auto-coder 0.1.269__tar.gz → 0.1.271__tar.gz

{auto_coder-0.1.269 → auto_coder-0.1.271}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: auto-coder
-Version: 0.1.269
+Version: 0.1.271
 Summary: AutoCoder: AutoCoder
 Author: allwefantasy
 Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence

{auto_coder-0.1.269 → auto_coder-0.1.271}/src/auto_coder.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: auto-coder
-Version: 0.1.269
+Version: 0.1.271
 Summary: AutoCoder: AutoCoder
 Author: allwefantasy
 Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence

{auto_coder-0.1.269 → auto_coder-0.1.271}/src/auto_coder.egg-info/SOURCES.txt

@@ -25,6 +25,7 @@ src/autocoder/agent/__init__.py
 src/autocoder/agent/auto_demand_organizer.py
 src/autocoder/agent/auto_filegroup.py
 src/autocoder/agent/auto_guess_query.py
+src/autocoder/agent/auto_learn_from_commit.py
 src/autocoder/agent/auto_review_commit.py
 src/autocoder/agent/auto_tool.py
 src/autocoder/agent/coder.py

auto_coder-0.1.271/src/autocoder/agent/auto_learn_from_commit.py

@@ -0,0 +1,209 @@
+from typing import Generator, List, Dict, Union, Tuple, Optional
+import os
+import yaml
+import byzerllm
+import pydantic
+import git
+from rich.console import Console
+from autocoder.common.printer import Printer
+from autocoder.common import AutoCoderArgs
+from autocoder.common.utils_code_auto_generate import stream_chat_with_continue
+import hashlib
+
+
+def load_yaml_config(yaml_file: str) -> Dict:
+    """加载YAML配置文件"""
+    try:
+        with open(yaml_file, 'r', encoding='utf-8') as f:
+            return yaml.safe_load(f)
+    except Exception as e:
+        printer = Printer()
+        printer.print_in_terminal("yaml_load_error", style="red", yaml_file=yaml_file, error=str(e))
+        return {}
+
+
+class AutoLearnFromCommit:
+    def __init__(self, llm: Union[byzerllm.ByzerLLM,byzerllm.SimpleByzerLLM],
+                 args:AutoCoderArgs,                 
+                 skip_diff: bool = False,                 
+                 console: Optional[Console] = None):
+        """
+        初始化 AutoLearnFromCommit
+
+        Args:
+            llm: ByzerLLM 实例，用于代码学习
+            project_dir: 项目根目录
+            skip_diff: 是否跳过获取 diff 信息            
+        """
+        self.project_dir = args.source_dir
+        self.actions_dir = os.path.join(args.source_dir, "actions")
+        self.llm = llm        
+        self.skip_diff = skip_diff
+        self.console = console or Console()
+
+    @byzerllm.prompt()
+    def learn(self, querie_with_urls_and_changes: List[Tuple[str, List[str], Dict[str, Tuple[str, str]]]], query: str) -> Generator[str,None,None]:
+        """
+        请根据下面的代码变更，总结出通用的代码调整模式，以帮助实现类似的需求。
+
+        下面用户对本次的目标以及你需要总结的要求：
+        <goal>
+        {{ query }}
+        </goal>
+
+        下面是本次提交的代码变更：
+        <changes>
+        {% for query,urls,changes in querie_with_urls_and_changes %}
+        ## 原始的任务需求
+        {{ query }}
+
+        修改的文件:
+        {% for url in urls %}
+        - {{ url }}
+        {% endfor %}
+
+        代码变更:
+        {% for file_path, (before, after) in changes.items() %}
+        ##File: {{ file_path }}
+        ##修改前:
+        
+        {{ before or "New file" }}        
+        
+        ##File: {{ file_path }}
+        ##修改后:
+        
+        {{ after or "File deleted" }}
+        
+        {% endfor %}
+        {% endfor %}
+        </changes>
+
+        请总结以下内容：
+        1. 代码调整模式：描述为了实现这个目标，通常需要对代码做出哪些通用调整
+        2. 关键修改点：指出本次提交中最关键或最具代表性的修改点
+        3. 潜在扩展：基于这些修改，未来类似需求可能需要进行哪些扩展
+        4. 注意事项：在实现类似功能时需要注意哪些问题
+
+        总结要求：
+        1. 总结应该抽象化，不要局限于具体实现细节
+        2. 对于每个模式都应该提供明确的适用场景
+        3. 应该考虑代码的可维护性和可扩展性
+        4. 应该指出在不同技术栈或框架下的通用性
+        """
+        pass
+
+    def parse_history_tasks(self) -> List[Dict]:
+        """
+        解析历史任务信息
+
+        Returns:
+            List[Dict]: 每个字典包含一个历史任务的信息
+        """
+        # 获取所有YAML文件
+        action_files = [
+            f for f in os.listdir(self.actions_dir)
+            if f[:3].isdigit() and "_" in f and f.endswith('.yml')
+        ]
+
+        # 按序号排序
+        def get_seq(name):
+            return int(name.split("_")[0])
+
+        # 获取最新的action文件列表
+        action_files = sorted(action_files, key=get_seq)
