server: (router) lazy-load startup_models after main setup

2026-10-02 19:07:25 -05:00 · 2026-08-20 01:56:18 +02:00
parent d59d455fd8
commit c47322cc8a
3 changed files with 34 additions and 4 deletions
@@ -423,6 +423,18 @@ int llama_server(common_params & params, int argc, char ** argv) {
            ctx_http.stop();
        };

+        try {
+            models_routes->models.load_startup_models();
+        } catch (const std::exception & e) {
+            SRV_ERR("failed to load models on startup: %s\n", e.what());
+            ctx_http.stop();
+            if (ctx_http.thread.joinable()) {
+                ctx_http.thread.join();
+            }
+            clean_up();
+            return 1;
+        }
+
    } else {
        // setup clean up function, to be called before exit
        clean_up = [&ctx_http, &ctx_server, &mcp_mgr]() {