[Frontend] Add per-request number of cached token stats (#10174)

vllm/entrypoints/openai/run_batch.py

@@ -78,6 +78,11 @@ def parse_args():
         help="Port number for the Prometheus metrics server "
         "(only needed if enable-metrics is set).",
     )
+    parser.add_argument(
+        "--enable-prompt-tokens-details",
+        action='store_true',
+        default=False,
+        help="If set to True, enable prompt_tokens_details in usage.")
 
     return parser.parse_args()
 
@@ -217,6 +222,7 @@ async def main(args):
         prompt_adapters=None,
         request_logger=request_logger,
         chat_template=None,
+        enable_prompt_tokens_details=args.enable_prompt_tokens_details,
     ) if model_config.task == "generate" else None
     openai_serving_embedding = OpenAIServingEmbedding(
         engine,