+        action_files.reverse()        
+
+        action_file = action_files[0]
+
+        querie_with_urls_and_changes = []
+        repo = git.Repo(self.project_dir)
+
+        # 收集所有query、urls和对应的文件变化
+        for yaml_file in [action_file]:
+            yaml_path = os.path.join(self.actions_dir, yaml_file)
+            config = load_yaml_config(yaml_path)
+
+            if not config:
+                continue
+
+            query = config.get('query', '')
+            urls = config.get('urls', [])
+
+            if query:
+                changes = {}
+                if not self.skip_diff:
+                    # 计算文件的MD5用于匹配commit   
+                    with open(yaml_path, 'r', encoding='utf-8') as f:
+                        yaml_content = f.read()                 
+                        file_md5 = hashlib.md5(yaml_content.encode("utf-8")).hexdigest()
+                    response_id = f"auto_coder_{yaml_file}_{file_md5}"
+                    # 查找对应的commit                   
+                    try:
+                        for commit in repo.iter_commits():
+                            if response_id in commit.message:
+                                if commit.parents:
+                                    parent = commit.parents[0]
+                                    # 获取所有文件的前后内容
+                                    for diff_item in parent.diff(commit):
+                                        file_path = diff_item.a_path if diff_item.a_path else diff_item.b_path
+                                        
+                                        # 获取变更前内容
+                                        before_content = None
+                                        try:
+                                            if diff_item.a_blob:
+                                                before_content = repo.git.show(f"{parent.hexsha}:{file_path}")
+                                        except git.exc.GitCommandError:
+                                            pass  # 文件可能是新增的
+
+                                        # 获取变更后内容
+                                        after_content = None
+                                        try:
+                                            if diff_item.b_blob:
+                                                after_content = repo.git.show(f"{commit.hexsha}:{file_path}")
+                                        except git.exc.GitCommandError:
+                                            pass  # 文件可能被删除
+
+                                        changes[file_path] = (before_content, after_content)
+                                break
+                    except git.exc.GitCommandError as e:
+                        printer = Printer()
+                        printer.print_in_terminal("git_command_error", style="red", error=str(e))
+                    except Exception as e:
+                        printer = Printer()
+                        printer.print_in_terminal("get_commit_changes_error", style="red", error=str(e))
+
+                querie_with_urls_and_changes.append((query, urls, changes))
+
+        return querie_with_urls_and_changes
+    
+
+    def learn_from_commit(self,query: str, conversations: List[Dict]) -> Generator[str,None,None]:
+        """
+        从最新的代码提交中学习通用模式
+
+        Returns:
+            Optional[Generator]: 学习结果生成器，如果出错则返回None
+        """
+        printer = Printer()
+        # 获取最新的提交信息
+        changes = self.parse_history_tasks()
+        if not changes:            
+            printer.print_in_terminal("no_latest_commit", style="red")
+            return None
+
+        # 调用LLM进行代码学习
+        try:
+            # 获取 prompt 内容            
+            query = self.learn.prompt(changes, query)
+            new_conversations = conversations.copy()[0:-1]
+            new_conversations.append({"role": "user", "content": query})
+            # 构造对话消息            
+            v = stream_chat_with_continue(
+                    llm=self.llm,
+                    conversations=new_conversations,
+                    llm_config={}
+            )
+            return v
+        except Exception as e:            
+            printer.print_in_terminal("code_learn_error", style="red", error=str(e))
+            return None

