Merge branch 'main' of https://github.com/nomic-ai/gpt4all into model-warnings

gpt4all-backend/llmodel_c.cpp

@@ -113,7 +113,8 @@ void llmodel_prompt(llmodel_model model, const char *prompt,
                     llmodel_response_callback response_callback,
                     llmodel_recalculate_callback recalculate_callback,
                     llmodel_prompt_context *ctx,
-                    bool special)
+                    bool special,
+                    const char *fake_reply)
 {
     LLModelWrapper *wrapper = reinterpret_cast<LLModelWrapper*>(model);
 
@@ -141,8 +142,13 @@ void llmodel_prompt(llmodel_model model, const char *prompt,
     wrapper->promptContext.repeat_last_n = ctx->repeat_last_n;
     wrapper->promptContext.contextErase = ctx->context_erase;
 
+    std::string fake_reply_str;
+    if (fake_reply) { fake_reply_str = fake_reply; }
+    auto *fake_reply_p = fake_reply ? &fake_reply_str : nullptr;
+
     // Call the C++ prompt method
-    wrapper->llModel->prompt(prompt, prompt_template, prompt_func, response_func, recalc_func, wrapper->promptContext, special);
+    wrapper->llModel->prompt(prompt, prompt_template, prompt_func, response_func, recalc_func, wrapper->promptContext,
+                             special, fake_reply_p);
 
     // Update the C context by giving access to the wrappers raw pointers to std::vector data
     // which involves no copies

gpt4all-backend/llmodel_c.h

@@ -169,6 +169,7 @@ uint64_t llmodel_restore_state_data(llmodel_model model, const uint8_t *src);
  * @param response_callback A callback function for handling the generated response.
  * @param recalculate_callback A callback function for handling recalculation requests.
  * @param special True if special tokens in the prompt should be processed, false otherwise.
+ * @param fake_reply A string to insert into context as the model's reply, or NULL to generate one.
  * @param ctx A pointer to the llmodel_prompt_context structure.
  */
 void llmodel_prompt(llmodel_model model, const char *prompt,
@@ -177,7 +178,8 @@ void llmodel_prompt(llmodel_model model, const char *prompt,
                     llmodel_response_callback response_callback,
                     llmodel_recalculate_callback recalculate_callback,
                     llmodel_prompt_context *ctx,
-                    bool special);
+                    bool special,
+                    const char *fake_reply);
 
 /**
  * Generate an embedding using the model.

gpt4all-bindings/python/gpt4all/_pyllmodel.py

@@ -100,6 +100,7 @@ llmodel.llmodel_prompt.argtypes = [
     RecalculateCallback,
     ctypes.POINTER(LLModelPromptContext),
     ctypes.c_bool,
+    ctypes.c_char_p,
 ]
 
 llmodel.llmodel_prompt.restype = None
@@ -361,6 +362,7 @@ class LLModel:
             RecalculateCallback(self._recalculate_callback),
             self.context,
             special,
+            ctypes.c_char_p(),
         )
 
 

gpt4all-bindings/python/gpt4all/gpt4all.py

@@ -158,9 +158,8 @@ class GPT4All:
                 if model_filename == m["filename"]:
                     config.update(m)
                     config["systemPrompt"] = config["systemPrompt"].strip()
-                    config["promptTemplate"] = config["promptTemplate"].replace(
-                        "%1", "{0}", 1
-                    )  # change to Python-style formatting
+                    # change to Python-style formatting
+                    config["promptTemplate"] = config["promptTemplate"].replace("%1", "{0}", 1).replace("%2", "{1}", 1)
                     break
 
         # Validate download directory
@@ -347,7 +346,7 @@ class GPT4All:
                     self.model.prompt_model(self.current_chat_session[0]["content"], "%1",
                                             _pyllmodel.empty_response_callback,
                                             n_batch=n_batch, n_predict=0, special=True)
-                prompt_template = self._current_prompt_template.format("%1")
+                prompt_template = self._current_prompt_template.format("%1", "%2")
             else:
                 warnings.warn(
                     "_format_chat_prompt_template is deprecated. Please use a chat session with a prompt template.",

gpt4all-chat/CMakeLists.txt

@@ -18,7 +18,7 @@ endif()
 
 set(APP_VERSION_MAJOR 2)
 set(APP_VERSION_MINOR 7)
-set(APP_VERSION_PATCH 2)
+set(APP_VERSION_PATCH 3)
 set(APP_VERSION "${APP_VERSION_MAJOR}.${APP_VERSION_MINOR}.${APP_VERSION_PATCH}")
 
 # Include the binary directory for the generated header file

gpt4all-chat/chat.h

@@ -46,6 +46,7 @@ public:
     explicit Chat(QObject *parent = nullptr);
     explicit Chat(bool isServer, QObject *parent = nullptr);
     virtual ~Chat();
+    void destroy() { m_llmodel->destroy(); }
     void connectLLM();
 
     QString id() const { return m_id; }

gpt4all-chat/chatgpt.cpp

@@ -86,13 +86,33 @@ void ChatGPT::prompt(const std::string &prompt,
     Q_UNUSED(promptCallback);
     Q_UNUSED(recalculateCallback);
     Q_UNUSED(special);
-    Q_UNUSED(fakeReply); // FIXME(cebtenzzre): I broke ChatGPT
 
     if (!isModelLoaded()) {
         std::cerr << "ChatGPT ERROR: prompt won't work with an unloaded model!\n";
         return;
     }
 
+    if (!promptCtx.n_past) { m_queuedPrompts.clear(); }
+    Q_ASSERT(promptCtx.n_past <= m_context.size());
+    m_context.resize(promptCtx.n_past);
+
+    // FIXME(cebtenzzre): We're assuming people don't try to use %2 with ChatGPT. What would that even mean?
+    m_queuedPrompts << QString::fromStdString(promptTemplate).arg(QString::fromStdString(prompt));
+
+    if (!promptCtx.n_predict && !fakeReply) {
+        return; // response explicitly suppressed, queue prompt for later
+    }
+
+    QString formattedPrompt = m_queuedPrompts.join("");
+    m_queuedPrompts.clear();
+
+    if (fakeReply) {
+        promptCtx.n_past += 1;
+        m_context.append(formattedPrompt);
+        m_context.append(QString::fromStdString(*fakeReply));
+        return;
+    }
+
     // FIXME: We don't set the max_tokens on purpose because in order to do so safely without encountering
     // an error we need to be able to count the tokens in our prompt. The only way to do this is to use
     // the OpenAI tiktokken library or to implement our own tokenization function that matches precisely
@@ -104,8 +124,9 @@ void ChatGPT::prompt(const std::string &prompt,
     root.insert("temperature", promptCtx.temp);
     root.insert("top_p", promptCtx.top_p);
 
+    // conversation history
     QJsonArray messages;
-    for (int i = 0; i < m_context.count() && i < promptCtx.n_past; ++i) {
+    for (int i = 0; i < m_context.count(); ++i) {
         QJsonObject message;
         message.insert("role", i % 2 == 0 ? "assistant" : "user");
         message.insert("content", m_context.at(i));
@@ -114,7 +135,7 @@ void ChatGPT::prompt(const std::string &prompt,
 
     QJsonObject promptObject;
     promptObject.insert("role", "user");
-    promptObject.insert("content", QString::fromStdString(promptTemplate).arg(QString::fromStdString(prompt)));
+    promptObject.insert("content", formattedPrompt);
     messages.append(promptObject);
     root.insert("messages", messages);
 
@@ -138,7 +159,7 @@ void ChatGPT::prompt(const std::string &prompt,
     workerThread.wait();
 
     promptCtx.n_past += 1;
-    m_context.append(QString::fromStdString(prompt));
+    m_context.append(formattedPrompt);
     m_context.append(worker.currentResponse());
     m_responseCallback = nullptr;
 

gpt4all-chat/chatgpt.h

@@ -3,11 +3,14 @@
 
 #include <stdexcept>
 
-#include <QObject>
+#include <QNetworkAccessManager>
 #include <QNetworkReply>
 #include <QNetworkRequest>
-#include <QNetworkAccessManager>
+#include <QObject>
+#include <QString>
+#include <QStringList>
 #include <QThread>
+
 #include "../gpt4all-backend/llmodel.h"
 
 class ChatGPT;
@@ -126,6 +129,7 @@ private:
     QString m_modelName;
     QString m_apiKey;
     QList<QString> m_context;
+    QStringList m_queuedPrompts;
 };
 
