server: fix remote preset handling, add test (#24938)

* server: add test for remote preset

* fix remote preset handling

* fix

* fix test

common/arg.cpp

@@ -301,6 +301,8 @@ static handle_model_result common_params_handle_model(struct common_params_model
                                                       const common_download_opts & opts) {
     handle_model_result result;
 
+    // TODO @ngxson : refactor this into a new common_model_download_context
+
     if (!model.docker_repo.empty()) {
         model.path = common_docker_resolve_model(model.docker_repo);
     } else if (!model.hf_repo.empty()) {
@@ -396,7 +398,7 @@ static bool parse_bool_value(const std::string & value) {
 // CLI argument parsing functions
 //
 
-bool common_params_handle_models(common_params & params, llama_example curr_ex, common_download_callback * callback) {
+bool common_params_handle_models(common_params & params, llama_example curr_ex, const common_params_handle_models_params & handle_params) {
     const bool spec_type_draft_mtp = std::find(params.speculative.types.begin(),
                                          params.speculative.types.end(),
                                          COMMON_SPECULATIVE_TYPE_DRAFT_MTP) != params.speculative.types.end();
@@ -407,9 +409,10 @@ bool common_params_handle_models(common_params & params, llama_example curr_ex,
     opts.skip_download   = params.skip_download;
     opts.download_mtp    = spec_type_draft_mtp;
     opts.download_mmproj = !params.no_mmproj && params.mmproj.path.empty() && params.mmproj.url.empty();
+    opts.preset_only     = handle_params.preset_only;
 
-    if (callback) {
-        opts.callback = callback;
+    if (handle_params.callback) {
+        opts.callback = handle_params.callback;
     }
 
     // sub-models (draft, mmproj, vocoder) are explicitly specified by the user,
@@ -596,7 +599,7 @@ static bool common_params_parse_ex(int argc, char ** argv, common_params_context
 
     if (!skip_model_download) {
         // handle model and download
-        common_params_handle_models(params, ctx_arg.ex);
+        common_params_handle_models(params, ctx_arg.ex, {});
 
         // model is required (except for server)
         // TODO @ngxson : maybe show a list of available models in CLI in this case

common/arg.h

@@ -130,6 +130,11 @@ bool common_params_to_map(int argc, char ** argv, llama_example ex, std::map<com
 // see: https://github.com/ggml-org/llama.cpp/issues/18163
 void common_params_add_preset_options(std::vector<common_arg> & args);
 
+struct common_params_handle_models_params {
+    common_download_callback * callback = nullptr;
+    bool preset_only = false; // if true, only check & download remote preset (for router mode)
+};
+
 // populate model paths (main model, mmproj, etc) from -hf if necessary
 // return true if the model is ready to use
 // throw an exception if there is an error that prevents the model from being used (e.g. network error, model not found, etc)
@@ -137,7 +142,7 @@ void common_params_add_preset_options(std::vector<common_arg> & args);
 bool common_params_handle_models(
     common_params & params,
     llama_example curr_ex,
-    common_download_callback * callback = nullptr);
+    const common_params_handle_models_params & handle_params);
 
 // initialize argument parser context - used by test-arg-parser and preset
 common_params_context common_params_parser_init(common_params & params, llama_example ex, void(*print_usage)(int, char **) = nullptr);

common/download.cpp

@@ -799,6 +799,7 @@ common_download_model_result common_download_model(const common_params_model  &
 
     bool download_mmproj = opts.download_mmproj;
     bool download_mtp = opts.download_mtp;
+    bool preset_only = opts.preset_only;
     bool is_hf = !model.hf_repo.empty();
 
     if (is_hf) {
@@ -806,7 +807,8 @@ common_download_model_result common_download_model(const common_params_model  &
         if (!hf.preset.path.empty()) {
             // if preset.ini exists, only download that file alone
             tasks.push_back({hf.preset.url, hf.preset.local_path});
-        } else {
+        } else if (!preset_only) {
+            // only add other files if we're NOT in preset-only mode (normal run, non-router)
             for (const auto & f : hf.model_files) {
                 tasks.push_back({f.url, f.local_path});
             }

common/download.h

@@ -55,6 +55,7 @@ struct common_download_opts {
     bool skip_download = false; // if true, only validation is performed, common_skip_download_exception may be thrown if the file is missing or invalid
     bool download_mmproj = false;
     bool download_mtp = false;
+    bool preset_only = false; // if true, only check & download remote preset (for router mode)
     common_download_callback * callback = nullptr;
 };