{auto_coder-0.1.269 → auto_coder-0.1.271}/src/autocoder/auto_coder.py

@@ -1339,6 +1339,10 @@ def main(input_args: Optional[List[str]] = None):
                 from autocoder.agent.auto_review_commit import AutoReviewCommit
                 reviewer = AutoReviewCommit(llm=chat_llm, args=args)
                 v = reviewer.review_commit(query=args.query,conversations=loaded_conversations)
+            elif "learn_from_commit" in args.action:
+                from autocoder.agent.auto_learn_from_commit import AutoLearnFromCommit
+                learner = AutoLearnFromCommit(llm=chat_llm, args=args)
+                v = learner.learn_from_commit(query=args.query,conversations=loaded_conversations)
             else:                
                 # 预估token数量
                 estimated_input_tokens = count_tokens(json.dumps(loaded_conversations, ensure_ascii=False))

{auto_coder-0.1.269 → auto_coder-0.1.271}/src/autocoder/auto_coder_runner.py

@@ -1,4 +1,6 @@
 from itertools import product
+from rich.console import Console
+from rich.panel import Panel
 from prompt_toolkit.formatted_text import HTML
 from prompt_toolkit.shortcuts import radiolist_dialog
 from prompt_toolkit import prompt
@@ -646,21 +648,28 @@ def revert():
         with redirect_stdout() as output:
             auto_coder_main(["revert", "--file", file_path])
         s = output.getvalue()
-        print(s, flush=True)
+        
+        console = Console()
+        panel = Panel(
+            Markdown(s),
+            title="Revert Result",
+            border_style="green" if "Successfully reverted changes" in s else "red",
+            padding=(1, 2),
+            expand=False
+        )
+        console.print(panel)
+        
         if "Successfully reverted changes" in s:
             result_manager.append(content=s, meta={"action": "revert","success":False, "input":{                
             }})
-            print(
-                "Reverted the last chat action successfully. Remove the yaml file {file_path}"
-            )
+                        
             os.remove(file_path)
         else:
             result_manager.append(content=s, meta={"action": "revert","success":False, "input":{                
             }})
     else:
         result_manager.append(content="No previous chat action found to revert.", meta={"action": "revert","success":False, "input":{                
-            }})
-        print("No previous chat action found to revert.")
+            }})        
 
 
 def add_files(args: List[str]):
@@ -1155,6 +1164,7 @@ def mcp(query: str):
         "silence": conf.get("silence", "true") == "true",
         "include_project_structure": conf.get("include_project_structure", "true")
         == "true",
+        "exclude_files": memory.get("exclude_files", []),
     }
     for key, value in conf.items():
         converted_value = convert_config_value(key, value)
@@ -1224,6 +1234,7 @@ def code_next(query: str):
         "silence": conf.get("silence", "true") == "true",
         "include_project_structure": conf.get("include_project_structure", "true")
         == "true",
+        "exclude_files": memory.get("exclude_files", []),
     }
     for key, value in conf.items():
         converted_value = convert_config_value(key, value)
@@ -1595,6 +1606,9 @@ def chat(query: str):
     if "/review" in query and "/commit" in query:
         yaml_config["action"].append("review_commit")
         query = query.replace("/review", "", 1).replace("/commit", "", 1).strip()
+    elif "/learn" in query:
+        yaml_config["action"].append("learn_from_commit")
+        query = query.replace("/learn", "", 1).strip()
     else:    
         is_review = query.strip().startswith("/review")
         if is_review:
@@ -2283,6 +2297,7 @@ def index_build():
     conf = memory.get("conf", {})
     yaml_config = {
         "include_file": ["./base/base.yml"],
+        "exclude_files": memory.get("exclude_files", []),
     }
 
     for key, value in conf.items():
@@ -2295,10 +2310,8 @@ def index_build():
 
