fix(consus): forward reasoning_effort + register in support manifest

Addresses two points of feedback on the open PR:

1. Override `custom_llm_provider` and `get_supported_openai_params`
   on `ConsusChatConfig` so `reasoning_effort` is forwarded for
   models with `supports_reasoning: true` instead of being silently
   dropped by the parent's default param filter. Mirrors xAI's
   pattern (Greptile review).

2. Add a `consus` entry to `provider_endpoints_support.json` so
   the documentation-coverage guard in CI's `code-quality` job
   passes.

Tests: new `TestConsusReasoningSupport` class with 3 cases
verifying `reasoning_effort` is advertised for Claude (reasoning),
suppressed for GPT-4.1 (non-reasoning), and that
`custom_llm_provider` returns `"consus"`. All 17 Consus tests pass.

litellm/llms/consus/chat/transformation.py

@@ -18,10 +18,25 @@ CONSUS_API_BASE = "https://api.consus.io/v1"
 
 
 class ConsusChatConfig(OpenAIGPTConfig):
+    @property
+    def custom_llm_provider(self) -> Optional[str]:
+        return "consus"
+
     @staticmethod
     def _resolve_api_key(api_key: Optional[str]) -> Optional[str]:
         return api_key or litellm.consus_key or get_secret_str("CONSUS_API_KEY")
 
+    def get_supported_openai_params(self, model: str) -> list:
+        base_params = super().get_supported_openai_params(model)
+        try:
+            if litellm.supports_reasoning(
+                model=model, custom_llm_provider=self.custom_llm_provider
+            ):
+                base_params.append("reasoning_effort")
+        except Exception:
+            pass
+        return base_params
+
     def _get_openai_compatible_provider_info(
         self, api_base: Optional[str], api_key: Optional[str]
     ) -> Tuple[Optional[str], Optional[str]]:

provider_endpoints_support.json

@@ -635,6 +635,24 @@
         "interactions": true
       }
     },
+    "consus": {
+      "display_name": "Consus (`consus`)",
+      "url": "https://docs.litellm.ai/docs/providers/consus",
+      "endpoints": {
+        "chat_completions": true,
+        "messages": true,
+        "responses": true,
+        "embeddings": false,
+        "image_generations": false,
+        "audio_transcriptions": false,
+        "audio_speech": false,
+        "moderations": false,
+        "batches": false,
+        "rerank": false,
+        "a2a": true,
+        "interactions": true
+      }
+    },
     "crusoe": {
       "display_name": "Crusoe (`crusoe`)",
       "url": "https://docs.litellm.ai/docs/providers/crusoe",

tests/test_litellm/llms/consus/chat/test_consus_chat_transformation.py

@@ -141,3 +141,26 @@ class TestConsusModelRouting:
         model, provider, _, _ = litellm.get_llm_provider("consus/gemini-2-5-pro:il5")
         assert provider == "consus"
         assert model == "gemini-2-5-pro:il5"
+
+
+class TestConsusReasoningSupport:
+    """Reasoning-capable Consus models must surface `reasoning_effort` as a
+    supported OpenAI param so it isn't silently filtered out before the
+    request leaves LiteLLM. Models without `supports_reasoning: true` in
+    the catalog must NOT advertise the param.
+    """
+
+    def test_reasoning_effort_supported_for_claude_models(self):
+        config = ConsusChatConfig()
+        params = config.get_supported_openai_params("claude-sonnet-4-5:il2")
+        assert "reasoning_effort" in params
+
+    def test_reasoning_effort_not_supported_for_gpt_4_1(self):
+        # gpt-4.1 is registered in the Consus catalog without
+        # `supports_reasoning`, so the param must NOT be advertised.
+        config = ConsusChatConfig()
+        params = config.get_supported_openai_params("gpt-4.1:il5+itar")
+        assert "reasoning_effort" not in params
+
+    def test_custom_llm_provider_returns_consus(self):
+        assert ConsusChatConfig().custom_llm_provider == "consus"