Optimize request handler

Palm1r · Sep 17, 2024 · 36f29ca · 36f29ca
1 parent 35c4ac6
commit 36f29ca
Show file tree

Hide file tree

Showing 12 changed files with 30 additions and 30 deletions.
diff --git a/LLMClientInterface.cpp b/LLMClientInterface.cpp
@@ -152,9 +152,14 @@ void LLMClientInterface::handleCompletion(const QJsonObject &request)
     config.promptTemplate = PromptTemplateManager::instance().getCurrentTemplate();
     config.url = QUrl(QString("%1%2").arg(Settings::generalSettings().url(),
                                           Settings::generalSettings().endPoint()));
-    config.modelName = Settings::generalSettings().modelName.value();
 
-    m_requestHandler.sendLLMRequest(config, request, updatedContext);
+    config.providerRequest = {{"model", Settings::generalSettings().modelName.value()},
+                              {"stream", true}};
+
+    config.promptTemplate->prepareRequest(config.providerRequest, updatedContext);
+    config.provider->prepareRequest(config.providerRequest);
+
+    m_requestHandler.sendLLMRequest(config, request);
 }
 
 ContextData LLMClientInterface::prepareContext(const QJsonObject &request,
@@ -183,9 +188,10 @@ ContextData LLMClientInterface::prepareContext(const QJsonObject &request,
 
 void LLMClientInterface::sendCompletionToClient(const QString &completion,
                                                 const QJsonObject &request,
-                                                const QJsonObject &position,
                                                 bool isComplete)
 {
+    QJsonObject position = request["params"].toObject()["doc"].toObject()["position"].toObject();
+
     QJsonObject response;
     response["jsonrpc"] = "2.0";
     response[LanguageServerProtocol::idKey] = request["id"];

diff --git a/LLMClientInterface.hpp b/LLMClientInterface.hpp
@@ -41,7 +41,6 @@ class LLMClientInterface : public LanguageClient::BaseClientInterface
 
     void sendCompletionToClient(const QString &completion,
                                 const QJsonObject &request,
-                                const QJsonObject &position,
                                 bool isComplete);
 
     void handleCompletion(const QJsonObject &request);

diff --git a/core/LLMRequestConfig.hpp b/core/LLMRequestConfig.hpp
@@ -12,7 +12,7 @@ struct LLMConfig
     QUrl url;
     Providers::LLMProvider *provider;
     Templates::PromptTemplate *promptTemplate;
-    QString modelName;
+    QJsonObject providerRequest;
 };
 
 } // namespace QodeAssist
diff --git a/core/LLMRequestHandler.cpp b/core/LLMRequestHandler.cpp
@@ -31,24 +31,18 @@ LLMRequestHandler::LLMRequestHandler(QObject *parent)
     , m_manager(new QNetworkAccessManager(this))
 {}
 
-void LLMRequestHandler::sendLLMRequest(const LLMConfig &config,
-                                       const QJsonObject &request,
-                                       const ContextData &prompt)
+void LLMRequestHandler::sendLLMRequest(const LLMConfig &config, const QJsonObject &request)
 {
-    QJsonObject providerRequest = {{"model", config.modelName}, {"stream", true}};
-
-    config.promptTemplate->prepareRequest(providerRequest, prompt);
-    config.provider->prepareRequest(providerRequest);
-
     logMessage(QString("Sending request to llm: \nurl: %1\nRequest body:\n%2")
                    .arg(config.url.toString(),
                         QString::fromUtf8(
-                            QJsonDocument(providerRequest).toJson(QJsonDocument::Indented))));
+                            QJsonDocument(config.providerRequest).toJson(QJsonDocument::Indented))));
 
     QNetworkRequest networkRequest(config.url);
-    prepareNetworkRequest(networkRequest, providerRequest);
+    prepareNetworkRequest(networkRequest, config.providerRequest);
 
-    QNetworkReply *reply = m_manager->post(networkRequest, QJsonDocument(providerRequest).toJson());
+    QNetworkReply *reply = m_manager->post(networkRequest,
+                                           QJsonDocument(config.providerRequest).toJson());
     if (!reply) {
         logMessage("Error: Failed to create network reply");
         return;
@@ -82,8 +76,6 @@ void LLMRequestHandler::handleLLMResponse(QNetworkReply *reply,
 
     bool isComplete = config.provider->handleResponse(reply, accumulatedResponse);
 
-    QJsonObject position = request["params"].toObject()["doc"].toObject()["position"].toObject();
-
     if (!Settings::generalSettings().multiLineCompletion()
         && processSingleLineCompletion(reply, request, accumulatedResponse, config)) {
         return;
@@ -93,9 +85,9 @@ void LLMRequestHandler::handleLLMResponse(QNetworkReply *reply,
         if (isComplete) {
             auto cleanedCompletion = removeStopWords(accumulatedResponse,
                                                      config.promptTemplate->stopWords());
-            emit completionReceived(cleanedCompletion, request, position, true);
+            emit completionReceived(cleanedCompletion, request, true);
         } else {
-            emit completionReceived(accumulatedResponse, request, position, false);
+            emit completionReceived(accumulatedResponse, request, false);
         }
         m_accumulatedResponses.remove(reply);
     }
@@ -137,9 +129,7 @@ bool LLMRequestHandler::processSingleLineCompletion(QNetworkReply *reply,
         singleLineCompletion = removeStopWords(singleLineCompletion,
                                                config.promptTemplate->stopWords());
 