     with open(yaml_file, "w",encoding="utf-8") as f:
         f.write(yaml_content)
-    try:
-        with redirect_stdout() as output:
-            auto_coder_main(["index", "--file", yaml_file])
-        print(output.getvalue(), flush=True)
+    try:        
+        auto_coder_main(["index", "--file", yaml_file])        
         completer.refresh_files()
     finally:
         os.remove(yaml_file)
@@ -2315,6 +2328,7 @@ def get_final_config()->AutoCoderArgs:
         "silence": conf.get("silence", "true") == "true",
         "include_project_structure": conf.get("include_project_structure", "true")
         == "true",
+        "exclude_files": memory.get("exclude_files", []),
     }
     for key, value in conf.items():
         converted_value = convert_config_value(key, value)
@@ -2380,10 +2394,8 @@ def index_query(query: str):
 
     with open(yaml_file, "w",encoding="utf-8") as f:
         f.write(yaml_content)
-    try:
-        with redirect_stdout() as output:
-            auto_coder_main(["index-query", "--file", yaml_file])
-        print(output.getvalue(), flush=True)
+    try:        
+        auto_coder_main(["index-query", "--file", yaml_file])        
     finally:
         os.remove(yaml_file)
 

{auto_coder-0.1.269 → auto_coder-0.1.271}/src/autocoder/commands/auto_command.py

@@ -171,15 +171,21 @@ class CommandAutoTuner:
         当前用户环境信息如下:
         <os_info>
         操作系统: {{ env_info.os_name }} {{ env_info.os_version }}
+        操作系统发行版: {{ os_distribution }}
         Python版本: {{ env_info.python_version }}
         终端类型: {{ env_info.shell_type }}
         终端编码: {{ env_info.shell_encoding }}
+        
+        {%- if shell_type %}
+        脚本类型：{{ shell_type }}
+        {%- endif %}
+
         {%- if env_info.conda_env %}
         Conda环境: {{ env_info.conda_env }}
         {%- endif %}
         {%- if env_info.virtualenv %}
         虚拟环境: {{ env_info.virtualenv }}
-        {%- endif %}  
+        {%- endif %}   
         </os_info>
         
         我们的目标是根据用户输入和当前上下文，组合多个函数来完成用户的需求。
@@ -221,6 +227,7 @@ class CommandAutoTuner:
         每个文件的tokens数，你可以根据这个信息来决定如何读取这个文件。比如对于很小的文件，那么可以直接全部读取，
         而对于分析一个超大文件推荐组合 read_files 带上 line_ranges 参数来读取，或者组合 read_file_withread_file_with_keyword_ranges 等来读取，
         每个函数你还可以使用多次来获取更多信息。
+        9. 根据操作系统，终端类型，脚本类型等各种信息，在涉及到路径或者脚本的时候，需要考虑平台差异性。
         </function_combination_readme>
 
 
@@ -260,7 +267,13 @@ class CommandAutoTuner:
         满足需求。
         """
 
-        env_info = detect_env()            
+        env_info = detect_env() 
+        shell_type = "bash"
+        if shells.is_running_in_cmd():
+            shell_type = "cmd"
+        elif shells.is_running_in_powershell():
+            shell_type = "powershell"
+
         return {
             "user_input": request.user_input,
             "current_files": self.memory_config.memory["current_files"]["files"],            
@@ -268,9 +281,10 @@ class CommandAutoTuner:
             "available_commands": self._command_readme.prompt(),
             "current_conf": json.dumps(self.memory_config.memory["conf"], indent=2),        
             "env_info": env_info,
-            "shell_type": shells.get_terminal_name(),
+            "shell_type": shell_type,
             "shell_encoding": shells.get_terminal_encoding(),
-            "conversation_safe_zone_tokens": self.args.conversation_prune_safe_zone_tokens
+            "conversation_safe_zone_tokens": self.args.conversation_prune_safe_zone_tokens,
+            "os_distribution": shells.get_os_distribution()
         }
     
     @byzerllm.prompt()
@@ -1179,6 +1193,18 @@ class CommandAutoTuner:
          此时会返回诸如 "ts,py,java,go,js,ts" 这样的字符串，表示项目类型。
         </usage>
         </command>
+
+        <command>
+        <name>response_user</name>
+        <description>响应用户。</description>
+        <usage>
+         如果你需要直接发送信息给用户，那么可以通过 response_user 函数来直接回复用户。
+         
+         比如用户问你是谁？
+         你可以通过如下方式来回答：
+         response_user(response="你好，我是 auto-coder")
+        </usage>
+        </command>
         </commands>
         
