jarvis-ai-assistant 0.1.159__py3-none-any.whl → 0.1.161__py3-none-any.whl

jarvis/jarvis_git_utils/git_commiter.py

@@ -10,11 +10,9 @@ import sys
 import argparse
 import os
 
-from jarvis.jarvis_utils.config import INPUT_WINDOW_REVERSE_SIZE, get_max_input_token_count
-from jarvis.jarvis_utils.embedding import get_context_token_count
 from jarvis.jarvis_utils.git_utils import find_git_root, has_uncommitted_changes
 from jarvis.jarvis_utils.output import OutputType, PrettyOutput
-from jarvis.jarvis_utils.utils import init_env, chat_with_files
+from jarvis.jarvis_utils.utils import init_env, is_context_overflow
 from jarvis.jarvis_utils.tag import ot, ct
 
 
@@ -107,7 +105,7 @@ class GitCommitTool:
                     diff = process.communicate()[0].decode()
                     spinner.write(f"✅ 获取差异 ({file_count} 个文件)")
                     try:
-                        temp_diff_file = None
+                        temp_diff_file_path = None
                         # 生成提交信息
                         spinner.text = "正在生成提交消息..."
                         
@@ -132,13 +130,12 @@ class GitCommitTool:
 
                         # 获取模型并尝试上传文件
                         platform = PlatformRegistry().get_normal_platform()
+                        upload_success = False
                         
-                        # 检查文件大小
-                        is_large_file = get_context_token_count(diff) > get_max_input_token_count() - INPUT_WINDOW_REVERSE_SIZE
+                        # Check if content is too large
+                        is_large_content = is_context_overflow(diff)
                         
-                        if is_large_file:
-                            # 文件较大，尝试上传
-                            upload_success = False
+                        if is_large_content and hasattr(platform, 'upload_files'):
                             spinner.text = "正在上传代码差异文件..."
                             try:
                                 with spinner.hidden():
@@ -148,75 +145,49 @@ class GitCommitTool:
                                         temp_diff_file.write(diff)
                                         temp_diff_file.flush()
                                         spinner.write(f"✅ 差异内容已写入临时文件")
-                                    
-                                    if hasattr(platform, 'upload_files'):
-                                        upload_success = platform.upload_files([temp_diff_file_path])
-                                        if upload_success:
-                                            spinner.write("✅ 成功上传代码差异文件")
-                                        else:
-                                            spinner.write("⚠️ 上传代码差异文件失败，将使用chat_with_files处理")
-                                    else:
-                                        spinner.write("⚠️ 模型不支持文件上传，将使用chat_with_files处理")
+                                    upload_success = platform.upload_files([temp_diff_file_path])
+                                if upload_success:
+                                    spinner.write("✅ 成功上传代码差异文件")
+                                else:
+                                    spinner.write("⚠️ 上传代码差异文件失败，将使用分块处理")
                             except Exception as e:
                                 spinner.write(f"⚠️ 上传文件时出错: {str(e)}")
                                 upload_success = False
-                            
-                            if upload_success:
-                                # 使用上传的文件
-                                prompt = base_prompt + f'''
+                        
+                        # 根据上传状态准备完整的提示
+                        if upload_success:
+                            # 使用上传的文件
+                            prompt = base_prompt + f'''
         # 变更概述
         - 变更文件数量: {file_count} 个文件
         - 已上传包含完整代码差异的文件
         
