Generate summary and save conversation WIP.

bin/assets/plugins/aiassistant.json

@@ -38,7 +38,8 @@
           "display_name": "Claude 3.5 Haiku",
           "max_output_tokens": 8192,
           "max_tokens": 200000,
-          "name": "claude-3-5-haiku-latest"
+          "name": "claude-3-5-haiku-latest",
+          "cheapest": true
         },
         {
           "cache_configuration": null,
@@ -79,7 +80,8 @@
           "display_name": "DeepSeek Chat",
           "max_output_tokens": 8192,
           "max_tokens": 64000,
-          "name": "deepseek-chat"
+          "name": "deepseek-chat",
+          "cheapest": true
         },
         {
           "display_name": "DeepSeek Reasoner",
@@ -121,7 +123,8 @@
         {
           "display_name": "Gemini 2.0 Flash Lite",
           "max_tokens": 1000000,
-          "name": "gemini-2.0-flash-lite-preview"
+          "name": "gemini-2.0-flash-lite-preview",
+          "cheapest": true
         }
       ]
     },
@@ -148,7 +151,8 @@
         {
           "display_name": "mistral-small-latest",
           "max_tokens": 32000,
-          "name": "mistral-small-latest"
+          "name": "mistral-small-latest",
+          "cheapest": true
         },
         {
           "display_name": "open-mistral-nemo",
@@ -190,7 +194,8 @@
         },
         {
           "max_tokens": 128000,
-          "name": "gpt-4o-mini"
+          "name": "gpt-4o-mini",
+          "cheapest": true
         },
         {
           "max_tokens": 200000,
@@ -216,7 +221,8 @@
       "display_name": "xAI",
       "models": [
         {
-          "name": "grok-2-latest"
+          "name": "grok-2-latest",
+          "cheapest": true
         },
         {
           "name": "grok-3-latest"

src/eepp/network/http.cpp

@@ -1109,9 +1109,12 @@ Http::Response Http::downloadRequest( const Http::Request& request, IOStream& wr
 		// Close the connection
 		if ( !mConnection->isKeepAlive() ) {
 			mConnection->disconnect();
-			HttpConnection* connection = mConnection;
-			eeSAFE_DELETE( connection );
-			mConnection = NULL;
+
+			if ( mConnection ) {
+				HttpConnection* connection = mConnection;
+				eeSAFE_DELETE( connection );
+				mConnection = NULL;
+			}
 		}
 	}
 
@@ -1177,9 +1180,11 @@ void Http::AsyncRequest::run() {
 	}
 
 	// The Async Request destroys the socket used to create the request
-	HttpConnection* connection = mHttp->mConnection;
-	eeSAFE_DELETE( connection );
-	mHttp->mConnection = NULL;
+	if ( mHttp->mConnection ) {
+		HttpConnection* connection = mHttp->mConnection;
+		eeSAFE_DELETE( connection );
+		mHttp->mConnection = NULL;
+	}
 
 	mRunning = false;
 }

src/tools/ecode/plugins/aiassistant/aiassistantplugin.cpp

@@ -57,15 +57,23 @@ static std::map<std::string, LLMProvider> parseLLMProviders( const nlohmann::jso
 				if ( modelJson.contains( "display_name" ) ) {
 					model.displayName = modelJson["display_name"].get<std::string>();
 				}
+
 				if ( modelJson.contains( "max_tokens" ) ) {
 					model.maxTokens = modelJson["max_tokens"].get<std::size_t>();
 				}
+
 				if ( modelJson.contains( "max_output_tokens" ) ) {
 					model.maxOutputTokens = modelJson["max_output_tokens"].get<std::size_t>();
 				}
+
 				if ( modelJson.contains( "default_temperature" ) ) {
 					model.defaultTemperature = modelJson["default_temperature"].get<double>();
 				}
+
+				if ( modelJson.contains( "cheapest" ) ) {
+					model.cheapest = modelJson.value( "cheapest", false );
+				}
+
 				if ( modelJson.contains( "cache_configuration" ) &&
 					 !modelJson["cache_configuration"].is_null() ) {
 					const auto& cacheJson = modelJson["cache_configuration"];

src/tools/ecode/plugins/aiassistant/chatui.cpp

@@ -1,6 +1,7 @@
 #include "aiassistantplugin.hpp"
 #include "chatui.hpp"
 
