jarvis-ai-assistant 0.1.149__py3-none-any.whl → 0.1.150__py3-none-any.whl

jarvis/jarvis_tools/file_operation.py

@@ -8,15 +8,14 @@ from jarvis.jarvis_utils.globals import add_read_file_record
 from jarvis.jarvis_utils.output import OutputType, PrettyOutput
 # 导入文件处理器
 from jarvis.jarvis_utils.file_processors import (
-    TextFileProcessor, PDFProcessor, DocxProcessor, 
-    PPTProcessor, ExcelProcessor
+    TextFileProcessor
 )
 
 
 
 class FileOperationTool:
     name = "file_operation"
-    description = "文件批量操作工具，可批量读写多个文件，支持文本、PDF、Word、Excel、PPT等格式，适用于需要同时处理多个文件的场景（读取配置文件、保存生成内容等）"
+    description = "文件批量操作工具，可批量读写多个文件，仅支持文本文件，适用于需要同时处理多个文件的场景（读取配置文件、保存生成内容等）"
     parameters = {
         "type": "object",
         "properties": {
@@ -44,10 +43,6 @@ class FileOperationTool:
     def _get_file_processor(self, file_path: str):
         """获取适合处理指定文件的处理器"""
         processors = [
-            PDFProcessor,    # PDF文件处理器
-            DocxProcessor,   # Word文档处理器
-            PPTProcessor,    # PowerPoint演示文稿处理器
-            ExcelProcessor,  # Excel表格处理器
             TextFileProcessor  # 文本文件处理器(放在最后作为兜底)
         ]
         
@@ -126,25 +121,11 @@ class FileOperationTool:
                                 "stderr": f"读取文本文件失败: {str(e)}"
                             }
                     else:
-                        # 使用专用处理器来提取非文本文件的内容
-                        try:
-                            spinner.text = f"使用 {processor.__name__} 提取 {abs_path} 的内容..."
-                            content = processor.extract_text(abs_path)
-                            # 获取文件类型友好名称
-                            file_type_names = {
-                                PDFProcessor: "PDF文档",
-                                DocxProcessor: "Word文档",
-                                PPTProcessor: "PowerPoint演示文稿",
-                                ExcelProcessor: "Excel表格"
-                            }
-                            file_type = file_type_names.get(processor, file_extension)
-                            file_info = f"\n文件: {abs_path} ({file_type})"
-                        except Exception as e:
-                            return {
-                                "success": False,
-                                "stdout": "",
-                                "stderr": f"提取 {file_extension} 文件内容失败: {str(e)}"
-                            }
+                        return {
+                            "success": False,
+                            "stdout": "",
+                            "stderr": f"不支持的文件类型: {file_extension}"
+                        }
                     
                     # 构建输出信息
                     output = f"{file_info}\n{content}" + "\n\n"

jarvis/jarvis_tools/methodology.py

@@ -3,6 +3,7 @@ import json
 import hashlib
 from typing import Dict, Any
 
+from jarvis.jarvis_utils.config import get_data_dir
 from jarvis.jarvis_utils.output import OutputType, PrettyOutput
 
 
@@ -35,7 +36,7 @@ class MethodologyTool:
 
     def __init__(self):
         """初始化经验管理工具"""
-        self.methodology_dir = os.path.expanduser("~/.jarvis/methodologies")
+        self.methodology_dir = os.path.join(get_data_dir(), "methodologies")
         self._ensure_dir_exists()
 
     def _ensure_dir_exists(self):

jarvis/jarvis_tools/registry.py

@@ -10,10 +10,12 @@ import yaml
 from jarvis.jarvis_agent.output_handler import OutputHandler
 from jarvis.jarvis_platform.registry import PlatformRegistry
 from jarvis.jarvis_tools.base import Tool
-from jarvis.jarvis_utils.config import INPUT_WINDOW_REVERSE_SIZE, get_max_input_token_count
+from jarvis.jarvis_utils.config import INPUT_WINDOW_REVERSE_SIZE, get_max_input_token_count, get_data_dir
 from jarvis.jarvis_utils.embedding import get_context_token_count
 from jarvis.jarvis_utils.output import OutputType, PrettyOutput
 from jarvis.jarvis_utils.utils import ct, ot, init_env
+from jarvis.jarvis_mcp.local_mcp_client import LocalMcpClient
+from jarvis.jarvis_mcp.remote_mcp_client import RemoteMcpClient
 
