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

autocoder/common/v2/agent/agentic_edit_tools/read_file_tool_resolver.py

@@ -1,18 +1,22 @@
 import os
-from typing import Dict, Any, Optional
+from typing import Dict, Any, Optional, List, Tuple
 from autocoder.common import AutoCoderArgs,SourceCode
+from autocoder.common.autocoderargs_parser import AutoCoderArgsParser
 from autocoder.common.v2.agent.agentic_edit_tools.base_tool_resolver import BaseToolResolver
 from autocoder.common.v2.agent.agentic_edit_types import ReadFileTool, ToolResult
-from autocoder.common.context_pruner import PruneContext
+from autocoder.common.pruner.context_pruner import PruneContext
 from autocoder.common import SourceCode
-from autocoder.rag.token_counter import count_tokens
+from autocoder.common.tokens import count_string_tokens as count_tokens
 from loguru import logger
 import typing
+import json
 from autocoder.rag.loaders import (
     extract_text_from_pdf,
     extract_text_from_docx,
     extract_text_from_ppt
 )
+from autocoder.common.llms.manager import LLMManager
+from autocoder.common.wrap_llm_hint.utils import add_hint_to_text
 
 if typing.TYPE_CHECKING:
     from autocoder.common.v2.agent.agentic_edit import AgenticEdit
@@ -22,13 +26,131 @@ class ReadFileToolResolver(BaseToolResolver):
     def __init__(self, agent: Optional['AgenticEdit'], tool: ReadFileTool, args: AutoCoderArgs):
         super().__init__(agent, tool, args)
         self.tool: ReadFileTool = tool  # For type hinting
-        self.shadow_manager = self.agent.shadow_manager if self.agent else None
+        
+        # Initialize AutoCoderArgs parser for flexible parameter parsing
+        self.args_parser = AutoCoderArgsParser()
+        
+        # 解析 context_prune_safe_zone_tokens 参数
+        parsed_safe_zone_tokens = self._get_parsed_safe_zone_tokens()
+        
         self.context_pruner = PruneContext(
-            max_tokens=self.args.context_prune_safe_zone_tokens,
+            max_tokens=parsed_safe_zone_tokens,
             args=self.args,
             llm=self.agent.context_prune_llm
+        ) if self.agent and self.agent.context_prune_llm else None
+
+        # LLM manager for model context window queries
+        self.llm_manager = LLMManager()
+
+    def _get_parsed_safe_zone_tokens(self) -> int:
+        """
+        解析 context_prune_safe_zone_tokens 参数，支持多种格式
+        
+        Returns:
+            解析后的 token 数量
+        """
+        return self.args_parser.parse_context_prune_safe_zone_tokens(
+            self.args.context_prune_safe_zone_tokens,
+            self.args.code_model
         )
 
