beswarm 0.1.57__py3-none-any.whl → 0.1.59__py3-none-any.whl

beswarm/aient/setup.py

@@ -4,7 +4,7 @@ from setuptools import setup, find_packages
 
 setup(
     name="aient",
-    version="1.1.12",
+    version="1.1.14",
     description="Aient: The Awakening of Agent.",
     long_description=Path.open(Path("README.md"), encoding="utf-8").read(),
     long_description_content_type="text/markdown",

beswarm/aient/src/aient/core/request.py

@@ -48,6 +48,7 @@ async def get_gemini_payload(request, engine, provider, api_key=None):
 
     messages = []
     systemInstruction = None
+    system_prompt = ""
     function_arguments = None
     for msg in request.messages:
         if msg.role == "assistant":
@@ -102,7 +103,8 @@ async def get_gemini_payload(request, engine, provider, api_key=None):
             messages.append({"role": msg.role, "parts": content})
         elif msg.role == "system":
             content[0]["text"] = re.sub(r"_+", "_", content[0]["text"])
-            systemInstruction = {"parts": content}
+            system_prompt = system_prompt + "\n\n" + content[0]["text"]
+    systemInstruction = {"parts": [{"text": system_prompt}]}
 
     if any(off_model in original_model for off_model in gemini_max_token_65k_models):
         safety_settings = "OFF"
@@ -212,23 +214,35 @@ async def get_gemini_payload(request, engine, provider, api_key=None):
         else:
             payload["generationConfig"]["maxOutputTokens"] = 8192
 
-    # 从请求模型名中检测思考预算设置
-    m = re.match(r".*-think-(-?\d+)", request.model)
-    if m:
-        try:
-            val = int(m.group(1))
-            if val < 0:
-                val = 0
-            elif val > 24576:
-                val = 24576
-            payload["generationConfig"]["thinkingConfig"] = {"thinkingBudget": val}
-        except ValueError:
-            # 如果转换为整数失败，忽略思考预算设置
-            pass
-
-    # 检测search标签
-    if request.model.endswith("-search"):
-        payload["tools"] = [{"googleSearch": {}}]
+    if "gemini-2.5" in original_model:
+        payload["generationConfig"]["thinkingConfig"] = {
+            "includeThoughts": True,
+        }
+        # 从请求模型名中检测思考预算设置
+        m = re.match(r".*-think-(-?\d+)", request.model)
+        if m:
+            try:
+                val = int(m.group(1))
+                if val < 0:
+                    val = 0
+                elif val > 24576:
+                    val = 24576
+                payload["generationConfig"]["thinkingConfig"]["thinkingBudget"] = val
+            except ValueError:
+                # 如果转换为整数失败，忽略思考预算设置
+                pass
+
+    # # 检测search标签
+    # if request.model.endswith("-search"):
+    #     payload["tools"] = [{"googleSearch": {}}]
+
+    if safe_get(provider, "preferences", "post_body_parameter_overrides", default=None):
+        for key, value in safe_get(provider, "preferences", "post_body_parameter_overrides", default={}).items():
+            if key == request.model:
+                for k, v in value.items():
+                    payload[k] = v
+            elif all(_model not in request.model.lower() for _model in ["gemini", "gpt", "claude"]):
+                payload[key] = value
 
     return url, headers, payload
 
@@ -303,16 +317,16 @@ async def get_vertex_gemini_payload(request, engine, provider, api_key=None):
         gemini_stream = "generateContent"
     model_dict = get_model_dict(provider)
     original_model = model_dict[request.model]
-    search_tool = None
+    # search_tool = None
 