 
 
@@ -131,6 +133,7 @@ class ToolRegistry(OutputHandler):
         # 加载内置工具和外部工具
         self._load_builtin_tools()
         self._load_external_tools()
+        self._load_mcp_tools()
         self.max_input_token_count = get_max_input_token_count() - INPUT_WINDOW_REVERSE_SIZE
 
     def use_tools(self, name: List[str]) -> None:
@@ -152,6 +155,16 @@ class ToolRegistry(OutputHandler):
         """
         self.tools = {name: tool for name, tool in self.tools.items() if name not in names}
 
+    def _load_mcp_tools(self) -> None:
+        """从jarvis_data/tools/mcp加载工具"""
+        mcp_tools_dir = Path(get_data_dir()) / 'mcp'
+        if not mcp_tools_dir.exists():
+            return
+
+        # 遍历目录中的所有.yaml文件
+        for file_path in mcp_tools_dir.glob("*.yaml"):
+            self.register_mcp_tool_by_file(str(file_path))
+
     def _load_builtin_tools(self) -> None:
         """从内置工具目录加载工具"""
         tools_dir = Path(__file__).parent
@@ -165,8 +178,8 @@ class ToolRegistry(OutputHandler):
             self.register_tool_by_file(str(file_path))
 
     def _load_external_tools(self) -> None:
-        """从~/.jarvis/tools加载外部工具"""
-        external_tools_dir = Path.home() / '.jarvis/tools'
+        """从jarvis_data/tools加载外部工具"""
+        external_tools_dir = Path(get_data_dir()) / 'tools'
         if not external_tools_dir.exists():
             return
 
@@ -178,6 +191,98 @@ class ToolRegistry(OutputHandler):
 
             self.register_tool_by_file(str(file_path))
 
+    def register_mcp_tool_by_file(self, file_path: str) -> bool:
+        """从指定文件加载并注册工具
+        
+        参数:
+            file_path: 工具文件的路径
+
+        返回:
+            bool: 工具是否加载成功
+        """
+        try:
+            config = yaml.safe_load(open(file_path, 'r', encoding='utf-8'))
+            if 'type' not in config:
+                PrettyOutput.print(f"文件 {file_path} 缺少type字段", OutputType.WARNING)
+                return False
+            if config['type'] == 'local':
+                if 'command' not in config:
+                    PrettyOutput.print(f"文件 {file_path} 缺少command字段", OutputType.WARNING)
+                    return False
+                
+                # 创建本地MCP客户端
+                mcp_client = LocalMcpClient(config)
+                
+                # 获取工具信息
+                tools = mcp_client.get_tool_list()
+                if not tools:
+                    PrettyOutput.print(f"从 {file_path} 获取工具列表失败", OutputType.WARNING)
+                    return False
+                
+                # 注册每个工具
+                for tool in tools:
+                    def create_local_execute_func(tool_name: str, client: LocalMcpClient):
+                        def execute(arguments: Dict[str, Any]) -> Dict[str, Any]:
+                            args = arguments.copy()
+                            args.pop('agent', None)
+                            args.pop('want', None)
+                            ret = client.execute(tool_name, args)
+                            PrettyOutput.print(f"MCP {tool_name} 执行结果:\n{yaml.safe_dump(ret)}", OutputType.TOOL)
+                            return ret
+                        return execute
+                    
+                    # 注册工具
+                    self.register_tool(
+                        name=tool['name'],
+                        description=tool['description'],
+                        parameters=tool['parameters'],
+                        func=create_local_execute_func(tool['name'], mcp_client)
+                    )
+                
+                return True
+                
+            elif config['type'] == 'remote':
+                if 'base_url' not in config:
+                    PrettyOutput.print(f"文件 {file_path} 缺少base_url字段", OutputType.WARNING)
+                    return False
+                
+                # 创建远程MCP客户端
+                mcp_client = RemoteMcpClient(config)
+                
+                # 获取工具信息
+                tools = mcp_client.get_tool_list()
+                if not tools:
+                    PrettyOutput.print(f"从 {file_path} 获取工具列表失败", OutputType.WARNING)
+                    return False
+                
+                # 注册每个工具
+                for tool in tools:
+                    def create_remote_execute_func(tool_name: str, client: RemoteMcpClient):
+                        def execute(arguments: Dict[str, Any]) -> Dict[str, Any]:
+                            args = arguments.copy()
+                            args.pop('agent', None)
+                            args.pop('want', None)
+                            ret = client.execute(tool_name, args)
+                            PrettyOutput.print(f"MCP {tool_name} 执行结果:\n{yaml.safe_dump(ret)}", OutputType.TOOL)
+                            return ret
+                        return execute
+                    
+                    # 注册工具
+                    self.register_tool(
+                        name=tool['name'],
+                        description=tool['description'],
+                        parameters=tool['parameters'],
+                        func=create_remote_execute_func(tool['name'], mcp_client)
+                    )
+                
+                return True
+            else:
+                PrettyOutput.print(f"文件 {file_path} 类型错误: {config['type']}", OutputType.WARNING)
+                return False
+        except Exception as e:
+            PrettyOutput.print(f"文件 {file_path} 加载失败: {str(e)}", OutputType.WARNING)
+            return False
+        
     def register_tool_by_file(self, file_path: str) -> bool:
         """从指定文件加载并注册工具
 