+    def _extract_lines_by_range(self, content: str, start_line: Optional[int], end_line: Optional[int]) -> str:
+        """根据行号范围提取文件内容
+        
+        参数说明:
+        - start_line: 起始行号。正数表示从第N行开始(1-based)，负数表示从倒数第N行开始
+        - end_line: 结束行号。正数表示到第N行结束(1-based)，-1表示到文件末尾
+        """
+        if start_line is None and end_line is None:
+            return content
+        
+        lines = content.split('\n')
+        total_lines = len(lines)
+        
+        # 处理行号参数（转换为0-based索引）
+        start_idx = 0
+        end_idx = total_lines
+        
+        if start_line is not None:
+            if start_line < 0:
+                # 负数表示从倒数第N行开始
+                start_idx = max(0, total_lines + start_line)
+            else:
+                # 正数表示从第N行开始(1-based)
+                start_idx = max(0, start_line - 1)
+        
+        if end_line is not None:
+            if end_line == -1:
+                # -1 表示到文件末尾
+                end_idx = total_lines
+            elif end_line < -1:
+                # 其他负数表示到倒数第N行结束
+                end_idx = max(0, total_lines + end_line + 1)
+            else:
+                # 正数表示到第N行结束(1-based)
+                end_idx = min(total_lines, end_line)
+        
+        # 验证行号范围
+        if start_line is not None and start_line > 0 and start_idx >= total_lines:
+            return f"Error: start_line {start_line} exceeds total lines {total_lines}"
+        
+        if start_line is not None and start_line < 0 and abs(start_line) > total_lines:
+            return f"Error: start_line {start_line} (倒数第{abs(start_line)}行) exceeds total lines {total_lines}"
+        
+        if start_idx >= end_idx and not (end_line == -1 and start_idx == total_lines):
+            if start_line is not None and start_line < 0:
+                start_desc = f"倒数第{abs(start_line)}行"
+            else:
+                start_desc = f"第{start_line}行"
+            
+            if end_line == -1:
+                end_desc = "文件末尾"
+            elif end_line is not None and end_line < -1:
+                end_desc = f"倒数第{abs(end_line)}行"
+            else:
+                end_desc = f"第{end_line}行"
+            
+            return f"Error: start_line ({start_desc}) should be before end_line ({end_desc})"
+        
+        # 提取指定行范围
+        extracted_lines = lines[start_idx:end_idx]
+        return '\n'.join(extracted_lines)
+
+    def _extract_content_by_query(self, content: str, query: str, file_path: str) -> str:
+        """根据查询描述提取最相关的内容"""
+        if not query or not self.context_pruner:
+            return content
+        
+        # 使用 context_pruner 的查询功能提取相关内容
+        try:
+            # 创建一个包含查询的伪对话
+            query_conversation = {
+                "role": "user",
+                "content": query
+            }
+            
+            source_code = SourceCode(
+                module_name=file_path,
+                source_code=content,
+                tokens=count_tokens(content)
+            )
+            
+            # 使用 context_pruner 根据查询提取相关内容
+            pruned_sources = self.context_pruner.handle_overflow(
+                file_sources=[source_code],
+                conversations=[query_conversation],
+                strategy=self.args.context_prune_strategy
+            )
+            
+            if pruned_sources:
+                return pruned_sources[0].source_code
+            else:
+                return content
+                
+        except Exception as e:
+            logger.warning(f"Error extracting content by query '{query}': {str(e)}")
+            return content
+
     def _prune_file_content(self, content: str, file_path: str) -> str:
         """对文件内容进行剪枝处理"""
         if not self.context_pruner:
@@ -36,7 +158,8 @@ class ReadFileToolResolver(BaseToolResolver):
 
         # 计算 token 数量
         tokens = count_tokens(content)
-        if tokens <= self.args.context_prune_safe_zone_tokens:
+        parsed_safe_zone_tokens = self._get_parsed_safe_zone_tokens()
+        if tokens <= parsed_safe_zone_tokens:
             return content
 