     # https://cloud.google.com/vertex-ai/generative-ai/docs/models/gemini/2-0-flash?hl=zh-cn
     pro_models = ["gemini-2.5", "gemini-2.0"]
     if any(pro_model in original_model for pro_model in pro_models):
         location = gemini2
-        search_tool = {"googleSearch": {}}
+        # search_tool = {"googleSearch": {}}
     else:
         location = gemini1
-        search_tool = {"googleSearchRetrieval": {}}
+        # search_tool = {"googleSearchRetrieval": {}}
 
     if "google-vertex-ai" in provider.get("base_url", ""):
         url = provider.get("base_url").rstrip('/') + "/v1/projects/{PROJECT_ID}/locations/{LOCATION}/publishers/google/models/{MODEL_ID}:{stream}".format(
@@ -334,6 +348,7 @@ async def get_vertex_gemini_payload(request, engine, provider, api_key=None):
 
     messages = []
     systemInstruction = None
+    system_prompt = ""
     function_arguments = None
     for msg in request.messages:
         if msg.role == "assistant":
@@ -387,7 +402,8 @@ async def get_vertex_gemini_payload(request, engine, provider, api_key=None):
         elif msg.role != "system":
             messages.append({"role": msg.role, "parts": content})
         elif msg.role == "system":
-            systemInstruction = {"parts": content}
+            system_prompt = system_prompt + "\n\n" + content[0]["text"]
+    systemInstruction = {"parts": [{"text": system_prompt}]}
 
     if any(off_model in original_model for off_model in gemini_max_token_65k_models):
         safety_settings = "OFF"
@@ -469,8 +485,34 @@ async def get_vertex_gemini_payload(request, engine, provider, api_key=None):
         else:
             payload["generationConfig"]["max_output_tokens"] = 8192
 
-    if request.model.endswith("-search"):
-        payload["tools"] = [search_tool]
+    if "gemini-2.5" in original_model:
+        payload["generationConfig"]["thinkingConfig"] = {
+            "includeThoughts": True,
+        }
+        # 从请求模型名中检测思考预算设置
+        m = re.match(r".*-think-(-?\d+)", request.model)
+        if m:
+            try:
+                val = int(m.group(1))
+                if val < 0:
+                    val = 0
+                elif val > 24576:
+                    val = 24576
+                payload["generationConfig"]["thinkingConfig"]["thinkingBudget"] = val
+            except ValueError:
+                # 如果转换为整数失败，忽略思考预算设置
+                pass
+
+    # if request.model.endswith("-search"):
+    #     payload["tools"] = [search_tool]
+
+    if safe_get(provider, "preferences", "post_body_parameter_overrides", default=None):
+        for key, value in safe_get(provider, "preferences", "post_body_parameter_overrides", default={}).items():
+            if key == request.model:
+                for k, v in value.items():
+                    payload[k] = v
+            elif all(_model not in request.model.lower() for _model in ["gemini", "gpt", "claude"]):
+                payload[key] = value
 
     return url, headers, payload
 
@@ -893,6 +935,9 @@ async def get_gpt_payload(request, engine, provider, api_key=None):
         headers['Authorization'] = f"Bearer {api_key}"
 
     url = provider['base_url']
+    if "openrouter.ai" in url:
+        headers['HTTP-Referer'] = "https://github.com/yym68686/uni-api"
+        headers['X-Title'] = "Uni API"
 
     messages = []
     for msg in request.messages:
@@ -1010,7 +1055,11 @@ async def get_gpt_payload(request, engine, provider, api_key=None):
 
     if safe_get(provider, "preferences", "post_body_parameter_overrides", default=None):
         for key, value in safe_get(provider, "preferences", "post_body_parameter_overrides", default={}).items():
-            payload[key] = value
+            if key == request.model:
+                for k, v in value.items():
+                    payload[k] = v
+            elif all(_model not in request.model.lower() for _model in ["gemini", "gpt", "claude"]):
+                payload[key] = value
 
     return url, headers, payload
 
@@ -1104,7 +1153,11 @@ async def get_azure_payload(request, engine, provider, api_key=None):
 
     if safe_get(provider, "preferences", "post_body_parameter_overrides", default=None):
         for key, value in safe_get(provider, "preferences", "post_body_parameter_overrides", default={}).items():
-            payload[key] = value
+            if key == request.model:
+                for k, v in value.items():
+                    payload[k] = v
+            elif all(_model not in request.model.lower() for _model in ["gemini", "gpt", "claude"]):
+                payload[key] = value
 
     return url, headers, payload
 
@@ -1118,6 +1171,9 @@ async def get_openrouter_payload(request, engine, provider, api_key=None):
         headers['Authorization'] = f"Bearer {api_key}"
 
     url = provider['base_url']
+    if "openrouter.ai" in url:
+        headers['HTTP-Referer'] = "https://github.com/yym68686/uni-api"
+        headers['X-Title'] = "Uni API"
 
     messages = []
     for msg in request.messages:
@@ -1165,6 +1221,14 @@ async def get_openrouter_payload(request, engine, provider, api_key=None):
         if field not in miss_fields and value is not None:
             payload[field] = value
 
+    if safe_get(provider, "preferences", "post_body_parameter_overrides", default=None):
+        for key, value in safe_get(provider, "preferences", "post_body_parameter_overrides", default={}).items():
+            if key == request.model:
+                for k, v in value.items():
+                    payload[k] = v
+            elif all(_model not in request.model.lower() for _model in ["gemini", "gpt", "claude"]):
+                payload[key] = value
+
     return url, headers, payload
 
 async def get_cohere_payload(request, engine, provider, api_key=None):
@@ -1433,9 +1497,13 @@ async def get_claude_payload(request, engine, provider, api_key=None):
             message_index = message_index + 1
 
     if "claude-3-7-sonnet" in original_model:
-        max_tokens = 20000
+        max_tokens = 128000
     elif "claude-3-5-sonnet" in original_model:
         max_tokens = 8192
+    elif "claude-sonnet-4" in original_model:
+        max_tokens = 64000
+    elif "claude-opus-4" in original_model:
+        max_tokens = 32000
     else:
         max_tokens = 4096
 

beswarm/aient/src/aient/core/response.py

@@ -535,15 +535,25 @@ async def fetch_response(client, url, headers, payload, engine, model):
         # print("parsed_data", json.dumps(parsed_data, indent=4, ensure_ascii=False))
         content = ""
         reasoning_content = ""
-        for item in parsed_data:
-            chunk = safe_get(item, "candidates", 0, "content", "parts", 0, "text")
-            is_think = safe_get(item, "candidates", 0, "content", "parts", 0, "thought", default=False)
+        parts_list = safe_get(parsed_data, 0, "candidates", 0, "content", "parts", default=[])
+        for item in parts_list:
+            chunk = safe_get(item, "text")
+            is_think = safe_get(item, "thought", default=False)
             # logger.info(f"chunk: {repr(chunk)}")
             if chunk:
                 if is_think:
                     reasoning_content += chunk
                 else:
                     content += chunk
+        # for item in parsed_data:
+        #     chunk = safe_get(item, "candidates", 0, "content", "parts", 0, "text")
+        #     is_think = safe_get(item, "candidates", 0, "content", "parts", 0, "thought", default=False)
+        #     # logger.info(f"chunk: {repr(chunk)}")
+        #     if chunk:
+        #         if is_think:
+        #             reasoning_content += chunk
+        #         else:
+        #             content += chunk
 
         usage_metadata = safe_get(parsed_data, -1, "usageMetadata")
         prompt_tokens = safe_get(usage_metadata, "promptTokenCount", default=0)

beswarm/aient/src/aient/core/utils.py

@@ -96,6 +96,7 @@ def get_engine(provider, endpoint=None, original_model=""):
     and "o3" not in original_model \
     and "o4" not in original_model \
     and "gemini" not in original_model \
+    and "gemma" not in original_model \
     and "learnlm" not in original_model \
     and "grok" not in original_model \
     and parsed_url.netloc != 'api.cloudflare.com' \

beswarm/aient/src/aient/plugins/excute_command.py

@@ -172,7 +172,8 @@ def excute_command(command):
             process.stderr.close()
 
