Dev

automation_ide/automation_editor_ui/extend_ai_gui/ai_gui_global_variable.py

@@ -4,8 +4,6 @@
     FIRST_CODE_REVIEW_TEMPLATE
 from automation_ide.automation_editor_ui.extend_ai_gui.prompt_edit_gui.cot_code_review_prompt_templates.first_summary_prompt import \
     FIRST_SUMMARY_TEMPLATE
-from automation_ide.automation_editor_ui.extend_ai_gui.prompt_edit_gui.cot_code_review_prompt_templates.global_rule import \
-    GLOBAL_RULE_TEMPLATE
 from automation_ide.automation_editor_ui.extend_ai_gui.prompt_edit_gui.cot_code_review_prompt_templates.linter import \
     LINTER_TEMPLATE
 from automation_ide.automation_editor_ui.extend_ai_gui.prompt_edit_gui.cot_code_review_prompt_templates.total_summary import \
@@ -16,7 +14,6 @@
     CODE_REVIEW_SKILL_TEMPLATE
 
 COT_TEMPLATE_FILES = [
-    "global_rule.md",
     "first_summary_prompt.md",
     "first_code_review.md",
     "linter.md",
@@ -25,7 +22,6 @@
 ]
 
 COT_TEMPLATE_RELATION = {
-    "global_rule.md": GLOBAL_RULE_TEMPLATE,
     "first_summary_prompt.md": FIRST_SUMMARY_TEMPLATE,
     "first_code_review.md": FIRST_CODE_REVIEW_TEMPLATE,
     "linter.md": LINTER_TEMPLATE,

automation_ide/automation_editor_ui/extend_ai_gui/cot_code_review_gui.py → automation_ide/automation_editor_ui/extend_ai_gui/code_review/code_review_thread.py

@@ -1,21 +1,13 @@
-import sys
-
+# Worker Thread 負責傳送資料
 import requests
 from PySide6.QtCore import QThread, Signal
-from PySide6.QtWidgets import (
-    QApplication, QWidget, QVBoxLayout, QHBoxLayout,
-    QPushButton, QTextEdit, QLabel, QLineEdit, QComboBox, QMessageBox
-)
 from je_editor import language_wrapper
 
-from automation_ide.automation_editor_ui.extend_ai_gui.ai_gui_global_variable import COT_TEMPLATE_FILES, \
-    COT_TEMPLATE_RELATION
+from automation_ide.automation_editor_ui.extend_ai_gui.ai_gui_global_variable import COT_TEMPLATE_RELATION
 from automation_ide.automation_editor_ui.extend_ai_gui.prompt_edit_gui.cot_code_review_prompt_templates.global_rule import \
     build_global_rule_template
-from automation_ide.automation_editor_ui.extend_multi_language.update_language_dict import update_language_dict
 
 
-# Worker Thread 負責傳送資料
 class SenderThread(QThread):
     update_response = Signal(str, str)  # (filename, response)
 
@@ -86,80 +78,4 @@ def run(self):
                 reply_text = f"{language_wrapper.language_word_dict.get("cot_gui_error_sending")} {file} {e}"
 