         # 创建 SourceCode 对象
@@ -58,6 +181,68 @@ class ReadFileToolResolver(BaseToolResolver):
 
         return pruned_sources[0].source_code
 
+    def _get_model_context_window(self) -> int:
+        """获取当前代码模型的上下文窗口大小，默认 128000"""
+        try:
+            model_name = self.args.code_model or self.args.model
+            model = self.llm_manager.get_model(model_name)
+            if model and model.context_window:
+                return int(model.context_window)
+        except Exception as e:
+            logger.warning(f"Failed to get model context window: {e}")
+        return 128000
+
+    def _get_pruned_conversations(self) -> List[Dict[str, Any]]:
+        """获取已按 agent 逻辑剪枝后的对话，用于计算 token 开销"""
+        try:
+            if not self.agent:
+                return []
+            conversations = self.agent.current_conversations or []
+            # 复用 Agentic 对话修剪器，保持与主循环一致
+            agentic_pruner = self.agent.agentic_pruner
+            if agentic_pruner and conversations:
+                from copy import deepcopy
+                return agentic_pruner.prune_conversations(deepcopy(conversations))
+            return conversations
+        except Exception as e:
+            logger.warning(f"Failed to get pruned conversations: {e}")
+            return []
+
+    def _prune_file_to_fit_context_window(self, content: str, file_path: str) -> Tuple[bool, str]:
+        """
+        如果 对话token + 文件token 可能超出模型上下文窗口，不在此处裁剪文件，
+        而是返回带有提示信息的结果，提示大模型应先对会话进行删减后再读取文件。
+        """
+        if not self.agent:
+            return False, content
+
+        try:
+            model_window = self._get_model_context_window()
+            pruned_convs = self._get_pruned_conversations()
+            conv_tokens = count_tokens(json.dumps(pruned_convs, ensure_ascii=False))
+            file_tokens = count_tokens(content)
+
+            if conv_tokens + file_tokens <= model_window:
+                return False, content
+
+            # 构造提示信息，指导大模型优先裁剪会话（如使用 conversation_message_ids_write），然后再调用 read_file
+            hint = (
+                "The combined size of current conversation and requested file likely exceeds the model context window. "
+                "Please prune the conversation first (e.g., delete unnecessary tool_result messages using conversation_message_ids_write), "
+                "then call read_file again. "
+                f"Details: conversation_tokens={conv_tokens}, file_tokens={file_tokens}, window={model_window}, file={file_path}"
+            )
+
+            logger.info(
+                f"Conversation + file tokens exceed window ({conv_tokens}+{file_tokens}>{model_window}). Returning hint instead of file content.")
+
+            # 返回包含提示的包装文本，前置文件前500字符作为上下文
+            prefix = content[:500] if content else ""
+            return True, add_hint_to_text(prefix, hint)
+        except Exception as e:
+            logger.warning(f"Failed to calculate context window overflow: {e}")
+            return False, content
+
     def _read_file_content(self, file_path_to_read: str) -> str:
         content = ""
         ext = os.path.splitext(file_path_to_read)[1].lower()
@@ -79,48 +264,64 @@ class ReadFileToolResolver(BaseToolResolver):
             with open(file_path_to_read, 'r', encoding='utf-8', errors='replace') as f:
                 content = f.read()
 
-        # 对内容进行剪枝处理
-        return self._prune_file_content(content, file_path_to_read)
-
-    def read_file_with_shadow(self, file_path: str, source_dir: str, abs_project_dir: str, abs_file_path: str) -> ToolResult:
-        """Read file using shadow manager for path translation"""
-        try:
-            # Check if shadow file exists
-            shadow_path = self.shadow_manager.to_shadow_path(abs_file_path)
-            if os.path.exists(shadow_path) and os.path.isfile(shadow_path):
-                try:
-                    content = self._read_file_content(shadow_path)
-                    logger.info(f"[Shadow] Successfully processed shadow file: {shadow_path}")
-                    return ToolResult(success=True, message=f"{file_path}", content=content)
-                except Exception as e_shadow:
-                    logger.warning(f"Failed to read shadow file {shadow_path} with _read_file_content: {str(e_shadow)}. Falling back to original file.")
-
-            # If shadow file doesn't exist or reading failed, try original file
-            if not os.path.exists(abs_file_path):
-                return ToolResult(success=False, message=f"Error: File not found at path: {file_path}")
-            if not os.path.isfile(abs_file_path):
-                return ToolResult(success=False, message=f"Error: Path is not a file: {file_path}")
-
-            content = self._read_file_content(abs_file_path)
-            logger.info(f"Successfully processed file: {file_path}")
-            return ToolResult(success=True, message=f"{file_path}", content=content)
+        # 处理新参数
+        # 1. 先根据行号范围提取内容
+        if self.tool.start_line is not None or self.tool.end_line is not None:
+            content = self._extract_lines_by_range(content, self.tool.start_line, self.tool.end_line)
+            
+        # 2. 如果有查询，根据查询提取相关内容
+        if self.tool.query:
+            content = self._extract_content_by_query(content, self.tool.query, file_path_to_read)
+        
+        # 3. 检查当前对话 + 文件是否超出模型窗口，若需要提示会话剪枝则直接返回带提示内容
+        should_prune, maybe_hinted_content = self._prune_file_to_fit_context_window(content, file_path_to_read)
+        if should_prune:
+            return maybe_hinted_content
 