         new_output_lines = []
-        output_lines = "".join(output_lines).strip().replace("\\u001b[A", "").replace("\\r", "\r").replace("\\\\", "").replace("\\n", "\n").replace("\r", "+++").replace("\n", "+++")
+        output_lines = "".join(output_lines).strip().replace("\\r", "\r").replace("\\\\", "").replace("\\n", "\n").replace("\r", "+++").replace("\n", "+++")
+        output_lines = re.sub(r'\\u001b\[[0-9;]*[a-zA-Z]', '', output_lines)
         for line in output_lines.split("+++"):
             if line.strip() == "":
                 continue

beswarm/aient/src/aient/utils/scripts.py

@@ -667,7 +667,6 @@ def convert_functions_to_xml(functions_list):
 
 if __name__ == "__main__":
 
-    # 运行本文件：python -m beswarm.aient.src.aient.utils.scripts
     os.system("clear")
     test_xml = """
 ✅ 好的，我现在读取 `README.md` 文件。
@@ -742,4 +741,6 @@ if __name__ == "__main__":
 请提供前两个 `excute_command` 的执行结果。
 """
 
-    print(parse_function_xml(test_xml))
+    print(parse_function_xml(test_xml))
+
+# 运行本文件：python -m beswarm.aient.src.aient.utils.scripts

beswarm/tools/__init__.py

@@ -1,7 +1,6 @@
 from .think import think
 from .edit_file import edit_file
 from .worker import worker
-from .UIworker import UIworker
 
 from .search_arxiv import search_arxiv
 from .repomap import get_code_repo_map
@@ -44,6 +43,5 @@ __all__ = [
     "find_and_click_element",
     "scroll_screen",
     "register_tool",
-    "UIworker",
     "search_web",
 ]

beswarm/tools/repomap.py

@@ -1,5 +1,6 @@
 import os
 import math
+import json
 import time
 import random
 import shutil
@@ -634,7 +635,23 @@ class RepoMap:
         return data
 
     def get_tags_raw(self, fname, rel_fname):
-        lang = filename_to_lang(str(self.root / Path(fname)))
+        # 检查是否为 .ipynb 文件，如果是则转换为 Python 代码再处理
+        if fname.endswith('.ipynb'):
+            # 读取 ipynb 文件内容
+            ipynb_content = self.io.read_text(str(self.root / Path(fname)))
+            if not ipynb_content:
+                return
+
+            # 转换为 Python 代码
+            py_content = self.convert_ipynb_to_py_content(ipynb_content)
+            if not py_content:
+                return
+
+            # 使用 Python 语言处理转换后的内容
+            lang = "python"
+        else:
+            lang = filename_to_lang(str(self.root / Path(fname)))
+
         # print(f"lang1: {lang}")
         if not lang:
             return
@@ -653,7 +670,11 @@ class RepoMap:
             return
         query_scm = query_scm.read_text()
 
-        code = self.io.read_text(str(self.root / Path(fname)))
+        # 根据文件类型选择代码内容
+        if fname.endswith('.ipynb'):
+            code = py_content
+        else:
+            code = self.io.read_text(str(self.root / Path(fname)))
         # print(f"code: {code}")
         if not code:
             return
@@ -1090,13 +1111,23 @@ class RepoMap:
             or self.tree_context_cache[rel_fname]["mtime"] != mtime
         ):
             # print(f"abs_fname: {abs_fname}")
-            code = self.io.read_text(abs_fname) or ""
+            # 处理 .ipynb 文件
+            if str(abs_fname).endswith('.ipynb'):
+                # 读取 ipynb 文件并转换
+                ipynb_content = self.io.read_text(abs_fname) or ""
+                code = self.convert_ipynb_to_py_content(ipynb_content) or ""
+                # 使用虚拟的 .py 文件名以便 TreeContext 能识别
+                context_filename = rel_fname.replace('.ipynb', '.py')
+            else:
+                code = self.io.read_text(abs_fname) or ""
+                context_filename = rel_fname
+
             # print(f"code: {code}")
             if not code.endswith("\n"):
                 code += "\n"
 
