changing to be more explicit

src/c++/perf_analyzer/client_backend/openai/openai_client.cc

@@ -63,13 +63,6 @@ namespace openai {
 void
 ChatCompletionRequest::SendResponse(bool is_final, bool is_null)
 {
-  // if final response has already been sent
-  // due to detecting the [DONE]
-  // ignore final response due to request completion
-  if (final_response_sent_) {
-    return;
-  }
-
   final_response_sent_ = is_final;
   response_callback_(new ChatCompletionResult(
       http_code_, std::move(response_buffer_), is_final, is_null, request_id_));
@@ -182,9 +175,11 @@ ChatCompletionClient::AsyncInfer(
     UpdateInferStat(request->timer_);
 
     // Send final response on request completion
-    // Ignored if final response has already been sent
-    // (in the case of seeing [DONE] in streaming case)
-    request->SendResponse(true /* is_final */, false /* is_null */);
+    // if it has not already been sent.
+    // (e.g. in the case of seeing [DONE] in streaming case)
+    if (!request->IsFinalResponseSent()) {
+      request->SendResponse(true /* is_final */, false /* is_null */);
+    }
   };
   std::unique_ptr<HttpRequest> request(new ChatCompletionRequest(
       std::move(completion_callback), std::move(callback), request_id,

src/c++/perf_analyzer/client_backend/openai/openai_client.h

@@ -121,6 +121,7 @@ class ChatCompletionRequest : public HttpRequest {
         request_id_(request_id)
   {
   }
+  bool IsFinalResponseSent() { return final_response_sent_; };
   void SendResponse(bool is_final, bool is_null);
   bool is_stream_{false};
   std::function<void(InferResult*)> response_callback_{nullptr};