fix: guardrails ports, disable auth (for now) for LLS LMEval tests, fix eval api (#780)

Author: adolfo-ab

tests/fixtures/inference.py

@@ -66,7 +66,7 @@ def qwen_isvc(
         storage_key=minio_data_connection.name,
         storage_path="Qwen2.5-0.5B-Instruct",
         wait_for_predictor_pods=False,
-        enable_auth=True,
+        enable_auth=False,
         resources={
             "requests": {"cpu": "2", "memory": "10Gi"},
             "limits": {"cpu": "2", "memory": "12Gi"},
@@ -77,7 +77,7 @@ def qwen_isvc(
 
 @pytest.fixture(scope="class")
 def qwen_isvc_url(qwen_isvc: InferenceService) -> str:
-    return f"http://{qwen_isvc.name}-predictor.{qwen_isvc.namespace}.svc.cluster.local:8032/v1"
+    return f"http://{qwen_isvc.name}-predictor.{qwen_isvc.namespace}.svc.cluster.local:80/v1"
 
 
 @pytest.fixture(scope="class")

tests/llama_stack/eval/test_lmeval_provider.py

@@ -55,7 +55,7 @@ def test_lmeval_register_benchmark(self, minio_pod, minio_data_connection, llama
     def test_llamastack_run_eval(
         self, minio_pod, minio_data_connection, patched_dsc_lmeval_allow_all, llama_stack_client
     ):
-        job = llama_stack_client.eval.run_eval(
+        job = llama_stack_client.alpha.eval.run_eval(
             benchmark_id=TRUSTYAI_LMEVAL_ARCEASY,
             benchmark_config={
                 "eval_candidate": {
@@ -65,14 +65,14 @@ def test_llamastack_run_eval(
                     "sampling_params": {"temperature": 0.7, "top_p": 0.9, "max_tokens": 10},
                 },
                 "scoring_params": {},
-                "num_examples": 2,
+                "num_examples": 1,
             },
         )
 
         samples = TimeoutSampler(
             wait_timeout=Timeout.TIMEOUT_10MIN,
             sleep=30,
-            func=lambda: llama_stack_client.eval.jobs.status(
+            func=lambda: llama_stack_client.alpha.eval.jobs.status(
                 job_id=job.job_id, benchmark_id=TRUSTYAI_LMEVAL_ARCEASY
             ).status,
         )

tests/model_explainability/guardrails/test_guardrails.py

@@ -244,7 +244,7 @@ def test_guardrails_builtin_detectors_negative_detection(
                                 "type": "text_contents",
                                 "service": {
                                     "hostname": f"{PROMPT_INJECTION_DETECTOR}-predictor",
-                                    "port": 8000,
+                                    "port": 80,
                                 },
                                 "chunker_id": "whole_doc_chunker",
                                 "default_threshold": 0.5,
@@ -253,7 +253,7 @@ def test_guardrails_builtin_detectors_negative_detection(
                                 "type": "text_contents",
                                 "service": {
                                     "hostname": f"{HAP_DETECTOR}-predictor",
-                                    "port": 8000,
+                                    "port": 80,
                                 },
                                 "chunker_id": "whole_doc_chunker",
                                 "default_threshold": 0.5,

utilities/constants.py

@@ -477,7 +477,7 @@ class LLMdInferenceSimConfig:
 
 
 LLM_D_CHAT_GENERATION_CONFIG: Dict[str, Any] = {
-    "service": {"hostname": f"{LLMdInferenceSimConfig.isvc_name}-predictor", "port": 8032}
+    "service": {"hostname": f"{LLMdInferenceSimConfig.isvc_name}-predictor", "port": 80}
 }