server : headers cleanup

tools/server/server-common.cpp

@@ -9,6 +9,9 @@
 
 #include "server-common.h"
 
+#include <random>
+#include <sstream>
+
 json format_error_response(const std::string & message, const enum error_type type) {
     std::string type_str;
     int code = 500;

tools/server/server-common.h

@@ -5,27 +5,14 @@
 #include "llama.h"
 #include "chat.h"
 #include "mtmd.h"
-#include "mtmd-helper.h"
 
 #define JSON_ASSERT GGML_ASSERT
 #include <nlohmann/json.hpp>
 
-#include <random>
-#include <sstream>
 #include <string>
 #include <vector>
-#include <memory>
 #include <cinttypes>
 
-// fix problem with std::min and std::max
-#if defined(_WIN32)
-#define WIN32_LEAN_AND_MEAN
-#ifndef NOMINMAX
-#   define NOMINMAX
-#endif
-#include <windows.h>
-#endif
-
 #define DEFAULT_OAICOMPAT_MODEL "gpt-3.5-turbo"
 
 const static std::string build_info("b" + std::to_string(LLAMA_BUILD_NUMBER) + "-" + LLAMA_COMMIT);

tools/server/server-queue.cpp

@@ -3,6 +3,8 @@
 
 #include "log.h"
 
+#include <chrono>
+
 #define QUE_INF(fmt, ...) LOG_INF("que  %12.*s: " fmt, 12, __func__, __VA_ARGS__)
 #define QUE_WRN(fmt, ...) LOG_WRN("que  %12.*s: " fmt, 12, __func__, __VA_ARGS__)
 #define QUE_ERR(fmt, ...) LOG_ERR("que  %12.*s: " fmt, 12, __func__, __VA_ARGS__)

tools/server/server-queue.h

@@ -2,7 +2,6 @@
 
 #include "server-task.h"
 
-#include <chrono>
 #include <condition_variable>
 #include <deque>
 #include <mutex>

tools/server/server.cpp

@@ -10,6 +10,7 @@
 #include "sampling.h"
 #include "speculative.h"
 #include "mtmd.h"
+#include "mtmd-helper.h"
 
 #include <atomic>
 #include <cstddef>
@@ -19,6 +20,15 @@
 #include <thread>
 #include <unordered_set>
 
+// fix problem with std::min and std::max
+#if defined(_WIN32)
+#define WIN32_LEAN_AND_MEAN
+#ifndef NOMINMAX
+#   define NOMINMAX
+#endif
+#include <windows.h>
+#endif
+
 using json = nlohmann::ordered_json;
 
 constexpr int HTTP_POLLING_SECONDS = 1;
@@ -3597,10 +3607,10 @@ private:
     }
 };
 
-std::function<void(int)> shutdown_handler;
+static std::function<void(int)> shutdown_handler;
-std::atomic_flag is_terminating = ATOMIC_FLAG_INIT;
+static std::atomic_flag is_terminating = ATOMIC_FLAG_INIT;
 
-inline void signal_handler(int signal) {
+static inline void signal_handler(int signal) {
     if (is_terminating.test_and_set()) {
         // in case it hangs, we can force terminate the server by hitting Ctrl+C twice
         // this is for better developer experience, we can remove when the server is stable enough
@@ -3769,6 +3779,7 @@ int main(int argc, char ** argv) {
         ctx_server.queue_tasks.terminate();
     };
 
+    // TODO: refactor in common/console
 #if defined (__unix__) || (defined (__APPLE__) && defined (__MACH__))
     struct sigaction sigint_action;
     sigint_action.sa_handler = signal_handler;