fix: apply ruff fixes to ray_serve package

[ray-serve only]

- Fix whitespace in docstrings
- Add strict=True to zip() calls
- Use ternary operators where appropriate
- Rename unused loop variables

ray-serve/ray_serve/serve_embeddings.py

@@ -4,9 +4,7 @@ Runs on: drizzt (Radeon 680M iGPU, ROCm)
 """
 
 import os
-import time
+from typing import Any
-import uuid
-from typing import Any, Dict, List, Union
 
 from ray import serve
 
@@ -14,8 +12,8 @@ from ray import serve
 @serve.deployment(name="EmbeddingsDeployment", num_replicas=1)
 class EmbeddingsDeployment:
     def __init__(self):
-        from sentence_transformers import SentenceTransformer
         import torch
+        from sentence_transformers import SentenceTransformer
 
         self.model_id = os.environ.get("MODEL_ID", "BAAI/bge-large-en-v1.5")
 
@@ -35,7 +33,7 @@ class EmbeddingsDeployment:
 
         print(f"Model loaded. Embedding dimension: {self.embedding_dim}")
 
-    async def __call__(self, request: Dict[str, Any]) -> Dict[str, Any]:
+    async def __call__(self, request: dict[str, Any]) -> dict[str, Any]:
         """
         Handle OpenAI-compatible embedding requests.
 
@@ -49,10 +47,7 @@ class EmbeddingsDeployment:
         input_data = request.get("input", "")
 
         # Handle both single string and list of strings
-        if isinstance(input_data, str):
+        texts = [input_data] if isinstance(input_data, str) else input_data
-            texts = [input_data]
-        else:
-            texts = input_data
 
         # Generate embeddings
         embeddings = self.model.encode(
@@ -64,12 +59,14 @@ class EmbeddingsDeployment:
         # Build response data
         data = []
         total_tokens = 0
-        for i, (text, embedding) in enumerate(zip(texts, embeddings)):
+        for i, (text, embedding) in enumerate(zip(texts, embeddings, strict=False)):
-            data.append({
+            data.append(
+                {
                     "object": "embedding",
                     "index": i,
                     "embedding": embedding.tolist(),
-            })
+                }
+            )
             total_tokens += len(text.split())
 
         # Return OpenAI-compatible response

ray-serve/ray_serve/serve_llm.py

@@ -6,7 +6,7 @@ Runs on: khelben (Strix Halo 64GB, ROCm)
 import os
 import time
 import uuid
-from typing import Any, Dict, List, Optional
+from typing import Any
 
 from ray import serve
 
@@ -33,7 +33,7 @@ class LLMDeployment:
         self.SamplingParams = SamplingParams
         print(f"Model {self.model_id} loaded successfully")
 
-    async def __call__(self, request: Dict[str, Any]) -> Dict[str, Any]:
+    async def __call__(self, request: dict[str, Any]) -> dict[str, Any]:
         """
         Handle OpenAI-compatible chat completion requests.
 
@@ -51,7 +51,7 @@ class LLMDeployment:
         temperature = request.get("temperature", 0.7)
         max_tokens = request.get("max_tokens", 256)
         top_p = request.get("top_p", 1.0)
-        stop = request.get("stop", None)
+        stop = request.get("stop")
 
         # Convert messages to prompt
         prompt = self._format_messages(messages)
@@ -89,7 +89,7 @@ class LLMDeployment:
             },
         }
 
-    def _format_messages(self, messages: List[Dict[str, str]]) -> str:
+    def _format_messages(self, messages: list[dict[str, str]]) -> str:
         """Format chat messages into a prompt string."""
         formatted = ""
         for msg in messages:

ray-serve/ray_serve/serve_reranker.py

@@ -4,9 +4,7 @@ Runs on: drizzt (Radeon 680M iGPU, ROCm) or danilo (Intel i915 iGPU, OpenVINO/IP
 """
 
 import os
-import time
+from typing import Any
-import uuid
-from typing import Any, Dict, List, Tuple
 
 from ray import serve
 
@@ -14,8 +12,8 @@ from ray import serve
 @serve.deployment(name="RerankerDeployment", num_replicas=1)
 class RerankerDeployment:
     def __init__(self):
-        from sentence_transformers import CrossEncoder
         import torch
+        from sentence_transformers import CrossEncoder
 
         self.model_id = os.environ.get("MODEL_ID", "BAAI/bge-reranker-v2-m3")
         self.use_ipex = False
@@ -24,6 +22,7 @@ class RerankerDeployment:
         # Detect device - check for Intel GPU first via IPEX
         try:
             import intel_extension_for_pytorch as ipex
+
             self.use_ipex = True
             if hasattr(torch, "xpu") and torch.xpu.is_available():
                 self.device = "xpu"
@@ -37,7 +36,7 @@ class RerankerDeployment:
         if not self.use_ipex:
             if torch.cuda.is_available():
                 self.device = "cuda"
-                print(f"Using CUDA/ROCm device")
+                print("Using CUDA/ROCm device")
             else:
                 print("No GPU detected, using CPU")
 
@@ -51,14 +50,15 @@ class RerankerDeployment:
         if self.use_ipex and self.device == "cpu":
             try:
                 import intel_extension_for_pytorch as ipex
+
                 self.model.model = ipex.optimize(self.model.model)
                 print("IPEX CPU optimization applied")
             except Exception as e:
                 print(f"IPEX optimization failed: {e}")
 
-        print(f"Reranker model loaded successfully")
+        print("Reranker model loaded successfully")
 
-    async def __call__(self, request: Dict[str, Any]) -> Dict[str, Any]:
+    async def __call__(self, request: dict[str, Any]) -> dict[str, Any]:
         """
         Handle reranking requests.
 
