Merge pull request Pythagora-io#693 from Pythagora-io/fixes

Fixes

pilot/const/messages.py

@@ -1,7 +1,7 @@
 CHECK_AND_CONTINUE = 'Is everything working? Let me know if something needs to be changed for this task or type "continue" to proceed.'
 WHEN_USER_DONE = 'Once you have completed, enter "continue"'
-AFFIRMATIVE_ANSWERS = ['', 'y', 'yes', 'ok', 'okay', 'sure', 'absolutely', 'indeed', 'correct', 'affirmative', 'Use GPT Pilot\'s code']
-NEGATIVE_ANSWERS = ['n', 'no', 'skip', 'negative', 'not now', 'cancel', 'decline', 'stop', 'Keep my changes']
+AFFIRMATIVE_ANSWERS = ['', 'y', 'yes', 'ok', 'okay', 'sure', 'absolutely', 'indeed', 'correct', 'affirmative']
+NEGATIVE_ANSWERS = ['n', 'no', 'skip', 'negative', 'not now', 'cancel', 'decline', 'stop']
 STUCK_IN_LOOP = 'I\'m stuck in loop'
 NONE_OF_THESE = 'none of these'
 MAX_PROJECT_NAME_LENGTH = 50

pilot/database/database.py

@@ -369,7 +369,7 @@ def save_command_run(project, command, cli_response, done_or_error_response, exi
     unique_data = {
         'app': project.args['app_id'],
         'previous_step': project.checkpoints['last_command_run'],
-        'high_level_step': project.current_step,
+        'high_level_step': str(project.checkpoints['last_development_step']['id']) if project.checkpoints['last_development_step'] else None,
     }
 
     data_fields = {
@@ -391,7 +391,7 @@ def save_user_input(project, query, user_input, hint):
     unique_data = {
         'app': project.args['app_id'],
         'previous_step': project.checkpoints['last_user_input'],
-        'high_level_step': project.current_step,
+        'high_level_step': str(project.checkpoints['last_development_step']['id']) if project.checkpoints['last_development_step'] else None,
     }
     data_fields = {
         'query': query,

pilot/helpers/Project.py

@@ -34,7 +34,7 @@
 
 from utils.telemetry import telemetry
 from utils.task import Task
-from utils.utils import remove_lines_with_string, is_extension_old_version
+from utils.utils import remove_lines_with_string
 
 
 class Project:
@@ -58,8 +58,6 @@ def __init__(
             current_step (str, optional): Current step in the project. Default is None.
         """
         self.args = args
-        # TODO remove everything related to is_extension_old_version once new version is released and everybody has to update core
-        self.is_extension_old_version = is_extension_old_version(args)
         self.llm_req_num = 0
         self.command_runs_count = 0
         self.user_inputs_count = 0
@@ -98,7 +96,7 @@ def __init__(
             File.update_paths()
 
         # start loading of project (since backwards compatibility)
-        self.should_overwrite_files = False
+        self.should_overwrite_files = None
         self.last_detailed_user_review_goal = None
         self.last_iteration = None
         self.tasks_to_load = []
@@ -118,22 +116,41 @@ def setup_loading(self):
             return
 
         self.skip_steps = True
-        should_overwrite_files = None
-        while should_overwrite_files is None or should_overwrite_files.lower() not in AFFIRMATIVE_ANSWERS + NEGATIVE_ANSWERS:
-            print('Use GPT Pilot\'s code/Keep my changes', type='buttons-only')
-            should_overwrite_files = styled_text(
+        while self.should_overwrite_files is None:
+            changes_made_question = f'Did you make any changes to "{self.args["name"]}" project files since last time you used Pythagora?'
+            print(changes_made_question, type='ipc', category='pythagora')
+            print('yes/no', type='buttons-only')
+            # must use styled_text() instead of ask_user() here to avoid finish_loading() call
+            changes_made = styled_text(
                 self,
-                "Can GPT Pilot overwrite code changes you made since last running GPT Pilot?",
-                ignore_user_input_count=True
+                changes_made_question,
+                ignore_user_input_count=True,
             )
 