jarvis/jarvis_utils/config.py

@@ -31,14 +31,6 @@ def get_max_input_token_count() -> int:
     """
     return int(os.getenv('JARVIS_MAX_INPUT_TOKEN_COUNT', '32000'))
 
-def get_thread_count() -> int:
-    """
-    获取用于并行处理的线程数。
-
-    返回：
-        int: 线程数，默认为1
-    """
-    return int(os.getenv('JARVIS_THREAD_COUNT', '1'))
 
 def is_auto_complete() -> bool:
     """
@@ -118,3 +110,17 @@ def get_max_tool_call_count() -> int:
         int: 最大连续工具调用次数，默认为20
     """
     return int(os.getenv('JARVIS_MAX_TOOL_CALL_COUNT', '20'))
+
+
+def get_data_dir() -> str:
+    """
+    获取Jarvis数据存储目录路径。
+    
+    返回:
+        str: 数据目录路径，优先从JARVIS_DATA_PATH环境变量获取，
+             如果未设置或为空，则使用~/.jarvis作为默认值
+    """
+    data_path = os.getenv('JARVIS_DATA_PATH', '').strip()
+    if not data_path:
+        return os.path.expanduser('~/.jarvis')
+    return data_path

jarvis/jarvis_utils/embedding.py

@@ -4,9 +4,9 @@ from typing import List
 import functools
 
 from jarvis.jarvis_utils.output import PrettyOutput, OutputType
+from jarvis.jarvis_utils.config import get_data_dir
 
 # 全局缓存，避免重复加载模型
-_global_models = {}
 _global_tokenizers = {}
 
 def get_context_token_count(text: str) -> int:
@@ -155,7 +155,7 @@ def load_tokenizer() -> AutoTokenizer:
         AutoTokenizer: 加载的分词器
     """
     model_name = "gpt2"
-    cache_dir = os.path.expanduser("~/.cache/huggingface/hub")
+    cache_dir = os.path.join(get_data_dir(), "huggingface", "hub")
 
     # 检查全局缓存
     if model_name in _global_tokenizers:

jarvis/jarvis_utils/file_processors.py

@@ -1,8 +1,3 @@
-from pathlib import Path
-import fitz  # PyMuPDF for PDF files
-from docx import Document as DocxDocument  # python-docx for DOCX files
-from pptx import Presentation
-import pandas as pd
 import unicodedata
 
 class FileProcessor:
@@ -84,260 +79,3 @@ class TextFileProcessor(FileProcessor):
 