             # 發送訊號更新 UI
-            self.update_response.emit(file, reply_text)
-
-
-class CoTCodeReviewGUI(QWidget):
-    def __init__(self):
-        super().__init__()
-        self.setWindowTitle(language_wrapper.language_word_dict.get("cot_gui_window_title"))
-
-        # 檔案清單
-        self.files = COT_TEMPLATE_FILES
-
-        # UI 元件
-        layout = QVBoxLayout()
-
-        # URL 輸入框
-        url_layout = QHBoxLayout()
-        url_layout.addWidget(QLabel(language_wrapper.language_word_dict.get("cot_gui_label_api_url")))
-        self.url_input = QLineEdit()
-        self.url_input.setPlaceholderText(language_wrapper.language_word_dict.get("cot_gui_placeholder_api_url"))
-        url_layout.addWidget(self.url_input)
-        layout.addLayout(url_layout)
-
-        # 傳送資料區域
-        self.code_paste_area = QTextEdit()
-        self.code_paste_area.setPlaceholderText(
-            language_wrapper.language_word_dict.get("cot_gui_placeholder_code_paste_area"))
-        layout.addWidget(QLabel(language_wrapper.language_word_dict.get("cot_gui_label_prompt_area")))
-        layout.addWidget(self.code_paste_area)
-
-        # 回傳區域
-        self.response_selector = QComboBox()  # 改用 ComboBox
-        self.response_view = QTextEdit()
-        self.response_view.setReadOnly(True)  # 可複製但不可編輯
-
-        hbox_layout = QHBoxLayout()
-        hbox_layout.addWidget(self.response_selector, 2)
-        hbox_layout.addWidget(self.response_view, 5)
-
-        layout.addWidget(QLabel(language_wrapper.language_word_dict.get("cot_gui_label_response_area")))
-        layout.addLayout(hbox_layout)
-
-        # 傳送按鈕
-        self.send_button = QPushButton(language_wrapper.language_word_dict.get("cot_gui_button_send"))
-        layout.addWidget(self.send_button)
-
-        self.setLayout(layout)
-
-        # 綁定事件
-        self.response_selector.currentTextChanged.connect(self.show_response)
-        self.send_button.clicked.connect(self.start_sending)
-
-        # 儲存回覆
-        self.responses = {}
-
-    def show_response(self, filename):
-        if filename in self.responses:
-            self.response_view.setPlainText(self.responses[filename])
-
-    def start_sending(self):
-        # 取得 URL
-        url = self.url_input.text().strip()
-        if not url:
-            message_box = QMessageBox()
-            message_box.warning(self, "Warning", language_wrapper.language_word_dict.get("cot_gui_error_no_url"))
-            message_box.exec_()
-            return
-
-        # 啟動傳送 Thread
-        self.thread = SenderThread(files=self.files, code=self.code_paste_area.toPlainText(), url=url)
-        self.thread.update_response.connect(self.handle_response)
-        self.thread.start()
-
-    def handle_response(self, filename, response):
-        self.responses[filename] = response
-        self.response_selector.addItem(filename)  # 加入 ComboBox
-        # 自動顯示最新回覆
-        self.response_selector.setCurrentText(filename)
+            self.update_response.emit(file, reply_text)

automation_ide/automation_editor_ui/extend_ai_gui/code_review/cot_code_review_gui.py

@@ -0,0 +1,82 @@
+from PySide6.QtWidgets import QWidget, QVBoxLayout, QHBoxLayout, QLabel, QLineEdit, QTextEdit, QComboBox, QPushButton, \
+    QMessageBox
+from je_editor import language_wrapper
+
+from automation_ide.automation_editor_ui.extend_ai_gui.ai_gui_global_variable import COT_TEMPLATE_FILES
+from automation_ide.automation_editor_ui.extend_ai_gui.code_review.code_review_thread import SenderThread
+
+
+class CoTCodeReviewGUI(QWidget):
+    def __init__(self):
+        super().__init__()
+        self.setWindowTitle(language_wrapper.language_word_dict.get("cot_gui_window_title"))
+
+        # 檔案清單
+        self.files = COT_TEMPLATE_FILES
+
+        # UI 元件
+        layout = QVBoxLayout()
+
+        # URL 輸入框
+        url_layout = QHBoxLayout()
+        url_layout.addWidget(QLabel(language_wrapper.language_word_dict.get("cot_gui_label_api_url")))
+        self.url_input = QLineEdit()
+        self.url_input.setPlaceholderText(language_wrapper.language_word_dict.get("cot_gui_placeholder_api_url"))
+        url_layout.addWidget(self.url_input)
+        layout.addLayout(url_layout)
+
+        # 傳送資料區域
+        self.code_paste_area = QTextEdit()
+        self.code_paste_area.setPlaceholderText(
+            language_wrapper.language_word_dict.get("cot_gui_placeholder_code_paste_area"))
+        layout.addWidget(QLabel(language_wrapper.language_word_dict.get("cot_gui_label_prompt_area")))
+        layout.addWidget(self.code_paste_area)
+
+        # 回傳區域
+        self.response_selector = QComboBox()  # 改用 ComboBox
+        self.response_view = QTextEdit()
+        self.response_view.setReadOnly(True)  # 可複製但不可編輯
+
+        hbox_layout = QHBoxLayout()
+        hbox_layout.addWidget(self.response_selector, 2)
+        hbox_layout.addWidget(self.response_view, 5)
+
+        layout.addWidget(QLabel(language_wrapper.language_word_dict.get("cot_gui_label_response_area")))
+        layout.addLayout(hbox_layout)
+
+        # 傳送按鈕
+        self.send_button = QPushButton(language_wrapper.language_word_dict.get("cot_gui_button_send"))
+        layout.addWidget(self.send_button)
+
+        self.setLayout(layout)
+
+        # 綁定事件
+        self.response_selector.currentTextChanged.connect(self.show_response)
+        self.send_button.clicked.connect(self.start_sending)
+
+        # 儲存回覆
+        self.responses = {}
+
+    def show_response(self, filename):
+        if filename in self.responses:
+            self.response_view.setPlainText(self.responses[filename])
+
+    def start_sending(self):
+        # 取得 URL
+        url = self.url_input.text().strip()
+        if not url:
+            message_box = QMessageBox()
+            message_box.warning(self, "Warning", language_wrapper.language_word_dict.get("cot_gui_error_no_url"))
+            message_box.exec_()
+            return
+
+        # 啟動傳送 Thread
+        self.thread = SenderThread(files=self.files, code=self.code_paste_area.toPlainText(), url=url)
+        self.thread.update_response.connect(self.handle_response)
+        self.thread.start()
+
+    def handle_response(self, filename, response):
+        self.responses[filename] = response
+        self.response_selector.addItem(filename)  # 加入 ComboBox
+        # 自動顯示最新回覆
+        self.response_selector.setCurrentText(filename)

