add tpot to inference gateway exposed metrics

Author: kaushikmitr

pkg/epp/handlers/streamingserver.go

@@ -227,6 +227,7 @@ func (s *StreamingServer) Process(srv extProcPb.ExternalProcessor_ProcessServer)
 						metrics.RecordResponseSizes(reqCtx.Model, reqCtx.ResolvedTargetModel, reqCtx.ResponseSize)
 						metrics.RecordInputTokens(reqCtx.Model, reqCtx.ResolvedTargetModel, reqCtx.Usage.PromptTokens)
 						metrics.RecordOutputTokens(reqCtx.Model, reqCtx.ResolvedTargetModel, reqCtx.Usage.CompletionTokens)
+						metrics.RecordLatencyPerOutputToken(ctx, reqCtx.Model, reqCtx.ResolvedTargetModel, reqCtx.RequestReceivedTimestamp, reqCtx.ResponseCompleteTimestamp, reqCtx.Usage.CompletionTokens)
 					}
 				}
 			}