         请详细分析已上传的代码差异文件，生成符合上述格式的提交信息。
         '''
-                            else:
-                                # 使用chat_with_files处理大文件
-                                spinner.write("使用chat_with_files处理大文件...")
-                                commit_message = chat_with_files(
-                                    [temp_diff_file_path],
-                                    platform,
-                                    base_prompt,
-                                    "请基于以上代码差异生成符合格式的提交信息。"
-                                )
-                                commit_message = self._extract_commit_message(commit_message)
-                                if commit_message:
-                                    spinner.write("✅ 生成提交消息")
-                                    # 执行提交
-                                    spinner.text = "正在准备提交..."
-                                    with tempfile.NamedTemporaryFile(mode='w', delete=True) as tmp_file:
-                                        tmp_file.write(commit_message)
-                                        tmp_file.flush()
-                                        spinner.text = "正在执行提交..."
-                                        commit_cmd = ["git", "commit", "-F", tmp_file.name]
-                                        subprocess.Popen(
-                                            commit_cmd,
-                                            stdout=subprocess.DEVNULL,
-                                            stderr=subprocess.DEVNULL
-                                        ).wait()
-                                        spinner.write("✅ 提交")
-                                    return {
-                                        "success": True,
-                                        "stdout": yaml.safe_dump({
-                                            "commit_hash": self._get_last_commit_hash(),
-                                            "commit_message": commit_message
-                                        }),
-                                        "stderr": ""
-                                    }
+                            commit_message = platform.chat_until_success(prompt)
                         else:
-                            # 文件较小，直接在提示中包含差异内容
-                            prompt = base_prompt + f'''
+                            # 如果上传失败但内容较大，使用chat_big_content
+                            if is_large_content and hasattr(platform, 'chat_big_content'):
+                                spinner.text = "正在使用分块处理生成提交信息..."
+                                commit_message = platform.chat_big_content(diff, base_prompt)
+                            else:
+                                # 直接在提示中包含差异内容
+                                prompt = base_prompt + f'''
         # 分析材料
         {diff}
         '''
+                                commit_message = platform.chat_until_success(prompt)
                         
                         # 尝试生成提交信息
                         spinner.text = "正在生成提交消息..."
                         while True:
-                            commit_message = platform.chat_until_success(prompt)
-                            commit_message = self._extract_commit_message(commit_message)
+                            # 只在特定情况下重新获取commit_message
+                            if not upload_success and not is_large_content and not commit_message:
+                                commit_message = platform.chat_until_success(prompt)
+                            extracted_message = self._extract_commit_message(commit_message)
                             # 如果成功提取，就跳出循环
-                            if commit_message:
+                            if extracted_message:
+                                commit_message = extracted_message
                                 break
                             prompt = f"""格式错误，请按照以下格式重新生成提交信息：
                             {ot("COMMIT_MESSAGE")}
@@ -225,6 +196,7 @@ class GitCommitTool:
         [可选] 详细描述变更内容和原因
         {ct("COMMIT_MESSAGE")}
                             """
+                            commit_message = platform.chat_until_success(prompt)
                         spinner.write("✅ 生成提交消息")
 
                         # 执行提交
@@ -246,7 +218,7 @@ class GitCommitTool:
                         spinner.ok("✅")
                     finally:
                         # 清理临时差异文件
-                        if temp_diff_file is not None and os.path.exists(temp_diff_file_path):
+                        if temp_diff_file_path is not None and os.path.exists(temp_diff_file_path):
                             try:
                                 os.unlink(temp_diff_file_path)
                             except Exception as e:

jarvis/jarvis_methodology/main.py

@@ -11,14 +11,14 @@ import hashlib
 import os
 import json
 import argparse
-import yaml
+import yaml # type: ignore
 from jarvis.jarvis_utils.methodology import (
     _get_methodology_directory,
     _load_all_methodologies
 )
 from jarvis.jarvis_platform.registry import PlatformRegistry
 from jarvis.jarvis_utils.output import PrettyOutput, OutputType
-from yaspin import yaspin
+from yaspin import yaspin # type: ignore
 
 def import_methodology(input_file):
     """导入方法论文件（合并策略）"""

jarvis/jarvis_platform/base.py

@@ -1,6 +1,8 @@
 from abc import ABC, abstractmethod
 import re
 from typing import List, Tuple
+from jarvis.jarvis_utils.config import get_max_input_token_count
+from jarvis.jarvis_utils.embedding import split_text_into_chunks
 from jarvis.jarvis_utils.globals import clear_read_file_record
 from jarvis.jarvis_utils.output import OutputType, PrettyOutput
 from jarvis.jarvis_utils.utils import get_context_token_count, while_success, while_true
@@ -37,38 +39,49 @@ class BasePlatform(ABC):
     @abstractmethod
     def upload_files(self, file_list: List[str]) -> bool:
         raise NotImplementedError("upload_files is not implemented")
+    
+    def chat_big_content(self, content: str, prompt: str) -> str:
+        prefix_prompt = f"""
+        我将分多次提供大量的上下文内容，在我明确告诉你内容已经全部提供完毕之前，每次仅需要输出“已收到”。
+        """
+        self.chat_until_success(prefix_prompt)
+        split_content = split_text_into_chunks(content, get_max_input_token_count() - 1024)
+        for chunk in split_content:
+            self.chat_until_success(f"<part_content>{chunk}</part_content>")
+        return self.chat_until_success(f"内容已经全部提供完毕\n\n{prompt}")
 
