don't allow accessing /models during sleep, it is not thread-safe

Author: ngxson

tools/server/README.md

@@ -1630,7 +1630,6 @@ When the server enters sleep mode, the model and its associated memory (includin
 Note that the following endpoints are exempt from being considered as incoming tasks. They do not trigger model reloading and do not reset the idle timer:
 - `GET /health`
 - `GET /props`
-- `GET /models` and `GET /v1/models`
 
 ## More examples
 

tools/server/server-context.cpp

@@ -3436,8 +3436,10 @@ void server_routes::init_routes() {
         return res;
     };
 
+    // TODO: this endpoint is unsafe to access during model reloading (i.e. wake up from sleeping)
+    // how to make it work even during load_model()?
     this->get_models = [this](const server_http_req &) {
-        auto res = std::make_unique<server_res_generator>(ctx_server, true);
+        auto res = std::make_unique<server_res_generator>(ctx_server);
         json model_meta = nullptr;
         if (is_ready()) {
             model_meta = ctx_server.json_server_model_meta;