-        except Exception as e:
-            logger.warning(f"Error processing file '{file_path}': {str(e)}")
-            logger.exception(e)
-            return ToolResult(success=False, message=f"An error occurred while processing the file '{file_path}': {str(e)}")
+        # 4. 最后进行常规的剪枝处理（基于 safe_zone_tokens 的兜底）
+        return self._prune_file_content(maybe_hinted_content, file_path_to_read)
 
-    def read_file_normal(self, file_path: str, source_dir: str, abs_project_dir: str, abs_file_path: str) -> ToolResult:
+    def read_file_normal(self, file_path: str) -> ToolResult:
         """Read file directly without using shadow manager"""
         try:
-            if not os.path.exists(abs_file_path):
+            if not os.path.exists(file_path):
                 return ToolResult(success=False, message=f"Error: File not found at path: {file_path}")
-            if not os.path.isfile(abs_file_path):
+            if not os.path.isfile(file_path):
                 return ToolResult(success=False, message=f"Error: Path is not a file: {file_path}")
 
-            content = self._read_file_content(abs_file_path)
-            logger.info(f"Successfully processed file: {file_path}")
-            return ToolResult(success=True, message=f"{file_path}", content=content)
+            content = self._read_file_content(file_path)
+            
+            # 构建详细的成功消息
+            message_parts = [f"{file_path}"]
+            if self.tool.start_line is not None or self.tool.end_line is not None:
+                # 构建起始行描述
+                if self.tool.start_line is not None:
+                    if self.tool.start_line < 0:
+                        start = f"-{abs(self.tool.start_line)}"  # 负数行号显示为 -N
+                    else:
+                        start = str(self.tool.start_line)
+                else:
+                    start = "1"
+                
+                # 构建结束行描述
+                if self.tool.end_line == -1:
+                    end = "end"
+                elif self.tool.end_line is not None:
+                    if self.tool.end_line < -1:
+                        end = str(self.tool.end_line)  # 其他负数行号直接显示
+                    else:
+                        end = str(self.tool.end_line)
+                else:
+                    end = "end"
+                
+                message_parts.append(f"lines {start}-{end}")
+            if self.tool.query:
+                message_parts.append(f"filtered by query: '{self.tool.query}'")
+            
+            message = " ".join(message_parts)
+            
+            logger.info(f"Successfully processed file: {message}")
+            return ToolResult(success=True, message=message, content=content)
 
         except Exception as e:
             logger.warning(f"Error processing file '{file_path}': {str(e)}")
@@ -129,13 +330,5 @@ class ReadFileToolResolver(BaseToolResolver):
 
     def resolve(self) -> ToolResult:
         """Resolve the read file tool by calling the appropriate implementation"""
-        file_path = self.tool.path
-        source_dir = self.args.source_dir or "."
-        abs_project_dir = os.path.abspath(source_dir)
-        abs_file_path = os.path.abspath(os.path.join(source_dir, file_path))
-
-        # Choose the appropriate implementation based on whether shadow_manager is available
-        if self.shadow_manager:
-            return self.read_file_with_shadow(file_path, source_dir, abs_project_dir, abs_file_path)
-        else:
-            return self.read_file_normal(file_path, source_dir, abs_project_dir, abs_file_path)
+        file_path = self.tool.path                        
+        return self.read_file_normal(file_path)