+    
+    def _chat(self, message: str):
+        import time
+        start_time = time.time()
+        response = self.chat(message)
+
+        end_time = time.time()
+        duration = end_time - start_time
+        char_count = len(response)
+
+        # Calculate token count and tokens per second
+        try:
+            token_count = get_context_token_count(response)
+            tokens_per_second = token_count / duration if duration > 0 else 0
+        except Exception as e:
+            PrettyOutput.print(f"Tokenization failed: {str(e)}", OutputType.WARNING)
+            token_count = 0
+            tokens_per_second = 0
+
+        # Print statistics
+        if not self.suppress_output:
+            PrettyOutput.print(
+                f"对话完成 - 耗时: {duration:.2f}秒, 输出字符数: {char_count}, 输出Token数量: {token_count}, 每秒Token数量: {tokens_per_second:.2f}",
+                OutputType.INFO,
+            )
+
+        # Keep original think tag handling
+        response = re.sub(ot("think")+r'.*?'+ct("think"), '', response, flags=re.DOTALL)
+        return response
+    
     def chat_until_success(self, message: str) -> str:
-        def _chat():
-            import time
-            start_time = time.time()
-            response = self.chat(message)
-
-            end_time = time.time()
-            duration = end_time - start_time
-            char_count = len(response)
-
-            # Calculate token count and tokens per second
-            try:
-                token_count = get_context_token_count(response)
-                tokens_per_second = token_count / duration if duration > 0 else 0
-            except Exception as e:
-                PrettyOutput.print(f"Tokenization failed: {str(e)}", OutputType.WARNING)
-                token_count = 0
-                tokens_per_second = 0
-
-            # Print statistics
-            if not self.suppress_output:
-                PrettyOutput.print(
-                    f"对话完成 - 耗时: {duration:.2f}秒, 输出字符数: {char_count}, 输出Token数量: {token_count}, 每秒Token数量: {tokens_per_second:.2f}",
-                    OutputType.INFO,
-                )
-
-            # Keep original think tag handling
-            response = re.sub(ot("think")+r'.*?'+ct("think"), '', response, flags=re.DOTALL)
-            return response
-
-        return while_true(lambda: while_success(lambda: _chat(), 5), 5)
+        return while_true(lambda: while_success(lambda: self._chat(message), 5), 5)
 
     @abstractmethod
     def name(self) -> str:

jarvis/jarvis_platform/kimi.py

@@ -1,5 +1,5 @@
 from typing import Dict, List, Tuple
-import requests
+import requests # type: ignore
 import json
 import os
 import mimetypes
@@ -178,7 +178,7 @@ class KimiModel(BasePlatform):
         if not file_list:
             return True
 
-        from yaspin import yaspin
+        from yaspin import yaspin # type: ignore
         
         if not self.chat_id:
             with yaspin(text="创建聊天会话...", color="yellow") as spinner:

jarvis/jarvis_tools/file_analyzer.py

@@ -3,8 +3,8 @@ import os
 
 from jarvis.jarvis_platform.registry import PlatformRegistry
 from jarvis.jarvis_utils.output import OutputType, PrettyOutput
-from yaspin import yaspin
-from yaspin.spinners import Spinners
+from yaspin import yaspin # type: ignore
+from yaspin.spinners import Spinners # type: ignore
 
 class FileAnalyzerTool:
     name = "file_analyzer"

jarvis/jarvis_tools/file_operation.py

@@ -2,7 +2,7 @@ from typing import Dict, Any
 import os
 from pathlib import Path
 
-from yaspin import yaspin
+from yaspin import yaspin # type: ignore
 
 from jarvis.jarvis_utils.globals import add_read_file_record
 from jarvis.jarvis_utils.output import OutputType, PrettyOutput

jarvis/jarvis_tools/read_webpage.py

@@ -31,7 +31,7 @@ class WebpageTool:
             
