/* * Copyright (C) 2024-2025 Petr Mironychev * * This file is part of QodeAssist. * * QodeAssist is free software: you can redistribute it and/or modify * it under the terms of the GNU General Public License as published by * the Free Software Foundation, either version 3 of the License, or * (at your option) any later version. * * QodeAssist is distributed in the hope that it will be useful, * but WITHOUT ANY WARRANTY; without even the implied warranty of * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the * GNU General Public License for more details. * * You should have received a copy of the GNU General Public License * along with QodeAssist. If not, see . */ #include "LlamaCppProvider.hpp" #include #include "logger/Logger.hpp" #include "settings/ChatAssistantSettings.hpp" #include "settings/CodeCompletionSettings.hpp" #include "settings/QuickRefactorSettings.hpp" #include "settings/GeneralSettings.hpp" #include "tools/ToolsRegistration.hpp" #include #include #include namespace QodeAssist::Providers { LlamaCppProvider::LlamaCppProvider(QObject *parent) : PluginLLMCore::Provider(parent) , m_client(new ::LLMCore::LlamaCppClient(url(), apiKey(), QString(), this)) { Tools::registerQodeAssistTools(m_client->tools()); } QString LlamaCppProvider::name() const { return "llama.cpp"; } QString LlamaCppProvider::url() const { return "http://localhost:8080"; } QString LlamaCppProvider::completionEndpoint() const { return "/infill"; } QString LlamaCppProvider::chatEndpoint() const { return "/v1/chat/completions"; } void LlamaCppProvider::prepareRequest( QJsonObject &request, PluginLLMCore::PromptTemplate *prompt, PluginLLMCore::ContextData context, PluginLLMCore::RequestType type, bool isToolsEnabled, bool isThinkingEnabled) { if (!prompt->isSupportProvider(providerID())) { LOG_MESSAGE(QString("Template %1 doesn't support %2 provider").arg(name(), prompt->name())); } prompt->prepareRequest(request, context); auto applyModelParams = [&request](const auto &settings) { request["max_tokens"] = settings.maxTokens(); request["temperature"] = settings.temperature(); if (settings.useTopP()) request["top_p"] = settings.topP(); if (settings.useTopK()) request["top_k"] = settings.topK(); if (settings.useFrequencyPenalty()) request["frequency_penalty"] = settings.frequencyPenalty(); if (settings.usePresencePenalty()) request["presence_penalty"] = settings.presencePenalty(); }; if (type == PluginLLMCore::RequestType::CodeCompletion) { applyModelParams(Settings::codeCompletionSettings()); } else if (type == PluginLLMCore::RequestType::QuickRefactoring) { applyModelParams(Settings::quickRefactorSettings()); } else { applyModelParams(Settings::chatAssistantSettings()); } if (isToolsEnabled) { auto toolsDefinitions = m_client->tools()->getToolsDefinitions(); if (!toolsDefinitions.isEmpty()) { request["tools"] = toolsDefinitions; LOG_MESSAGE(QString("Added %1 tools to llama.cpp request").arg(toolsDefinitions.size())); } } } QFuture> LlamaCppProvider::getInstalledModels(const QString &) { return QtFuture::makeReadyFuture(QList{}); } QString LlamaCppProvider::apiKey() const { return {}; } void LlamaCppProvider::prepareNetworkRequest(QNetworkRequest &networkRequest) const { networkRequest.setHeader(QNetworkRequest::ContentTypeHeader, "application/json"); } PluginLLMCore::ProviderID LlamaCppProvider::providerID() const { return PluginLLMCore::ProviderID::LlamaCpp; } PluginLLMCore::ProviderCapabilities LlamaCppProvider::capabilities() const { return PluginLLMCore::ProviderCapability::Tools | PluginLLMCore::ProviderCapability::Image; } ::LLMCore::BaseClient *LlamaCppProvider::client() const { return m_client; } } // namespace QodeAssist::Providers