@@ -167,21 +167,21 @@ describe("formatAssistantErrorText", () => {
    expect(result).toBe("⚠️ Your quota has been exhausted, try again in 24 hours");
  });

-  it("falls back to generic copy for HTML quota pages", () => {
+  it("returns upstream HTML copy for HTML quota pages", () => {
    const msg = makeAssistantError(
      "429 <!DOCTYPE html><html><body>Your quota is exhausted</body></html>",
    );
    expect(formatAssistantErrorText(msg)).toBe(
-      "⚠️ API rate limit reached. Please try again later.",
+      "The provider returned an HTML error page instead of an API response. This usually means a CDN or gateway (e.g. Cloudflare) blocked the request. Retry in a moment or check provider status.",
    );
  });

-  it("falls back to generic copy for prefixed HTML rate-limit pages", () => {
+  it("returns upstream HTML copy for prefixed HTML rate-limit pages", () => {
    const msg = makeAssistantError(
      "Error: 521 <!DOCTYPE html><html><body>rate limit</body></html>",
    );
    expect(formatAssistantErrorText(msg)).toBe(
-      "⚠️ API rate limit reached. Please try again later.",
+      "The provider returned an HTML error page instead of an API response. This usually means a CDN or gateway (e.g. Cloudflare) blocked the request. Retry in a moment or check provider status.",
    );
  });

--- a/src/agents/pi-embedded-helpers/errors.ts
+++ b/src/agents/pi-embedded-helpers/errors.ts
@@ -257,6 +257,7 @@ export type ProviderRuntimeFailureKind =
  | "auth_scope"
  | "auth_refresh"
  | "auth_html_403"
+  | "upstream_html"
  | "proxy"
  | "rate_limit"
  | "dns"
@@ -324,11 +325,15 @@ const REPLAY_INVALID_RE =
 const SANDBOX_BLOCKED_RE =
  /\bapproval is required\b|\bapproval timed out\b|\bapproval was denied\b|\bblocked by sandbox\b|\bsandbox\b.*\b(?:blocked|denied|forbidden|disabled|not allowed)\b/i;

+function stripErrorPrefix(raw: string): string {
+  return raw.replace(/^error:\s*/i, "").trim();
+}
+
 function inferSignalStatus(signal: FailoverSignal): number | undefined {
  if (typeof signal.status === "number" && Number.isFinite(signal.status)) {
    return signal.status;
  }
-  return extractLeadingHttpStatus(signal.message?.trim() ?? "")?.code;
+  return extractLeadingHttpStatus(stripErrorPrefix(signal.message?.trim() ?? ""))?.code;
 }

 function isHtmlErrorResponse(raw: string, status?: number): boolean {
@@ -336,17 +341,22 @@ function isHtmlErrorResponse(raw: string, status?: number): boolean {
  if (!trimmed) {
    return false;
  }
+  const candidate = extractLeadingHttpStatus(trimmed) ? trimmed : stripErrorPrefix(trimmed);
  const inferred =
    typeof status === "number" && Number.isFinite(status)
      ? status
-      : extractLeadingHttpStatus(trimmed)?.code;
+      : extractLeadingHttpStatus(candidate)?.code;
  if (typeof inferred !== "number" || inferred < 400) {
    return false;
  }
-  const rest = extractLeadingHttpStatus(trimmed)?.rest ?? trimmed;
+  const rest = extractLeadingHttpStatus(candidate)?.rest ?? candidate;
  return HTML_BODY_RE.test(rest) && HTML_CLOSE_RE.test(rest);
 }

+function isTransportHtmlErrorStatus(status: number | undefined): boolean {
+  return status !== 401 && status !== 403 && status !== 407;
+}
+
 function isOpenAICodexScopeContext(raw: string, provider?: string): boolean {
  const normalizedProvider = normalizeLowercaseStringOrEmpty(provider);
  return (
@@ -669,7 +679,9 @@ function isOpenRouterKeyLimitExceededError(raw: string, provider?: string): bool
 }

 function isExactUnknownNoDetailsError(raw: string): boolean {
-  return normalizeOptionalLowercaseString(raw)?.trim() === "unknown error (no error details in response)";
+  return (
+    normalizeOptionalLowercaseString(raw)?.trim() === "unknown error (no error details in response)"
+  );
 }

 function classifyFailoverClassificationFromMessage(
@@ -757,6 +769,13 @@ function classifyFailoverClassificationFromMessage(

 export function classifyFailoverSignal(signal: FailoverSignal): FailoverClassification | null {
  const inferredStatus = inferSignalStatus(signal);
+  if (
+    signal.message &&
+    isTransportHtmlErrorStatus(inferredStatus) &&
+    isHtmlErrorResponse(signal.message, inferredStatus)
+  ) {
+    return toReasonClassification("timeout");
+  }
  const messageClassification = signal.message
    ? classifyFailoverClassificationFromMessage(signal.message, signal.provider)
    : null;
@@ -791,12 +810,12 @@ export function classifyProviderRuntimeFailureKind(
  if (message && isAuthScopeErrorMessage(message, status, normalizedSignal.provider)) {
    return "auth_scope";
  }
-  if (message && status === 403 && isHtmlErrorResponse(message, status)) {
-    return "auth_html_403";
-  }
  if (message && isProxyErrorMessage(message, status)) {
    return "proxy";
  }
+  if (message && isHtmlErrorResponse(message, status)) {
+    return status === 403 ? "auth_html_403" : "upstream_html";
+  }
  const failoverClassification = classifyFailoverSignal({
    ...normalizedSignal,
    status,
@@ -885,6 +904,14 @@ export function formatAssistantErrorText(
    );
  }