 #endif // CHATGPT_H

gpt4all-chat/chatlistmodel.h

@@ -192,13 +192,8 @@ public:
 
     int count() const { return m_chats.size(); }
 
-    void clearChats() {
-        m_newChat = nullptr;
-        m_serverChat = nullptr;
-        m_currentChat = nullptr;
-        for (auto * chat: m_chats) { delete chat; }
-        m_chats.clear();
-    }
+    // stop ChatLLM threads for clean shutdown
+    void destroyChats() { for (auto *chat: m_chats) { chat->destroy(); } }
 
     void removeChatFile(Chat *chat) const;
     Q_INVOKABLE void saveChats();

gpt4all-chat/chatllm.cpp

@@ -95,6 +95,10 @@ ChatLLM::ChatLLM(Chat *parent, bool isServer)
 
 ChatLLM::~ChatLLM()
 {
+    destroy();
+}
+
+void ChatLLM::destroy() {
     m_stopGenerating = true;
     m_llmThread.quit();
     m_llmThread.wait();
@@ -487,7 +491,7 @@ void ChatLLM::resetResponse()
 
 void ChatLLM::resetContext()
 {
-    regenerateResponse();
+    resetResponse();
     m_processedSystemPrompt = false;
     m_ctx = LLModel::PromptContext();
 }

gpt4all-chat/chatllm.h

@@ -72,6 +72,7 @@ public:
     ChatLLM(Chat *parent, bool isServer = false);
     virtual ~ChatLLM();
 
+    void destroy();
     bool isModelLoaded() const;
     void regenerateResponse();
     void resetResponse();

gpt4all-chat/download.cpp

@@ -130,7 +130,7 @@ void Download::downloadModel(const QString &modelFile)
 
     ModelList::globalInstance()->updateDataByFilename(modelFile, ModelList::DownloadingRole, true);
     ModelInfo info = ModelList::globalInstance()->modelInfoByFilename(modelFile);
-    QString url = !info.url.isEmpty() ? info.url : "http://gpt4all.io/models/gguf/" + modelFile;
+    QString url = !info.url().isEmpty() ? info.url() : "http://gpt4all.io/models/gguf/" + modelFile;
     Network::globalInstance()->sendDownloadStarted(modelFile);
     QNetworkRequest request(url);
     request.setAttribute(QNetworkRequest::User, modelFile);
@@ -201,6 +201,8 @@ void Download::removeModel(const QString &modelFile)
 
     QFile file(filePath);
     if (file.exists()) {
+        const ModelInfo info = ModelList::globalInstance()->modelInfoByFilename(modelFile);
+        ModelList::globalInstance()->removeInstalled(info);
         Network::globalInstance()->sendRemoveModel(modelFile);
         file.remove();
     }
@@ -364,8 +366,8 @@ HashAndSaveFile::HashAndSaveFile()
     m_hashAndSaveThread.start();
 }
 
-void HashAndSaveFile::hashAndSave(const QString &expectedHash, const QString &saveFilePath,
-        QFile *tempFile, QNetworkReply *modelReply)
+void HashAndSaveFile::hashAndSave(const QString &expectedHash, QCryptographicHash::Algorithm a,
+    const QString &saveFilePath, QFile *tempFile, QNetworkReply *modelReply)
 {
     Q_ASSERT(!tempFile->isOpen());
     QString modelFilename = modelReply->request().attribute(QNetworkRequest::User).toString();
@@ -379,13 +381,16 @@ void HashAndSaveFile::hashAndSave(const QString &expectedHash, const QString &sa
         return;
     }
 
-    QCryptographicHash hash(QCryptographicHash::Md5);
+    QCryptographicHash hash(a);
     while(!tempFile->atEnd())
         hash.addData(tempFile->read(16384));
-    if (hash.result().toHex() != expectedHash) {
+    if (hash.result().toHex() != expectedHash.toLatin1()) {
         tempFile->close();
         const QString error
-            = QString("ERROR: Download error MD5SUM did not match: %1 != %2 for %3").arg(hash.result().toHex()).arg(expectedHash).arg(modelFilename);
+            = QString("ERROR: Download error hash did not match: %1 != %2 for %3")
+                .arg(hash.result().toHex())
+                .arg(expectedHash.toLatin1())
+                .arg(modelFilename);
         qWarning() << error;
         tempFile->remove();
         emit hashAndSaveFinished(false, error, tempFile, modelReply);
@@ -472,9 +477,12 @@ void Download::handleModelDownloadFinished()
 
     // Notify that we are calculating hash
     ModelList::globalInstance()->updateDataByFilename(modelFilename, ModelList::CalcHashRole, true);
-    QByteArray md5sum =  ModelList::globalInstance()->modelInfoByFilename(modelFilename).md5sum;
+    QByteArray hash =  ModelList::globalInstance()->modelInfoByFilename(modelFilename).hash;
+    ModelInfo::HashAlgorithm hashAlgorithm =  ModelList::globalInstance()->modelInfoByFilename(modelFilename).hashAlgorithm;
     const QString saveFilePath = MySettings::globalInstance()->modelPath() + modelFilename;
-    emit requestHashAndSave(md5sum, saveFilePath, tempFile, modelReply);
+    emit requestHashAndSave(hash,
+        (hashAlgorithm == ModelInfo::Md5 ? QCryptographicHash::Md5 : QCryptographicHash::Sha256),
+        saveFilePath, tempFile, modelReply);
 }
 
 void Download::handleHashAndSaveFinished(bool success, const QString &error,
@@ -489,10 +497,14 @@ void Download::handleHashAndSaveFinished(bool success, const QString &error,
     tempFile->deleteLater();
 
     ModelList::globalInstance()->updateDataByFilename(modelFilename, ModelList::DownloadingRole, false);
-    if (!success)
+    if (!success) {
         ModelList::globalInstance()->updateDataByFilename(modelFilename, ModelList::DownloadErrorRole, error);
-    else
+    } else {
+        ModelInfo info = ModelList::globalInstance()->modelInfoByFilename(modelFilename);
+        if (info.isDiscovered())
+            ModelList::globalInstance()->updateDiscoveredInstalled(info);
         ModelList::globalInstance()->updateDataByFilename(modelFilename, ModelList::DownloadErrorRole, QString());
+    }
 }
 
 void Download::handleReadyRead()

gpt4all-chat/download.h

@@ -28,7 +28,7 @@ public:
     HashAndSaveFile();
 
 public Q_SLOTS:
-    void hashAndSave(const QString &hash, const QString &saveFilePath,
+    void hashAndSave(const QString &hash, QCryptographicHash::Algorithm a, const QString &saveFilePath,
         QFile *tempFile, QNetworkReply *modelReply);
 
 Q_SIGNALS:
@@ -72,7 +72,7 @@ private Q_SLOTS:
 Q_SIGNALS:
     void releaseInfoChanged();
     void hasNewerReleaseChanged();
-    void requestHashAndSave(const QString &hash, const QString &saveFilePath,
+    void requestHashAndSave(const QString &hash, QCryptographicHash::Algorithm a, const QString &saveFilePath,
         QFile *tempFile, QNetworkReply *modelReply);
 
 private:

gpt4all-chat/llm.cpp

@@ -42,8 +42,11 @@ LLM::LLM()
     m_compatHardware = minimal;
 
