feat: 支持小猫猫做 codeReview (#260)

- 支持生成摘要
- 支持 CR

server/agent/base.py

@@ -6,7 +6,7 @@
 from langchain.agents.format_scratchpad.openai_tools import (
     format_to_openai_tool_messages,
 )
-from langchain_core.messages import AIMessage, FunctionMessage, HumanMessage
+from langchain_core.messages import AIMessage, FunctionMessage, HumanMessage, SystemMessage
 from langchain.agents.output_parsers.openai_tools import OpenAIToolsAgentOutputParser
 from langchain.prompts import MessagesPlaceholder
 from langchain_core.prompts import ChatPromptTemplate
@@ -92,12 +92,15 @@ def get_prompt(self):
     def chat_history_transform(self, messages: list[Message]):
         transformed_messages = []
         for message in messages:
-            if message.role == "user":
-                transformed_messages.append(HumanMessage(self.chat_model.parse_content(content=message.content)))
-            elif message.role == "assistant":
-                transformed_messages.append(AIMessage(content=message.content))
-            else:
-                transformed_messages.append(FunctionMessage(content=message.content))
+            match message.role:
+                case "user":
+                    transformed_messages.append(HumanMessage(self.chat_model.parse_content(content=message.content)))
+                case "assistant":
+                    transformed_messages.append(AIMessage(content=message.content))
+                case "system":
+                    transformed_messages.append(SystemMessage(content=message.content))
+                case _:
+                    transformed_messages.append(FunctionMessage(content=message.content))
         return transformed_messages
 
     async def run_stream_chat(self, input_data: ChatData) -> AsyncIterator[str]:

server/agent/bot/get_bot.py

@@ -0,0 +1,9 @@
+
+from core.dao.botDAO import BotDAO
+from core.models.bot import Bot
+from petercat_utils.data_class import ChatData
+
+def get_bot(input_data: ChatData) -> Bot:
+    bot_dao = BotDAO()
+    bot = bot_dao.get_bot(input_data.bot_id)
+    return bot

server/agent/prompts/issue_helper.py

@@ -1,7 +1,7 @@
 ISSUE_PROMPT = """
 # Task
-You have required to resolve an issue {issue_url} now:
-
+Introduce yourself and briefly explain the assistance you can provide. 
+If users need further support, prompt them to @ you for help. Generate a response suitable for this scenario.
 ## Issue Content:
 {issue_content}
 

server/agent/prompts/pull_request.py

@@ -0,0 +1,56 @@
+PULL_REQUEST_ROLE = """
+# Character Description
+You are an experienced Code Reviewer, specializing in identifying critical issues and potentially incorrect code in Pull Requests (PRs).
+
+# Skills Description
+## Skill 1: Pull Request Summarize
+You excel at analyzing users' code changes and get summaries.
+Offering specific and targeted in pinpointing code that may lead to errors, security vulnerabilities, or significant performance issues. 
+You focus only on identifying and addressing severe or fundamentally flawed code practices.
+You are equipped with two powerful tool2, used to leave a summary and code review comments:
+- create_pr_summary; This tools is used to create summary of PR.
+- create_review_comment: This tool is used to leave review comment of file.
+
+Constraints
+- Focus exclusively on identifying and reviewing highly inappropriate code usage or potential errors.
+- Avoid reviewing minor style inconsistencies or non-critical issues.
+- Respect the language of the PR's title and description when providing feedback, ensuring that all comments and summarize are given in the same language.
+- Provide concise, clear, and actionable feedback, directly related to improving the correctness and reliability of the code.
+"""
+
+PULL_REQUEST_SUMMARY = """
+# Task
+You have two Pull Requst review task with basic infomation:
+```
+repo_name: {repo_name}
+pull_number: {pull_number}
+title: {title}
+description: {description}
+```
+
+## Task 1: Summarize the Pull Request
+Using `create_pr_summary` tool to create PR summary.
+Provider your response in markdown with the following content. follow the user's language.
+  - **Walkthrough**:  A high-level summary of the overall change instead of specific files within 80 words.
+  - **Changes**: A markdown table of files and their summaries. Group files with similar changes together into a single row to save space.
+
+## Task 2: using `create_review_comment` tool to Create code review comments for every new_hunk file that may lead to errors, vulnerabilities. 
+
+## File Diff:
+{file_diff}
+
+# Constraints
+- After completing the tasks, only output "All task finished".
+"""
+
+def get_role_prompt(repo_name: str, ref: str):
+  return PULL_REQUEST_ROLE.format(repo_name=repo_name, ref=ref)
+
+def get_pr_summary(repo_name: str, pull_number: int, title: str, description: str, file_diff: str):
+  return PULL_REQUEST_SUMMARY.format(
+    repo_name=repo_name,
+    pull_number={pull_number},
+    title={title},
+    description={description},
+    file_diff=file_diff
+  )

server/agent/qa_chat.py

@@ -2,29 +2,31 @@
 from github import Auth
 from agent.base import AgentBuilder
 from agent.llm import get_llm
-from core.dao.botDAO import BotDAO
+
 from core.models.bot import Bot
 from agent.prompts.bot_template import generate_prompt_by_repo_name
 from petercat_utils.data_class import ChatData
 