@@ -81,11 +81,13 @@ class RerankerDeployment:
             scores = self.model.predict(pairs)
 
             results = []
-            for i, (pair, score) in enumerate(zip(pairs, scores)):
+            for i, (_pair, score) in enumerate(zip(pairs, scores, strict=False)):
-                results.append({
+                results.append(
+                    {
                         "index": i,
                         "score": float(score),
-                })
+                    }
+                )
 
             return {
                 "object": "list",
@@ -114,7 +116,7 @@ class RerankerDeployment:
 
         # Create results with indices and scores
         results = []
-        for i, (doc, score) in enumerate(zip(documents, scores)):
+        for i, (doc, score) in enumerate(zip(documents, scores, strict=False)):
             result = {
                 "index": i,
                 "score": float(score),

ray-serve/ray_serve/serve_tts.py

@@ -3,12 +3,10 @@ Ray Serve deployment for Coqui TTS.
 Runs on: elminster (RTX 2070 8GB, CUDA)
 """
 
-import os
-import io
-import time
-import uuid
 import base64
-from typing import Any, Dict, Optional
+import io
+import os
+from typing import Any
 
 from ray import serve
 
@@ -16,8 +14,8 @@ from ray import serve
 @serve.deployment(name="TTSDeployment", num_replicas=1)
 class TTSDeployment:
     def __init__(self):
-        from TTS.api import TTS
         import torch
+        from TTS.api import TTS
 
         self.model_name = os.environ.get("MODEL_NAME", "tts_models/en/ljspeech/tacotron2-DDC")
 
@@ -32,9 +30,9 @@ class TTSDeployment:
         if self.use_gpu:
             self.tts = self.tts.to("cuda")
 
-        print(f"TTS model loaded successfully")
+        print("TTS model loaded successfully")
 
-    async def __call__(self, request: Dict[str, Any]) -> Dict[str, Any]:
+    async def __call__(self, request: dict[str, Any]) -> dict[str, Any]:
         """
         Handle text-to-speech requests.
 
@@ -52,8 +50,8 @@ class TTSDeployment:
         from scipy.io import wavfile
 
         text = request.get("text", "")
-        speaker = request.get("speaker", None)
+        speaker = request.get("speaker")
-        language = request.get("language", None)
+        language = request.get("language")
         speed = request.get("speed", 1.0)
         output_format = request.get("output_format", "wav")
         return_base64 = request.get("return_base64", True)
@@ -79,7 +77,11 @@ class TTSDeployment:
             wav_int16 = (wav * 32767).astype(np.int16)
 
             # Get sample rate from model config
-            sample_rate = self.tts.synthesizer.output_sample_rate if hasattr(self.tts, 'synthesizer') else 22050
+            sample_rate = (
+                self.tts.synthesizer.output_sample_rate
+                if hasattr(self.tts, "synthesizer")
+                else 22050
+            )
 
             # Write to buffer
             buffer = io.BytesIO()
@@ -106,10 +108,10 @@ class TTSDeployment:
                 "model": self.model_name,
             }
 
-    def list_speakers(self) -> Dict[str, Any]:
+    def list_speakers(self) -> dict[str, Any]:
         """List available speakers for multi-speaker models."""
         speakers = []
-        if hasattr(self.tts, 'speakers') and self.tts.speakers:
+        if hasattr(self.tts, "speakers") and self.tts.speakers:
             speakers = self.tts.speakers
 
         return {

ray-serve/ray_serve/serve_whisper.py

@@ -3,12 +3,10 @@ Ray Serve deployment for faster-whisper STT.
 Runs on: elminster (RTX 2070 8GB, CUDA)
 """
 
-import os
-import io
-import time
-import uuid
 import base64
-from typing import Any, Dict, Optional
+import io
+import os
+from typing import Any
 
 from ray import serve
 
@@ -16,8 +14,8 @@ from ray import serve
 @serve.deployment(name="WhisperDeployment", num_replicas=1)
 class WhisperDeployment:
     def __init__(self):
-        from faster_whisper import WhisperModel
         import torch
+        from faster_whisper import WhisperModel
 
         self.model_size = os.environ.get("MODEL_SIZE", "large-v3")
 
@@ -38,9 +36,9 @@ class WhisperDeployment:
             compute_type=self.compute_type,
         )
 
-        print(f"Whisper model loaded successfully")
+        print("Whisper model loaded successfully")
 
-    async def __call__(self, request: Dict[str, Any]) -> Dict[str, Any]:
+    async def __call__(self, request: dict[str, Any]) -> dict[str, Any]:
         """
         Handle transcription requests.
 
@@ -60,10 +58,8 @@ class WhisperDeployment:
             ...
         }
         """
-        import numpy as np
-        from scipy.io import wavfile
 
-        language = request.get("language", None)
+        language = request.get("language")
         task = request.get("task", "transcribe")  # transcribe or translate
         response_format = request.get("response_format", "json")
         word_timestamps = request.get("word_timestamps", False)