fix: [codegeex] optimized efficiency and functionality

1. The speed of inserting vector data into the database has become faster.
2. fix some bugs related multithreads.
3. Ui and tip

log: as title

src/plugins/codegeex/codegeex/askapi.cpp

@@ -53,17 +53,21 @@ class AskApiPrivate : public QObject
     QString locale = "zh";
     bool codebaseEnabled = false;
     bool networkEnabled = false;
+    bool terminated = false;
     QStringList referenceFiles;
 };
 
 AskApiPrivate::AskApiPrivate(AskApi *qq)
     : q(qq),
       manager(new QNetworkAccessManager(qq))
 {
+    connect(q, &AskApi::stopReceive, this, [=](){ terminated = true; });
 }
 
 QNetworkReply *AskApiPrivate::postMessage(const QString &url, const QString &token, const QByteArray &body)
 {
+    if (terminated)
+        return nullptr;
     QNetworkRequest request(url);
     request.setHeader(QNetworkRequest::ContentTypeHeader, "application/json");
     request.setRawHeader("code-token", token.toUtf8());
@@ -192,7 +196,7 @@ QByteArray AskApiPrivate::assembleSSEChatBody(const QString &prompt, const QStri
         if (!chunks.isEmpty()) {
             CodeGeeXManager::instance()->cleanHistoryMessage();   // incase history is too big
             if (result["Completed"].toBool() == false)
-                CodeGeeXManager::instance()->notify(0, CodeGeeXManager::tr("The indexing of project %1 has not been completed, which may cause the results to be inaccurate.").arg(currentProjectPath));
+                emit q->notify(0, CodeGeeXManager::tr("The indexing of project %1 has not been completed, which may cause the results to be inaccurate.").arg(currentProjectPath));
             jsonObject["history"] = QJsonArray();
             QString context;
             context += prompt;
@@ -283,6 +287,11 @@ AskApi::AskApi(QObject *parent)
       d(new AskApiPrivate(this))
 {
     connect(this, &AskApi::syncSendMessage, this, &AskApi::slotSendMessage);
+    connect(this, &AskApi::notify, this, [](int type, const QString &message) {
+        using namespace dpfservice;
+        WindowService *windowService = dpfGetService(WindowService);
+        windowService->notify(type, "Ai", message, QStringList {});
+    });
 }
 
 AskApi::~AskApi()
@@ -373,6 +382,7 @@ void AskApi::postSSEChat(const QString &url,
                          const QMultiMap<QString, QString> &history,
                          const QString &talkId)
 {
+    d->terminated = false;
     QJsonArray jsonArray = convertHistoryToJSONArray(history);
 
 #ifdef SUPPORTMINIFORGE

src/plugins/codegeex/codegeex/askapi.h

@@ -108,6 +108,7 @@ class AskApi : public QObject
     void stopReceive();
     void syncSendMessage(const QString url, const QString &token, const QByteArray &body);
     void noChunksFounded();
+    void notify(int type, const QString &message);
 
 public slots:
     void slotSendMessage(const QString url, const QString &token, const QByteArray &body);

src/plugins/codegeex/widgets/inputeditwidget.cpp

@@ -227,6 +227,11 @@ void InputEdit::onTextChanged()
     QSet<QString> tagList;
     int last_pos = 0;
 
+    // bug: tag will removed when send message. and causes the tag to be reset before it is used.
+    // update tag when codegeex is not running
+    if (CodeGeeXManager::instance()->checkRunningState(true))
+        return;
+
     cursor.setPosition(0);
     formatList.clear();
 

src/plugins/codegeex/widgets/intropage.cpp

@@ -68,7 +68,8 @@ void IntroPage::initIntroContent()
     introLayout->setSpacing(10);
     qobject_cast<QVBoxLayout *>(layout())->addLayout(introLayout);
 