-            logger.info('should_overwrite_files: %s', should_overwrite_files)
-            if should_overwrite_files in NEGATIVE_ANSWERS:
-                self.should_overwrite_files = False
-                break
-            elif should_overwrite_files in AFFIRMATIVE_ANSWERS:
+            # if there were no changes just load files from db
+            if changes_made.lower() in NEGATIVE_ANSWERS:
                 self.should_overwrite_files = True
                 break
+            # otherwise ask user if they want to use those changes
+            elif changes_made.lower() in AFFIRMATIVE_ANSWERS:
+                use_changes_question = 'Do you want to use those changes you made?'
+                use_changes_msg = 'yes'
+                dont_use_changes_msg = 'no, restore last pythagora state'
+                print(use_changes_question, type='ipc', category='pythagora')
+                print(f'{use_changes_msg}/{dont_use_changes_msg}', type='buttons-only')
+                print(f'"{dont_use_changes_msg}" means Pythagora will restore (overwrite) all files to last stored state.\n'
+                      f'"{use_changes_msg}" means Pythagora will continue working on project using current state of files.', type='hint')
+                use_changes = styled_text(
+                    self,
+                    use_changes_question,
+                    ignore_user_input_count=True
+                )
+
+                logger.info('Use changes: %s', use_changes)
+                if use_changes.lower() in NEGATIVE_ANSWERS + [dont_use_changes_msg]:
+                    self.should_overwrite_files = True
+                elif use_changes.lower() in AFFIRMATIVE_ANSWERS + [use_changes_msg]:
+                    self.should_overwrite_files = False
 
         load_step_before_coding = ('step' in self.args and
                                    self.args['step'] is not None and

pilot/helpers/agents/CodeMonkey.py

@@ -103,9 +103,10 @@ def implement_code_changes(
                 # There are no changes or there was problem talking with the LLM, we're done here
                 break
 
+            print('Sending code for review...', type='verbose', category='agent:code-monkey')
             print('', type='verbose', category='agent:reviewer')
             content, rework_feedback = self.review_change(convo, code_change_description, file_name, file_content, content)
-            print('', type='verbose', category='agent:code-monkey')
+            print('Review finished. Continuing...', type='verbose', category='agent:code-monkey')
             if not rework_feedback:
                 # No rework needed, we're done here
                 break

pilot/helpers/agents/Developer.py

@@ -45,14 +45,13 @@ def __init__(self, project):
         self.debugger = Debugger(self)
 
     def start_coding(self, task_source):
-        print('', type='verbose', category='agent:developer')
+        print('Starting development...', type='verbose', category='agent:developer')
         if not self.project.finished:
             self.project.current_step = 'coding'
             update_app_status(self.project.args['app_id'], self.project.current_step)
 
         # DEVELOPMENT
         if not self.project.skip_steps:
-            print(color_green_bold("🚀 Now for the actual development...\n"))
             logger.info("Starting to create the actual code...")
 
         total_tasks = len(self.project.development_plan)
@@ -123,7 +122,8 @@ def implement_task(self, i, task_source, development_task=None):
         :param task_source: The source of the task, one of: 'app', 'feature', 'debugger', 'iteration'.
         :param development_task: The task to implement.
         """
-        print(color_green_bold(f'Implementing task #{i + 1}: ') + color_green(f' {development_task["description"]}\n'), category='agent:developer')
+        print(color_green_bold(f'Implementing task #{i + 1}: ') + color_green(f' {development_task["description"]}\n'), category='pythagora')
+        print(f'Starting task #{i + 1} implementation...', type='verbose', category='agent:developer')
         self.project.dot_pilot_gpt.chat_log_folder(i + 1)
 
         convo_dev_task = AgentConvo(self)
@@ -636,7 +636,7 @@ def continue_development(self, iteration_convo, last_branch_name, continue_descr
                                                                   return_cli_response=True, is_root_task=True)},
                 convo=iteration_convo,
                 is_root_task=True,
-                add_loop_button=iteration_count > 3,
+                add_loop_button=iteration_count > 2,
                 category='human-test')
 
             logger.info('response: %s', response)
@@ -667,7 +667,7 @@ def continue_development(self, iteration_convo, last_branch_name, continue_descr
 
                     tried_alternative_solutions_to_current_issue.append(next_solution_to_try)
                 else:
-                    user_feedback = self.bug_report_generator(user_feedback)
+                    user_feedback = self.bug_report_generator(user_feedback, user_description)
 
                 print_task_progress(1, 1, development_task['description'], 'troubleshooting', 'in_progress')
                 iteration_convo = AgentConvo(self)
@@ -713,11 +713,12 @@ def continue_development(self, iteration_convo, last_branch_name, continue_descr
                 self.execute_task(iteration_convo, task_steps, is_root_task=True, task_source='troubleshooting')
                 print_task_progress(1, 1, development_task['description'], 'troubleshooting', 'done')
 
