diff --git a/controllers/llamaCPP.cc b/controllers/llamaCPP.cc
index cf7bbd328..981712463 100644
--- a/controllers/llamaCPP.cc
+++ b/controllers/llamaCPP.cc
@@ -383,7 +383,7 @@ bool llamaCPP::loadModelImpl(const Json::Value &jsonBody) {
   gpt_params params;
 
   // By default will setting based on number of handlers
-  int drogon_thread = drogon::app().getThreadNum();
+  int drogon_thread = drogon::app().getThreadNum() - 1;
   LOG_INFO << "Drogon thread is:" << drogon_thread;
   if (jsonBody) {
     params.model = jsonBody["llama_model_path"].asString();
diff --git a/controllers/processManager.cc b/controllers/processManager.cc
new file mode 100644
index 000000000..13849ac93
--- /dev/null
+++ b/controllers/processManager.cc
@@ -0,0 +1,11 @@
+#include "processManager.h"
+#include <cstdlib>
+#include <trantor/utils/Logger.h>
+
+void processManager::destroy(const HttpRequestPtr &req,
+             std::function<void(const HttpResponsePtr &)> &&callback) {
+  LOG_INFO << "Program is exitting, goodbye!";
+  exit(0);
+  return;
+};
+// Add definition of your processing function here
diff --git a/controllers/processManager.h b/controllers/processManager.h
new file mode 100644
index 000000000..dcc4b47f6
--- /dev/null
+++ b/controllers/processManager.h
@@ -0,0 +1,23 @@
+#pragma once
+
+#include <drogon/HttpController.h>
+#include <drogon/HttpTypes.h>
+
+using namespace drogon;
+
+class processManager : public drogon::HttpController<processManager> {
+public:
+  METHOD_LIST_BEGIN
+  // use METHOD_ADD to add your custom processing function here;
+  // METHOD_ADD(processManager::get, "/{2}/{1}", Get); // path is
+  // /processManager/{arg2}/{arg1}
+  METHOD_ADD(processManager::destroy, "/destroy",
+             Delete); // path is /processManager/{arg1}/{arg2}/list
+
+  METHOD_LIST_END
+  // your declaration of processing function maybe like this:
+  void destroy(const HttpRequestPtr &req,
+               std::function<void(const HttpResponsePtr &)> &&callback);
+  // void your_method_name(const HttpRequestPtr& req, std::function<void (const
+  // HttpResponsePtr &)> &&callback, double p1, int p2) const;
+};
diff --git a/llama.cpp b/llama.cpp
index fbbc42827..fecac4565 160000
--- a/llama.cpp
+++ b/llama.cpp
@@ -1 +1 @@
-Subproject commit fbbc42827b2949b95bcde23ce47bb47d006c895d
+Subproject commit fecac45658a99eddc4d6e36ba0310ca8f87a77f0
diff --git a/main.cc b/main.cc
index 40cbc1375..a1c6187fd 100644
--- a/main.cc
+++ b/main.cc
@@ -39,7 +39,7 @@ int main(int argc, char *argv[]) {
   LOG_INFO << "Server started, listening at: " << host << ":" << port;
   LOG_INFO << "Please load your model";
   drogon::app().addListener(host, port);
-  drogon::app().setThreadNum(thread_num);
+  drogon::app().setThreadNum(thread_num + 1);
   LOG_INFO << "Number of thread is:" << drogon::app().getThreadNum();
 
   drogon::app().run();