auto-coder 0.1.400__py3-none-any.whl → 2.0.0__py3-none-any.whl

autocoder/common/v2/agent/agentic_edit_types.py

@@ -1,113 +1,465 @@
 from pydantic import BaseModel
-from typing import List, Dict, Any, Callable, Optional, Type
+from typing import List, Dict, Any, Callable, Optional, Type, Union
 from pydantic import SkipValidation
+from enum import Enum
+
+# Import for Pull Request support and Git operations
+try:
+    from autocoder.common.pull_requests.models import PRResult, PlatformType
+except ImportError:
+    # Fallback if pull_requests module is not available
+    PRResult = None
+    PlatformType = None
+
+try:
+    from autocoder.common.git_utils import CommitResult
+except ImportError:
+    # Fallback if git_utils is not available
+    CommitResult = None
+
+
+class ConversationAction(str, Enum):
+    """Enumeration for conversation actions"""
+
+    RESUME = "resume"
+    NEW = "new"
+    CONTINUE = "continue"
+    LIST = "list"
+    COMMAND = "command"
+
 
 # Result class used by Tool Resolvers
 class ToolResult(BaseModel):
     success: bool
     message: str
-    content: Any = None # Can store file content, command output, etc.
+    content: Any = None  # Can store file content, command output, etc.
+
 
 # Pydantic Models for Tools
 class BaseTool(BaseModel):
     pass
 
+
 class ExecuteCommandTool(BaseTool):
     command: str
     requires_approval: bool
+    timeout: Optional[int] = 60  # 超时时间（秒），默认60秒
+    background: Optional[bool] = False  # 是否后台运行，默认为False
+
 
 class ReadFileTool(BaseTool):
     path: str
+    start_line: Optional[int] = None
+    end_line: Optional[int] = None
+    query: Optional[str] = None
+
 
 class WriteToFileTool(BaseTool):
     path: str
     content: str
+    mode: str = "write"  # write or append
+
+
+class CountTokensTool(BaseTool):
+    """Token统计工具类型"""
+
+    path: str  # 文件或目录路径
+    recursive: Optional[bool] = True  # 是否递归统计目录
+    include_summary: Optional[bool] = True  # 是否包含分类汇总
+
+
+class ExtractToTextTool(BaseTool):
+    """提取各种格式文件（包括URL）的文本内容到目标文件"""
+
+    source_path: str  # 源路径，可以是本地文件路径或URL
+    target_path: str  # 目标文本文件路径
+
+
+class SessionStartTool(BaseTool):
+    """启动交互式会话工具"""
+
+    command: str  # 要执行的命令，如 "python -i"
+    timeout: Optional[int] = 60  # 启动超时时间（秒）
+    cwd: Optional[str] = None  # 工作目录
+    env: Optional[Dict[str, str]] = None  # 环境变量
+
+
+class SessionInteractiveTool(BaseTool):
+    """与交互式会话交互的工具"""
+
+    session_id: str  # 会话ID
+    input: str  # 发送到stdin的内容
+    read_timeout: Optional[int] = 5  # 等待读取输出的时间（秒）
+    max_bytes: Optional[int] = 40960  # 每次读取的最大字节数
+    expect_prompt: Optional[bool] = False  # 是否等待提示符出现再返回
+    prompt_regex: Optional[str] = None  # 提示符的正则表达式
+
+
+class SessionStopTool(BaseTool):
+    """停止交互式会话工具"""
+
+    session_id: str  # 会话ID
+    force: Optional[bool] = False  # True=强制杀死，False=优雅终止
+
 
 class ReplaceInFileTool(BaseTool):
     path: str
     diff: str
+    fence_0: Optional[str] = "```"
+    fence_1: Optional[str] = "```"
+
 
 class SearchFilesTool(BaseTool):
     path: str
     regex: str
     file_pattern: Optional[str] = None
 
+
 class ListFilesTool(BaseTool):
     path: str
     recursive: Optional[bool] = False
 
+
 class ListCodeDefinitionNamesTool(BaseTool):
     path: str
 
+
 class AskFollowupQuestionTool(BaseTool):
     question: str
     options: Optional[List[str]] = None
 
+
 class AttemptCompletionTool(BaseTool):
     result: str
     command: Optional[str] = None
 
+
 class PlanModeRespondTool(BaseTool):
     response: str
     options: Optional[List[str]] = None
 