-    def bug_report_generator(self, user_feedback):
+    def bug_report_generator(self, user_feedback, task_review_description):
         """
         Generate a bug report from the user feedback.
 
         :param user_feedback: The user feedback.
+        :param task_review_description: The task review description.
         :return: The bug report.
         """
         bug_report_convo = AgentConvo(self)
@@ -729,6 +730,7 @@ def bug_report_generator(self, user_feedback):
                 "user_feedback": user_feedback,
                 "app_summary": self.project.project_description,
                 "files": self.project.get_all_coded_files(),
+                "task_review_description": task_review_description,
                 "questions_and_answers": questions_and_answers,
             }, GET_BUG_REPORT_MISSING_DATA)
 
@@ -762,6 +764,7 @@ def bug_report_generator(self, user_feedback):
             bug_report_summary_convo = AgentConvo(self)
             user_feedback = bug_report_summary_convo.send_message('development/bug_report_summary.prompt', {
                 "app_summary": self.project.project_description,
+                "task_review_description": task_review_description,
                 "user_feedback": user_feedback,
                 "questions_and_answers": questions_and_answers,
             })
@@ -773,7 +776,7 @@ def review_task(self):
         Review all task changes and refactor big files.
         :return: bool - True if the task changes passed review, False if not
         """
-        print('', type='verbose', category='agent:reviewer')
+        print('Starting review of all changes made in this task...', type='verbose', category='agent:reviewer')
         self.review_count += 1
         review_result = self.review_code_changes()
         refactoring_done = self.refactor_code()

pilot/prompts/development/bug_report.prompt

@@ -5,12 +5,19 @@ Here is description of app that you are working on:
 
 {{ files_list }}
 
-Client wrote this feedback:
+{% if task_review_description %}
+User was given instructions on how to test if the app is working correctly. Here are the instructions:
+```
+{{ task_review_description }}
+```
+{% endif %}
+
+User wrote this feedback:
 ```
 {{ user_feedback }}
 ```
 {% if questions_and_answers|length > 0 %}
-Here are questions and answers that you already asked the client:
+Here are questions and answers that you already asked the user:
 ```{% for row in questions_and_answers %}
 Q: {{ row.question }}
 A: {{ row.answer }}
@@ -23,13 +30,15 @@ If you have enough information don't ask any questions.
 When thinking of questions, consider the following:
 - After getting answers to your questions, you must be able to solve the problem.
 - Category on which you need more information ("general", "frontend", "backend", "database", "devops", "other")
-- Client is not very technical person but understands basics. It is crucial that your questions can be answered without looking into code or knowing codebase. Do not ask how something is implemented, how code works or for code snippets.
+- User is not very technical person but understands basics. It is crucial that your questions can be answered without looking into code or knowing codebase. Do not ask how something is implemented, how code works or for code snippets.
+- Ask questions having in mind that you have access to whole codebase related to this issue.
 - Make sure to cover all categories that you need to solve the problem.
 - Ask only specific information that you need to solve the problem.
 - Ask clear, short and concise questions.
 - Ask only crucial questions. Do not ask for information that you do not need to solve the problem.
 - Ask least amount of questions to get the most information.
 - Ask least amount of questions to solve the problem.
+- Do not ask any questions about recent changes in code. You should be able to solve the problem without knowing recent changes in code.
 
 Here are some examples of good questions:
 "Are there any logs in browser?"

pilot/prompts/development/bug_report_summary.prompt

@@ -3,6 +3,13 @@ You are working on this app:
 {{ app_summary }}
 ```
 