-    appendDescLabel(introLayout, tr("CodeGeeX provides code completion suggestions in editor, Press %1 Ctrl + T %2 to accept.").arg("<font style='color:dodgerblue;'>", "</font>"));
+    appendDescLabel(introLayout, tr("CodeGeeX provides code completion suggestions in editor, Press %1 Tab %2 to accept.").arg("<font style='color:dodgerblue;'>", "</font>"));
+    appendDescLabel(introLayout, tr("CodeGeeX provides inline chat functionality in editor, Press %1 Ctrl + T %2 to use it.").arg("<font style='color:dodgerblue;'>", "</font>"));
     appendDescLabel(introLayout, tr("Select code and %1 right-click %2 to add comments or translate code.").arg("<font style='color:dodgerblue;'>", "</font>"));
     appendDescLabel(introLayout, tr("Also, you can directly %1 ask CodeGeeX any questions %2.").arg("<font style='color:dodgerblue;'>", "</font>"));
 }

src/scripts/rag/generate.py

@@ -258,7 +258,7 @@ def _insert_chunks_sync(db: sqlite3.Connection, tag_string: str, chunks: List[Di
     finally:
         cursor.close()
 
-def _insert_embedding_sync(db: sqlite3.Connection, vector: bytes, chunk: Dict[str, Any]):
+def _insert_embedding_sync(db: sqlite3.Connection, embedding_map: Dict[str, bytes], chunks: List[Dict[str, Any]]):
     cursor = db.db.cursor()
     try:
         cursor.execute("BEGIN")
@@ -267,21 +267,20 @@ def _insert_embedding_sync(db: sqlite3.Connection, vector: bytes, chunk: Dict[st
             INSERT INTO lance_db_cache (uuid, cacheKey, path, artifact_id, vector, startLine, endLine, contents) 
             VALUES (?, ?, ?, ?, ?, ?, ?, ?)
         """
-
-        cursor.execute(embedding_sql, (
-            uuid.uuid4().hex,
-            chunk["digest"],
-            chunk["filepath"],
-            #chunk["index"],
-            "all-MiniLM-L6-v2", 
-            vector,
-            chunk["startLine"],
-            chunk["endLine"],
-            chunk["content"]
-        ))
+        for chunk in chunks:
+            cursor.execute(embedding_sql, (
+                str(uuid.uuid4()),
+                chunk["digest"],
+                chunk["filepath"],
+                "lance",
+                embedding_map[chunk["digest"]],
+                chunk["startLine"],
+                chunk["endLine"],
+                chunk["content"]
+            ))
 
         if cursor.rowcount == 0:
-            raise Exception("Failed to insert into embeddings table")
+            raise Exception("Failed to insert any embeddings into the table")
 
         db.db.commit()
     except Exception as e:
@@ -324,6 +323,7 @@ def initDb(db: sqlite3.Connection):
 
 def embeddingDirectory(dir: str, provider: ONNXEmbeddingsProvider, db: sqlite3.Connection):
     for entry in os.listdir(dir):
+        embedding_map = {}
         full_path = os.path.join(dir, entry)
         if "/3rdparty/" in full_path or "/.unioncode/" in full_path:
             continue
@@ -335,14 +335,13 @@ def embeddingDirectory(dir: str, provider: ONNXEmbeddingsProvider, db: sqlite3.C
             code = file.read()
             max_chunk_size = 1024
             chunks = list(chunk_document(full_path, code, max_chunk_size))
-            _insert_chunks_sync(db, "test", chunks)
-
             for chunk in chunks:
                 content = chunk['content']
                 embedding = provider.embed_single(content)
                 provider.code_blocks.append(content)
-                _insert_embedding_sync(db, embedding.tobytes(), chunk)
-
+                embedding_map[chunk['digest']] = embedding.tobytes()
+            _insert_chunks_sync(db, "chunk", chunks)
+            _insert_embedding_sync(db, embedding_map, chunks)
         elif os.path.isdir(full_path):
             embeddingDirectory(full_path, provider, db)