include llama.cpp since there were some edits

llama.cpp/tools/server/server-connect.h

@@ -0,0 +1,39 @@
+#pragma once
+
+#include "common.h"
+
+#include <atomic>
+#include <memory>
+#include <string>
+#include <thread>
+
+struct common_subproc;
+
+// spawns llama-connect, which exposes this server to a remote browser over WebRTC
+// core binary is in a separate project: https://github.com/ggml-org/llama-connect
+struct server_connect {
+    server_connect();
+    ~server_connect();
+
+    server_connect(const server_connect &) = delete;
+    server_connect & operator=(const server_connect &) = delete;
+
+    // path of the llama-connect binary, empty if not found
+    static std::string find_binary();
+
+    // why --connect cannot work here, empty if it can
+    static std::string unavailable_reason(const common_params & params);
+
+    static std::string resolve_code(const std::string & code);
+
+    // port is the bound one, it differs from params.port with --port 0 or the GCP override
+    bool start(const common_params & params, int port);
+
+    // idempotent, also called by the destructor
+    void stop();
+
+private:
+    std::unique_ptr<common_subproc> proc;
+    std::thread log_thread;
+    std::atomic<bool> stopping{false}; // tells the log thread the exit is expected
+};