     QNetworkInformation::loadDefaultBackend();
-    connect(QNetworkInformation::instance(), &QNetworkInformation::reachabilityChanged,
-        this, &LLM::isNetworkOnlineChanged);
+    auto * netinfo = QNetworkInformation::instance();
+    if (netinfo) {
+        connect(netinfo, &QNetworkInformation::reachabilityChanged,
+            this, &LLM::isNetworkOnlineChanged);
+    }
 }
 
 bool LLM::hasSettingsAccess() const
@@ -108,8 +111,6 @@ QString LLM::systemTotalRAMInGBString() const
 
 bool LLM::isNetworkOnline() const
 {
-    if (!QNetworkInformation::instance())
-        return false;
-
-    return QNetworkInformation::instance()->reachability() == QNetworkInformation::Reachability::Online;
+    auto * netinfo = QNetworkInformation::instance();
+    return !netinfo || netinfo->reachability() == QNetworkInformation::Reachability::Online;
 }

gpt4all-chat/main.cpp

@@ -67,7 +67,7 @@ int main(int argc, char *argv[])
 
     // Make sure ChatLLM threads are joined before global destructors run.
     // Otherwise, we can get a heap-use-after-free inside of llama.cpp.
-    ChatListModel::globalInstance()->clearChats();
+    ChatListModel::globalInstance()->destroyChats();
 
     return res;
 }

gpt4all-chat/metadata/models3.json

@@ -1,6 +1,22 @@
 [
   {
     "order": "a",
+    "md5sum": "a5f6b4eabd3992da4d7fb7f020f921eb",
+    "name": "Nous Hermes 2 Mistral DPO",
+    "filename": "Nous-Hermes-2-Mistral-7B-DPO.Q4_0.gguf",
+    "filesize": "4108928000",
+    "requires": "2.7.1",
+    "ramrequired": "8",
+    "parameters": "7 billion",
+    "quant": "q4_0",
+    "type": "Mistral",
+    "description": "<strong>Best overall fast chat model</strong><br><ul><li>Fast responses</li><li>Chat based model</li><li>Accepts system prompts in ChatML format</li><li>Trained by Mistral AI<li>Finetuned by Nous Research on the OpenHermes-2.5 dataset<li>Licensed for commercial use</ul>",
+    "url": "https://huggingface.co/NousResearch/Nous-Hermes-2-Mistral-7B-DPO-GGUF/resolve/main/Nous-Hermes-2-Mistral-7B-DPO.Q4_0.gguf",
+    "promptTemplate": "<|im_start|>user\n%1<|im_end|>\n<|im_start|>assistant\n%2<|im_end|>\n",
+    "systemPrompt": ""
+  },
+  {
+    "order": "b",
     "md5sum": "f692417a22405d80573ac10cb0cd6c6a",
     "name": "Mistral OpenOrca",
     "filename": "mistral-7b-openorca.gguf2.Q4_0.gguf",
@@ -10,13 +26,13 @@
     "parameters": "7 billion",
     "quant": "q4_0",
     "type": "Mistral",
-    "description": "<strong>Best overall fast chat model</strong><br><ul><li>Fast responses</li><li>Chat based model</li><li>Trained by Mistral AI<li>Finetuned on OpenOrca dataset curated via <a href=\"https://atlas.nomic.ai/\">Nomic Atlas</a><li>Licensed for commercial use</ul>",
+    "description": "<strong>Strong overall fast chat model</strong><br><ul><li>Fast responses</li><li>Chat based model</li><li>Trained by Mistral AI<li>Finetuned on OpenOrca dataset curated via <a href=\"https://atlas.nomic.ai/\">Nomic Atlas</a><li>Licensed for commercial use</ul>",
     "url": "https://gpt4all.io/models/gguf/mistral-7b-openorca.gguf2.Q4_0.gguf",
     "promptTemplate": "<|im_start|>user\n%1<|im_end|>\n<|im_start|>assistant\n%2<|im_end|>\n",
     "systemPrompt": "<|im_start|>system\nYou are MistralOrca, a large language model trained by Alignment Lab AI. For multi-step problems, write out your reasoning for each step.\n<|im_end|>"
   },
   {
-    "order": "b",
+    "order": "c",
     "md5sum": "97463be739b50525df56d33b26b00852",
     "name": "Mistral Instruct",
     "filename": "mistral-7b-instruct-v0.1.Q4_0.gguf",
@@ -27,12 +43,12 @@
     "quant": "q4_0",
     "type": "Mistral",
     "systemPrompt": " ",
-    "description": "<strong>Best overall fast instruction following model</strong><br><ul><li>Fast responses</li><li>Trained by Mistral AI<li>Uncensored</li><li>Licensed for commercial use</li></ul>",
+    "description": "<strong>Strong overall fast instruction following model</strong><br><ul><li>Fast responses</li><li>Trained by Mistral AI<li>Uncensored</li><li>Licensed for commercial use</li></ul>",
     "url": "https://gpt4all.io/models/gguf/mistral-7b-instruct-v0.1.Q4_0.gguf",
     "promptTemplate": "[INST] %1 [/INST]"
   },
   {
-    "order": "c",
+    "order": "d",
     "md5sum": "c4c78adf744d6a20f05c8751e3961b84",
     "name": "GPT4All Falcon",
     "filename": "gpt4all-falcon-newbpe-q4_0.gguf",
@@ -89,7 +105,7 @@
     "quant": "q4_0",
     "type": "LLaMA2",
     "systemPrompt": " ",
-    "description": "<strong>Best overall larger model</strong><br><ul><li>Instruction based<li>Gives very long responses<li>Finetuned with only 1k of high-quality data<li>Trained by Microsoft and Peking University<li>Cannot be used commercially</ul>",
+    "description": "<strong>Strong overall larger model</strong><br><ul><li>Instruction based<li>Gives very long responses<li>Finetuned with only 1k of high-quality data<li>Trained by Microsoft and Peking University<li>Cannot be used commercially</ul>",
     "url": "https://gpt4all.io/models/gguf/wizardlm-13b-v1.2.Q4_0.gguf"
   },
   {

gpt4all-chat/metadata/release.json

@@ -705,6 +705,25 @@
 * Jared Van Bortel (Nomic AI)
 * Adam Treat (Nomic AI)
 * Community (beta testers, bug reporters, bindings authors)
+"
+  },
+  {
+    "version": "2.7.2",
+    "notes":
+"
+* New support for model search/discovery using huggingface search in downloads
+* Support for more model architectures for GPU acceleration
+* Three different crash fixes for corner case settings
+* Add a minp sampling parameter
+* Bert layoer norm epsilon value
+* Fix problem with blank lines between reply and next prompt
+",
+    "contributors":
+"
+* Christopher Barrera
+* Jared Van Bortel (Nomic AI)
+* Adam Treat (Nomic AI)
+* Community (beta testers, bug reporters, bindings authors)
 "
   }
 ]

