make llama server configurable

plugins/plugin_llama-cpp-web.cpp

@@ -28,11 +28,11 @@ SOLANA_PLUGIN_EXPORT uint32_t solana_plugin_start(struct SolanaAPI* solana_api)
 	}
 
 	try {
-		//auto* conf = PLUG_RESOLVE_INSTANCE(ConfigModelI);
+		auto* conf = PLUG_RESOLVE_INSTANCE(ConfigModelI);
 
 		// static store, could be anywhere tho
 		// construct with fetched dependencies
-		g_lcw = std::make_unique<LlamaCppWeb>();
+		g_lcw = std::make_unique<LlamaCppWeb>(*conf);
 
 		// register types
 		PLUG_PROVIDE_INSTANCE(LlamaCppWeb, plugin_name, g_lcw.get());

src/solanaceae/llama-cpp-web/llama_cpp_web_impl.cpp

@@ -18,6 +18,11 @@ static std::string convertToSafeGrammarString(std::string_view input) {
 	return res;
 }
 
+LlamaCppWeb::LlamaCppWeb(
+	ConfigModelI& conf
+) : _conf(conf), _cli(_conf.get_string("LlamaCppWeb", "server").value_or("localhost:8080")) {
+}
+
 LlamaCppWeb::~LlamaCppWeb(void) {
 }
 

src/solanaceae/llama-cpp-web/llama_cpp_web_impl.hpp

@@ -2,6 +2,8 @@
 
 #include "./text_completion_interface.hpp"
 
+#include <solanaceae/util/config_model.hpp>
+
 #include <httplib.h>
 #include <nlohmann/json_fwd.hpp>
 
@@ -9,14 +11,19 @@
 #include <atomic>
 
 struct LlamaCppWeb : public TextCompletionI {
-	// this mutex locks internally
-	httplib::Client _cli{"http://localhost:8080"};
+	ConfigModelI& _conf;
+
+	// this mutex-locks internally
+	httplib::Client _cli;
 
 	// this is a bad idea
 	static std::minstd_rand thread_local _rng;
 
 	std::atomic<bool> _use_server_cache {true};
 
+	LlamaCppWeb(
+		ConfigModelI& conf
+	);
 	~LlamaCppWeb(void);
 
 	bool isGood(void) override;

src/test1.cpp

@@ -1,5 +1,7 @@
 #include <solanaceae/llama-cpp-web/llama_cpp_web_impl.hpp>
 
+#include <solanaceae/util/simple_config_model.hpp>
+
 #include <nlohmann/json.hpp>
 
 #include <iostream>
@@ -9,7 +11,8 @@
 #include <cstdint>
 
 int main(void) {
-	LlamaCppWeb lcw;
+	SimpleConfigModel scm;
+	LlamaCppWeb lcw{scm};
 
 	if (!lcw.isGood()) {
 		std::cerr << lcw._cli.host() << " " << lcw._cli.port() << " endpoint not healthy\n";