         except Exception as e:
             raise Exception(f"Failed to read file: {str(e)}")
-
-class PDFProcessor(FileProcessor):
-    """PDF file processor"""
-    @staticmethod
-    def can_handle(file_path: str) -> bool:
-        return Path(file_path).suffix.lower() == '.pdf'
-
-    @staticmethod
-    def extract_text(file_path: str) -> str:
-        """提取PDF文件中的所有文本内容，包括页码、图片描述等"""
-        try:
-            text_parts = []
-            with fitz.open(file_path) as doc:  # type: ignore
-                # 添加文档信息
-                info = doc.metadata
-                if info:
-                    meta_text = []
-                    if info.get("title"):
-                        meta_text.append(f"标题: {info['title']}")
-                    if info.get("author"):
-                        meta_text.append(f"作者: {info['author']}")
-                    if info.get("subject"):
-                        meta_text.append(f"主题: {info['subject']}")
-                    if info.get("keywords"):
-                        meta_text.append(f"关键词: {info['keywords']}")
-                    
-                    if meta_text:
-                        text_parts.append("=== 文档信息 ===")
-                        text_parts.append("\n".join(meta_text))
-                
-                # 提取目录结构（如果有）
-                toc = doc.get_toc()  # type: ignore
-                if toc:
-                    text_parts.append("\n=== 目录结构 ===")
-                    for level, title, page in toc:
-                        indent = "  " * (level - 1)
-                        text_parts.append(f"{indent}- {title} (第{page}页)")
-                
-                # 处理各页内容
-                text_parts.append("\n=== 页面内容 ===")
-                for page_index in range(len(doc)):  # 使用范围遍历而不是直接枚举文档对象
-                    # 添加页码标记
-                    text_parts.append(f"\n--- 第{page_index+1}页 ---")
-                    
-                    # 获取页面
-                    page = doc[page_index]
-                    
-                    # 提取页面文本（包括结构信息）
-                    try:
-                        # 尝试使用结构化提取（保留段落和块结构）
-                        text = page.get_text("text")  # type: ignore
-                        text = text.strip()
-                        if text:
-                            text_parts.append(text)
-                    except Exception:
-                        # 如果结构化提取失败，回退到简单文本提取
-                        text = page.get_text()  # type: ignore
-                        if text.strip():
-                            text_parts.append(text.strip())
-                    
-                    # 提取图像信息（如果需要）
-                    # 注意：这可能会增加处理时间，可根据需要启用
-                    """
-                    image_list = page.get_images()
-                    if image_list:
-                        text_parts.append(f"本页包含 {len(image_list)} 个图像")
-                    """
-            
-            # 合并所有文本
-            return "\n".join(text_parts)
-        
-        except Exception as e:
-            # 处理可能的异常
-            return f"PDF处理错误: {str(e)}"
-
-class DocxProcessor(FileProcessor):
-    """DOCX file processor"""
-    @staticmethod
-    def can_handle(file_path: str) -> bool:
-        return Path(file_path).suffix.lower() == '.docx'
-
-    @staticmethod
-    def extract_text(file_path: str) -> str:
-        """提取 DOCX 文件中的所有文本内容，包括段落、表格、页眉页脚等"""
-        doc = DocxDocument(file_path)
-        full_text = []
-        
-        # 提取段落文本
-        for para in doc.paragraphs:
-            if para.text.strip():  # 跳过空段落
-                full_text.append(para.text)
-        
-        # 提取表格文本
-        for table in doc.tables:
-            for row in table.rows:
-                row_texts = []
-                for cell in row.cells:
-                    # 每个单元格可能包含多个段落
-                    cell_text = "\n".join([p.text for p in cell.paragraphs if p.text.strip()])
-                    if cell_text:
-                        row_texts.append(cell_text)
-                if row_texts:
-                    full_text.append(" | ".join(row_texts))
-        
-        # 提取页眉页脚（如果有节）
-        try:
-            for section in doc.sections:
-                # 提取页眉
-                if section.header:
-                    header_text = "\n".join([p.text for p in section.header.paragraphs if p.text.strip()])
-                    if header_text:
-                        full_text.append(f"页眉: {header_text}")
-                
-                # 提取页脚
-                if section.footer:
-                    footer_text = "\n".join([p.text for p in section.footer.paragraphs if p.text.strip()])
-                    if footer_text:
-                        full_text.append(f"页脚: {footer_text}")
-        except:
-            # 如果提取页眉页脚失败，忽略错误继续
-            pass
-        
-        # 合并所有文本
-        return "\n\n".join(full_text)
-
-class PPTProcessor(FileProcessor):
-    """PPT file processor"""
-    @staticmethod
-    def can_handle(file_path: str) -> bool:
-        return Path(file_path).suffix.lower() in ['.ppt', '.pptx']
-
-    @staticmethod