gpt4all-chat/modellist.h

@@ -11,16 +11,17 @@ struct ModelInfo {
     Q_PROPERTY(QString filename READ filename WRITE setFilename)
     Q_PROPERTY(QString dirpath MEMBER dirpath)
     Q_PROPERTY(QString filesize MEMBER filesize)
-    Q_PROPERTY(QByteArray md5sum MEMBER md5sum)
+    Q_PROPERTY(QByteArray hash MEMBER hash)
+    Q_PROPERTY(HashAlgorithm hashAlgorithm MEMBER hashAlgorithm)
     Q_PROPERTY(bool calcHash MEMBER calcHash)
     Q_PROPERTY(bool installed MEMBER installed)
     Q_PROPERTY(bool isDefault MEMBER isDefault)
     Q_PROPERTY(bool disableGUI MEMBER disableGUI)
     Q_PROPERTY(bool isOnline MEMBER isOnline)
-    Q_PROPERTY(QString description MEMBER description)
+    Q_PROPERTY(QString description READ description WRITE setDescription)
     Q_PROPERTY(QString requiresVersion MEMBER requiresVersion)
-    Q_PROPERTY(QString deprecatedVersion MEMBER deprecatedVersion)
-    Q_PROPERTY(QString url MEMBER url)
+    Q_PROPERTY(QString versionRemoved MEMBER versionRemoved)
+    Q_PROPERTY(QString url READ url WRITE setUrl)
     Q_PROPERTY(qint64 bytesReceived MEMBER bytesReceived)
     Q_PROPERTY(qint64 bytesTotal MEMBER bytesTotal)
     Q_PROPERTY(qint64 timestamp MEMBER timestamp)
@@ -31,9 +32,10 @@ struct ModelInfo {
     Q_PROPERTY(QString order MEMBER order)
     Q_PROPERTY(int ramrequired MEMBER ramrequired)
     Q_PROPERTY(QString parameters MEMBER parameters)
-    Q_PROPERTY(QString quant MEMBER quant)
-    Q_PROPERTY(QString type MEMBER type)
-    Q_PROPERTY(bool isClone MEMBER isClone)
+    Q_PROPERTY(QString quant READ quant WRITE setQuant)
+    Q_PROPERTY(QString type READ type WRITE setType)
+    Q_PROPERTY(bool isClone READ isClone WRITE setIsClone)
+    Q_PROPERTY(bool isDiscovered READ isDiscovered WRITE setIsDiscovered)
     Q_PROPERTY(double temperature READ temperature WRITE setTemperature)
     Q_PROPERTY(double topP READ topP WRITE setTopP)
     Q_PROPERTY(double minP READ minP WRITE setMinP)
@@ -48,8 +50,16 @@ struct ModelInfo {
     Q_PROPERTY(int repeatPenaltyTokens READ repeatPenaltyTokens WRITE setRepeatPenaltyTokens)
     Q_PROPERTY(QString promptTemplate READ promptTemplate WRITE setPromptTemplate)
     Q_PROPERTY(QString systemPrompt READ systemPrompt WRITE setSystemPrompt)
+    Q_PROPERTY(int likes READ likes WRITE setLikes)
+    Q_PROPERTY(int downloads READ downloads WRITE setDownloads)
+    Q_PROPERTY(QDateTime recency READ recency WRITE setRecency)
 
 public:
+    enum HashAlgorithm {
+        Md5,
+        Sha256
+    };
+
     QString id() const;
     void setId(const QString &id);
 
@@ -59,18 +69,44 @@ public:
     QString filename() const;
     void setFilename(const QString &name);
 
+    QString description() const;
+    void setDescription(const QString &d);
+
+    QString url() const;
+    void setUrl(const QString &u);
+
+    QString quant() const;
+    void setQuant(const QString &q);
+
+    QString type() const;
+    void setType(const QString &t);
+
+    bool isClone() const;
+    void setIsClone(bool b);
+
+    bool isDiscovered() const;
+    void setIsDiscovered(bool b);
+
+    int likes() const;
+    void setLikes(int l);
+
+    int downloads() const;
+    void setDownloads(int d);
+
+    QDateTime recency() const;
+    void setRecency(const QDateTime &r);
+
     QString dirpath;
     QString filesize;
-    QByteArray md5sum;
+    QByteArray hash;
+    HashAlgorithm hashAlgorithm;
     bool calcHash = false;
     bool installed = false;
     bool isDefault = false;
     bool isOnline = false;
     bool disableGUI = false;
-    QString description;
     QString requiresVersion;
-    QString deprecatedVersion;
-    QString url;
+    QString versionRemoved;
     qint64 bytesReceived = 0;
     qint64 bytesTotal = 0;
     qint64 timestamp = 0;
@@ -79,11 +115,8 @@ public:
     bool isIncomplete = false;
     QString downloadError;
     QString order;
-    int ramrequired = 0;
+    int ramrequired = -1;
     QString parameters;
-    QString quant;
-    QString type;
-    bool isClone = false;
 
     bool operator==(const ModelInfo &other) const {
         return  m_id == other.m_id;
@@ -116,10 +149,21 @@ public:
     QString systemPrompt() const;
     void setSystemPrompt(const QString &p);
 
+    bool shouldSaveMetadata() const;
+
 private:
     QString m_id;
     QString m_name;
     QString m_filename;
+    QString m_description;
+    QString m_url;
+    QString m_quant;
+    QString m_type;
+    bool    m_isClone              = false;
+    bool    m_isDiscovered         = false;
+    int     m_likes                = -1;
+    int     m_downloads            = -1;
+    QDateTime m_recency;
     double  m_temperature          = 0.7;
     double  m_topP                 = 0.4;
     double  m_minP                 = 0.0;
@@ -183,6 +227,8 @@ public:
     bool isExpanded() const;
     void setExpanded(bool expanded);
 
+    Q_INVOKABLE void discoverAndFilter(const QString &discover);
+
 Q_SIGNALS:
     void countChanged();
 
@@ -195,6 +241,7 @@ Q_SIGNALS:
 private:
     bool m_expanded;
     int m_limit;
+    QString m_discoverFilter;
 };
 
 class ModelList : public QAbstractListModel
@@ -207,17 +254,30 @@ class ModelList : public QAbstractListModel
     Q_PROPERTY(DownloadableModels* downloadableModels READ downloadableModels NOTIFY downloadableModelsChanged)
     Q_PROPERTY(QList<QString> userDefaultModelList READ userDefaultModelList NOTIFY userDefaultModelListChanged)
     Q_PROPERTY(bool asyncModelRequestOngoing READ asyncModelRequestOngoing NOTIFY asyncModelRequestOngoingChanged)
+    Q_PROPERTY(int discoverLimit READ discoverLimit WRITE setDiscoverLimit NOTIFY discoverLimitChanged)
+    Q_PROPERTY(int discoverSortDirection READ discoverSortDirection WRITE setDiscoverSortDirection NOTIFY discoverSortDirectionChanged)
+    Q_PROPERTY(DiscoverSort discoverSort READ discoverSort WRITE setDiscoverSort NOTIFY discoverSortChanged)
+    Q_PROPERTY(float discoverProgress READ discoverProgress NOTIFY discoverProgressChanged)
+    Q_PROPERTY(bool discoverInProgress READ discoverInProgress NOTIFY discoverInProgressChanged)
 
 public:
     static ModelList *globalInstance();
 
+    enum DiscoverSort {
+        Default,
+        Likes,
+        Downloads,
+        Recent
+    };
+
     enum Roles {
         IdRole = Qt::UserRole + 1,
         NameRole,
         FilenameRole,
         DirpathRole,
         FilesizeRole,
-        Md5sumRole,
+        HashRole,
+        HashAlgorithmRole,
         CalcHashRole,
         InstalledRole,
         DefaultRole,
@@ -225,7 +285,7 @@ public:
         DisableGUIRole,
         DescriptionRole,
         RequiresVersionRole,
-        DeprecatedVersionRole,
+        VersionRemovedRole,
         UrlRole,
         BytesReceivedRole,
         BytesTotalRole,
@@ -240,6 +300,7 @@ public:
         QuantRole,
         TypeRole,
         IsCloneRole,
+        IsDiscoveredRole,
         TemperatureRole,
         TopPRole,
         TopKRole,
@@ -252,6 +313,9 @@ public:
         PromptTemplateRole,
         SystemPromptRole,
         MinPRole,
+        LikesRole,
+        DownloadsRole,
+        RecencyRole
     };
 
     QHash<int, QByteArray> roleNames() const override
@@ -262,7 +326,8 @@ public:
         roles[FilenameRole] = "filename";
         roles[DirpathRole] = "dirpath";
         roles[FilesizeRole] = "filesize";
-        roles[Md5sumRole] = "md5sum";
+        roles[HashRole] = "hash";
+        roles[HashAlgorithmRole] = "hashAlgorithm";
         roles[CalcHashRole] = "calcHash";
         roles[InstalledRole] = "installed";
         roles[DefaultRole] = "isDefault";
@@ -270,7 +335,7 @@ public:
         roles[DisableGUIRole] = "disableGUI";
         roles[DescriptionRole] = "description";
         roles[RequiresVersionRole] = "requiresVersion";
-        roles[DeprecatedVersionRole] = "deprecatedVersion";
+        roles[VersionRemovedRole] = "versionRemoved";
         roles[UrlRole] = "url";
         roles[BytesReceivedRole] = "bytesReceived";
         roles[BytesTotalRole] = "bytesTotal";
@@ -285,6 +350,7 @@ public:
         roles[QuantRole] = "quant";
         roles[TypeRole] = "type";
         roles[IsCloneRole] = "isClone";
+        roles[IsDiscoveredRole] = "isDiscovered";
         roles[TemperatureRole] = "temperature";
         roles[TopPRole] = "topP";
         roles[MinPRole] = "minP";
@@ -297,6 +363,9 @@ public:
         roles[RepeatPenaltyTokensRole] = "repeatPenaltyTokens";
         roles[PromptTemplateRole] = "promptTemplate";
         roles[SystemPromptRole] = "systemPrompt";
+        roles[LikesRole] = "likes";
+        roles[DownloadsRole] = "downloads";
+        roles[RecencyRole] = "recency";
         return roles;
     }
 