+
 class UseMcpTool(BaseTool):
     server_name: str
     tool_name: str
-    query:str
+    query: str
+
 
 class UseRAGTool(BaseTool):
     server_name: str
     query: str
 
-class ListPackageInfoTool(BaseTool):
+
+class RunNamedSubagentsTool(BaseTool):
+    """
+    Tool for running a named subagent with a specific query.
+    Gets agent details from the agents module and executes via auto-coder.run.
+    """
+
+    subagents: str
+    background: Optional[bool] = False  # 是否后台运行，默认为False
+
+
+class ACModReadTool(BaseTool):
     path: str  # 源码包目录，相对路径或绝对路径
 
+
+class ACModWriteTool(BaseTool):
+    """
+    Tool for creating or updating an AC Module's .ac.mod.md file.
+    """
+
+    path: str  # AC Module directory path, relative or absolute path
+    diff: str  # diff content to edit the .ac.mod.md file
+
+
+class ACModListTool(BaseTool):
+    """
+    Tool for listing all directories containing .ac.mod.md files (AC Modules).
+    """
+
+    path: Optional[str] = None  # Optional root path to search, defaults to source_dir
+
+
+class TodoReadTool(BaseTool):
+    """
+    Tool for reading the current todo list.
+    Takes no parameters.
+    """
+
+    pass  # No parameters needed
+
+
+class TodoWriteTool(BaseTool):
+    """
+    Tool for creating and managing a structured task list.
+    """
+
+    action: str  # 'create', 'update', 'mark_progress', 'mark_completed', 'add_task'
+    task_id: Optional[str] = None  # Task ID for update/mark operations
+    content: Optional[str] = None  # Task content for create/add operations
+    priority: Optional[str] = None  # 'high', 'medium', 'low'
+    status: Optional[str] = None  # 'pending', 'in_progress', 'completed'
+    notes: Optional[str] = None  # Additional notes for the task
+
+
+class ConversationMessageIdsWriteTool(BaseTool):
+    """
+    Tool for writing conversation message IDs for pruning.
+    Manages message IDs that should be DELETED during conversation pruning.
+    """
+
+    message_ids: str  # Message IDs specification like "9226b3a4,204e1cd8"
+    action: str  # Action to perform: "create", "append", or "delete"
+
+
+class ConversationMessageIdsReadTool(BaseTool):
+    """
+    Tool for reading existing conversation message IDs configuration.
+    Returns the current message IDs configuration for the active conversation.
+    """
+
+    pass  # No parameters needed, uses current conversation ID
+
+
+class BackgroundTaskTool(BaseTool):
+    """后台任务管理工具
+
+    支持对后台运行的命令和子代理任务进行统一管理，包括：
+    - list: 列出当前会话的后台任务
+    - monitor: 监控任务输出内容
+    - cleanup: 清理已完成的任务记录
+    - kill: 终止正在运行的任务
+
+    任务标识支持 task_id、pid 和 process_uniq_id 三种方式。
+    """
+
+    action: str  # 操作类型: "list", "monitor", "cleanup", "kill"
+
+    # list 操作参数
+    show_completed: Optional[bool] = False  # 是否显示已完成的任务
+    task_type: Optional[str] = None  # 过滤任务类型: "command", "subagent", None(全部)
+
+    # monitor 操作参数
+    task_id: Optional[str] = None  # 任务ID（UUID格式）
+    pid: Optional[int] = None  # 进程ID（系统PID）
+    process_uniq_id: Optional[str] = None  # 进程唯一ID（显示友好的短ID）
+    lines: Optional[int] = 100  # 显示的行数，默认100行
+    output_type: Optional[str] = "both"  # "stdout", "stderr", "both"
+
+    # cleanup 操作参数
+    status_filter: Optional[str] = None  # "completed", "failed", None(全部已结束的)
+    older_than_minutes: Optional[int] = None  # 清理多少分钟前完成的任务
+    task_ids: Optional[List[str]] = None  # 指定要清理的任务ID列表
+
+    # kill 操作参数
+    pids: Optional[List[int]] = None  # 批量终止多个进程（系统PID）
+    force: Optional[bool] = False  # 是否强制终止（SIGKILL vs SIGTERM）
+    kill_children: Optional[bool] = True  # 是否同时终止子进程
+
+
+class WebCrawlTool(BaseTool):
+    """网页爬取工具，使用 Firecrawl、Metaso 或 BochaAI 进行网页爬取
+
+    支持多个API key配置时的并发爬取，结果会合并返回。
+    使用线程池进行并发处理以提高效率。
+    """
+
+    url: str  # 要爬取的URL
+    limit: Optional[int] = 10  # 爬取页面数量限制
+    scrape_options: Optional[str] = None  # 抓取选项，JSON字符串格式
+    exclude_paths: Optional[str] = None  # 排除的路径，逗号分割
+    include_paths: Optional[str] = None  # 包含的路径，逗号分割
+    max_depth: Optional[int] = None  # 最大爬取深度
+    allow_subdomains: Optional[str] = "false"  # 是否允许子域名，true/false
+    crawl_entire_domain: Optional[str] = "false"  # 是否爬取整个域名，true/false
+
+
+class WebSearchTool(BaseTool):
+    """网页搜索工具，使用 Firecrawl、Metaso 或 BochaAI 进行网页搜索
+
+    支持多个API key配置时的并发搜索，结果会合并返回。
+    使用线程池进行并发处理以提高效率。
+    """
+
+    query: str  # 搜索查询
+    limit: Optional[int] = 5  # 返回结果数量限制
+    sources: Optional[str] = None  # 搜索源类型，逗号分割：web,news,images
+    scrape_options: Optional[str] = None  # 抓取选项，JSON字符串格式
+    location: Optional[str] = None  # 搜索位置
+    tbs: Optional[str] = None  # 时间过滤参数
+
+
+class LoadExtraDocumentTool(BaseTool):
+    """按名称加载内置额外文档（以 prompt 文本形式返回）
+
+    例如：
+    <load_extra_document>
+        <name>workflow_subagents</name>
+    </load_extra_document>
+    """
+
+    name: str
+
+
+class ExecuteWorkflowTool(BaseTool):
+    """执行 workflow 工具
+
+    按名称执行定义在 .autocoderworkflow/ 等目录的 workflow YAML 文件。
+
+    例如：
+    <execute_workflow>
+        <name>coder</name>
+    </execute_workflow>
+    """
+
+    name: str
+    vars_override: Optional[str] = None  # JSON 字符串，可选的变量覆盖
+
+
 # Event Types for Rich Output Streaming
 class LLMOutputEvent(BaseModel):
     """Represents plain text output from the LLM."""
+
     text: str
 
