Merge pull request #65 from Codium-ai/tr/agent_logic

pr_information_from_user_prompts

pr_agent/cli.py

@@ -5,6 +5,7 @@ import os
 
 from pr_agent.tools.pr_code_suggestions import PRCodeSuggestions
 from pr_agent.tools.pr_description import PRDescription
+from pr_agent.tools.pr_information_from_user import PRInformationFromUser
 from pr_agent.tools.pr_questions import PRQuestions
 from pr_agent.tools.pr_reviewer import PRReviewer
 
@@ -28,7 +29,8 @@ improve / improve_code - Suggest improvements to the code in the PR as pull requ
     parser.add_argument('command', type=str, help='The', choices=['review', 'review_pr',
                                                                   'ask', 'ask_question',
                                                                   'describe', 'describe_pr',
-                                                                  'improve', 'improve_code'], default='review')
+                                                                  'improve', 'improve_code',
+                                                                  'user_questions'], default='review')
     parser.add_argument('rest', nargs=argparse.REMAINDER, default=[])
     args = parser.parse_args()
     logging.basicConfig(level=os.environ.get("LOGLEVEL", "INFO"))
@@ -54,6 +56,10 @@ improve / improve_code - Suggest improvements to the code in the PR as pull requ
         print(f"Reviewing PR: {args.pr_url}")
         reviewer = PRReviewer(args.pr_url, cli_mode=True)
         asyncio.run(reviewer.review())
+    elif command in ['user_questions']:
+        print(f"Asking the PR author questions: {args.pr_url}")
+        reviewer = PRInformationFromUser(args.pr_url)
+        asyncio.run(reviewer.generate_questions())
     else:
         print(f"Unknown command: {command}")
         parser.print_help()

pr_agent/config_loader.py

@@ -13,6 +13,7 @@ settings = Dynaconf(
          "settings/pr_questions_prompts.toml",
          "settings/pr_description_prompts.toml",
          "settings/pr_code_suggestions_prompts.toml",
+         "settings/pr_information_from_user_prompts.toml",
          "settings_prod/.secrets.toml"
         ]]
 )

pr_agent/settings/configuration.toml

@@ -2,7 +2,7 @@
 model="gpt-4-0613"
 git_provider="github"
 publish_review=true
-verbosity_level=0 # 0,1,2
+verbosity_level=2 # 0,1,2
 
 [pr_reviewer]
 require_focused_review=true

pr_agent/settings/pr_information_from_user_prompts.toml

@@ -0,0 +1,33 @@
+[pr_information_from_user_prompt]
+system="""You are CodiumAI-PR-Reviewer, a language model designed to review git pull requests.
+Given the PR Info and the PR Git Diff, generate 4 questions about the PR for the PR author.
+The goal of the questions is to help the language model understand the PR better, so the questions should be insightful, informative, non-trivial, and relevant to the PR.
+Prefer yes\\no or multiple choice questions. If you have to ask open-ended questions, make sure they are not too difficult, and can be answered in a sentence or two.
+
+
+Example output:
+'
+Questions to better understand the PR:
+1. ...
+2. ...
+...
+"""
+
+user="""PR Info:
+Title: '{{title}}'
+Branch: '{{branch}}'
+Description: '{{description}}'
+{%- if language %}
+Main language: {{language}}
+{%- endif %}
+
+
+The PR Git Diff:
+```
+{{diff}}
+```
+Note that lines in the diff body are prefixed with a symbol that represents the type of change: '-' for deletions, '+' for additions, and ' ' (a space) for unchanged lines
+
+
+Response:
+"""

pr_agent/tools/pr_information_from_user.py

@@ -0,0 +1,71 @@
+import copy
+import logging
+
+from jinja2 import Environment, StrictUndefined
+
+from pr_agent.algo.ai_handler import AiHandler
+from pr_agent.algo.pr_processing import get_pr_diff
+from pr_agent.algo.token_handler import TokenHandler
+from pr_agent.config_loader import settings
+from pr_agent.git_providers import get_git_provider
+from pr_agent.git_providers.git_provider import get_main_pr_language
+
+
+class PRInformationFromUser:
+    def __init__(self, pr_url: str):
+        self.git_provider = get_git_provider()(pr_url)
+        self.main_pr_language = get_main_pr_language(
+            self.git_provider.get_languages(), self.git_provider.get_files()
+        )
+        self.ai_handler = AiHandler()
+        self.vars = {
+            "title": self.git_provider.pr.title,
+            "branch": self.git_provider.get_pr_branch(),
+            "description": self.git_provider.get_description(),
+            "language": self.main_pr_language,
+            "diff": "",  # empty diff for initial calculation
+        }
+        self.token_handler = TokenHandler(self.git_provider.pr,
+                                          self.vars,
+                                          settings.pr_information_from_user_prompt.system,
+                                          settings.pr_information_from_user_prompt.user)
+        self.patches_diff = None
+        self.prediction = None
+
+    async def generate_questions(self):
+        logging.info('Generating question to the user...')
+        if settings.config.publish_review:
+            self.git_provider.publish_comment("Preparing answer...", is_temporary=True)
+        logging.info('Getting PR diff...')
+        self.patches_diff = get_pr_diff(self.git_provider, self.token_handler)
+        logging.info('Getting AI prediction...')
+        self.prediction = await self._get_prediction()
+        logging.info('Preparing questions...')
+        pr_comment = self._prepare_pr_answer()
+        if settings.config.publish_review:
+            logging.info('Pushing questions...')
+            self.git_provider.publish_comment(pr_comment)
+            self.git_provider.remove_initial_comment()
+        return ""
+
+    async def _get_prediction(self):
+        variables = copy.deepcopy(self.vars)
+        variables["diff"] = self.patches_diff  # update diff
+        environment = Environment(undefined=StrictUndefined)
+        system_prompt = environment.from_string(settings.pr_information_from_user_prompt.system).render(variables)
+        user_prompt = environment.from_string(settings.pr_information_from_user_prompt.user).render(variables)
+        if settings.config.verbosity_level >= 2:
+            logging.info(f"\nSystem prompt:\n{system_prompt}")
+            logging.info(f"\nUser prompt:\n{user_prompt}")
+        model = settings.config.model
+        response, finish_reason = await self.ai_handler.chat_completion(model=model, temperature=0.2,
+                                                                        system=system_prompt, user=user_prompt)
+        return response
+
+    def _prepare_pr_answer(self) -> str:
+        model_output = self.prediction.strip()
+        if settings.config.verbosity_level >= 2:
+            logging.info(f"answer_str:\n{model_output}")
+        answer_str = f"{model_output}\n\n Please respond to the question above in the following format:\n\n" + \
+                     f"/answer <question_id> <answer>\n\n" + f"Example:\n'\n/answer\n1. Yes, because ...\n2. No, because ...\n'"
+        return answer_str