server: log prompts to directory (#22031)

* server: log prompts to directory

Add `--log-prompts-dir` to write each prompt to a separate text file in
the specified directory.

* Apply suggestion from @ngxson

---------

Co-authored-by: Xuan-Son Nguyen <thichthat@gmail.com>

common/arg.cpp

@@ -3333,6 +3333,13 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
             common_log_set_file(common_log_main(), value.c_str());
         }
     ).set_env("LLAMA_ARG_LOG_FILE"));
+    add_opt(common_arg(
+        {"--log-prompts-dir"}, "PATH",
+        "Log prompts to directory (only used for debugging, default: disabled)",
+        [](common_params & params, const std::string & value) {
+            params.path_prompts_log_dir = value;
+        }
+    ).set_examples({LLAMA_EXAMPLE_SERVER, LLAMA_EXAMPLE_CLI}));
     add_opt(common_arg(
         {"--log-colors"}, "[on|off|auto]",
         "Set colored logging ('on', 'off', or 'auto', default: 'auto')\n"

common/common.h

@@ -489,6 +489,7 @@ struct common_params {
     std::string input_prefix         = ""; // string to prefix user inputs with                             // NOLINT
     std::string input_suffix         = ""; // string to suffix user inputs with                             // NOLINT
     std::string logits_file          = ""; // file for saving *all* logits                                  // NOLINT
+    std::string path_prompts_log_dir = ""; // directory with logged prompts                                 // NOLINT
 
     // llama-debug specific options
     std::string logits_output_dir = "data"; // directory for saving logits output files                     // NOLINT

tools/server/server-context.cpp

@@ -27,6 +27,7 @@
 #include <memory>
 #include <filesystem>
 #include <utility>
+#include <fstream>
 
 // fix problem with std::min and std::max
 #if defined(_WIN32)
@@ -3719,6 +3720,16 @@ std::unique_ptr<server_res_generator> server_routes::handle_completions_impl(
         // TODO: this log can become very long, put it behind a flag or think about a more compact format
         //SRV_DBG("Prompt: %s\n", prompt.is_string() ? prompt.get<std::string>().c_str() : prompt.dump(2).c_str());
 
+        if (!params.path_prompts_log_dir.empty()) {
+            const auto file_path = std::filesystem::path(params.path_prompts_log_dir) / string_format("%012" PRId64 ".txt", ggml_time_ms());
+            std::ofstream f(file_path);
+            if (f) {
+                f << (prompt.is_string() ? prompt.get<std::string>().c_str() : prompt.dump(2).c_str());
+            } else {
+                SRV_ERR("failed to create %s\n", file_path.string().c_str());
+            }
+        }
+
         // process prompt
         std::vector<server_tokens> inputs;
 

tools/ui/scripts/vite-plugin-llama-cpp-build.ts

@@ -46,10 +46,12 @@ export function llamaCppBuildPlugin(): Plugin {
 
 					content = content.replace(/\r/g, '');
 					content = GUIDE_FOR_FRONTEND + '\n' + content;
-					content = content.replace(/\/_app\/immutable\/bundle\.[^"]+\.js/g, './bundle.js');
+
+					// Keep the Vite hash as a query string so each build busts the browser cache
+					content = content.replace(/\/_app\/immutable\/bundle\.([^".]+)\.js/g, './bundle.js?$1');
 					content = content.replace(
-						/\/_app\/immutable\/assets\/bundle\.[^"]+\.css/g,
-						'./bundle.css'
+						/\/_app\/immutable\/assets\/bundle\.([^".]+)\.css/g,
+						'./bundle.css?$1'
 					);
 					content = content.replace(/__sveltekit_[a-z0-9]+/g, '__sveltekit__');
 

tools/ui/src/routes/+layout.svelte

@@ -254,7 +254,7 @@
 	/>
 
 	<Sidebar.Provider bind:open={sidebarOpen}>
-		<div class="flex h-screen w-full">
+		<div class="flex h-dvh w-full">
 			<Sidebar.Root variant="floating" class="h-full"
 				><SidebarNavigation bind:this={chatSidebar} /></Sidebar.Root
 			>