fix: honor hyde_enabled toggle and record ping failures before client creation (Task 6 review)

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

tests/test_status_models_routes.py

@@ -58,6 +58,23 @@ def test_get_models_reflects_recorded_usage(client):
     assert main_row["status"] == "ok"
 
 
+def test_get_models_hyde_llm_disabled_forces_disabled_status(client):
+    """settings.hyde_enabled=False must force hyde_llm to enabled=False/status='disabled',
+    mirroring the reranker override, even if HyDE previously ran successfully."""
+    from app.config.settings import settings
+    get_model_usage_tracker().record(
+        provider=settings.hyde_llm_provider or settings.llm_provider,
+        model=settings.hyde_llm_model or settings.llm_model,
+        success=True,
+    )
+    with patch.object(settings, "hyde_enabled", False):
+        resp = client.get("/api/v1/status/models")
+    assert resp.status_code == 200
+    hyde_row = next(m for m in resp.json()["models"] if m["role"] == "hyde_llm")
+    assert hyde_row["enabled"] is False
+    assert hyde_row["status"] == "disabled"
+
+
 def test_ping_models_calls_each_enabled_model_once(client):
     """POST /status/models/ping must invoke chat()/embed_query() and return fresh statuses."""
     mock_llm_response = LLMResponse(content="pong", model="test-model", usage={"total_tokens": 1})
@@ -90,3 +107,27 @@ def test_ping_models_survives_one_model_failing(client):
 
     assert resp.status_code == 200
     mock_embedding.embed_query.assert_called_once()
+
+
+def test_ping_records_get_llm_client_failure_instead_of_dropping_it(client):
+    """A get_llm_client() failure (raised before any TrackedLLMClient exists) must still
+    be recorded into the tracker, so it is visible afterwards via _build_model_status()
+    instead of being silently discarded by asyncio.gather(return_exceptions=True)."""
+    mock_embedding = MagicMock()
+    mock_embedding.embed_query.return_value = [0.1]
+
+    with patch("app.api.routes.status.get_llm_client", side_effect=RuntimeError("missing api key")), \
+         patch("app.api.routes.status.get_embedding_provider", return_value=mock_embedding), \
+         patch("app.api.routes.status.get_reranker", return_value=None):
+        resp = client.post("/api/v1/status/models/ping")
+
+    assert resp.status_code == 200
+    body = resp.json()
+    main_row = next(m for m in body["models"] if m["role"] == "main_llm")
+    hyde_row = next(m for m in body["models"] if m["role"] == "hyde_llm")
+    assert main_row["status"] == "error"
+    assert main_row["call_count_error"] == 1
+    assert main_row["last_error"] == "missing api key"
+    assert hyde_row["status"] == "error"
+    assert hyde_row["call_count_error"] == 1
+    assert hyde_row["last_error"] == "missing api key"