             # Create Yuanbao model instance
             model = PlatformRegistry().get_normal_platform()
-            model.web = True
+            model.set_web(True)
             model.set_suppress_output(False)  # type: ignore
 
             # Construct prompt based on want parameter

jarvis/jarvis_tools/registry.py

@@ -3,6 +3,7 @@ from pathlib import Path
 import re
 import sys
 import tempfile
+import os
 from typing import Any, Callable, Dict, List, Optional, Tuple
 
 import yaml
@@ -10,14 +11,9 @@ 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,
-    get_data_dir,
-)
-from jarvis.jarvis_utils.embedding import get_context_token_count
+from jarvis.jarvis_utils.config import get_data_dir
 from jarvis.jarvis_utils.output import OutputType, PrettyOutput
-from jarvis.jarvis_utils.utils import init_env
+from jarvis.jarvis_utils.utils import init_env, is_context_overflow
 from jarvis.jarvis_utils.tag import ot, ct
 from jarvis.jarvis_mcp.stdio_mcp_client import StdioMcpClient
 from jarvis.jarvis_mcp.sse_mcp_client import SSEMcpClient
@@ -175,9 +171,6 @@ class ToolRegistry(OutputHandler):
         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:
         """使用指定工具
@@ -618,46 +611,45 @@ class ToolRegistry(OutputHandler):
                 result["stdout"], result.get("stderr", "")
             )
 
-            # 处理结果
-            if get_context_token_count(output) > self.max_input_token_count:
-                with tempfile.NamedTemporaryFile(
-                    mode="w", suffix=".txt", delete=False
-                ) as tmp_file:
+            # 检查内容是否过大
+            is_large_content = is_context_overflow(output)
+            
+            if is_large_content:
+                # 创建临时文件
+                with tempfile.NamedTemporaryFile(mode="w", suffix=".txt", delete=False) as tmp_file:
                     output_file = tmp_file.name
                     tmp_file.write(output)
                     tmp_file.flush()
-                platform: Any = PlatformRegistry().get_normal_platform()
-                if platform:
-                    platform.set_suppress_output(False)
-                    try:
+                
+                try:
+                    # 获取平台实例
+                    platform = PlatformRegistry().get_normal_platform()
+                    if platform and hasattr(platform, 'upload_files'):
+                        platform.set_suppress_output(False)
                         # 尝试上传文件
-                        if platform.upload_files([output_file]):
-                            prompt = f"""该文件为工具执行结果，请阅读文件内容，并根据文件提取出以下信息：{want}
-
-注意：
-1. 请仔细阅读文件内容，确保提取的信息准确完整
-2. 如果文件内容包含错误信息，请特别关注并提取
-3. 提取的信息应该直接回答用户的问题，不要包含无关内容"""
+                        upload_success = platform.upload_files([output_file])
+                        
+                        if upload_success:
+                            # 使用上传的文件生成摘要
+                            prompt = f"该文件为工具执行结果，请阅读文件内容，并根据文件提取出以下信息：{want}"
                             return f"""工具调用原始输出过长，以下是根据输出提出的信息：
 
 {platform.chat_until_success(prompt)}"""
-                        else:
-                            # 上传失败，使用chat_with_files
-                            PrettyOutput.print("⚠️ 上传文件失败，将使用chat_with_files处理", OutputType.WARNING)
-                            from jarvis.jarvis_utils.utils import chat_with_files
-                            prompt = f"""请阅读以下文件内容，并根据文件提取出以下信息：{want}
-
-注意：
-1. 请仔细阅读文件内容，确保提取的信息准确完整
-2. 如果文件内容包含错误信息，请特别关注并提取
-3. 提取的信息应该直接回答用户的问题，不要包含无关内容"""
+                        elif hasattr(platform, 'chat_big_content'):
+                            # 如果上传失败但支持大内容处理，使用chat_big_content
+                            prompt = f"以下内容为工具执行结果，请阅读内容，并根据内容提取出以下信息：{want}\n\n{output}"
                             return f"""工具调用原始输出过长，以下是根据输出提出的信息：
 