-    def extract_text(file_path: str) -> str:
-        """提取PPT文件中的所有文本内容，包括标题、文本框、备注等"""
-        prs = Presentation(file_path)
-        all_text = []
-        
-        # 遍历所有幻灯片
-        for slide_index, slide in enumerate(prs.slides, 1):
-            slide_text = []
-            
-            # 添加幻灯片编号
-            slide_text.append(f"=== 幻灯片 {slide_index} ===")
-            
-            # 提取幻灯片中所有形状的文本
-            for shape in slide.shapes:
-                # 提取带有文本的形状
-                try:
-                    if hasattr(shape, "text_frame") and shape.text_frame:  # type: ignore
-                        for paragraph in shape.text_frame.paragraphs:  # type: ignore
-                            text = paragraph.text.strip()
-                            if text:
-                                slide_text.append(text)
-                except AttributeError:
-                    pass
-                
-                # 提取表格内容
-                try:
-                    if hasattr(shape, "table") and shape.table:  # type: ignore
-                        for row in shape.table.rows:  # type: ignore
-                            row_texts = []
-                            for cell in row.cells:
-                                if hasattr(cell, "text_frame") and cell.text_frame:
-                                    cell_paragraphs = cell.text_frame.paragraphs  # type: ignore
-                                    cell_text = " ".join([p.text.strip() for p in cell_paragraphs if p.text.strip()])
-                                    if cell_text:
-                                        row_texts.append(cell_text)
-                            if row_texts:
-                                slide_text.append(" | ".join(row_texts))
-                except AttributeError:
-                    pass
-            
-            # 提取幻灯片备注
-            try:
-                if hasattr(slide, "has_notes_slide") and slide.has_notes_slide:
-                    notes_slide = slide.notes_slide
-                    if notes_slide and hasattr(notes_slide, "notes_text_frame") and notes_slide.notes_text_frame:
-                        notes_text = notes_slide.notes_text_frame.text.strip()  # type: ignore
-                        if notes_text:
-                            slide_text.append(f"备注: {notes_text}")
-            except AttributeError:
-                pass
-            
-            # 合并当前幻灯片的所有文本
-            if len(slide_text) > 1:  # 如果除了幻灯片编号外还有其他内容
-                all_text.append("\n".join(slide_text))
-        
-        # 返回所有幻灯片的文本内容
-        return "\n\n".join(all_text)
-
-class ExcelProcessor(FileProcessor):
-    """Excel file processor"""
-    @staticmethod
-    def can_handle(file_path: str) -> bool:
-        return Path(file_path).suffix.lower() in ['.xls', '.xlsx']
-
-    @staticmethod
-    def extract_text(file_path: str) -> str:
-        """提取 Excel 文件中的所有文本内容，包括多个工作表及格式化内容"""
-        try:
-            # 读取所有工作表
-            excel_file = pd.ExcelFile(file_path)
-            sheets_text = []
-            
-            # 处理每个工作表
-            for sheet_name in excel_file.sheet_names:
-                # 读取当前工作表
-                df = pd.read_excel(file_path, sheet_name=sheet_name)
-                
-                # 如果是空表格，跳过
-                if df.empty:
-                    continue
-                
-                # 添加工作表标题
-                sheet_text = [f"=== 工作表: {sheet_name} ==="]
-                
-                # 填充空单元格，避免NaN显示
-                df = df.fillna("")
-                
-                # 提取表格头信息
-                if not df.columns.empty:
-                    headers = [str(col) for col in df.columns]
-                    sheet_text.append("列标题: " + " | ".join(headers))
-                
-                # 尝试提取表格中可能的关键信息
-                # 1. 表格内容概述
-                row_count, col_count = df.shape
-                sheet_text.append(f"表格大小: {row_count}行 x {col_count}列")
-                
-                # 2. 表格数据，使用更友好的格式
-                try:
-                    # 转换数据框为字符串表示
-                    # 设置最大行数和列数，避免过大的表格
-                    max_rows = min(500, row_count)  # 最多显示500行
-                    if row_count > max_rows:
-                        sheet_text.append(f"注意: 表格太大，仅显示前{max_rows}行")
-                    
-                    # 将DataFrame转换为字符串表格
-                    table_str = df.head(max_rows).to_string(index=True, max_rows=max_rows, max_cols=None)
-                    sheet_text.append(table_str)
-                    
-                except Exception as e:
-                    sheet_text.append(f"表格数据提取错误: {str(e)}")
-                
-                # 合并当前工作表的文本
-                sheets_text.append("\n".join(sheet_text))
-            
-            # 如果没有提取到任何内容，返回一个提示信息
-            if not sheets_text:
-                return "Excel文件为空或无法提取内容"
-                
-            # 合并所有工作表的文本
-            return "\n\n".join(sheets_text)
-            
-        except Exception as e:
-            # 处理可能的异常，返回错误信息
-            return f"Excel文件处理错误: {str(e)}"