@@ -306,6 +375,7 @@ public:
     QVariant dataByFilename(const QString &filename, int role) const;
     void updateData(const QString &id, int role, const QVariant &value);
     void updateDataByFilename(const QString &filename, int role, const QVariant &value);
+    void updateData(const QString &id, const QVector<QPair<int, QVariant>> &data);
 
     int count() const { return m_models.size(); }
 
@@ -315,7 +385,8 @@ public:
     Q_INVOKABLE ModelInfo modelInfoByFilename(const QString &filename) const;
     Q_INVOKABLE bool isUniqueName(const QString &name) const;
     Q_INVOKABLE QString clone(const ModelInfo &model);
-    Q_INVOKABLE void remove(const ModelInfo &model);
+    Q_INVOKABLE void removeClone(const ModelInfo &model);
+    Q_INVOKABLE void removeInstalled(const ModelInfo &model);
     ModelInfo defaultModelInfo() const;
     int defaultEmbeddingModelIndex() const;
 
@@ -345,6 +416,21 @@ public:
     bool asyncModelRequestOngoing() const { return m_asyncModelRequestOngoing; }
 
     void updateModelsFromDirectory();
+    void updateDiscoveredInstalled(const ModelInfo &info);
+
+    int discoverLimit() const;
+    void setDiscoverLimit(int limit);
+
+    int discoverSortDirection() const;
+    void setDiscoverSortDirection(int direction); // -1 or 1
+
+    DiscoverSort discoverSort() const;
+    void setDiscoverSort(DiscoverSort sort);
+
+    float discoverProgress() const;
+    bool discoverInProgress() const;
+
+    Q_INVOKABLE void discoverSearch(const QString &discover);
 
 Q_SIGNALS:
     void countChanged();
@@ -354,22 +440,35 @@ Q_SIGNALS:
     void userDefaultModelListChanged();
     void asyncModelRequestOngoingChanged();
     void defaultEmbeddingModelIndexChanged();
+    void discoverLimitChanged();
+    void discoverSortDirectionChanged();
+    void discoverSortChanged();
+    void discoverProgressChanged();
+    void discoverInProgressChanged();
 
 private Q_SLOTS:
+    void resortModel();
     void updateModelsFromJson();
     void updateModelsFromJsonAsync();
     void updateModelsFromSettings();
     void updateDataForSettings();
     void handleModelsJsonDownloadFinished();
     void handleModelsJsonDownloadErrorOccurred(QNetworkReply::NetworkError code);
+    void handleDiscoveryFinished();
+    void handleDiscoveryErrorOccurred(QNetworkReply::NetworkError code);
+    void handleDiscoveryItemFinished();
+    void handleDiscoveryItemErrorOccurred(QNetworkReply::NetworkError code);
     void handleSslErrors(QNetworkReply *reply, const QList<QSslError> &errors);
 
 private:
+    void removeInternal(const ModelInfo &model);
+    void clearDiscoveredModels();
     QString modelDirPath(const QString &modelName, bool isOnline);
     int indexForModel(ModelInfo *model);
     QVariant dataInternal(const ModelInfo *info, int role) const;
-    static bool lessThan(const ModelInfo* a, const ModelInfo* b);
+    static bool lessThan(const ModelInfo* a, const ModelInfo* b, DiscoverSort s, int d);
     void parseModelsJsonFile(const QByteArray &jsonData, bool save);
+    void parseDiscoveryJsonFile(const QByteArray &jsonData);
     QString uniqueModelName(const ModelInfo &model) const;
 
 private:
@@ -381,8 +480,14 @@ private:
     QList<ModelInfo*> m_models;
     QHash<QString, ModelInfo*> m_modelMap;
     bool m_asyncModelRequestOngoing;
+    int m_discoverLimit;
+    int m_discoverSortDirection;
+    DiscoverSort m_discoverSort;
+    int m_discoverNumberOfResults;
+    int m_discoverResultsCompleted;
+    bool m_discoverInProgress;
 
-private:
+protected:
     explicit ModelList();
     ~ModelList() {}
     friend class MyModelList;

gpt4all-chat/mysettings.cpp

@@ -140,12 +140,10 @@ void MySettings::setModelName(const ModelInfo &m, const QString &name, bool forc
         return;
 
     QSettings setting;
-    if ((m.m_name == name || m.m_filename == name) && !m.isClone)
+    if ((m.m_name == name || m.m_filename == name) && !m.shouldSaveMetadata())
         setting.remove(QString("model-%1").arg(m.id()) + "/name");
     else
         setting.setValue(QString("model-%1").arg(m.id()) + "/name", name);
-    if (m.isClone)
-        setting.setValue(QString("model-%1").arg(m.id()) + "/isClone", "true");
     setting.sync();
     if (!force)
         emit nameChanged(m);
@@ -164,7 +162,7 @@ void MySettings::setModelFilename(const ModelInfo &m, const QString &filename, b
         return;
 
     QSettings setting;
-    if (m.m_filename == filename && !m.isClone)
+    if (m.m_filename == filename && !m.shouldSaveMetadata())
         setting.remove(QString("model-%1").arg(m.id()) + "/filename");
     else
         setting.setValue(QString("model-%1").arg(m.id()) + "/filename", filename);
@@ -173,6 +171,186 @@ void MySettings::setModelFilename(const ModelInfo &m, const QString &filename, b
         emit filenameChanged(m);
 }
 
