Pass headers to openai requests (#575)

Author: nv-hwoo

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

@@ -159,7 +159,8 @@ ChatCompletionClient::ResponseHandler(
 Error
 ChatCompletionClient::AsyncInfer(
     std::function<void(InferResult*)> callback,
-    std::string& serialized_request_body, const std::string& request_id)
+    std::string& serialized_request_body, const std::string& request_id,
+    const Headers& headers)
 {
   if (callback == nullptr) {
     return Error(
@@ -186,7 +187,7 @@ ChatCompletionClient::AsyncInfer(
       serialized_request_body.size());
 
   CURL* multi_easy_handle = curl_easy_init();
-  Error err = PreRunProcessing(multi_easy_handle, raw_request);
+  Error err = PreRunProcessing(multi_easy_handle, raw_request, headers);
   if (!err.IsOk()) {
     curl_easy_cleanup(multi_easy_handle);
     return err;
@@ -200,7 +201,7 @@ ChatCompletionClient::AsyncInfer(
 
 Error
 ChatCompletionClient::PreRunProcessing(
-    CURL* curl, ChatCompletionRequest* request)
+    CURL* curl, ChatCompletionRequest* request, const Headers& headers)
 {
   curl_easy_setopt(curl, CURLOPT_URL, url_.c_str());
   curl_easy_setopt(curl, CURLOPT_USERAGENT, "libcurl-agent/1.0");
@@ -235,6 +236,12 @@ ChatCompletionClient::PreRunProcessing(
   struct curl_slist* list = nullptr;
   list = curl_slist_append(list, "Expect:");
   list = curl_slist_append(list, "Content-Type: application/json");
+
+  for (const auto& pr : headers) {
+    std::string hdr = pr.first + ": " + pr.second;
+    list = curl_slist_append(list, hdr.c_str());
+  }
+
   curl_easy_setopt(curl, CURLOPT_HTTPHEADER, list);
 
   // The list will be freed when the request is destructed

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

@@ -156,13 +156,15 @@ class ChatCompletionClient : public HttpClient {
   /// with a OpenAI-compatible server in both streaming and non-streaming case.
   Error AsyncInfer(
       std::function<void(InferResult*)> callback,
-      std::string& serialized_request_body, const std::string& request_id);
+      std::string& serialized_request_body, const std::string& request_id,
+      const Headers& headers);
 
   const InferStat& ClientInferStat() { return infer_stat_; }
 
  private:
   // setup curl handle
-  Error PreRunProcessing(CURL* curl, ChatCompletionRequest* request);
+  Error PreRunProcessing(
+      CURL* curl, ChatCompletionRequest* request, const Headers& headers);
 
   static size_t ResponseHandler(
       void* contents, size_t size, size_t nmemb, void* userp);

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

@@ -67,7 +67,8 @@ OpenAiClientBackend::AsyncInfer(
   auto raw_input = dynamic_cast<OpenAiInferInput*>(inputs[0]);
   raw_input->PrepareForRequest();
   RETURN_IF_CB_ERROR(http_client_->AsyncInfer(
-      callback, raw_input->GetRequestBody(), options.request_id_));
+      callback, raw_input->GetRequestBody(), options.request_id_,
+      *http_headers_));
   return Error::Success;
 }