         
@@ -1224,7 +1250,9 @@ class CommandAutoTuner:
             "get_project_related_files": self.tools.get_project_related_files,
             "ask_user":self.tools.ask_user,
             "read_file_with_keyword_ranges": self.tools.read_file_with_keyword_ranges,
-            "get_project_type": self.project_type_analyzer.analyze,                                    
+            "get_project_type": self.project_type_analyzer.analyze,
+            "response_user": self.tools.response_user,
+                                    
         }
 
         if command not in command_map:

{auto_coder-0.1.269 → auto_coder-0.1.271}/src/autocoder/commands/tools.py

@@ -28,6 +28,10 @@ from autocoder.utils.queue_communicate import (
 import sys
 import io
 from autocoder.common import files as files_utils
+from autocoder.common.printer import Printer
+from prompt_toolkit import PromptSession
+from prompt_toolkit.styles import Style
+
 
 @byzerllm.prompt()
 def detect_rm_command(command: str) -> Bool:
@@ -56,6 +60,7 @@ class AutoCommandTools:
         self.args = args
         self.llm = llm
         self.result_manager = ResultManager()
+        self.printer = Printer()
 
     def ask_user(self,question:str) -> str:
         '''
@@ -91,35 +96,43 @@ class AutoCommandTools:
         )
 
         # 显示问题面板
-        console.print(question_panel)
+        console.print(question_panel)        
 
-        # 创建一个自定义提示符
-        prompt = Prompt.ask(
-            "\n[bold green]Your Answer[/bold green]",
-            console=console
-        )
+        session = PromptSession( message=self.printer.get_message_from_key('tool_ask_user'))
+        try:
+            answer = session.prompt()
+        except KeyboardInterrupt:
+            answer = ""        
 
-        # 获取用户的回答
-        answer = prompt
+        self.result_manager.append(content=answer, meta = {
+            "action": "ask_user",
+            "input": {
+                "question": question
+            }
+        })
 
-        # 显示用户的回答
-        answer_text = Text(answer, style="italic")
+        return answer
+    
+    def response_user(self, response: str):
+        console = Console()
+        answer_text = Text(response, style="italic")
         answer_panel = Panel(
             answer_text,
-            title="[bold yellow]Your Response[/bold yellow]",
+            title="",
             border_style="green",
             expand=False
         )
         console.print(answer_panel)
 
-        self.result_manager.append(content=answer, meta = {
-            "action": "ask_user",
+        self.result_manager.append(content=response, meta = {
+            "action": "response_user",
             "input": {
-                "question": question
+                "response": response
             }
         })
 
-        return answer
+        return response
+        
 
     def run_python_code(self, code: str) -> str:
         """

{auto_coder-0.1.269 → auto_coder-0.1.271}/src/autocoder/common/auto_coder_lang.py