+QString MySettings::modelDescription(const ModelInfo &m) const
+{
+    QSettings setting;
+    setting.sync();
+    return setting.value(QString("model-%1").arg(m.id()) + "/description", m.m_description).toString();
+}
+
+void MySettings::setModelDescription(const ModelInfo &m, const QString &d, bool force)
+{
+    if ((modelDescription(m) == d || m.id().isEmpty()) && !force)
+        return;
+
+    QSettings setting;
+    if (m.m_description == d && !m.shouldSaveMetadata())
+        setting.remove(QString("model-%1").arg(m.id()) + "/description");
+    else
+        setting.setValue(QString("model-%1").arg(m.id()) + "/description", d);
+    setting.sync();
+}
+
+QString MySettings::modelUrl(const ModelInfo &m) const
+{
+    QSettings setting;
+    setting.sync();
+    return setting.value(QString("model-%1").arg(m.id()) + "/url", m.m_url).toString();
+}
+
+void MySettings::setModelUrl(const ModelInfo &m, const QString &u, bool force)
+{
+    if ((modelUrl(m) == u || m.id().isEmpty()) && !force)
+        return;
+
+    QSettings setting;
+    if (m.m_url == u && !m.shouldSaveMetadata())
+        setting.remove(QString("model-%1").arg(m.id()) + "/url");
+    else
+        setting.setValue(QString("model-%1").arg(m.id()) + "/url", u);
+    setting.sync();
+}
+
+QString MySettings::modelQuant(const ModelInfo &m) const
+{
+    QSettings setting;
+    setting.sync();
+    return setting.value(QString("model-%1").arg(m.id()) + "/quant", m.m_quant).toString();
+}
+
+void MySettings::setModelQuant(const ModelInfo &m, const QString &q, bool force)
+{
+    if ((modelUrl(m) == q || m.id().isEmpty()) && !force)
+        return;
+
+    QSettings setting;
+    if (m.m_quant == q && !m.shouldSaveMetadata())
+        setting.remove(QString("model-%1").arg(m.id()) + "/quant");
+    else
+        setting.setValue(QString("model-%1").arg(m.id()) + "/quant", q);
+    setting.sync();
+}
+
+QString MySettings::modelType(const ModelInfo &m) const
+{
+    QSettings setting;
+    setting.sync();
+    return setting.value(QString("model-%1").arg(m.id()) + "/type", m.m_type).toString();
+}
+
+void MySettings::setModelType(const ModelInfo &m, const QString &t, bool force)
+{
+    if ((modelType(m) == t || m.id().isEmpty()) && !force)
+        return;
+
+    QSettings setting;
+    if (m.m_type == t && !m.shouldSaveMetadata())
+        setting.remove(QString("model-%1").arg(m.id()) + "/type");
+    else
+        setting.setValue(QString("model-%1").arg(m.id()) + "/type", t);
+    setting.sync();
+}
+
+bool MySettings::modelIsClone(const ModelInfo &m) const
+{
+    QSettings setting;
+    setting.sync();
+    return setting.value(QString("model-%1").arg(m.id()) + "/isClone", m.m_isClone).toBool();
+}
+
+void MySettings::setModelIsClone(const ModelInfo &m, bool b, bool force)
+{
+    if ((modelIsClone(m) == b || m.id().isEmpty()) && !force)
+        return;
+
+    QSettings setting;
+    if (m.m_isClone == b && !m.shouldSaveMetadata())
+        setting.remove(QString("model-%1").arg(m.id()) + "/isClone");
+    else
+        setting.setValue(QString("model-%1").arg(m.id()) + "/isClone", b);
+    setting.sync();
+}
+
+bool MySettings::modelIsDiscovered(const ModelInfo &m) const
+{
+    QSettings setting;
+    setting.sync();
+    return setting.value(QString("model-%1").arg(m.id()) + "/isDiscovered", m.m_isDiscovered).toBool();
+}
+
+void MySettings::setModelIsDiscovered(const ModelInfo &m, bool b, bool force)
+{
+    if ((modelIsDiscovered(m) == b || m.id().isEmpty()) && !force)
+        return;
+
+    QSettings setting;
+    if (m.m_isDiscovered == b && !m.shouldSaveMetadata())
+        setting.remove(QString("model-%1").arg(m.id()) + "/isDiscovered");
+    else
+        setting.setValue(QString("model-%1").arg(m.id()) + "/isDiscovered", b);
+    setting.sync();
+}
+
+int MySettings::modelLikes(const ModelInfo &m) const
+{
+    QSettings setting;
+    setting.sync();
+    return setting.value(QString("model-%1").arg(m.id()) + "/likes", m.m_likes).toInt();
+}
+
+void MySettings::setModelLikes(const ModelInfo &m, int l, bool force)
+{
+    if ((modelLikes(m) == l || m.id().isEmpty()) && !force)
+        return;
+
+    QSettings setting;
+    if (m.m_likes == l && !m.shouldSaveMetadata())
+        setting.remove(QString("model-%1").arg(m.id()) + "/likes");
+    else
+        setting.setValue(QString("model-%1").arg(m.id()) + "/likes", l);
+    setting.sync();
+}
+
+int MySettings::modelDownloads(const ModelInfo &m) const
+{
+    QSettings setting;
+    setting.sync();
+    return setting.value(QString("model-%1").arg(m.id()) + "/downloads", m.m_downloads).toInt();
+}
+
+void MySettings::setModelDownloads(const ModelInfo &m, int d, bool force)
+{
+    if ((modelDownloads(m) == d || m.id().isEmpty()) && !force)
+        return;
+
+    QSettings setting;
+    if (m.m_downloads == d && !m.shouldSaveMetadata())
+        setting.remove(QString("model-%1").arg(m.id()) + "/downloads");
+    else
+        setting.setValue(QString("model-%1").arg(m.id()) + "/downloads", d);
+    setting.sync();
+}
+
+QDateTime MySettings::modelRecency(const ModelInfo &m) const
+{
+    QSettings setting;
+    setting.sync();
+    return setting.value(QString("model-%1").arg(m.id()) + "/recency", m.m_recency).toDateTime();
+}
+
+void MySettings::setModelRecency(const ModelInfo &m, const QDateTime &r, bool force)
+{
+    if ((modelRecency(m) == r || m.id().isEmpty()) && !force)
+        return;
+
+    QSettings setting;
+    if (m.m_recency == r && !m.shouldSaveMetadata())
+        setting.remove(QString("model-%1").arg(m.id()) + "/recency");
+    else
+        setting.setValue(QString("model-%1").arg(m.id()) + "/recency", r);
+    setting.sync();
+}
+
 double MySettings::modelTemperature(const ModelInfo &m) const
 {
     QSettings setting;
@@ -186,7 +364,7 @@ void MySettings::setModelTemperature(const ModelInfo &m, double t, bool force)
         return;
 
     QSettings setting;
-    if (m.m_temperature == t && !m.isClone)
+    if (m.m_temperature == t && !m.shouldSaveMetadata())
         setting.remove(QString("model-%1").arg(m.id()) + "/temperature");
     else
         setting.setValue(QString("model-%1").arg(m.id()) + "/temperature", t);
@@ -215,7 +393,7 @@ void MySettings::setModelTopP(const ModelInfo &m, double p, bool force)
         return;
 
     QSettings setting;
-    if (m.m_topP == p && !m.isClone)
+    if (m.m_topP == p && !m.shouldSaveMetadata())
         setting.remove(QString("model-%1").arg(m.id()) + "/topP");
     else
         setting.setValue(QString("model-%1").arg(m.id()) + "/topP", p);
@@ -230,7 +408,7 @@ void MySettings::setModelMinP(const ModelInfo &m, double p, bool force)
         return;
 
     QSettings setting;
-    if (m.m_minP == p && !m.isClone)
+    if (m.m_minP == p && !m.shouldSaveMetadata())
         setting.remove(QString("model-%1").arg(m.id()) + "/minP");
     else
         setting.setValue(QString("model-%1").arg(m.id()) + "/minP", p);
@@ -252,7 +430,7 @@ void MySettings::setModelTopK(const ModelInfo &m, int k, bool force)
         return;
 
     QSettings setting;
-    if (m.m_topK == k && !m.isClone)
+    if (m.m_topK == k && !m.shouldSaveMetadata())
         setting.remove(QString("model-%1").arg(m.id()) + "/topK");
     else
         setting.setValue(QString("model-%1").arg(m.id()) + "/topK", k);
@@ -274,7 +452,7 @@ void MySettings::setModelMaxLength(const ModelInfo &m, int l, bool force)
         return;
 
     QSettings setting;
-    if (m.m_maxLength == l && !m.isClone)
+    if (m.m_maxLength == l && !m.shouldSaveMetadata())
         setting.remove(QString("model-%1").arg(m.id()) + "/maxLength");
     else
         setting.setValue(QString("model-%1").arg(m.id()) + "/maxLength", l);
@@ -296,7 +474,7 @@ void MySettings::setModelPromptBatchSize(const ModelInfo &m, int s, bool force)
         return;
 
     QSettings setting;
-    if (m.m_promptBatchSize == s && !m.isClone)
+    if (m.m_promptBatchSize == s && !m.shouldSaveMetadata())
         setting.remove(QString("model-%1").arg(m.id()) + "/promptBatchSize");
     else
         setting.setValue(QString("model-%1").arg(m.id()) + "/promptBatchSize", s);
@@ -318,7 +496,7 @@ void MySettings::setModelContextLength(const ModelInfo &m, int l, bool force)
         return;
 
     QSettings setting;
-    if (m.m_contextLength == l && !m.isClone)
+    if (m.m_contextLength == l && !m.shouldSaveMetadata())
         setting.remove(QString("model-%1").arg(m.id()) + "/contextLength");
     else
         setting.setValue(QString("model-%1").arg(m.id()) + "/contextLength", l);
@@ -340,7 +518,7 @@ void MySettings::setModelGpuLayers(const ModelInfo &m, int l, bool force)
         return;
 
