server : (webui) let server send locally-defined default webui settings

common/arg.cpp

@@ -3465,6 +3465,14 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
         }
     ).set_examples({LLAMA_EXAMPLE_SERVER}));
 
+    add_opt(common_arg(
+        {"--default-client-config"}, "JSON_FNAME",
+        string_format("JSON file containing the default client config"),
+        [](common_params & params, const std::string & value) {
+            params.public_default_client_config = value;
+        }
+    ).set_examples({LLAMA_EXAMPLE_SERVER}).set_env("LLAMA_ARG_DEFAULT_CLIENT_CONFIG"));
+
     add_opt(common_arg(
         { "--diffusion-steps" }, "N",
         string_format("number of diffusion steps (default: %d)", params.diffusion.steps),
@@ -3510,6 +3518,5 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
         [](common_params & params, const std::string & value) { params.diffusion.add_gumbel_noise = std::stof(value); }
     ).set_examples({ LLAMA_EXAMPLE_DIFFUSION }));
 
-
     return ctx_arg;
 }

common/common.h

@@ -390,6 +390,7 @@ struct common_params {
 
     std::string hostname      = "127.0.0.1";
     std::string public_path   = "";                                                                         // NOLINT
+    std::string public_default_client_config = "";                                                          // NOLINT
     std::string api_prefix    = "";                                                                         // NOLINT
     std::string chat_template = "";                                                                         // NOLINT
     bool use_jinja = false;                                                                                 // NOLINT

tools/server/server.cpp

@@ -24,6 +24,7 @@
 #include <cstddef>
 #include <cinttypes>
 #include <deque>
+#include <fstream>
 #include <memory>
 #include <mutex>
 #include <signal.h>
@@ -1945,6 +1946,7 @@ struct server_context {
     // slots / clients
     std::vector<server_slot> slots;
     json default_generation_settings_for_props;
+    json default_client_config = json::object();
 
     server_queue    queue_tasks;
     server_response queue_results;
@@ -2138,6 +2140,15 @@ struct server_context {
 
         default_generation_settings_for_props = slots[0].to_json();
 
+        if (!params_base.public_default_client_config.empty()) {
+            std::ifstream file(params_base.public_default_client_config);
+            LOG_INF("%s: Loading default client config from %s\n", __func__, params_base.public_default_client_config.c_str());
+            if (!file.is_open()) {
+                throw std::runtime_error("Error: default client config file not open");
+            }
+            file >> default_client_config;
+        }
+
         // the update_slots() logic will always submit a maximum of n_batch or n_parallel tokens
         // note that n_batch can be > n_ctx (e.g. for non-causal attention models such as BERT where the KV cache is not used)
         {
@@ -4157,6 +4168,7 @@ int main(int argc, char ** argv) {
             { "bos_token",                   common_token_to_piece(ctx_server.ctx, llama_vocab_bos(ctx_server.vocab), /* special= */ true)},
             { "eos_token",                   common_token_to_piece(ctx_server.ctx, llama_vocab_eos(ctx_server.vocab), /* special= */ true)},
             { "build_info",                  build_info },
+            { "default_client_config",       ctx_server.default_client_config },
         };
         if (ctx_server.params_base.use_jinja) {
             if (auto tool_use_src = common_chat_templates_source(ctx_server.chat_templates.get(), "tool_use")) {

tools/server/webui/src/components/SettingDialog.tsx

@@ -1,4 +1,4 @@
-import { useState } from 'react';
+import { useState, useEffect } from 'react';
 import { useAppContext } from '../utils/app.context';
 import { CONFIG_DEFAULT, CONFIG_INFO } from '../Config';
 import { isDev } from '../Config';
@@ -276,7 +276,7 @@ export default function SettingDialog({
   show: boolean;
   onClose: () => void;
 }) {
-  const { config, saveConfig } = useAppContext();
+  const { config, saveConfig, serverProps } = useAppContext();
   const [sectionIdx, setSectionIdx] = useState(0);
 
   // clone the config object to prevent direct mutation
@@ -285,6 +285,25 @@ export default function SettingDialog({
   );
   const { showConfirm, showAlert } = useModals();
 
+  // get default client settings
+  useEffect(() => {
+    if (
+      serverProps &&
+      serverProps.default_client_config &&
+      Object.keys(serverProps.default_client_config).length > 0
+    ) {
+      if (StorageUtils.setDefaultConfig(serverProps.default_client_config)) {
+        console.log(
+          'Setting default config:',
+          serverProps.default_client_config
+        );
+        const newConfig = StorageUtils.getConfig();
+        saveConfig(newConfig);
+        setLocalConfig(JSON.parse(JSON.stringify(newConfig)));
+      }
+    }
+  }, [serverProps, saveConfig]);
+
   const resetConfig = async () => {
     if (await showConfirm('Are you sure you want to reset all settings?')) {
       setLocalConfig(CONFIG_DEFAULT);

tools/server/webui/src/utils/storage.ts

@@ -213,6 +213,20 @@ const StorageUtils = {
       localStorage.setItem('theme', theme);
     }
   },
+  setDefaultConfig(defaultConfig: object | null | undefined): boolean {
+    if (localStorage.getItem('config') === null) {
+      try {
+        // Ensure there still is no config when we overwrite it
+        if (localStorage.getItem('config') === null) {
+          localStorage.setItem('config', JSON.stringify(defaultConfig));
+        }
+        return true;
+      } catch (e) {
+        console.error(e);
+      }
+    }
+    return false;
+  },
 };
 
 export default StorageUtils;

tools/server/webui/src/utils/types.ts

@@ -134,5 +134,6 @@ export interface LlamaCppServerProps {
     vision: boolean;
     audio: boolean;
   };
+  default_client_config: Record<string, number | string | boolean>;
   // TODO: support params
 }