-from agent.tools import issue, sourcecode, knowledge, git_info
+from agent.tools import issue, pull_request, sourcecode, knowledge, git_info
 
 
 def get_tools(bot: Bot, token: Optional[Auth.Token]):
     issue_tools = issue.factory(token=token)
+    pull_request_tools = pull_request.factory(token=token)
+
     return {
         "search_knowledge": knowledge.factory(bot_id=bot.id),
         "create_issue": issue_tools["create_issue"],
         "get_issues": issue_tools["get_issues"],
+        "get_file_content": pull_request_tools["get_file_content"],
+        "create_review_comment": pull_request_tools["create_review_comment"],
+        "create_pr_summary": pull_request_tools["create_pr_summary"],
         "search_issues": issue_tools["search_issues"],
         "search_code": sourcecode.search_code,
         "search_repo": git_info.search_repo,
     }
 
-def agent_stream_chat(input_data: ChatData, token: Auth.Token) -> AsyncIterator[str]:
-    bot_dao = BotDAO()
-    bot = bot_dao.get_bot(input_data.bot_id)
-
+def agent_stream_chat(input_data: ChatData, token: Auth.Token, bot: Bot) -> AsyncIterator[str]:
     agent = AgentBuilder(
         chat_model=get_llm(bot.llm),
         prompt=bot.prompt or generate_prompt_by_repo_name("ant-design"),
@@ -33,14 +35,12 @@ def agent_stream_chat(input_data: ChatData, token: Auth.Token) -> AsyncIterator[
     return agent.run_stream_chat(input_data)
 
 
-def agent_chat(input_data: ChatData, token: Auth.Token) -> AsyncIterator[str]:
-    bot_dao = BotDAO()
-    bot = bot_dao.get_bot(input_data.bot_id)
+def agent_chat(input_data: ChatData, token: Auth.Token, bot: Bot) -> AsyncIterator[str]:
 
     prompt = bot.prompt or generate_prompt_by_repo_name("ant-design")
+
     if input_data.prompt is not None:
         prompt = f"{prompt}\n\n{input_data.prompt}"
-    print(f"agent_chat: prompt={prompt}")
 
     agent = AgentBuilder(
         chat_model=get_llm(bot.llm),

server/agent/tools/pull_request.py

@@ -0,0 +1,83 @@
+
+from typing import Optional
+from github import Github, Auth
+import json
+
+from langchain.tools import tool
+from agent.tools.helper import need_github_login
+
+def factory(token: Optional[Auth.Token]):
+
+    @tool
+    def get_file_content(repo_name: str, path: str, ref: str):
+        """
+        Get content of specified pull requst file
+        :param repo_name: The name of the repository, e.g., "ant-design/ant-design"
+        :param path: The path of file, e.g., /contents/file1.txt
+        :param ref: The name of the commit/branch/tag. Default: the repository’s default branch.
+        """
+        if token is None:
+            return need_github_login()
+
+        g = Github(auth=token)
+        try:
+            repo = g.get_repo(repo_name)
+            content = repo.get_contents(path=path, ref=ref)
+            print(f"get_content: f{content}")
+            return json.dumps([])
+        except Exception as e:
+            print(f"An error occurred: {e}")
+            return json.dumps([])
+
+    @tool
+    def create_pr_summary(repo_name: str, pull_number: int, summary: str):
+        """
+        Create a code review of specified pull requst file
+        :param repo_name: The name of the repository, e.g., "ant-design/ant-design"
+        :param pull_number: The number of pull requst: e.g., 123
+        :param summary: markdown content of PR summary 
+        """
+        if token is None:
+            return need_github_login()
+
+        g = Github(auth=token)
+        repo = g.get_repo(repo_name)
+        pull_request = repo.get_pull(pull_number)
+        pull_request.create_issue_comment(summary)
+
+    @tool
+    def create_review_comment(repo_name: str, pull_number: int, sha: str, path: str, line: int, comment: str):
+        """
+        Create a code review of specified pull requst file
+        :param repo_name: The name of the repository, e.g., "ant-design/ant-design"
+        :param pull_number: The number of pull requst: e.g., 123
+        :param sha: The sha of file. e.g., 6dcb09b5b57875f334f61aebed695e2e4193db5e
+        :param path: The path of file, e.g., /contents/file1.txt
+        :param line: The line number to create comment at. e.g., 19
+        :param comment: Content of review comments 
+        """
+        if token is None:
+            return need_github_login()
+
+        g = Github(auth=token)
+        try:
+            repo = g.get_repo(repo_name)
+            pull_request = repo.get_pull(pull_number)
+            commit = repo.get_commit(sha=sha)
+            # print(f"create_review_comment, pull_request={pull_request}, commit={commit}, comment={comment}")
+            pull_request.create_review_comment(
+                body=comment,
+                path=path,
+                commit=commit,
+                line=line,
+            )
+
+        except Exception as e:
+            print(f"An error occurred: {e}")
+            return json.dumps([])
+
+    return {
+        "get_file_content": get_file_content,
+        "create_pr_summary": create_pr_summary,
+        "create_review_comment": create_review_comment,
+    }

server/auth/get_user_info.py

@@ -1,14 +1,12 @@
 from fastapi import Request
 import httpx
 import secrets
-import random
-import string
+
+from utils.random_str import random_str
 
 from .get_oauth_token import get_oauth_token
 from petercat_utils import get_client, get_env_variable
 