@@ -130,7 +130,7 @@ MESSAGES = {
         "quick_filter_too_long": "⚠️ index file is too large ({{ tokens_len }}/{{ max_tokens }}). The query will be split into {{ split_size }} chunks.",
         "quick_filter_tokens_len": "📊 Current index size: {{ tokens_len }} tokens",
         "estimated_chat_input_tokens": "Estimated chat input tokens: {{ estimated_input_tokens }}",
-        "estimated_input_tokens_in_generate": "Estimated input tokens in generate ({{ generate_mode }}): {{ estimated_input_tokens }}",
+        "estimated_input_tokens_in_generate": "Estimated input tokens in generate ({{ generate_mode }}): {{ estimated_input_tokens }}",        
         "model_has_access_restrictions": "{{model_name}} has access restrictions, cannot use the current function",
         "auto_command_not_found": "Auto command not found: {{command}}. Please check your input and try again.",
         "auto_command_failed": "Auto command failed: {{error}}. Please check your input and try again.",
@@ -163,11 +163,22 @@ MESSAGES = {
         "diff_blocks_title":"diff blocks",
         "index_exclude_files_error": "index filter exclude files fail: {{ error }}",
         "file_sliding_window_processing": "File {{ file_path }} is too large ({{ tokens }} tokens), processing with sliding window...",
-        "file_snippet_processing": "Processing file {{ file_path }} with code snippet extraction..."
+        "file_snippet_processing": "Processing file {{ file_path }} with code snippet extraction...",
+        "context_pruning_start": "⚠️ Context pruning started. Total tokens: {{ total_tokens }} (max allowed: {{ max_tokens }}). Applying strategy: {{ strategy }}.",
+        "context_pruning_reason": "Context length exceeds maximum limit ({{ total_tokens }} > {{ max_tokens }}). Pruning is required to fit within the model's context window.",
+        "rank_code_modification_title": "{{model_name}} ranking codes",
+        "sorted_files_message": "Reordered files:\n{% for file in files %}- {{ file }}\n{% endfor %}",
+        "estimated_input_tokens_in_ranking": "estimate input token {{ estimated_input_tokens }} when ranking",
+        "file_snippet_procesed": "{{ file_path }} processed with tokens: {{ tokens }} => {{ snippet_tokens }}. Current total tokens: {{ total_tokens }}",
+        "tool_ask_user": "Your Reply: ",
+        "tool_ask_user_accept":"Your Response received",
+        
     },
     "zh": {
         "file_sliding_window_processing": "文件 {{ file_path }} 过大 ({{ tokens }} tokens)，正在使用滑动窗口处理...",
         "file_snippet_processing": "正在对文件 {{ file_path }} 进行代码片段提取...",
+        "context_pruning_start": "⚠️ 开始上下文剪枝。总token数: {{ total_tokens }} (最大允许: {{ max_tokens }})。正在应用策略: {{ strategy }}。",
+        "context_pruning_reason": "上下文长度超过最大限制 ({{ total_tokens }} > {{ max_tokens }})。需要进行剪枝以适配模型的上下文窗口。",
         "file_scored_message": "文件评分: {{file_path}} - 分数: {{score}}",
         "invalid_file_pattern": "无效的文件模式: {{file_pattern}}. 例如: regex://.*/package-lock\\.json",
         "conf_not_found": "未找到配置文件: {{path}}",
@@ -308,7 +319,7 @@ MESSAGES = {
         "quick_filter_title": "{{ model_name }} 正在分析如何筛选上下文...",
         "quick_filter_failed": "❌ 快速过滤器失败: {{ error }}. ",
         "estimated_chat_input_tokens": "对话输入token预估为: {{ estimated_input_tokens }}",
-        "estimated_input_tokens_in_generate": "生成代码({{ generate_mode }})预计输入token数: {{ estimated_input_tokens_in_generate }}",
+        "estimated_input_tokens_in_generate": "生成代码({{ generate_mode }})预计输入token数: {{ estimated_input_tokens }}",        
         "model_has_access_restrictions": "{{model_name}} 有访问限制，无法使用当前功能",
         "auto_command_not_found": "未找到自动命令: {{command}}。请检查您的输入并重试。",
         "auto_command_failed": "自动命令执行失败: {{error}}。请检查您的输入并重试。",
@@ -324,7 +335,13 @@ MESSAGES = {
         "index_import_success": "索引导入成功: {{path}}",
         "edits_title": "编辑块",
         "diff_blocks_title": "差异块",
-        "index_exclude_files_error": "索引排除文件时出错: {{error}}"
+        "index_exclude_files_error": "索引排除文件时出错: {{error}}",
+        "rank_code_modification_title": "模型{{model_name}}对代码打分",
+        "sorted_files_message": "重新排序后的文件路径:\n{% for file in files %}- {{ file }}\n{% endfor %}",
+        "estimated_input_tokens_in_ranking": "排序预计输入token数: {{ estimated_input_tokens }}",
+        "file_snippet_procesed": "文件 {{ file_path }} 处理后token数: {{ tokens }} => {{ snippet_tokens }} 当前总token数: {{ total_tokens }}",
+        "tool_ask_user": "您的回复: ",
+        "tool_ask_user_accept":"收到您的回复",
     }}