fix(aux): trigger fallback on 429 rate-limit errors in auxiliary client

When a provider returns a 429 rate-limit error (not billing-related),
the auxiliary client's call_llm/async_call_llm previously did NOT trigger
the fallback chain. This caused auxiliary tasks like session_search to
exhaust all 3 retries against the same rate-limited endpoint, losing
session metadata that depended on the summarization completing.

Root cause: `_is_payment_error()` only matched 429s containing billing
keywords ("credits", "insufficient funds", etc.). Provider-specific
rate-limit messages like Nous's "Hold up for a bit, you've exceeded the
rate limit on your API key" didn't match, so `_is_payment_error` returned
False, `_is_connection_error` returned False, and `should_fallback` was
False — all retries hit the same rate-limited provider.

Fix:
- New `_is_rate_limit_error()` function that detects 429 + rate-limit
  keywords, generic 429 without billing keywords, and OpenAI SDK
  `RateLimitError` class instances (which may omit .status_code).
- Updated `should_fallback` in both `call_llm` and `async_call_llm` to
  include `_is_rate_limit_error`.
- Updated the max_tokens retry path to also check for rate-limit errors.
- Updated the reason string to include "rate limit".

This complements the Nous rate guard (PR #10568) which prevents new calls
to Nous when already rate-limited — this fix handles the case where a
request is already in flight when the 429 arrives.

Related: #8023, #12554, #11034
Co-authored-by: Zeejay <zjtan1@gmail.com>

tests/agent/test_auxiliary_client.py

@@ -20,6 +20,7 @@ from agent.auxiliary_client import (
     _read_codex_access_token,
     _get_provider_chain,
     _is_payment_error,
+    _is_rate_limit_error,
     _normalize_aux_provider,
     _try_payment_fallback,
     _resolve_auto,
@@ -789,6 +790,65 @@ class TestIsPaymentError:
         assert _is_payment_error(exc) is False
 
 
+class TestIsRateLimitError:
+    """_is_rate_limit_error detects 429 rate-limit errors warranting fallback."""
+
+    def test_429_with_rate_limit_message(self):
+        exc = Exception("Rate limit exceeded, try again in 2 seconds")
+        exc.status_code = 429
+        assert _is_rate_limit_error(exc) is True
+
+    def test_429_with_resets_in_message(self):
+        """Nous-style 429: 'resets in 3508s'."""
+        exc = Exception("Hold up for a bit, you've exceeded the rate limit on your API key")
+        exc.status_code = 429
+        assert _is_rate_limit_error(exc) is True
+
+    def test_429_with_too_many_requests(self):
+        exc = Exception("Too many requests")
+        exc.status_code = 429
+        assert _is_rate_limit_error(exc) is True
+
+    def test_429_without_billing_keywords_is_rate_limit(self):
+        """Generic 429 without billing keywords = likely a rate limit."""
+        exc = Exception("Something went wrong")
+        exc.status_code = 429
+        assert _is_rate_limit_error(exc) is True
+
+    def test_429_with_credits_message_is_not_rate_limit(self):
+        """Billing-related 429 should NOT be classified as rate limit."""
+        exc = Exception("insufficient credits remaining")
+        exc.status_code = 429
+        assert _is_rate_limit_error(exc) is False
+
+    def test_429_with_billing_message_is_not_rate_limit(self):
+        exc = Exception("you can only afford 1000 tokens")
+        exc.status_code = 429
+        assert _is_rate_limit_error(exc) is False
+
+    def test_402_is_not_rate_limit(self):
+        exc = Exception("Payment Required")
+        exc.status_code = 402
+        assert _is_rate_limit_error(exc) is False
+
+    def test_500_is_not_rate_limit(self):
+        exc = Exception("Internal Server Error")
+        exc.status_code = 500
+        assert _is_rate_limit_error(exc) is False
+
+    def test_openai_ratelimiterror_classname(self):
+        """OpenAI SDK RateLimitError may omit .status_code — detect by class name."""
+        class RateLimitError(Exception):
+            pass
+        exc = RateLimitError("rate limit exceeded")
+        # No status_code set, but class name matches
+        assert _is_rate_limit_error(exc) is True
+
+    def test_no_status_code_no_keywords_is_not_rate_limit(self):
+        exc = Exception("connection reset")
+        assert _is_rate_limit_error(exc) is False
+
+
 class TestGetProviderChain:
     """_get_provider_chain() resolves functions at call time (testable)."""
 
@@ -860,13 +920,18 @@ class TestTryPaymentFallback:
 
 
 class TestCallLlmPaymentFallback:
-    """call_llm() retries with a different provider on 402 / payment errors."""
+    """call_llm() retries with a different provider on 402 / payment / rate-limit errors."""
 
     def _make_402_error(self, msg="Payment Required: insufficient credits"):
         exc = Exception(msg)
         exc.status_code = 402
         return exc
 
+    def _make_429_rate_limit_error(self, msg="Rate limit exceeded, try again in 60 seconds"):
+        exc = Exception(msg)
+        exc.status_code = 429
+        return exc
+
     def test_non_payment_error_not_caught(self, monkeypatch):
         """Non-payment/non-connection errors (500) should NOT trigger fallback."""
         monkeypatch.setenv("OPENROUTER_API_KEY", "or-key")
@@ -886,6 +951,32 @@ class TestCallLlmPaymentFallback:
                     messages=[{"role": "user", "content": "hello"}],
                 )
 
+    def test_429_rate_limit_triggers_fallback(self, monkeypatch):
+        """429 rate-limit errors should trigger fallback to next provider."""
+        monkeypatch.setenv("OPENROUTER_API_KEY", "or-key")
+
+        primary_client = MagicMock()
+        rate_err = self._make_429_rate_limit_error()
+        primary_client.chat.completions.create.side_effect = rate_err
+
+        fallback_client = MagicMock()
+        fallback_client.chat.completions.create.return_value = MagicMock(choices=[
+            MagicMock(message=MagicMock(content="fallback response"))
+        ])
+
+        with patch("agent.auxiliary_client._get_cached_client",
+                    return_value=(primary_client, "xiaomi/mimo-v2-pro")), \
+             patch("agent.auxiliary_client._resolve_task_provider_model",
+                    return_value=("auto", "xiaomi/mimo-v2-pro", None, None, None)), \
+             patch("agent.auxiliary_client._try_payment_fallback",
+                    return_value=(fallback_client, "fallback-model", "openrouter")):
+            result = call_llm(
+                task="session_search",
+                messages=[{"role": "user", "content": "hello"}],
+            )
+        # Fallback client should have been used
+        assert fallback_client.chat.completions.create.called
+
 # ---------------------------------------------------------------------------
 # Gate: _resolve_api_key_provider must skip anthropic when not configured
 # ---------------------------------------------------------------------------