fix: make mlflow_logger import optional with no-op fallback

The strixhalo LLM worker uses py_executable pointing to the Docker
image venv which doesn't have the updated ray-serve-apps package.
Wrap all InferenceLogger imports in try/except and guard usage with
None checks so apps degrade gracefully without MLflow logging.

ray_serve/serve_embeddings.py

@@ -9,7 +9,10 @@ from typing import Any
 
 from ray import serve
 
-from ray_serve.mlflow_logger import InferenceLogger
+try:
+    from ray_serve.mlflow_logger import InferenceLogger
+except ImportError:
+    InferenceLogger = None
 
 
 @serve.deployment(name="EmbeddingsDeployment", num_replicas=1)
@@ -37,6 +40,7 @@ class EmbeddingsDeployment:
         print(f"Model loaded. Embedding dimension: {self.embedding_dim}")
 
         # MLflow metrics
+        if InferenceLogger is not None:
             self._mlflow = InferenceLogger(
                 experiment_name="ray-serve-embeddings",
                 run_name=f"embeddings-{self.model_id.split('/')[-1]}",
@@ -46,6 +50,8 @@ class EmbeddingsDeployment:
             self._mlflow.initialize(
                 params={"model_id": self.model_id, "embedding_dim": str(self.embedding_dim), "device": self.device}
             )
+        else:
+            self._mlflow = None
 
     async def __call__(self, request: dict[str, Any]) -> dict[str, Any]:
         """
@@ -86,6 +92,7 @@ class EmbeddingsDeployment:
             total_tokens += len(text.split())
 
         # Log to MLflow
+        if self._mlflow:
             self._mlflow.log_request(
                 latency_s=time.time() - _start,
                 batch_size=len(texts),

ray_serve/serve_llm.py

@@ -10,7 +10,10 @@ from typing import Any
 
 from ray import serve
 
-from ray_serve.mlflow_logger import InferenceLogger
+try:
+    from ray_serve.mlflow_logger import InferenceLogger
+except ImportError:
+    InferenceLogger = None
 
 
 @serve.deployment(name="LLMDeployment", num_replicas=1)
@@ -40,6 +43,7 @@ class LLMDeployment:
         print(f"Model {self.model_id} async engine created")
 
         # MLflow metrics
+        if InferenceLogger is not None:
             self._mlflow = InferenceLogger(
                 experiment_name="ray-serve-llm",
                 run_name=f"llm-{self.model_id.split('/')[-1]}",
@@ -53,6 +57,8 @@ class LLMDeployment:
                     "gpu_memory_utilization": str(self.gpu_memory_utilization),
                 }
             )
+        else:
+            self._mlflow = None
 
     async def __call__(self, request: dict[str, Any]) -> dict[str, Any]:
         """
@@ -96,6 +102,7 @@ class LLMDeployment:
         completion_tokens = len(generated_text.split())
 
         # Log to MLflow
+        if self._mlflow:
             self._mlflow.log_request(
                 latency_s=latency,
                 prompt_tokens=prompt_tokens,

ray_serve/serve_reranker.py

@@ -9,7 +9,10 @@ from typing import Any
 
 from ray import serve
 
-from ray_serve.mlflow_logger import InferenceLogger
+try:
+    from ray_serve.mlflow_logger import InferenceLogger
+except ImportError:
+    InferenceLogger = None
 
 
 @serve.deployment(name="RerankerDeployment", num_replicas=1)
@@ -62,6 +65,7 @@ class RerankerDeployment:
         print("Reranker model loaded successfully")
 
         # MLflow metrics
+        if InferenceLogger is not None:
             self._mlflow = InferenceLogger(
                 experiment_name="ray-serve-reranker",
                 run_name=f"reranker-{self.model_id.split('/')[-1]}",
@@ -71,6 +75,8 @@ class RerankerDeployment:
             self._mlflow.initialize(
                 params={"model_id": self.model_id, "device": self.device, "use_ipex": str(self.use_ipex)}
             )
+        else:
+            self._mlflow = None
 
     async def __call__(self, request: dict[str, Any]) -> dict[str, Any]:
         """
@@ -105,6 +111,7 @@ class RerankerDeployment:
                     }
                 )
 
+            if self._mlflow:
                 self._mlflow.log_request(
                     latency_s=time.time() - _start,
                     num_pairs=len(pairs),
@@ -153,6 +160,7 @@ class RerankerDeployment:
         results = results[:top_k]
 
         # Log to MLflow
+        if self._mlflow:
             self._mlflow.log_request(
                 latency_s=time.time() - _start,
                 num_pairs=len(pairs),

ray_serve/serve_tts.py

@@ -11,7 +11,10 @@ from typing import Any
 
 from ray import serve
 
-from ray_serve.mlflow_logger import InferenceLogger
+try:
+    from ray_serve.mlflow_logger import InferenceLogger
+except ImportError:
+    InferenceLogger = None
 
 
 @serve.deployment(name="TTSDeployment", num_replicas=1)
@@ -36,6 +39,7 @@ class TTSDeployment:
         print("TTS model loaded successfully")
 
         # MLflow metrics
+        if InferenceLogger is not None:
             self._mlflow = InferenceLogger(
                 experiment_name="ray-serve-tts",
                 run_name=f"tts-{self.model_name.split('/')[-1]}",
@@ -43,6 +47,8 @@ class TTSDeployment:
                 flush_every=5,
             )
             self._mlflow.initialize(params={"model_name": self.model_name, "use_gpu": str(self.use_gpu)})
+        else:
+            self._mlflow = None
 
     async def __call__(self, request: dict[str, Any]) -> dict[str, Any]:
         """
@@ -104,6 +110,7 @@ class TTSDeployment:
             duration = len(wav) / sample_rate
 
             # Log to MLflow
+            if self._mlflow:
                 self._mlflow.log_request(
                     latency_s=time.time() - _start,
                     audio_duration_s=duration,

ray_serve/serve_whisper.py

@@ -11,7 +11,10 @@ from typing import Any
 
 from ray import serve
 
-from ray_serve.mlflow_logger import InferenceLogger
+try:
+    from ray_serve.mlflow_logger import InferenceLogger
+except ImportError:
+    InferenceLogger = None
 
 
 @serve.deployment(name="WhisperDeployment", num_replicas=1)
@@ -42,6 +45,7 @@ class WhisperDeployment:
         print("Whisper model loaded successfully")
 
         # MLflow metrics
+        if InferenceLogger is not None:
             self._mlflow = InferenceLogger(
                 experiment_name="ray-serve-whisper",
                 run_name=f"whisper-{self.model_size}",
@@ -51,6 +55,8 @@ class WhisperDeployment:
             self._mlflow.initialize(
                 params={"model_size": self.model_size, "device": self.device, "compute_type": self.compute_type}
             )
+        else:
+            self._mlflow = None
 
     async def __call__(self, request: dict[str, Any]) -> dict[str, Any]:
         """
@@ -146,6 +152,7 @@ class WhisperDeployment:
             }
 
         # Log to MLflow
+        if self._mlflow:
             self._mlflow.log_request(
                 latency_s=time.time() - _start,
                 audio_duration_s=info.duration,