+
 class LLMThinkingEvent(BaseModel):
     """Represents text within <thinking> tags from the LLM."""
+
     text: str
 
+
 class ToolCallEvent(BaseModel):
     """Represents the LLM deciding to call a tool."""
-    tool: SkipValidation[BaseTool] # Use SkipValidation as BaseTool itself is complex
+
+    tool: SkipValidation[BaseTool]  # Use SkipValidation as BaseTool itself is complex
     tool_xml: str
 
+
 class ToolResultEvent(BaseModel):
     """Represents the result of executing a tool."""
+
     tool_name: str
     result: ToolResult
 
+
 class TokenUsageEvent(BaseModel):
     """Represents the result of executing a tool."""
+
     usage: Any
 
 
 class ConversationIdEvent(BaseModel):
     """Represents the conversation id."""
+
     conversation_id: str
 
+
 class PlanModeRespondEvent(BaseModel):
     """Represents the LLM attempting to complete the task."""
-    completion: SkipValidation[PlanModeRespondTool] # Skip validation
+
+    completion: SkipValidation[PlanModeRespondTool]  # Skip validation
     completion_xml: str
 
+
 class CompletionEvent(BaseModel):
     """Represents the LLM attempting to complete the task."""
-    completion: SkipValidation[AttemptCompletionTool] # Skip validation
+
+    completion: SkipValidation[AttemptCompletionTool]  # Skip validation
     completion_xml: str
 
+
+class PreCommitEvent(BaseModel):
+    """Represents the LLM attempting to complete the task."""
+
+    commit_result: CommitResult
+    tpe: str = "pre_commit"
+
+
+class CommitEvent(BaseModel):
+    """Represents the LLM attempting to complete the task."""
+
+    commit_result: CommitResult
+    tpe: str = "commit"
+
+
 class ErrorEvent(BaseModel):
     """Represents an error during the process."""
+
+    message: str
+
+
+class RetryEvent(BaseModel):
+    """Represents a retry event."""
+
     message: str
 