+{% if task_review_description %}
+User was given instructions on how to test if the app is working correctly. Here are the instructions:
+```
+{{ task_review_description }}
+```
+{% endif %}
+
 User came to you with this bug report:
 ```
 {{ user_feedback }}
@@ -17,7 +24,7 @@ A: {{ row.answer }}
 
 Your job is now to write issue explanation that will be sent to developer, strictly following these rules:
 - If there are some logs you MUST copy all logs in whole! Do not omit any logs! This is also true for code snippets or stack traces!
-- Explanation must be very clear and factual, keep it as short as possible.
+- Explanation must be very clear and factual, keep it as short as possible. Do not make assumptions, only state the facts.
 - When writing explanation of the issue, it is good to cover all categories that you have information on. If you don't have information about one category, then don't mention it. Here are categories: "general", "frontend", "backend", "database", "devops", "other".
 - Omit all information that turns out to be irrelevant for this issue (e.g. after asking additional questions it turns out that first user message is not relevant then you should ignore it)
 - Write issue explanation as if you are talking directly to developer (in first person). Do not mention "user", talk as if you found the the issue.
@@ -26,3 +33,4 @@ Your job is now to write issue explanation that will be sent to developer, stric
 - Do not write any new code, only if something is provided by user.
 - Have in mind that developer is smart and he will understand everything as long as you provide all information that you have and that is needed to fix this issue.
 - Have in mind that issue might not be related to your current development task.
+- Do not use user instructions when creating issue explanation, that is only for you to understand the issue better.

pilot/prompts/prompts.py

@@ -55,10 +55,7 @@ def ask_for_main_app_definition(project):
 def ask_user(project, question: str, require_some_input=True, hint: str = None, ignore_user_input_count: bool = False):
     while True:
         if hint is not None:
-            if project.is_extension_old_version:
-                print(color_white_bold(hint) + '\n', type='hint')
-            else:
-                print(color_white_bold(hint) + '\n')
+            print(color_white_bold(hint) + '\n')
         project.finish_loading()
         answer = styled_text(project, question, hint=hint, ignore_user_input_count=ignore_user_input_count)
 

pilot/utils/custom_print.py

@@ -2,11 +2,9 @@
 from helpers.ipc import IPCClient
 from const.ipc import MESSAGE_TYPE, LOCAL_IGNORE_MESSAGE_TYPES
 from utils.print import remove_ansi_codes
-from utils.utils import is_extension_old_version
 
 
 def get_custom_print(args):
-    is_old_version = is_extension_old_version(args)
     built_in_print = builtins.print
 
     def print_to_external_process(*args, **kwargs):
@@ -19,13 +17,10 @@ def print_to_external_process(*args, **kwargs):
             local_print(*args, **kwargs)
             return
 
-        if is_old_version and not message and kwargs['type'] != MESSAGE_TYPE['loadingFinished']:
-            return
-
         ipc_client_instance.send({
             'type': MESSAGE_TYPE[kwargs['type']],
             'category': kwargs['category'] if 'category' in kwargs else '',
-            'content': message if is_old_version else remove_ansi_codes(message),
+            'content': remove_ansi_codes(message),
         })
         if kwargs['type'] == MESSAGE_TYPE['user_input_request']:
             return ipc_client_instance.listen()

pilot/utils/questionary.py

@@ -24,8 +24,6 @@ def styled_text(project, question, ignore_user_input_count=False, style=None, hi
 
     if project is not None and project.check_ipc():
         response = print(question, type='user_input_request')
-        if project.is_extension_old_version:
-            print(response)
     else:
         used_style = style if style is not None else style_config.get_style()
         question = remove_ansi_codes(question)  # Colorama and questionary are not compatible and styling doesn't work

pilot/utils/utils.py

@@ -210,18 +210,3 @@ def remove_lines_with_string(file_content, matching_string):
     lines = file_content.split('\n')
     new_lines = [line for line in lines if matching_string not in line.lower()]
     return '\n'.join(new_lines)
-
-
-def is_extension_old_version(args):
-    is_old_version = True
-    if "extension_version" in args:
-        arg_version_tuple = tuple(map(int, args["extension_version"].split('.')))
-        compare_version_tuple = (0, 0, 47)
-        exception_version = (0, 0, 46)
-        if arg_version_tuple == exception_version:
-            return False
-
-        is_old_version = arg_version_tuple <= compare_version_tuple
-        print(is_old_version)
-
-    return is_old_version