     QSettings setting;
-    if (m.m_gpuLayers == l && !m.isClone)
+    if (m.m_gpuLayers == l && !m.shouldSaveMetadata())
         setting.remove(QString("model-%1").arg(m.id()) + "/gpuLayers");
     else
         setting.setValue(QString("model-%1").arg(m.id()) + "/gpuLayers", l);
@@ -362,7 +540,7 @@ void MySettings::setModelRepeatPenalty(const ModelInfo &m, double p, bool force)
         return;
 
     QSettings setting;
-    if (m.m_repeatPenalty == p && !m.isClone)
+    if (m.m_repeatPenalty == p && !m.shouldSaveMetadata())
         setting.remove(QString("model-%1").arg(m.id()) + "/repeatPenalty");
     else
         setting.setValue(QString("model-%1").arg(m.id()) + "/repeatPenalty", p);
@@ -384,7 +562,7 @@ void MySettings::setModelRepeatPenaltyTokens(const ModelInfo &m, int t, bool for
         return;
 
     QSettings setting;
-    if (m.m_repeatPenaltyTokens == t && !m.isClone)
+    if (m.m_repeatPenaltyTokens == t && !m.shouldSaveMetadata())
         setting.remove(QString("model-%1").arg(m.id()) + "/repeatPenaltyTokens");
     else
         setting.setValue(QString("model-%1").arg(m.id()) + "/repeatPenaltyTokens", t);
@@ -406,7 +584,7 @@ void MySettings::setModelPromptTemplate(const ModelInfo &m, const QString &t, bo
         return;
 
     QSettings setting;
-    if (m.m_promptTemplate == t && !m.isClone)
+    if (m.m_promptTemplate == t && !m.shouldSaveMetadata())
         setting.remove(QString("model-%1").arg(m.id()) + "/promptTemplate");
     else
         setting.setValue(QString("model-%1").arg(m.id()) + "/promptTemplate", t);
@@ -428,7 +606,7 @@ void MySettings::setModelSystemPrompt(const ModelInfo &m, const QString &p, bool
         return;
 
     QSettings setting;
-    if (m.m_systemPrompt == p && !m.isClone)
+    if (m.m_systemPrompt == p && !m.shouldSaveMetadata())
         setting.remove(QString("model-%1").arg(m.id()) + "/systemPrompt");
     else
         setting.setValue(QString("model-%1").arg(m.id()) + "/systemPrompt", p);

gpt4all-chat/mysettings.h

@@ -43,6 +43,26 @@ public:
     Q_INVOKABLE void setModelName(const ModelInfo &m, const QString &name, bool force = false);
     QString modelFilename(const ModelInfo &m) const;
     Q_INVOKABLE void setModelFilename(const ModelInfo &m, const QString &filename, bool force = false);
+
+    QString modelDescription(const ModelInfo &m) const;
+    void setModelDescription(const ModelInfo &m, const QString &d, bool force = false);
+    QString modelUrl(const ModelInfo &m) const;
+    void setModelUrl(const ModelInfo &m, const QString &u, bool force = false);
+    QString modelQuant(const ModelInfo &m) const;
+    void setModelQuant(const ModelInfo &m, const QString &q, bool force = false);
+    QString modelType(const ModelInfo &m) const;
+    void setModelType(const ModelInfo &m, const QString &t, bool force = false);
+    bool modelIsClone(const ModelInfo &m) const;
+    void setModelIsClone(const ModelInfo &m, bool b, bool force = false);
+    bool modelIsDiscovered(const ModelInfo &m) const;
+    void setModelIsDiscovered(const ModelInfo &m, bool b, bool force = false);
+    int modelLikes(const ModelInfo &m) const;
+    void setModelLikes(const ModelInfo &m, int l, bool force = false);
+    int modelDownloads(const ModelInfo &m) const;
+    void setModelDownloads(const ModelInfo &m, int d, bool force = false);
+    QDateTime modelRecency(const ModelInfo &m) const;
+    void setModelRecency(const ModelInfo &m, const QDateTime &r, bool force = false);
+
     double modelTemperature(const ModelInfo &m) const;
     Q_INVOKABLE void setModelTemperature(const ModelInfo &m, double t, bool force = false);
     double modelTopP(const ModelInfo &m) const;
@@ -112,7 +132,6 @@ public:
     bool networkUsageStatsActive() const;
     void setNetworkUsageStatsActive(bool b);
 
-
     QVector<QString> deviceList() const;
     void setDeviceList(const QVector<QString> &deviceList);
 

gpt4all-chat/qml/ModelDownloaderDialog.qml

@@ -41,17 +41,222 @@ MyDialog {
 
         Label {
             id: listLabel
-            text: qsTr("Available Models")
-            visible: false
-            Layout.alignment: Qt.AlignLeft
+            text: qsTr("Discover and Download Models")
+            visible: true
             Layout.fillWidth: true
+            horizontalAlignment: Qt.AlignHCenter
+            verticalAlignment: Qt.AlignVCenter
             color: theme.titleTextColor
-            font.pixelSize: theme.fontSizeLarge
+            font.pixelSize: theme.fontSizeLargest
             font.bold: true
         }
 