+
 class WindowLengthChangeEvent(BaseModel):
     """Represents the token usage in the conversation window."""
+
     tokens_used: int
+    pruned_tokens_used: int
+    conversation_round: int
+
+
+# Base event class for all agent events
+class AgentEvent(BaseModel):
+    """Base class for all agent events."""
+
+    pass
+
+
+# Metadata for token usage tracking
+class AgentSingleOutputMeta(BaseModel):
+    """Metadata for tracking token usage for a single LLM output."""
+
+    model_name: str
+    input_tokens: int
+    output_tokens: int
+    input_cost: float
+    output_cost: float
+
+
+# Pull Request Result type - compatible with PRResult from pull_requests module
+class PullRequestResult(BaseModel):
+    """Pull Request operation result for agentic edit events"""
+
+    success: bool
+    pr_number: Optional[int] = None
+    pr_url: Optional[str] = None
+    pr_id: Optional[str] = None
+    error_message: Optional[str] = None
+    error_code: Optional[str] = None
+    platform: Optional[str] = None  # Using str instead of PlatformType for flexibility
+    raw_response: Optional[Dict[str, Any]] = None
+    retry_after: Optional[int] = None
+
+    @classmethod
+    def from_pr_result(cls, pr_result) -> "PullRequestResult":
+        """Create PullRequestResult from PRResult instance"""
+        if pr_result is None:
+            return cls(success=False, error_message="No PR result provided")
+
+        return cls(
+            success=pr_result.success,
+            pr_number=pr_result.pr_number,
+            pr_url=pr_result.pr_url,
+            pr_id=pr_result.pr_id,
+            error_message=pr_result.error_message,
+            error_code=pr_result.error_code,
+            platform=str(pr_result.platform) if pr_result.platform else None,
+            raw_response=pr_result.raw_response,
+            retry_after=pr_result.retry_after,
+        )
+
+
+class PullRequestEvent(BaseModel):
+    """Represents a Pull Request creation event."""
+
+    pull_request_result: PullRequestResult
+
 
 # Deprecated: Will be replaced by specific Event types
 # class PlainTextOutput(BaseModel):
@@ -128,9 +480,27 @@ TOOL_MODEL_MAP: Dict[str, Type[BaseTool]] = {
     "plan_mode_respond": PlanModeRespondTool,
     "use_mcp_tool": UseMcpTool,
     "use_rag_tool": UseRAGTool,
-    "list_package_info": ListPackageInfoTool,    
+    "run_named_subagents": RunNamedSubagentsTool,
+    "todo_read": TodoReadTool,
+    "todo_write": TodoWriteTool,
+    "ac_mod_read": ACModReadTool,
+    "ac_mod_write": ACModWriteTool,
+    "ac_mod_list": ACModListTool,
+    "count_tokens": CountTokensTool,
+    "extract_to_text": ExtractToTextTool,
+    "session_start": SessionStartTool,
+    "session_interactive": SessionInteractiveTool,
+    "session_stop": SessionStopTool,
+    "conversation_message_ids_write": ConversationMessageIdsWriteTool,
+    "conversation_message_ids_read": ConversationMessageIdsReadTool,
+    "background_task": BackgroundTaskTool,
+    "web_crawl": WebCrawlTool,
+    "web_search": WebSearchTool,
+    "load_extra_document": LoadExtraDocumentTool,
+    "execute_workflow": ExecuteWorkflowTool,
 }
 
+
 class FileChangeEntry(BaseModel):
     type: str  # 'added' or 'modified'
     diffs: List[str] = []
@@ -144,6 +514,8 @@ class AgenticEditRequest(BaseModel):
 class FileOperation(BaseModel):
     path: str
     operation: str  # e.g., "MODIFY", "REFERENCE", "ADD", "REMOVE"
+
+
 class MemoryConfig(BaseModel):
     """
     A model to encapsulate memory configuration and operations.
@@ -182,9 +554,13 @@ class CommandConfig(BaseModel):
     index_import: SkipValidation[Callable]
     exclude_files: SkipValidation[Callable]
 