-def random_str(N):
-    return ''.join(random.SystemRandom().choice(string.ascii_uppercase + string.ascii_lowercase + string.digits) for _ in range(N))
 
 AUTH0_DOMAIN = get_env_variable("AUTH0_DOMAIN")
 

server/chat/router.py

@@ -2,13 +2,15 @@
 from github import Auth
 from fastapi import APIRouter, Depends
 from fastapi.responses import StreamingResponse
+from agent.bot import bot_builder
+from agent.bot.get_bot import get_bot
+from core.models.bot import Bot
 from petercat_utils.data_class import ChatData
 
-from agent import qa_chat, bot_builder
+from agent import qa_chat
 from auth.rate_limit import verify_rate_limit
 from auth.get_user_info import get_user_access_token, get_user_id
 
-
 router = APIRouter(
     prefix="/api/chat",
     tags=["chat"],
@@ -29,14 +31,15 @@ async def generate_auth_failed_stream():
 def run_qa_chat(
     input_data: ChatData,
     user_access_token: Annotated[str | None, Depends(get_user_access_token)] = None,
+    bot: Annotated[Bot | None, Depends(get_bot)] = None,
 ):
     print(
         f"run_qa_chat: input_data={input_data}, user_access_token={user_access_token}"
     )
 
     token = Auth.Token(user_access_token) if user_access_token is not None else None
     result = qa_chat.agent_stream_chat(
-        input_data=input_data, token=token
+        input_data=input_data, token=token, bot=bot
     )
     return StreamingResponse(result, media_type="text/event-stream")
 
@@ -45,9 +48,10 @@ def run_qa_chat(
 async def run_issue_helper(
     input_data: ChatData,
     user_access_token: Annotated[str | None, Depends(get_user_access_token)] = None,
+    bot: Annotated[Bot | None, Depends(get_bot)] = None,
 ):
     token = Auth.Token(user_access_token) if user_access_token is not None else None
-    result = await qa_chat.agent_chat(input_data, token)
+    result = await qa_chat.agent_chat(input_data, token, bot)
     return result
 
 

server/core/models/bot.py

@@ -10,4 +10,4 @@ class Bot(BaseModel):
     prompt: Optional[str] = ""
     name: str
     llm: Optional[str] = "openai"
-    created_at: datetime 
+    created_at: datetime = datetime.now()

server/event_handler/issue.py

@@ -5,13 +5,13 @@
 from github import GithubException
 from agent.prompts.issue_helper import generate_issue_comment_prompt, generate_issue_prompt
 
+from core.dao.botDAO import BotDAO
 from core.dao.repositoryConfigDAO import RepositoryConfigDAO
 from petercat_utils.data_class import ChatData, Message, TextContentBlock
 
 from agent.qa_chat import agent_chat
 
 
-
 class IssueEventHandler:
     event: Any
     auth: Auth.AppAuth
@@ -45,13 +45,15 @@ async def execute(self):
 
                 repository_config = RepositoryConfigDAO()
                 repo_config = repository_config.get_by_repo_name(repo.full_name)
-
+                bot_dao = BotDAO()
+                bot = bot_dao.get_bot(repo_config.robot_id)
+
                 analysis_result = await agent_chat(
                     ChatData(
                         prompt=prompt,
                         messages=[message],
                         bot_id=repo_config.robot_id
-                    ), self.auth)
+                    ), self.auth, bot)
 
                 issue.create_comment(analysis_result["output"])
 
@@ -61,13 +63,20 @@ async def execute(self):
             return {"success": False, "error": str(e)}
 
 class IssueCommentEventHandler(IssueEventHandler):
+    def not_mentioned_me(self):
+        return "@petercat-bot" not in self.event["comment"]["body"]
+
     async def execute(self):
         try:
             print(f"actions={self.event['action']},sender={self.event['sender']}")
             # 忽略机器人回复
             if self.event['sender']['type'] == "Bot":
                 return {"success": True}
             if self.event["action"] == "created":
+                # 如果没有 AT 我。就算了
+                if self.not_mentioned_me():
+                    return {"success": True}
+
                 issue, repo = self.get_issue()
                 issue_comments = issue.get_comments()
 
@@ -78,7 +87,6 @@ async def execute(self):
                     ) for comment in issue_comments
                 ]
 
-                print(f"messages={messages}")
                 issue_content = f"{issue.title}: {issue.body}"
                 prompt = generate_issue_comment_prompt(issue_url=issue.url, issue_content=issue_content)