+  if (providerRuntimeFailureKind === "upstream_html") {
+    return (
+      "The provider returned an HTML error page instead of an API response. " +
+      "This usually means a CDN or gateway (e.g. Cloudflare) blocked the request. " +
+      "Retry in a moment or check provider status."
+    );
+  }
+
  if (providerRuntimeFailureKind === "proxy") {
    return "LLM request failed: proxy or tunnel configuration blocked the provider request.";
  }
--- a/src/agents/pi-embedded-helpers/provider-error-patterns.test.ts
+++ b/src/agents/pi-embedded-helpers/provider-error-patterns.test.ts
@@ -16,7 +16,11 @@ vi.mock("../../plugins/provider-runtime.js", async () => {
  };
 });

-import { classifyFailoverReason, isContextOverflowError } from "./errors.js";
+import {
+  classifyFailoverReason,
+  classifyProviderRuntimeFailureKind,
+  isContextOverflowError,
+} from "./errors.js";
 import {
  classifyProviderSpecificError,
  matchesProviderContextOverflow,
@@ -146,3 +150,69 @@ describe("classifyFailoverReason with provider patterns", () => {
    );
  });
 });
+
+describe("Cloudflare / CDN HTML error page classification (#67517)", () => {
+  const cloudflareHtml502 =
+    "<!doctype html><html><head><title>502 Bad Gateway</title></head>" +
+    "<body><h1>502 Bad Gateway</h1><p>cloudflare-nginx</p></body></html>";
+  const cloudflareHtml503 =
+    "<!doctype html><html><head><title>503</title></head>" +
+    "<body><h1>Service Unavailable</h1><p>Please try again. Rate limit exceeded.</p></body></html>";
+  const html401 =
+    "<!doctype html><html><head><title>401 Unauthorized</title></head>" +
+    "<body><h1>Unauthorized</h1></body></html>";
+  const html403 =
+    "<!doctype html><html><head><title>403 Forbidden</title></head>" +
+    "<body><h1>Forbidden</h1></body></html>";
+  const html407 =
+    "<!doctype html><html><head><title>407 Proxy Authentication Required</title></head>" +
+    "<body><h1>Proxy Authentication Required</h1></body></html>";
+  const prefixedHtml401 = `Error: 401 ${html401}`;
+  const prefixedHtml407 = `Error: 407 ${html407}`;
+
+  it("classifies Cloudflare HTML 502 as timeout", () => {
+    expect(classifyFailoverReason(`502 ${cloudflareHtml502}`)).toBe("timeout");
+  });
+
+  it("classifies Cloudflare HTML 503 with rate-limit text as timeout", () => {
+    expect(classifyFailoverReason(`503 ${cloudflareHtml503}`)).toBe("timeout");
+  });
+
+  it("preserves auth classification for 401 HTML", () => {
+    expect(classifyFailoverReason(`401 ${html401}`)).toBe("auth");
+  });
+
+  it("preserves auth classification for 403 HTML", () => {
+    expect(classifyFailoverReason(`403 ${html403}`)).toBe("auth");
+  });
+
+  it("preserves auth classification for Error-prefixed 401 HTML", () => {
+    expect(classifyFailoverReason(prefixedHtml401)).toBe("auth");
+  });
+
+  it("classifies runtime failure kind as upstream_html for non-auth HTML", () => {
+    expect(classifyProviderRuntimeFailureKind({ status: 502, message: cloudflareHtml502 })).toBe(
+      "upstream_html",
+    );
+  });
+
+  it("classifies 403 HTML runtime failures as auth_html_403", () => {
+    expect(classifyProviderRuntimeFailureKind({ status: 403, message: html403 })).toBe(
+      "auth_html_403",
+    );
+  });
+
+  it("classifies 407 HTML runtime failures as proxy", () => {
+    expect(classifyProviderRuntimeFailureKind({ status: 407, message: html407 })).toBe("proxy");
+  });
+
+  it("classifies Error-prefixed 407 HTML runtime failures as proxy", () => {
+    expect(classifyProviderRuntimeFailureKind(prefixedHtml407)).toBe("proxy");
+  });
+
+  it("does not misclassify JSON API rate-limit responses as HTML", () => {
+    const jsonRateLimit =
+      '429 {"error":{"type":"rate_limit_error","message":"Rate limit exceeded"}}';
+    expect(classifyFailoverReason(jsonRateLimit)).toBe("rate_limit");
+  });
+});