-class AgenticEditConversationConfig(BaseModel):     
+
+class AgenticEditConversationConfig(BaseModel):
     conversation_name: Optional[str] = "current"
-    conversation_id: Optional[str] = None 
+    conversation_id: Optional[str] = None
     action: Optional[str] = None
     query: Optional[str] = None
     pull_request: bool = False
+    pre_commit: bool = False
+    post_commit: bool = False
+    is_sub_agent: bool = False

autocoder/common/v2/agent/runner/__init__.py

@@ -0,0 +1,31 @@
+"""
+Runner 模块提供了多种运行模式，用于在不同环境下执行 AgenticEdit 代理。
+
+这个模块包含三种主要的运行器:
+1. TerminalRunner: 在终端环境中运行代理，提供格式化输出
+2. FileBasedEventRunner: 将代理事件转换为标准事件系统格式
+3. SdkRunner: 提供生成器接口，适用于SDK环境
+
+使用示例:
+```python
+from autocoder.common.v2.agent.runner import TerminalRunner
+from autocoder.common.v2.agent.agentic_edit_types import AgenticEditRequest
+
+runner = TerminalRunner(llm=llm, args=args, ...)
+runner.run(AgenticEditRequest(user_input="请帮我实现一个HTTP服务器"))
+```
+"""
+
+from .base_runner import BaseRunner
+from .terminal_runner import TerminalRunner
+from .file_based_event_runner import FileBasedEventRunner
+from .sdk_runner import SdkRunner
+from .tool_display import get_tool_display_message
+
+__all__ = [
+    "BaseRunner",
+    "TerminalRunner",
+    "FileBasedEventRunner",
+    "SdkRunner",
+    "get_tool_display_message"
+]

autocoder/common/v2/agent/runner/base_runner.py

@@ -0,0 +1,92 @@
+"""
+BaseRunner 提供了所有运行器的基础实现和接口定义。
+
+这个模块定义了代理运行的核心流程，包括请求处理、事件流生成和变更应用等。
+各种具体运行器通过继承 BaseRunner 并实现特定的运行逻辑来提供不同的运行模式。
+"""
+
+import logging
+from typing import Generator, Any, Dict, Optional, Callable, List
+
+from autocoder.common.v2.agent.agentic_edit_types import (
+    AgenticEditRequest, AgentEvent,
+    AgenticEditConversationConfig
+)
+from autocoder.common import AutoCoderArgs
+from autocoder.common.printer import Printer
+
+logger = logging.getLogger(__name__)
+
+class BaseRunner:
+    """
+    代理运行器的基类，定义了运行代理的核心流程和接口。
+    
+    所有具体的运行器实现（终端、事件系统、SDK）都应该继承这个类，
+    并根据特定需求实现 run 方法。
+    """
+    
+    def __init__(
+        self,
+        llm: Any,
+        args: AutoCoderArgs,
+        conversation_config: Optional[AgenticEditConversationConfig] = None,
+        cancel_token: Optional[str] = None,
+        system_prompt: Optional[str] = None
+    ):
+        """
+        初始化代理运行器。
+        
+        Args:
+            llm: 大语言模型实例
+            args: 自动编码器参数
+            conversation_config: 会话配置（可选）
+        """
+        from autocoder.common.v2.agent.agentic_edit import AgenticEdit
+        
+        self.llm = llm
+        self.args = args
+        self.conversation_config = conversation_config  
+        self.system_prompt = system_prompt
+        self.printer = Printer()              
+        
+        self.agent = AgenticEdit(
+            llm=llm,
+            args=args,
+            conversation_config=conversation_config,
+            cancel_token=cancel_token,
+            custom_system_prompt=system_prompt
+        )
+    
+    def run(self, request: AgenticEditRequest) -> Any:
+        """
+        运行代理处理请求。
+        
+        这是一个抽象方法，需要由子类实现。
+        
+        Args:
+            request: 代理编辑请求
+            
+        Returns:
+            根据具体实现返回不同类型的结果
+        """
+        raise NotImplementedError("子类必须实现run方法")
+    
+    def apply_pre_changes(self) -> None:
+        """应用预处理变更"""
+        self.agent.apply_pre_changes()
+    
+    def apply_changes(self) -> None:
+        """应用代理执行后的变更"""
+        self.agent.apply_changes()
+    
+    def analyze(self, request: AgenticEditRequest) -> Generator[AgentEvent, None, None]:
+        """
+        分析请求并生成事件流。
+        
+        Args:
+            request: 代理编辑请求
+            
+        Returns:
+            事件生成器
+        """
+        return self.agent.analyze(request)