jarvis-ai-assistant 0.1.130__py3-none-any.whl → 0.1.132__py3-none-any.whl

jarvis/jarvis_tools/search_web.py

@@ -1,5 +1,5 @@
 from typing import Dict, Any, List
-
+import concurrent.futures
 from regex import W
 from yaspin import yaspin
 from jarvis.jarvis_platform.registry import PlatformRegistry
@@ -7,7 +7,7 @@ from jarvis.jarvis_tools.read_webpage import WebpageTool
 from playwright.sync_api import sync_playwright
 from urllib.parse import quote
 
-from jarvis.jarvis_utils.config import get_max_token_count
+from jarvis.jarvis_utils.config import get_max_token_count, get_thread_count
 from jarvis.jarvis_utils.embedding import get_context_token_count
 from jarvis.jarvis_utils.output import OutputType, PrettyOutput
 
@@ -104,7 +104,7 @@ class SearchTool:
             "max_results": {
                 "type": "integer",
                 "description": "最大搜索结果数量",
-                "default": 3
+                "default": 10
             }
         },
         "required": ["query", "question"]
@@ -168,22 +168,20 @@ class SearchTool:
             # Process each batch
             batch_results = []
             for i, batch in enumerate(batches, 1):
-                PrettyOutput.print(f"正在处理批次 {i}/{len(batches)}...", OutputType.PROGRESS)
-                
-                prompt = f"""Please analyze these search results to answer the question: {question}
+                prompt = f"""请根据以下搜索结果回答以下问题：{question}
 
-Search results content (Batch {i}/{len(batches)}):
+搜索结果内容 (第 {i} 批/{len(batches)}):
 {'-' * 40}
 {''.join(batch)}
 {'-' * 40}
 
-Please extract key information related to the question. Focus on:
-1. Relevant facts and details
-2. Maintaining objectivity
-3. Citing sources when appropriate
-4. Noting any uncertainties
+请提取与问题相关的关键信息。重点关注：
+1. 相关事实和细节
+2. 保持客观性
+3. 在适当的时候引用来源
+4. 注明任何不确定性
 
-Format your response as a clear summary of findings from this batch."""
+请将您的回答格式化为对本批次搜索结果的清晰总结。"""
 
                 response = self.model.chat_until_success(prompt)
                 batch_results.append(response)
@@ -196,27 +194,27 @@ Format your response as a clear summary of findings from this batch."""
             batch_findings = '\n\n'.join(f'Batch {i+1}:\n{result}' for i, result in enumerate(batch_results))
             separator = '-' * 40
             
-            synthesis_prompt = f"""Please provide a comprehensive answer to the original question by synthesizing the findings from multiple batches of search results.
+            synthesis_prompt = f"""请通过综合多个批次的搜索结果，为原始问题提供一个全面的回答。
 
-Original Question: {question}
+原始问题: {question}
 
-Findings from each batch:
+各批次的发现:
 {separator}
 {batch_findings}
 {separator}
 
-Please synthesize a final answer that:
-1. Combines key insights from all batches
-2. Resolves any contradictions between sources
-3. Maintains clear source attribution
-4. Acknowledges any remaining uncertainties
-5. Provides a coherent and complete response to the original question"""
+请综合出一个最终答案，要求：
+1. 整合所有批次的关键见解
+2. 解决不同来源之间的矛盾
+3. 保持清晰的来源归属
+4. 承认任何剩余的不确定性
+5. 提供对原始问题的连贯完整的回答"""
 
             final_response = self.model.chat_until_success(synthesis_prompt)
             return final_response
 
         except Exception as e:
-            return f"Information extraction failed: {str(e)}"
+            return f"信息提取失败：{str(e)}"
 
     def execute(self, args: Dict) -> Dict[str, Any]:
         """Execute search and extract information"""
@@ -238,18 +236,60 @@ Please synthesize a final answer that:
                     "stderr": "No search results found"
                 }
             
