[Feat][CLI] enforce-include-usage (#19695)

Signed-off-by: Max Wittig <max.wittig@siemens.com>

vllm/entrypoints/openai/serving_chat.py

@@ -64,12 +64,14 @@ class OpenAIServingChat(OpenAIServing):
         enable_auto_tools: bool = False,
         tool_parser: Optional[str] = None,
         enable_prompt_tokens_details: bool = False,
+        enable_force_include_usage: bool = False,
     ) -> None:
         super().__init__(engine_client=engine_client,
                          model_config=model_config,
                          models=models,
                          request_logger=request_logger,
-                         return_tokens_as_token_ids=return_tokens_as_token_ids)
+                         return_tokens_as_token_ids=return_tokens_as_token_ids,
+                         enable_force_include_usage=enable_force_include_usage)
 
         self.response_role = response_role
         self.chat_template = chat_template
@@ -110,6 +112,7 @@ class OpenAIServingChat(OpenAIServing):
                                 "been registered") from e
 
         self.enable_prompt_tokens_details = enable_prompt_tokens_details
+        self.enable_force_include_usage = enable_force_include_usage
         self.default_sampling_params = (
             self.model_config.get_diff_sampling_param())
         if self.default_sampling_params:
@@ -261,8 +264,14 @@ class OpenAIServingChat(OpenAIServing):
         # Streaming response
         if request.stream:
             return self.chat_completion_stream_generator(
-                request, result_generator, request_id, model_name,
-                conversation, tokenizer, request_metadata)
+                request,
+                result_generator,
+                request_id,
+                model_name,
+                conversation,
+                tokenizer,
+                request_metadata,
+                enable_force_include_usage=self.enable_force_include_usage)
 
         try:
             return await self.chat_completion_full_generator(
@@ -405,6 +414,7 @@ class OpenAIServingChat(OpenAIServing):
         conversation: list[ConversationMessage],
         tokenizer: AnyTokenizer,
         request_metadata: RequestResponseMetadata,
+        enable_force_include_usage: bool,
     ) -> AsyncGenerator[str, None]:
         created_time = int(time.time())
         chunk_object_type: Final = "chat.completion.chunk"
@@ -471,7 +481,8 @@ class OpenAIServingChat(OpenAIServing):
 
         stream_options = request.stream_options
         if stream_options:
-            include_usage = stream_options.include_usage
+            include_usage = stream_options.include_usage \
+                            or enable_force_include_usage
             include_continuous_usage = include_usage and \
                                        stream_options.continuous_usage_stats
         else: