mirror/QodeAssist
1
0
Fork 0
mirror of https://github.com/Palm1r/QodeAssist.git synced 2025-07-18 21:14:34 -04:00
Code Issues Packages Projects Releases Wiki Activity

♻️ refactor: Improve response handler for LMStudio

Browse Source
This commit is contained in:
Petr Mironychev
2024-12-10 17:13:56 +01:00
parent b692402897
commit 882047d7b2
8 changed files with 238 additions and 81 deletions
Download Patch File Download Diff File Expand all files Collapse all files

49
providers/LMStudioProvider.cpp
View File

@ -25,6 +25,7 @@
#include <QJsonObject>
#include <QNetworkReply>
#include "llmcore/OpenAIMessage.hpp"
#include "logger/Logger.hpp"
#include "settings/ChatAssistantSettings.hpp"
#include "settings/CodeCompletionSettings.hpp"
@ -101,43 +102,19 @@ void LMStudioProvider::prepareRequest(QJsonObject &request, LLMCore::RequestType
bool LMStudioProvider::handleResponse(QNetworkReply *reply, QString &accumulatedResponse)
{
bool isComplete = false;
while (reply->canReadLine()) {
QByteArray line = reply->readLine().trimmed();
if (line.isEmpty()) {
continue;
}
if (line == "data: [DONE]") {
isComplete = true;
break;
}
if (line.startsWith("data: ")) {
line = line.mid(6); // Remove "data: " prefix
}
QJsonDocument jsonResponse = QJsonDocument::fromJson(line);
if (jsonResponse.isNull()) {
qWarning() << "Invalid JSON response from LM Studio:" << line;
continue;
}
QJsonObject responseObj = jsonResponse.object();
if (responseObj.contains("choices")) {
QJsonArray choices = responseObj["choices"].toArray();
if (!choices.isEmpty()) {
QJsonObject choice = choices.first().toObject();
QJsonObject delta = choice["delta"].toObject();
if (delta.contains("content")) {
QString completion = delta["content"].toString();
accumulatedResponse += completion;
}
if (choice["finish_reason"].toString() == "stop") {
isComplete = true;
break;
}
}
}
QByteArray data = reply->readAll();
if (data.isEmpty()) {
return false;
}
return isComplete;
auto message = LLMCore::OpenAIMessage::fromJson(data);
if (message.hasError()) {
LOG_MESSAGE("Error in OpenAI response: " + message.error);
return false;
}
accumulatedResponse += message.getContent();
return message.isDone();
}
QList<QString> LMStudioProvider::getInstalledModels(const QString &url)

76
providers/OllamaMessage.cpp
View File

@ -1,76 +0,0 @@
/*
* Copyright (C) 2024 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 <https://www.gnu.org/licenses/>.
*/
#include "OllamaMessage.hpp"
namespace QodeAssist::Providers {
OllamaMessage OllamaMessage::fromJson(const QJsonObject &obj, Type type)
{
OllamaMessage msg;
msg.model = obj["model"].toString();
msg.createdAt = QDateTime::fromString(obj["created_at"].toString(), Qt::ISODate);
msg.done = obj["done"].toBool();
msg.doneReason = obj["done_reason"].toString();
msg.error = obj["error"].toString();
if (type == Type::Generate) {
auto &genResponse = msg.response.emplace<GenerateResponse>();
genResponse.response = obj["response"].toString();
if (msg.done && obj.contains("context")) {
const auto array = obj["context"].toArray();
genResponse.context.reserve(array.size());
for (const auto &val : array) {
genResponse.context.append(val.toInt());
}
}
} else {
auto &chatResponse = msg.response.emplace<ChatResponse>();
const auto msgObj = obj["message"].toObject();
chatResponse.role = msgObj["role"].toString();
chatResponse.content = msgObj["content"].toString();
}
if (msg.done) {
msg.metrics
= {obj["total_duration"].toVariant().toLongLong(),
obj["load_duration"].toVariant().toLongLong(),
obj["prompt_eval_count"].toVariant().toLongLong(),
obj["prompt_eval_duration"].toVariant().toLongLong(),
obj["eval_count"].toVariant().toLongLong(),
obj["eval_duration"].toVariant().toLongLong()};
}
return msg;
}
QString OllamaMessage::getContent() const
{
if (std::holds_alternative<GenerateResponse>(response)) {
return std::get<GenerateResponse>(response).response;
}
return std::get<ChatResponse>(response).content;
}
bool OllamaMessage::hasError() const
{
return !error.isEmpty();
}
} // namespace QodeAssist::Providers

71
providers/OllamaMessage.hpp
View File

@ -1,71 +0,0 @@
/*
* Copyright (C) 2024 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 <https://www.gnu.org/licenses/>.
*/
#pragma once
#include <QDateTime>
#include <QJsonArray>
#include <QJsonObject>
#include <QObject>
namespace QodeAssist::Providers {
class OllamaMessage
{
public:
enum class Type { Generate, Chat };
struct Metrics
{
qint64 totalDuration{0};
qint64 loadDuration{0};
qint64 promptEvalCount{0};
qint64 promptEvalDuration{0};
qint64 evalCount{0};
qint64 evalDuration{0};
};
struct GenerateResponse
{
QString response;
QVector<int> context;
};
struct ChatResponse
{
QString role;
QString content;
};
QString model;
QDateTime createdAt;
std::variant<GenerateResponse, ChatResponse> response;
bool done{false};
QString doneReason;
Metrics metrics;
QString error;
static OllamaMessage fromJson(const QJsonObject &obj, Type type);
QString getContent() const;
bool hasError() const;
};
} // namespace QodeAssist::Providers

50
providers/OllamaProvider.cpp
View File

@ -25,7 +25,7 @@
#include <QNetworkReply>
#include <QtCore/qeventloop.h>
#include "OllamaMessage.hpp"
#include "llmcore/OllamaMessage.hpp"
#include "logger/Logger.hpp"
#include "settings/ChatAssistantSettings.hpp"
#include "settings/CodeCompletionSettings.hpp"
@ -88,41 +88,23 @@ void OllamaProvider::prepareRequest(QJsonObject &request, LLMCore::RequestType t
bool OllamaProvider::handleResponse(QNetworkReply *reply, QString &accumulatedResponse)
{
const QString endpoint = reply->url().path();
auto messageType = endpoint == completionEndpoint() ? OllamaMessage::Type::Generate
: OllamaMessage::Type::Chat;
auto processMessage =
[&accumulatedResponse](const QJsonDocument &doc, OllamaMessage::Type messageType) {
if (doc.isNull()) {
LOG_MESSAGE("Invalid JSON response from Ollama");
return false;
}
auto message = OllamaMessage::fromJson(doc.object(), messageType);
if (message.hasError()) {
LOG_MESSAGE("Error in Ollama response: " + message.error);
return false;
}
accumulatedResponse += message.getContent();
return message.done;
};
if (reply->canReadLine()) {
while (reply->canReadLine()) {
QByteArray line = reply->readLine().trimmed();
if (line.isEmpty())
continue;
if (processMessage(QJsonDocument::fromJson(line), messageType)) {
return true;
}
}
QByteArray data = reply->readAll();
if (data.isEmpty()) {
return false;
} else {
return processMessage(QJsonDocument::fromJson(reply->readAll()), messageType);
}
const QString endpoint = reply->url().path();
auto messageType = endpoint == completionEndpoint() ? LLMCore::OllamaMessage::Type::Generate
: LLMCore::OllamaMessage::Type::Chat;
auto message = LLMCore::OllamaMessage::fromJson(data, messageType);
if (message.hasError()) {
LOG_MESSAGE("Error in Ollama response: " + message.error);
return false;
}
accumulatedResponse += message.getContent();
return message.done;
}
QList<QString> OllamaProvider::getInstalledModels(const QString &url)