+            # Read webpages in parallel using ThreadPoolExecutor
             contents = []
-            for i, result in enumerate(results, 1):
+            
+            # Print starting message
+            PrettyOutput.print(f"开始并行读取 {len(results)} 个网页结果...", OutputType.INFO)
+            
+            def fetch_webpage(result_info):
+                """Function to fetch a single webpage in a separate thread"""
+                idx, result = result_info
                 try:
-                    PrettyOutput.print(f"正在读取结果 {i}/{len(results)}... {result['title']} - {result['href']}", OutputType.PROGRESS)
+                    # Removed progress print here to avoid mixed output
                     webpage_result = self.webpage_tool.execute({"url": result["href"]})
                     if webpage_result["success"]:
-                        contents.append(f"\nSource {i}: {result['href']}\n")
-                        contents.append(webpage_result["stdout"])
+                        return idx, result, webpage_result["stdout"], True
+                    return idx, result, None, False
                 except Exception as e:
-                    PrettyOutput.print(f"读取结果失败 {i}: {str(e)}", OutputType.WARNING)
-                    continue
+                    return idx, result, str(e), False
+            
+            # Use ThreadPoolExecutor for parallel processing
+            processed_results = []
+            with yaspin(text="正在并行读取网页内容...", color="cyan") as spinner:
+                with concurrent.futures.ThreadPoolExecutor(max_workers=get_thread_count()) as executor:
+                    # Submit all webpage fetch tasks
+                    future_to_result = {
+                        executor.submit(fetch_webpage, (i, result)): i 
+                        for i, result in enumerate(results)
+                    }
+                    
+                    # Collect results as they complete
+                    for future in concurrent.futures.as_completed(future_to_result):
+                        processed_results.append(future.result())
+                        # Update spinner with current progress
+                        spinner.text = f"正在并行读取网页内容... ({len(processed_results)}/{len(results)})"
+                
+                spinner.text = "网页内容读取完成"
+                spinner.ok("✅")
             
+            # Sort results by original index to maintain ordering
+            processed_results.sort(key=lambda x: x[0])
+            
+            # Print results in order and add to contents
+            PrettyOutput.section("搜索结果概览", OutputType.INFO)
+            
+            output = []
+            for idx, result, content, success in processed_results:
+                if success:
+                    output.append(f"✅ 读取结果 {idx+1}/{len(results)} 完成: {result['title']} - {result['href']}")
+                    contents.append(f"\nSource {idx+1}: {result['href']}\n")
+                    contents.append(content)
+                else:
+                    output.append(f"❌ 读取结果 {idx+1}/{len(results)} 失败: {result['title']} - {result['href']}")
+            
+            PrettyOutput.print("\n".join(output), OutputType.INFO)
+
             if not contents:
                 return {
                     "success": False,
@@ -262,10 +302,13 @@ Please synthesize a final answer that:
                 analysis = self._extract_info(contents, question)
                 spinner.text = "信息提取完成"
                 spinner.ok("✅")
+
+            output = f"分析结果:\n\n{analysis}"
+            PrettyOutput.print(output, OutputType.SUCCESS)
             
             return {
                 "success": True,
-                "stdout": f"Search analysis results:\n\n{analysis}",
+                "stdout": output,
                 "stderr": ""
             }
 

jarvis/jarvis_tools/tool_generator.py

@@ -7,6 +7,7 @@ from typing import Dict, Any
 
 from yaspin import yaspin
 from jarvis.jarvis_platform.registry import PlatformRegistry
+from jarvis.jarvis_utils.utils import ct, ot
 
 class ToolGenerator:
     """工具生成器类，用于自动创建与Jarvis系统集成的新工具"""
@@ -84,7 +85,7 @@ class ToolGenerator:
                 tools_dir.mkdir(parents=True, exist_ok=True)
                 tool_file = tools_dir / f"{tool_name}.py"
                 