automation_ide/automation_editor_ui/extend_ai_gui/prompt_edit_gui/cot_code_review_prompt_templates/judge.py

@@ -0,0 +1,91 @@
+JUDGE_TEMPLATE = """
+# Code Review Comment Evaluation Template (Enhanced)
+
+Please evaluate the review comments focusing on how well they address important issues in the code, especially leveraging code smell and linter messages.  
+The input may contain multiple independent review reports.
+
+Score range: 1–100
+
+Five evaluation dimensions:
+### 1. Readability
+    - 1–20: Comments are very hard to understand, poorly structured, confusing language.
+    - 21–40: Some parts are readable, but many unclear sections remain.
+    - 41–60: Comments are generally clear, but structure or phrasing needs improvement.
+    - 61–80: Comments are well-structured, consistent, and easy to follow.
+    - 81–100: Comments are highly readable, elegantly phrased, and well-organized.
+
+### 2. Constructiveness (Maintainability)
+    - 1–20: Comments lack constructive suggestions, no improvement direction.
+    - 21–40: Comments provide partial suggestions, but vague or impractical.
+    - 41–60: Comments offer basic improvement ideas, somewhat helpful.
+    - 61–80: Comments are specific and actionable, clearly guiding improvements.
+    - 81–100: Comments are highly constructive, offering clear and practical improvement paths.
+
+### 3. Correctness
+    - 1–20: Comments contain errors or misleading advice.
+    - 21–40: Mostly correct, but important issues are overlooked.
+    - 41–60: Largely correct, with only minor gaps.
+    - 61–80: Correct and reasonable, with small room for refinement.
+    - 81–100: Fully correct, logically sound, and precise in identifying issues.
+
+### 4. Multi-Review Coverage, Structural Independence & Extractability
+
+    > Evaluate how well the comments cover important issues across multiple review reports, and whether each comment block is structurally independent, self-contained, and understandable on its own.
+
+    #### Scoring Criteria
+
+    - **1–20**  
+      Rarely addresses important issues; structure is disorganized; comments heavily depend on surrounding context; blocks cannot be read independently; unclear linkage to specific code smells or linter messages.
+
+    - **21–40**  
+      Addresses some issues but misses many key points; unclear boundaries between sections; frequent cross-references required for understanding; difficult to isolate specific issue blocks.
+
+    - **41–60**  
+      Covers many important issues; basic structure is present; some blocks can be read independently, but certain sections still rely on context or lack completeness.
+
+    - **61–80**  
+      Most key issues are addressed; comment blocks are clearly structured with defined themes; most sections can be independently read and understood (problem + reasoning + suggestion); code smell / linter-related blocks are reasonably extractable.
+
+    - **81–100**  
+      Thoroughly addresses key issues; each comment block is **independent, complete, and self-contained** (including problem description, impact explanation, and actionable improvement suggestions);  
+      clearly segmented structure; any block can be extracted without losing meaning;  
+      strongly aligned with specific code smells or linter messages; highly readable and maintainable.
+    ---
+    ### Additional Evaluation Criteria
+
+    When scoring this dimension, explicitly check whether each comment block:
+    - Has a clear title or thematic focus  
+    - Clearly identifies the issue source (e.g., specific code smell or linter message)  
+    - Explains impact or risk  
+    - Provides concrete and actionable improvement suggestions  
+    - Does not rely on other sections for comprehension  
+
+    Also assess whether:
+    - A single comment block can be extracted and shared independently without losing clarity  
+    - Code smell or linter-related sections can be isolated for structured analysis  
+    - Summary sections and detailed comments are clearly distinguished  
+
+### 5. Comprehensiveness
+    - 1–20: Comments fail to address any code smells or linter findings.
+    - 21–40: Comments mention at least one code smell or linter warning.
+    - 41–60: Comments cover some code smells or linter findings.
+    - 61–80: Comments cover most code smells and linter findings.
+    - 81–100: Comments comprehensively address all code smells and linter findings, with improvement suggestions.
+
+Reviewers should:
+- Assign a score (1–100) for each dimension.
+- Provide brief reasoning for each score.
+- Conclude with an average score and overall recommendation.
+
+## Review Comment:
+{review_comment}
+
+## Code Smells:
+{code_smell_detector_messages}
+
+## Linter Messages:
+{linter_messages}
+
+## Origin code
+{code_diff}
+"""

automation_ide/automation_editor_ui/extend_ai_gui/prompt_edit_gui/cot_code_review_prompt_templates/judge_single_review.py