-{chat_with_files([output_file], platform, prompt, "")}"""
-                    except Exception as e:
-                        PrettyOutput.print(f"处理大文件时出错: {str(e)}", OutputType.ERROR)
-                        return self._truncate_output(output)
-                return self._truncate_output(output)
+{platform.chat_big_content(output, prompt)}"""
+                    
+                    # 如果都不支持，返回截断的输出
+                    return self._truncate_output(output)
+                finally:
+                    # 清理临时文件
+                    try:
+                        os.unlink(output_file)
+                    except Exception:
+                        pass
 
             return output
 

jarvis/jarvis_tools/search_web.py

@@ -16,7 +16,7 @@ class SearchWebTool:
     def execute(self, args: Dict[str, Any]) -> Dict[str, Any]: # type: ignore
         query = args.get("query")
         model = PlatformRegistry().get_normal_platform()
-        model.web = True
+        model.set_web(True)
         model.set_suppress_output(False) # type: ignore
         return {
             "stdout": model.chat_until_success(query), # type: ignore

jarvis/jarvis_utils/input.py

@@ -82,29 +82,20 @@ class FileCompleter(Completer):
         # 计算替换长度
         replace_length = len(text_after_at) + 1
 
-        # 从replace_map获取建议列表
-        default_suggestions = [
-            (ot(tag), desc) for tag, desc in [
-                (tag, self._get_description(tag)) 
-                for tag in self.replace_map.keys()
-            ]
-        ]
-        # 添加特殊标记
-        default_suggestions.extend([
+        # 获取所有可能的补全项
+        all_completions = []
+        
+        # 1. 添加特殊标记
+        all_completions.extend([
+            (ot(tag), self._get_description(tag))
+            for tag in self.replace_map.keys()
+        ])
+        all_completions.extend([
             (ot("Summary"), '总结'),
             (ot("Clear"), '清除历史'),
         ])
-        for name, desc in default_suggestions:
-            yield Completion(
-                text=f"'{name}'",
-                start_position=-replace_length,
-                display=name,
-                display_meta=desc
-            ) # type: ignore
-
-
-        # 使用git ls-files获取所有可能的文件
-        all_files = []
+        
+        # 2. 添加文件列表
         try:
             import subprocess
             result = subprocess.run(['git', 'ls-files'],
@@ -112,28 +103,40 @@ class FileCompleter(Completer):
                                  stderr=subprocess.PIPE,
                                  text=True)
             if result.returncode == 0:
-                all_files = [line.strip() for line in result.stdout.splitlines() if line.strip()]
+                all_completions.extend([
+                    (path, "File") 
+                    for path in result.stdout.splitlines() 
+                    if path.strip()
+                ])
         except Exception:
             pass
-        # 生成补全建议
-        if not file_path:
-            scored_files = [(path, 100) for path in all_files[:self.max_suggestions]]
-        else:
-            scored_files_data = process.extract(file_path, all_files, limit=self.max_suggestions)
-            scored_files = [(m[0], m[1]) for m in scored_files_data]
-            scored_files.sort(key=lambda x: x[1], reverse=True)
-            scored_files = scored_files[:self.max_suggestions]
-        # 生成补全项
-        for path, score in scored_files:
-            if not file_path or score > self.min_score:
-                display_text = path
-                if file_path and score < 100:
-                    display_text = f"{path} ({score}%)"
+            
+        # 统一过滤和排序
+        if file_path:
+            # 使用模糊匹配过滤
+            scored_items = process.extract(file_path, [item[0] for item in all_completions], limit=self.max_suggestions)
+            scored_items = [(item[0], item[1]) for item in scored_items if item[1] > self.min_score]
+            # 创建映射以便查找描述
+            completion_map = {item[0]: item[1] for item in all_completions}
+            # 生成补全项
+            for text, score in scored_items:
+                display_text = text
+                if score < 100:
+                    display_text = f"{text} ({score}%)"
                 yield Completion(
-                    text=f"'{path}'",
+                    text=f"'{text}'",
                     start_position=-replace_length,
                     display=display_text,
-                    display_meta="File"
+                    display_meta=completion_map.get(text, "")
+                ) # type: ignore
+        else:
+            # 没有输入时返回前max_suggestions个建议
+            for text, desc in all_completions[:self.max_suggestions]:
+                yield Completion(
+                    text=f"'{text}'",
+                    start_position=-replace_length,
+                    display=text,
+                    display_meta=desc
                 ) # type: ignore
     
     def _get_description(self, tag: str) -> str: