[Misc] Streamline unique id generation (#29375)

Signed-off-by: Nick Hill <nhill@redhat.com>

vllm/entrypoints/openai/protocol.py

@@ -661,7 +661,7 @@ class ChatCompletionRequest(OpenAIBaseModel):
         ),
     )
     request_id: str = Field(
-        default_factory=lambda: f"{random_uuid()}",
+        default_factory=random_uuid,
         description=(
             "The request_id related to this request. If the caller does "
             "not set it, a random_uuid will be generated. This id is used "
@@ -1078,7 +1078,7 @@ class CompletionRequest(OpenAIBaseModel):
         ),
     )
     request_id: str = Field(
-        default_factory=lambda: f"{random_uuid()}",
+        default_factory=random_uuid,
         description=(
             "The request_id related to this request. If the caller does "
             "not set it, a random_uuid will be generated. This id is used "
@@ -1375,7 +1375,7 @@ class EmbeddingCompletionRequest(OpenAIBaseModel):
         ),
     )
     request_id: str = Field(
-        default_factory=lambda: f"{random_uuid()}",
+        default_factory=random_uuid,
         description=(
             "The request_id related to this request. If the caller does "
             "not set it, a random_uuid will be generated. This id is used "
@@ -1470,7 +1470,7 @@ class EmbeddingChatRequest(OpenAIBaseModel):
         ),
     )
     request_id: str = Field(
-        default_factory=lambda: f"{random_uuid()}",
+        default_factory=random_uuid,
         description=(
             "The request_id related to this request. If the caller does "
             "not set it, a random_uuid will be generated. This id is used "
@@ -1892,7 +1892,7 @@ class ClassificationCompletionRequest(OpenAIBaseModel):
         ),
     )
     request_id: str = Field(
-        default_factory=lambda: f"{random_uuid()}",
+        default_factory=random_uuid,
         description=(
             "The request_id related to this request. If the caller does "
             "not set it, a random_uuid will be generated. This id is used "
@@ -1983,7 +1983,7 @@ class ClassificationChatRequest(OpenAIBaseModel):
     )
 
     request_id: str = Field(
-        default_factory=lambda: f"{random_uuid()}",
+        default_factory=random_uuid,
         description=(
             "The request_id related to this request. If the caller does "
             "not set it, a random_uuid will be generated. This id is used "
@@ -3094,7 +3094,7 @@ class TranslationResponseVerbose(OpenAIBaseModel):
 ####### Tokens IN <> Tokens OUT #######
 class GenerateRequest(BaseModel):
     request_id: str = Field(
-        default_factory=lambda: f"{random_uuid()}",
+        default_factory=random_uuid,
         description=(
             "The request_id related to this request. If the caller does "
             "not set it, a random_uuid will be generated. This id is used "
@@ -3151,7 +3151,7 @@ class GenerateResponseChoice(BaseModel):
 
 class GenerateResponse(BaseModel):
     request_id: str = Field(
-        default_factory=lambda: f"{random_uuid()}",
+        default_factory=random_uuid,
         description=(
             "The request_id related to this request. If the caller does "
             "not set it, a random_uuid will be generated. This id is used "

vllm/entrypoints/openai/serving_engine.py

@@ -1349,11 +1349,12 @@ class OpenAIServing:
         raw_request: Request | None, default: str | None = None
     ) -> str | None:
         """Pulls the request id to use from a header, if provided"""
-        default = default or random_uuid()
-        if raw_request is None:
-            return default
+        if raw_request is not None and (
+            (req_id := raw_request.headers.get("X-Request-Id")) is not None
+        ):
+            return req_id
 
-        return raw_request.headers.get("X-Request-Id", default)
+        return random_uuid() if default is None else default
 
     @staticmethod
     def _get_data_parallel_rank(raw_request: Request | None) -> int | None:

vllm/utils/__init__.py

@@ -52,9 +52,11 @@ STR_FLASHINFER_ATTN_VAL: str = "FLASHINFER"
 STR_FLASH_ATTN_VAL: str = "FLASH_ATTN"
 STR_INVALID_VAL: str = "INVALID"
 
+MASK_64_BITS = (1 << 64) - 1
+
 
 def random_uuid() -> str:
-    return str(uuid.uuid4().hex)
+    return f"{uuid.uuid4().int & MASK_64_BITS:016x}"  # 16 hex chars
 
 
 def length_from_prompt_token_ids_or_embeds(