Keep embedding retry on GCP when 111 fallback disabled

tests/test_ollama_retry_chain.py

@@ -378,7 +378,32 @@ def test_embedding_can_disable_111_fallback_for_background_rag_work():
 
     posted_hosts = [call.args[0].split('/api/embed')[0] for call in mock_post.call_args_list]
     assert vec == []
-    assert posted_hosts == [oss.OLLAMA_HOST_SECONDARY]
+    assert posted_hosts == [oss.OLLAMA_HOST_SECONDARY, oss.OLLAMA_HOST_PRIMARY]
+    assert oss.OLLAMA_HOST_FALLBACK not in posted_hosts
+
+
+def test_embedding_fallback_disabled_uses_gcp_chain_when_resolver_returns_111():
+    """resolver 若因 unhealthy cache 回 111，背景 embedding 仍要嘗試 GCP-A/GCP-B。"""
+    import requests
+    from services import ollama_service as oss
+    from services.ollama_service import OllamaService
+
+    svc = OllamaService()
+
+    with patch(
+        'services.ollama_service.resolve_ollama_host',
+        side_effect=[oss.OLLAMA_HOST_FALLBACK, oss.OLLAMA_HOST_FALLBACK],
+    ), patch.dict('os.environ', {}, clear=False), patch(
+        'services.ollama_service.requests.post',
+        side_effect=requests.Timeout('gcp timeout'),
+    ) as mock_post:
+        import os
+        os.environ.pop('EMBEDDING_HOST', None)
+        vec = svc.generate_embedding('test text', allow_111_fallback=False)
+
+    posted_hosts = [call.args[0].split('/api/embed')[0] for call in mock_post.call_args_list]
+    assert vec == []
+    assert posted_hosts == [oss.OLLAMA_HOST_PRIMARY, oss.OLLAMA_HOST_SECONDARY]
     assert oss.OLLAMA_HOST_FALLBACK not in posted_hosts