-        QJsonObject position = request["params"].toObject()["doc"].toObject()["position"].toObject();
-
-        emit completionReceived(singleLineCompletion, request, position, true);
+        emit completionReceived(singleLineCompletion, request, true);
         m_accumulatedResponses.remove(reply);
         reply->abort();
 

diff --git a/core/LLMRequestHandler.hpp b/core/LLMRequestHandler.hpp
@@ -37,19 +37,14 @@ class LLMRequestHandler : public QObject
 public:
     explicit LLMRequestHandler(QObject *parent = nullptr);
 
-    void sendLLMRequest(const LLMConfig &config,
-                        const QJsonObject &request,
-                        const ContextData &prompt);
+    void sendLLMRequest(const LLMConfig &config, const QJsonObject &request);
     void handleLLMResponse(QNetworkReply *reply,
                            const QJsonObject &request,
                            const LLMConfig &config);
     bool cancelRequest(const QString &id);
 
 signals:
-    void completionReceived(const QString &completion,
-                            const QJsonObject &request,
-                            const QJsonObject &position,
-                            bool isComplete);
+    void completionReceived(const QString &completion, const QJsonObject &request, bool isComplete);
     void requestFinished(const QString &requestId, bool success, const QString &errorString);
     void requestCancelled(const QString &id);
 

diff --git a/providers/LLMProvider.hpp b/providers/LLMProvider.hpp
@@ -37,6 +37,7 @@ class LLMProvider
     virtual QString completionEndpoint() const = 0;
 
     virtual void prepareRequest(QJsonObject &request) = 0;
+    virtual void prepareChatRequest(QJsonObject &request) = 0;
     virtual bool handleResponse(QNetworkReply *reply, QString &accumulatedResponse) = 0;
     virtual QList<QString> getInstalledModels(const Utils::Environment &env) = 0;
 };

diff --git a/providers/LMStudioProvider.cpp b/providers/LMStudioProvider.cpp
@@ -73,6 +73,8 @@ void LMStudioProvider::prepareRequest(QJsonObject &request)
         request["presence_penalty"] = settings.presencePenalty();
 }
 
+void LMStudioProvider::prepareChatRequest(QJsonObject &request) {}
+
 bool LMStudioProvider::handleResponse(QNetworkReply *reply, QString &accumulatedResponse)
 {
     bool isComplete = false;

diff --git a/providers/LMStudioProvider.hpp b/providers/LMStudioProvider.hpp
@@ -32,6 +32,7 @@ class LMStudioProvider : public LLMProvider
     QString url() const override;
     QString completionEndpoint() const override;
     void prepareRequest(QJsonObject &request) override;
+    void prepareChatRequest(QJsonObject &request) override;
     bool handleResponse(QNetworkReply *reply, QString &accumulatedResponse) override;
     QList<QString> getInstalledModels(const Utils::Environment &env) override;
 };

diff --git a/providers/OllamaProvider.cpp b/providers/OllamaProvider.cpp
@@ -71,6 +71,8 @@ void OllamaProvider::prepareRequest(QJsonObject &request)
     request["options"] = options;
 }
 
+void OllamaProvider::prepareChatRequest(QJsonObject &request) {}
+
 bool OllamaProvider::handleResponse(QNetworkReply *reply, QString &accumulatedResponse)
 {
     bool isComplete = false;

diff --git a/providers/OllamaProvider.hpp b/providers/OllamaProvider.hpp
@@ -32,6 +32,7 @@ class OllamaProvider : public LLMProvider
     QString url() const override;
     QString completionEndpoint() const override;
     void prepareRequest(QJsonObject &request) override;
+    void prepareChatRequest(QJsonObject &request) override;
     bool handleResponse(QNetworkReply *reply, QString &accumulatedResponse) override;
     QList<QString> getInstalledModels(const Utils::Environment &env) override;
 };

diff --git a/providers/OpenAICompatProvider.cpp b/providers/OpenAICompatProvider.cpp
@@ -77,6 +77,8 @@ void OpenAICompatProvider::prepareRequest(QJsonObject &request)
     }
 }
 
+void OpenAICompatProvider::prepareChatRequest(QJsonObject &request) {}
+
 bool OpenAICompatProvider::handleResponse(QNetworkReply *reply, QString &accumulatedResponse)
 {
     bool isComplete = false;

diff --git a/providers/OpenAICompatProvider.hpp b/providers/OpenAICompatProvider.hpp
@@ -32,6 +32,7 @@ class OpenAICompatProvider : public LLMProvider
     QString url() const override;
     QString completionEndpoint() const override;
     void prepareRequest(QJsonObject &request) override;
+    void prepareChatRequest(QJsonObject &request) override;
     bool handleResponse(QNetworkReply *reply, QString &accumulatedResponse) override;
     QList<QString> getInstalledModels(const Utils::Environment &env) override;
 };