-                with open(tool_file, "w") as f:
+                with open(tool_file, "w", errors="ignore") as f:
                     f.write(implementation)
                 spinner.text = "工具保存完成"
                 spinner.ok("✅")
@@ -112,8 +113,7 @@ class ToolGenerator:
         Returns:
             格式化后的提示字符串
         """
-        example_code = '''
-<TOOL>
+        example_code = ot("TOOL")+'''
 from typing import Dict, Any
 from jarvis.utils import OutputType, PrettyOutput
 from jarvis.jarvis_platform.registry import PlatformRegistry
@@ -163,8 +163,7 @@ class CustomTool:
                 "stdout": "",
                 "stderr": str(e)
             }
-</TOOL>
-'''
+''' + ct("TOOL")
 
         return f'''创建一个与Jarvis系统集成的Python工具类。请遵循以下要求：
 1. 类名: {tool_name.capitalize()}Tool
@@ -183,9 +182,9 @@ class CustomTool:
 8. 仅返回Python实现代码
 9. 代码应该是完整且可直接使用的
 10. 按照以下格式输出代码：
-<TOOL>
+{ot("TOOL")}
 {example_code}
-</TOOL>
+{ct("TOOL")}
 
 示例：
 {example_code}
@@ -199,8 +198,7 @@ class CustomTool:
         Returns:
             提取到的Python代码字符串
         """
-        # 查找第一个<TOOL>和</TOOL>标签之间的内容
-        sm = re.search(r'<TOOL>(.*?)</TOOL>', response, re.DOTALL)
+        sm = re.search(ot("TOOL")+r'(.*?)'+ct("TOOL"), response, re.DOTALL)
         if sm:
             return sm.group(1)
         return ""

jarvis/jarvis_utils/__init__.py

@@ -22,3 +22,4 @@ colorama.init()
 os.environ["TOKENIZERS_PARALLELISM"] = "false"
 # 安装rich traceback处理器以获得更好的错误信息
 install_rich_traceback()
+

jarvis/jarvis_utils/config.py

@@ -11,10 +11,10 @@ import os
 """
 def get_max_token_count() -> int:
     """
-    获取API请求的最大token数量。
+    获取模型允许的最大token数量。
     
-    返回：
-        int: 最大token数量，默认为131072（128k）
+    返回:
+        int: 模型能处理的最大token数量。
     """
     return int(os.getenv('JARVIS_MAX_TOKEN_COUNT', '131072'))  # 默认128k
     
@@ -172,3 +172,67 @@ def is_confirm_before_apply_patch() -> bool:
         bool: 如果需要确认则返回True，默认为False
     """
     return os.getenv('JARVIS_CONFIRM_BEFORE_APPLY_PATCH', 'false') == 'true'
+
+def get_rag_ignored_paths() -> list:
+    """
+    获取RAG索引时需要忽略的路径列表。
+    
+    首先尝试从.jarvis/rag_ignore.txt文件中读取，
+    如果该文件不存在，则返回默认忽略列表。
+    
+    返回：
+        list: 忽略路径的列表，默认包含常见忽略路径
+    """
+    # 默认忽略路径
+    default_ignored = [
+        '.git', 
+        '__pycache__', 
+        'node_modules', 
+        '.jarvis',
+        '.jarvis-*',
+        'target',
+        'venv',
+        'env',
+        '.env',
+        '.venv',
+        '.idea',
+        '.vscode',
+        'dist',
+        'build',
+        '*.pyc',
+        '*.pyo',
+        '*.so',
+        '*.o',
+        '*.a',
+        '*.pyd',
+        '*.dll',
+        '*.exe',
+        '*.bin',
+        '*.obj',
+        '*.out',
+        '*.jpg',
+        '*.jpeg',
+        '*.png',
+        '*.gif',
+        '*.tiff',
+        '*.pdf',
+        '*.zip',
+        '*.tar',
+        '*.tar.gz',
+        '*.gz',
+        '*.bz2',
+        '*.xz',
+        '*.rar'
+    ]
+    
+    # 尝试从配置文件中读取
+    try:
+        config_path = os.path.join('.jarvis', 'rag_ignore.txt')
+        if os.path.exists(config_path):
+            with open(config_path, 'r') as f:
+                custom_ignored = [line.strip() for line in f if line.strip() and not line.startswith('#')]
+                return custom_ignored
+    except Exception:
+        pass
+        
+    return default_ignored