+#include <eepp/system/filesystem.hpp>
 #include <eepp/ui/doc/syntaxdefinitionmanager.hpp>
 #include <eepp/ui/uicodeeditor.hpp>
 #include <eepp/ui/uidropdownlist.hpp>
@@ -14,6 +15,7 @@
 
 #include <nlohmann/json.hpp>
 
+using namespace EE::System;
 using namespace EE::Window;
 
 namespace ecode {
@@ -398,22 +400,36 @@ Uint32 LLMChatUI::onMessage( const NodeMessage* msg ) {
 	return 0;
 }
 
-nlohmann::json LLMChatUI::serializeChat() {
+nlohmann::json LLMChatUI::serializeChat( const LLMModel& model ) {
 	nlohmann::json j = {
-		{ "model", mCurModel.name }, { "stream", true }, { "messages", chatToJson() } };
-	if ( mCurModel.maxOutputTokens )
-		j["max_tokens"] = *mCurModel.maxOutputTokens;
+		{ "model", model.name }, { "stream", true }, { "messages", chatToJson() } };
+	if ( model.maxOutputTokens )
+		j["max_tokens"] = *model.maxOutputTokens;
 	return j;
 }
 
 nlohmann::json LLMChatUI::serialize() {
 	nlohmann::json j;
 	j["uuid"] = mUUID.toString();
-	j["chat"] = serializeChat();
+	j["chat"] = serializeChat( mCurModel );
 	return j;
 }
 
-void unserialize( const nlohmann::json& /*payload*/ ) {}
+void LLMChatUI::unserialize( const nlohmann::json& /*payload*/ ) {}
+
+void LLMChatUI::saveChat() {
+	auto plugin = getPlugin();
+	if ( plugin == nullptr )
+		return;
+
+	std::string pluginsStatePath( plugin->getManager()->getPluginsPath() + "plugins_state" +
+								  FileSystem::getOSSlash() + "aiassistant" +
+								  FileSystem::getOSSlash() );
+	if ( !FileSystem::fileExists( pluginsStatePath ) )
+		FileSystem::makeDir( pluginsStatePath, true );
+	std::string path( pluginsStatePath + mSummary + ".json" );
+	FileSystem::fileWrite( path, serialize().dump( 2 ) );
+}
 
 std::string LLMChatUI::prepareApiUrl( const std::string& apiKey ) {
 	const auto& provider = mProviders[mCurModel.provider];
@@ -441,9 +457,11 @@ void LLMChatUI::doRequest() {
 	UIWidget* chat = addChatUI( LLMChat::Role::Assistant );
 	toggleEnableChats( false );
 
+	auto model = mCurModel;
 	auto* editor = chat->findByClass<UICodeEditor>( "data_ui" );
+	std::string apiUrl( prepareApiUrl( apiKeyStr ) );
 	mRequest = std::make_unique<LLMChatCompletionRequest>(
-		prepareApiUrl( apiKeyStr ), apiKeyStr, serializeChat().dump(), mCurModel.provider );
+		apiUrl, apiKeyStr, serializeChat( model ).dump(), model.provider );
 	mRequest->streamedResponseCb = [this, editor]( const std::string& chunk ) {
 		auto conversation = chunk;
 		editor->runOnMainThread( [this, conversation = std::move( conversation ), editor] {
@@ -452,11 +470,14 @@ void LLMChatUI::doRequest() {
 			resizeToFit( editor );
 		} );
 	};
-	mRequest->doneCb = [this, editor]( const LLMChatCompletionRequest&, Http::Response& response ) {
+	mRequest->doneCb = [this, editor, apiUrl = std::move( apiUrl ),
+						apiKeyStr = std::move( apiKeyStr ), model = std::move( model )](
+						   const LLMChatCompletionRequest&, Http::Response& response ) {
 		auto status = response.getStatus();
 		auto statusDesc = response.getStatusDescription();
 
-		runOnMainThread( [this, editor, status, statusDesc] {
+		runOnMainThread( [this, editor, status, statusDesc, apiUrl = std::move( apiUrl ),
+						  apiKeyStr = std::move( apiKeyStr ), model = std::move( model )] {
 			if ( status != Http::Response::Ok ) {
 				auto resp = nlohmann::json::parse( mRequest->getStream(), nullptr, false );
 				if ( resp.contains( "error" ) && resp["error"].contains( "message" ) ) {
@@ -477,6 +498,38 @@ void LLMChatUI::doRequest() {
 