-        Item {
-            height: 0 // for visible space between close button and rest of dialog
+        RowLayout {
+            Layout.fillWidth: true
+            Layout.alignment: Qt.AlignCenter
+            Layout.margins: 0
+            spacing: 10
+            MyTextField {
+                id: discoverField
+                property string textBeingSearched: ""
+                readOnly: ModelList.discoverInProgress
+                Layout.alignment: Qt.AlignCenter
+                Layout.preferredWidth: 720
+                Layout.preferredHeight: 90
+                font.pixelSize: theme.fontSizeLarger
+                placeholderText: qsTr("Discover and download models by keyword search...")
+                Accessible.role: Accessible.EditableText
+                Accessible.name: placeholderText
+                Accessible.description: qsTr("Text field for discovering and filtering downloadable models")
+                Connections {
+                    target: ModelList
+                    function onDiscoverInProgressChanged() {
+                        if (ModelList.discoverInProgress) {
+                            discoverField.textBeingSearched = discoverField.text;
+                            discoverField.text = qsTr("Searching \u00B7 ") + discoverField.textBeingSearched;
+                        } else {
+                            discoverField.text = discoverField.textBeingSearched;
+                            discoverField.textBeingSearched = "";
+                        }
+                    }
+                }
+                background: ProgressBar {
+                    id: discoverProgressBar
+                    indeterminate: ModelList.discoverInProgress && ModelList.discoverProgress === 0.0
+                    value: ModelList.discoverProgress
+                    background: Rectangle {
+                        color: theme.controlBackground
+                        radius: 10
+                    }
+                    contentItem: Item {
+                        Rectangle {
+                            visible: ModelList.discoverInProgress
+                            anchors.bottom: parent.bottom
+                            width: discoverProgressBar.visualPosition * parent.width
+                            height: 10
+                            radius: 2
+                            color: theme.progressForeground
+                        }
+                    }
+                }
+
+                Keys.onReturnPressed: (event)=> {
+                    if (event.modifiers & Qt.ControlModifier || event.modifiers & Qt.ShiftModifier)
+                        event.accepted = false;
+                    else {
+                        editingFinished();
+                        sendDiscovery()
+                    }
+                }
+                function sendDiscovery() {
+                    ModelList.downloadableModels.discoverAndFilter(discoverField.text);
+                }
+                RowLayout {
+                    spacing: 0
+                    anchors.right: discoverField.right
+                    anchors.verticalCenter: discoverField.verticalCenter
+                    anchors.rightMargin: 15
+                    visible: !ModelList.discoverInProgress
+                    MyMiniButton {
+                        id: clearDiscoverButton
+                        backgroundColor: theme.textColor
+                        backgroundColorHovered: theme.iconBackgroundDark
+                        visible: discoverField.text !== ""
+                        contentItem: Text {
+                            color: clearDiscoverButton.hovered ? theme.iconBackgroundDark : theme.textColor
+                            text: "\u2715"
+                            font.pixelSize: theme.fontSizeLarge
+                        }
+                        onClicked: {
+                            discoverField.text = ""
+                            discoverField.sendDiscovery() // should clear results
+                        }
+                    }
+                    MyMiniButton {
+                        backgroundColor: theme.textColor
+                        backgroundColorHovered: theme.iconBackgroundDark
+                        source: "qrc:/gpt4all/icons/settings.svg"
+                        onClicked: {
+                            discoveryTools.visible = !discoveryTools.visible
+                        }
+                    }
+                    MyMiniButton {
+                        id: sendButton
+                        enabled: !ModelList.discoverInProgress
+                        backgroundColor: theme.textColor
+                        backgroundColorHovered: theme.iconBackgroundDark
+                        source: "qrc:/gpt4all/icons/send_message.svg"
+                        Accessible.name: qsTr("Initiate model discovery and filtering")
+                        Accessible.description: qsTr("Triggers discovery and filtering of models")
+                        onClicked: {
+                            discoverField.sendDiscovery()
+                        }
+                    }
+                }
+            }
+        }
+
+        RowLayout {
+            id: discoveryTools
+            Layout.fillWidth: true
+            Layout.alignment: Qt.AlignCenter
+            Layout.margins: 0
+            spacing: 20
+            visible: false
+            MyComboBox {
+                id: comboSort
+                model: [qsTr("Default"), qsTr("Likes"), qsTr("Downloads"), qsTr("Recent")]
+                currentIndex: ModelList.discoverSort
+                contentItem: Text {
+                    anchors.horizontalCenter: parent.horizontalCenter
+                    rightPadding: 30
+                    color: theme.textColor
+                    text: {
+                        return qsTr("Sort by: ") + comboSort.displayText
+                    }
+                    font.pixelSize: theme.fontSizeLarger
+                    verticalAlignment: Text.AlignVCenter
+                    horizontalAlignment: Text.AlignHCenter
+                    elide: Text.ElideRight
+                }
+                onActivated: function (index) {
+                    ModelList.discoverSort = index;
+                }
+            }
+            MyComboBox {
+                id: comboSortDirection
+                model: [qsTr("Asc"), qsTr("Desc")]
+                currentIndex: {
+                    if (ModelList.discoverSortDirection === 1)
+                        return 0
+                    else
+                        return 1;
+                }
+                contentItem: Text {
+                    anchors.horizontalCenter: parent.horizontalCenter
+                    rightPadding: 30
+                    color: theme.textColor
+                    text: {
+                        return qsTr("Sort dir: ") + comboSortDirection.displayText
+                    }
+                    font.pixelSize: theme.fontSizeLarger
+                    verticalAlignment: Text.AlignVCenter
+                    horizontalAlignment: Text.AlignHCenter
+                    elide: Text.ElideRight
+                }
+                onActivated: function (index) {
+                    if (index === 0)
+                        ModelList.discoverSortDirection = 1;
+                    else
+                        ModelList.discoverSortDirection = -1;
+                }
+            }
+            MyComboBox {
+                id: comboLimit
+                model: ["5", "10", "20", "50", "100", qsTr("None")]
+                currentIndex: {
+                    if (ModelList.discoverLimit === 5)
+                        return 0;
+                    else if (ModelList.discoverLimit === 10)
+                        return 1;
+                    else if (ModelList.discoverLimit === 20)
+                        return 2;
+                    else if (ModelList.discoverLimit === 50)
+                        return 3;
+                    else if (ModelList.discoverLimit === 100)
+                        return 4;
+                    else if (ModelList.discoverLimit === -1)
+                        return 5;
+                }
+                contentItem: Text {
+                    anchors.horizontalCenter: parent.horizontalCenter
+                    rightPadding: 30
+                    color: theme.textColor
+                    text: {
+                        return qsTr("Limit: ") + comboLimit.displayText
+                    }
+                    font.pixelSize: theme.fontSizeLarger
+                    verticalAlignment: Text.AlignVCenter
+                    horizontalAlignment: Text.AlignHCenter
+                    elide: Text.ElideRight
+                }
+                onActivated: function (index) {
+                    switch (index) {
+                    case 0:
+                        ModelList.discoverLimit = 5; break;
+                    case 1:
+                        ModelList.discoverLimit = 10; break;
+                    case 2:
+                        ModelList.discoverLimit = 20; break;
+                    case 3:
+                        ModelList.discoverLimit = 50; break;
+                    case 4:
+                        ModelList.discoverLimit = 100; break;
+                    case 5:
+                        ModelList.discoverLimit = -1; break;
+                    }
+                }
+            }
         }
 
         Label {
@@ -60,7 +265,7 @@ MyDialog {
             Layout.fillHeight: true
             horizontalAlignment: Qt.AlignHCenter
             verticalAlignment: Qt.AlignVCenter
-            text: qsTr("Network error: could not retrieve http://gpt4all.io/models/models2.json")
+            text: qsTr("Network error: could not retrieve http://gpt4all.io/models/models3.json")
             font.pixelSize: theme.fontSizeLarge
             color: theme.mutedTextColor
         }
@@ -213,15 +418,11 @@ MyDialog {
                                         Layout.leftMargin: 20
                                         textFormat: Text.StyledText
                                         text: "<strong><font size=\"1\">"
-                                            + (qsTr("Download size: ") + filesize)
-                                            + "<br>"
-                                            + (qsTr("RAM required: ") + (ramrequired > 0 ? ramrequired + " GB" : qsTr("minimal")))
-                                            + "<br>"
-                                            + (qsTr("Parameters: ") + parameters)
-                                            + "<br>"
-                                            + (qsTr("Quantization: ") + quant)
-                                            + "<br>"
-                                            + (qsTr("Type: ") + type)
+                                            + (qsTr("File size: ") + filesize)
+                                            + (ramrequired < 0 ? "" : "<br>" + (qsTr("RAM required: ") + (ramrequired > 0 ? ramrequired + " GB" : qsTr("minimal"))))
+                                            + (parameters === "" ? "" : "<br>" + qsTr("Parameters: ") + parameters)
+                                            + (quant === "" ? "" : "<br>" + (qsTr("Quantization: ") + quant))
+                                            + (type === "" ? "" : "<br>" + (qsTr("Type: ") + type))
                                             + "</strong></font>"
                                         color: theme.textColor
                                         font.pixelSize: theme.fontSizeLarge
@@ -350,11 +551,6 @@ MyDialog {
                                     Accessible.role: Accessible.EditableText
                                     Accessible.name: placeholderText
                                     Accessible.description: qsTr("Whether the file hash is being calculated")
-                                    TextMetrics {
-                                        id: textMetrics
-                                        font: apiKey.font
-                                        text: apiKey.placeholderText
-                                    }
                                 }
                             }
                         }

gpt4all-chat/qml/ModelSettings.qml

@@ -82,7 +82,7 @@ MySettingsTab {
                 enabled: root.currentModelInfo.isClone
                 text: qsTr("Remove")
                 onClicked: {
-                    ModelList.remove(root.currentModelInfo);
+                    ModelList.removeClone(root.currentModelInfo);
                     comboBox.currentIndex = 0;
                 }
             }
@@ -453,7 +453,7 @@ MySettingsTab {
                 Accessible.description: ToolTip.text
             }
             MySettingsLabel {
-                id: minPLabel 
+                id: minPLabel
                 text: qsTr("Min P")
                 Layout.row: 3
                 Layout.column: 0

gpt4all-chat/qml/MyMiniButton.qml

@@ -11,8 +11,8 @@ Button {
     property color backgroundColorHovered: theme.iconBackgroundHovered
     property alias source: image.source
     property alias fillMode: image.fillMode
-    width: 30
-    height: 30
+    implicitWidth: 30
+    implicitHeight: 30
     contentItem: Text {
         text: myButton.text
         horizontalAlignment: Text.AlignHCenter