feat: add MLflow inference logging to all Ray Serve apps

- Add mlflow_logger.py: lightweight REST-based MLflow logger (no mlflow dep)
- Instrument serve_llm.py with latency, token counts, tokens/sec metrics
- Instrument serve_embeddings.py with latency, batch_size, total_tokens
- Instrument serve_whisper.py with latency, audio_duration, realtime_factor
- Instrument serve_tts.py with latency, audio_duration, text_chars
- Instrument serve_reranker.py with latency, num_pairs, top_k

ray_serve/serve_reranker.py

@@ -4,10 +4,13 @@ Runs on: drizzt (Radeon 680M iGPU, ROCm) or danilo (Intel i915 iGPU, OpenVINO/IP
 """
 
 import os
+import time
 from typing import Any
 
 from ray import serve
 
+from ray_serve.mlflow_logger import InferenceLogger
+
 
 @serve.deployment(name="RerankerDeployment", num_replicas=1)
 class RerankerDeployment:
@@ -58,6 +61,17 @@ class RerankerDeployment:
 
         print("Reranker model loaded successfully")
 
+        # MLflow metrics
+        self._mlflow = InferenceLogger(
+            experiment_name="ray-serve-reranker",
+            run_name=f"reranker-{self.model_id.split('/')[-1]}",
+            tags={"model.name": self.model_id, "model.framework": "sentence-transformers", "device": self.device},
+            flush_every=10,
+        )
+        self._mlflow.initialize(
+            params={"model_id": self.model_id, "device": self.device, "use_ipex": str(self.use_ipex)}
+        )
+
     async def __call__(self, request: dict[str, Any]) -> dict[str, Any]:
         """
         Handle reranking requests.
@@ -75,6 +89,8 @@ class RerankerDeployment:
             "pairs": [["query", "doc1"], ["query", "doc2"]]
         }
         """
+        _start = time.time()
+
         # Handle pairs format
         if "pairs" in request:
             pairs = request["pairs"]
@@ -89,6 +105,11 @@ class RerankerDeployment:
                     }
                 )
 
+            self._mlflow.log_request(
+                latency_s=time.time() - _start,
+                num_pairs=len(pairs),
+            )
+
             return {
                 "object": "list",
                 "results": results,
@@ -131,6 +152,14 @@ class RerankerDeployment:
         # Apply top_k
         results = results[:top_k]
 
+        # Log to MLflow
+        self._mlflow.log_request(
+            latency_s=time.time() - _start,
+            num_pairs=len(pairs),
+            num_documents=len(documents),
+            top_k=top_k,
+        )
+
         return {
             "object": "list",
             "results": results,