 			if ( editor->hasFocus() )
 				mChatInput->setFocus();
+
+			if ( !mSummaryRequest && mSummary.empty() && status == Http::Response::Ok ) {
+				static const std::string SummaryPrompt =
+					"Generate a concise 3-7 word title for this conversation, omitting "
+					"punctuation. Go "
+					"straight to the title, without any preamble and prefix like `Here's a concise "
+					"suggestion:...` or `Title:`. Ignore this message for the summary generation.";
+
+				auto jchat = serializeChat( getCheapestModelFromCurrentProvider() );
+
+				jchat["messages"].push_back(
+					{ { "role", LLMChat::roleToString( LLMChat::Role::User ) },
+					  { "content", SummaryPrompt } } );
+
+				auto chatstr = jchat.dump();
+
+				mSummaryRequest = std::make_unique<LLMChatCompletionRequest>(
+					apiUrl, apiKeyStr, chatstr, model.provider );
+
+				mSummaryRequest->doneCb = [this]( const LLMChatCompletionRequest& req,
+												  Http::Response& response ) {
+					auto status = response.getStatus();
+					if ( status == Http::Response::Ok ) {
+						mSummary = req.getResponse();
+						saveChat();
+					}
+					runOnMainThread( [this] { mSummaryRequest.reset(); } );
+				};
+				mSummaryRequest->requestAsync();
+			} else {
+				saveChat();
+			}
 		} );
 	};
 	mRequest->requestAsync();
@@ -612,4 +665,14 @@ AIAssistantPlugin* LLMChatUI::getPlugin() {
 	return nullptr;
 }
 
+const LLMModel& LLMChatUI::getCheapestModelFromCurrentProvider() const {
+	auto providerIt = mProviders.find( mCurModel.provider );
+	if ( providerIt != mProviders.end() ) {
+		for ( const auto& model : providerIt->second.models )
+			if ( model.cheapest )
+				return model;
+	}
+	return mCurModel;
+}
+
 } // namespace ecode

src/tools/ecode/plugins/aiassistant/chatui.hpp

@@ -53,6 +53,7 @@ class LLMChatUI : public UILinearLayout {
 
   protected:
 	UUID mUUID;
+	std::string mSummary;
 	PluginManager* mManager{ nullptr };
 	UIWidget* mChatsList{ nullptr };
 	UICodeEditor* mChatInput{ nullptr };
@@ -62,6 +63,7 @@ class LLMChatUI : public UILinearLayout {
 	UIScrollView* mChatScrollView{ nullptr };
 	UIDropDownList* mModelDDL{ nullptr };
 	std::unique_ptr<LLMChatCompletionRequest> mRequest;
+	std::unique_ptr<LLMChatCompletionRequest> mSummaryRequest;
 	LLMProviders mProviders;
 	LLMModel mCurModel;
 	std::unordered_map<String::HashType, LLMModel> mModelsMap;
@@ -72,7 +74,7 @@ class LLMChatUI : public UILinearLayout {
 
 	void showMsg( String msg );
 
-	nlohmann::json serializeChat();
+	nlohmann::json serializeChat( const LLMModel& model );
 
 	nlohmann::json chatToJson();
 
@@ -101,6 +103,10 @@ class LLMChatUI : public UILinearLayout {
 	void setProviders( LLMProviders&& providers );
 
 	virtual Uint32 onMessage( const NodeMessage* );
+
+	const LLMModel& getCheapestModelFromCurrentProvider() const;
+
+	void saveChat();
 };
 
 } // namespace ecode

src/tools/ecode/plugins/aiassistant/llmchatcompletionrequest.cpp

@@ -103,7 +103,7 @@ void LLMChatCompletionRequest::cancel() {
 	mCancel = true;
 }
 
-const std::string& LLMChatCompletionRequest::getStream() {
+const std::string& LLMChatCompletionRequest::getStream() const {
 	return mStream.getStream();
 }
 

src/tools/ecode/plugins/aiassistant/llmchatcompletionrequest.hpp

@@ -32,7 +32,9 @@ class LLMChatCompletionRequest {
 
 	void cancel();
 
-	const std::string& getStream();
+	const std::string& getStream() const;
+
+	const std::string& getResponse() const { return mResponse; }
 
   protected:
 	URI mUrl;

src/tools/ecode/plugins/aiassistant/protocol.hpp

@@ -22,6 +22,7 @@ struct LLMModel {
 	std::optional<double> defaultTemperature;
 	std::optional<LLMCacheConfiguration> cacheConfiguration;
 	bool isEphemeral{ false };
+	bool cheapest{ false };
 };
 
 struct LLMProvider {