jarvis/jarvis_utils/input.py

@@ -158,8 +158,14 @@ def get_multiline_input(tip: str) -> str:
         'prompt': 'ansicyan',
     })
     try:
+        from prompt_toolkit.history import FileHistory
+        from jarvis.jarvis_utils.config import get_data_dir
+        import os
+        # 获取数据目录路径
+        history_dir = get_data_dir()
+        # 初始化带历史记录的会话
         session = PromptSession(
-            history=None,
+            history=FileHistory(os.path.join(history_dir, 'multiline_input_history')),
             completer=FileCompleter(),
             key_bindings=bindings,
             complete_while_typing=True,

jarvis/jarvis_utils/methodology.py

@@ -11,7 +11,7 @@ import json
 import tempfile
 from typing import Dict, Optional
 
-from jarvis.jarvis_utils.config import INPUT_WINDOW_REVERSE_SIZE, get_max_input_token_count
+from jarvis.jarvis_utils.config import INPUT_WINDOW_REVERSE_SIZE, get_max_input_token_count, get_data_dir
 from jarvis.jarvis_utils.embedding import get_context_token_count
 from jarvis.jarvis_utils.output import PrettyOutput, OutputType
 from jarvis.jarvis_platform.registry import PlatformRegistry
@@ -23,7 +23,7 @@ def _get_methodology_directory() -> str:
     返回：
         str: 方法论目录的路径
     """
-    methodology_dir = os.path.expanduser("~/.jarvis/methodologies")
+    methodology_dir = os.path.join(get_data_dir(), "methodologies")
     if not os.path.exists(methodology_dir):
         try:
             os.makedirs(methodology_dir, exist_ok=True)

jarvis/jarvis_utils/utils.py

@@ -3,22 +3,22 @@ import time
 import hashlib
 from pathlib import Path
 from typing import List, Any, Callable
-from jarvis.jarvis_utils.config import get_max_input_token_count
+from jarvis.jarvis_utils.config import get_max_input_token_count, get_data_dir
 from jarvis.jarvis_utils.embedding import get_context_token_count
 from jarvis.jarvis_utils.input import get_single_line_input
 from jarvis.jarvis_utils.output import PrettyOutput, OutputType
 def init_env() -> None:
-    """初始化环境变量从~/.jarvis/env文件
+    """初始化环境变量从jarvis_data/env文件
 
     功能：
-    1. 创建不存在的.jarvis目录
+    1. 创建不存在的jarvis_data目录
     2. 加载环境变量到os.environ
     3. 处理文件读取异常
     """
-    jarvis_dir = Path.home() / ".jarvis"
+    jarvis_dir = Path(get_data_dir())
     env_file = jarvis_dir / "env"
 
-    # 检查~/.jarvis目录是否存在
+    # 检查jarvis_data目录是否存在
     if not jarvis_dir.exists():
         jarvis_dir.mkdir(parents=True)
     if env_file.exists():