@@ -0,0 +1,85 @@
+JUDGE_SINGLE_REVIEW_TEMPLATE = """
+# Code Review Comment Evaluation Template (Enhanced)
+
+Please evaluate the review comments focusing on how well they address important issues in the code, especially leveraging code smell and linter messages.  
+The input may contain multiple independent review reports.
+
+Score range: 1–100
+
+Five evaluation dimensions:
+### 1. Readability
+    - 1–20: Comments are very hard to understand, poorly structured, confusing language.
+    - 21–40: Some parts are readable, but many unclear sections remain.
+    - 41–60: Comments are generally clear, but structure or phrasing needs improvement.
+    - 61–80: Comments are well-structured, consistent, and easy to follow.
+    - 81–100: Comments are highly readable, elegantly phrased, and well-organized.
+
+### 2. Constructiveness (Maintainability)
+    - 1–20: Comments lack constructive suggestions, no improvement direction.
+    - 21–40: Comments provide partial suggestions, but vague or impractical.
+    - 41–60: Comments offer basic improvement ideas, somewhat helpful.
+    - 61–80: Comments are specific and actionable, clearly guiding improvements.
+    - 81–100: Comments are highly constructive, offering clear and practical improvement paths.
+
+### 3. Correctness
+    - 1–20: Comments contain errors or misleading advice.
+    - 21–40: Mostly correct, but important issues are overlooked.
+    - 41–60: Largely correct, with only minor gaps.
+    - 61–80: Correct and reasonable, with small room for refinement.
+    - 81–100: Fully correct, logically sound, and precise in identifying issues.
+
+### 4. Multi-Review Coverage, Structural Independence & Extractability
+
+    > Evaluate how well the comments cover important issues across multiple review reports, and whether each comment block is structurally independent, self-contained, and understandable on its own.
+
+    #### Scoring Criteria
+
+    - **1–20**  
+      Rarely addresses important issues; structure is disorganized; comments heavily depend on surrounding context; blocks cannot be read independently; unclear linkage to specific code smells or linter messages.
+
+    - **21–40**  
+      Addresses some issues but misses many key points; unclear boundaries between sections; frequent cross-references required for understanding; difficult to isolate specific issue blocks.
+
+    - **41–60**  
+      Covers many important issues; basic structure is present; some blocks can be read independently, but certain sections still rely on context or lack completeness.
+
+    - **61–80**  
+      Most key issues are addressed; comment blocks are clearly structured with defined themes; most sections can be independently read and understood (problem + reasoning + suggestion); code smell / linter-related blocks are reasonably extractable.
+
+    - **81–100**  
+      Thoroughly addresses key issues; each comment block is **independent, complete, and self-contained** (including problem description, impact explanation, and actionable improvement suggestions);  
+      clearly segmented structure; any block can be extracted without losing meaning;  
+      strongly aligned with specific code smells or linter messages; highly readable and maintainable.
+    ---
+    ### Additional Evaluation Criteria
+
+    When scoring this dimension, explicitly check whether each comment block:
+    - Has a clear title or thematic focus  
+    - Clearly identifies the issue source (e.g., specific code smell or linter message)  
+    - Explains impact or risk  
+    - Provides concrete and actionable improvement suggestions  
+    - Does not rely on other sections for comprehension  
+
+    Also assess whether:
+    - A single comment block can be extracted and shared independently without losing clarity  
+    - Code smell or linter-related sections can be isolated for structured analysis  
+    - Summary sections and detailed comments are clearly distinguished  
+
+### 5. Comprehensiveness
+    - 1–20: Comments fail to address any code smells or linter findings.
+    - 21–40: Comments mention at least one code smell or linter warning.
+    - 41–60: Comments cover some code smells or linter findings.
+    - 61–80: Comments cover most code smells and linter findings.
+    - 81–100: Comments comprehensively address all code smells and linter findings, with improvement suggestions.
+
+Reviewers should:
+- Assign a score (1–100) for each dimension.
+- Provide brief reasoning for each score.
+- Conclude with an average score and overall recommendation.
+
+## Review Comment:
+{review_comment}
+
+## Origin code
+{code_diff}
+"""