             context = TreeContext(
-                rel_fname,
+                context_filename,
                 code,
                 color=False,
                 line_number=False,
@@ -1161,6 +1192,73 @@ class RepoMap:
 
         return output
 
+    def convert_ipynb_to_py_content(self, ipynb_content):
+        """
+        将 .ipynb 文件内容转换为 Python 代码字符串
+        Markdown cells 转换为注释
+        Code cells 保持为 Python 代码
+        """
+        try:
+            notebook_data = json.loads(ipynb_content)
+        except json.JSONDecodeError:
+            return None
+
+        py_lines = []
+
+        for cell in notebook_data.get('cells', []):
+            cell_type = cell.get('cell_type')
+            source = cell.get('source', [])
+
+            if not isinstance(source, list):
+                source = [source]
+
+            source_lines = "".join(source).splitlines()
+
+            if cell_type == 'markdown':
+                for line in source_lines:
+                    py_lines.append(f"# {line}")
+                py_lines.append("")
+            elif cell_type == 'code':
+                for line in source_lines:
+                    if line.startswith("!") or line.startswith("%"):
+                        py_lines.append(f"# {line}")
+                    else:
+                        py_lines.append(line)
+
+                outputs = cell.get('outputs', [])
+                has_output_comment = False
+                for output in outputs:
+                    output_type = output.get('output_type')
+                    if output_type == 'stream':
+                        if not has_output_comment:
+                            py_lines.append("# --- Output ---")
+                            has_output_comment = True
+                        text_output = output.get('text', [])
+                        if isinstance(text_output, list):
+                            for line in "".join(text_output).splitlines():
+                                py_lines.append(f"# {line}")
+                        else:
+                            for line in text_output.splitlines():
+                                 py_lines.append(f"# {line}")
+                    elif output_type == 'execute_result':
+                        data = output.get('data', {})
+                        if 'text/plain' in data:
+                            if not has_output_comment:
+                                py_lines.append("# --- Output ---")
+                                has_output_comment = True
+                            text_output = data['text/plain']
+                            if isinstance(text_output, list):
+                                for line in "".join(text_output).splitlines():
+                                    py_lines.append(f"# {line}")
+                            else:
+                                for line in text_output.splitlines():
+                                    py_lines.append(f"# {line}")
+                if has_output_comment:
+                     py_lines.append("# --- End Output ---")
+                py_lines.append("")
+
+        return '\n'.join(py_lines)
+
 
 def find_src_files(directory):
     if not os.path.isdir(directory):
@@ -1287,6 +1385,7 @@ if __name__ == "__main__":
     # print(get_code_repo_map("."))
     # print(get_code_repo_map("/Users/yanyuming/Downloads/GitHub/uni-api"))
     # print(get_code_repo_map("/Users/yanyuming/Downloads/GitHub/text-to-motion"))
-    print(get_code_repo_map("/Users/yanyuming/Downloads/GitHub/beswarm/work/secretary/secretary"))
+    # print(get_code_repo_map("/Users/yanyuming/Downloads/GitHub/beswarm/work/secretary/secretary"))
+    print(get_code_repo_map("/Users/yanyuming/Downloads/GitHub/beswarm/work/fer/fer"))
 
 # python -m beswarm.tools.repomap

beswarm/tools/worker.py

@@ -2,13 +2,11 @@ import os
 import copy
 import platform
 from datetime import datetime
-from ..aient.src.aient.plugins import register_tool
 
 from ..aient.src.aient.models import chatgpt
-from ..aient.src.aient.plugins import get_function_call_list
+from ..aient.src.aient.plugins import register_tool, get_function_call_list
 from ..aient.src.aient.prompt import system_prompt, instruction_system_prompt
-
-from ..utils import extract_xml_content
+from ..utils import extract_xml_content, get_current_screen_image_message
 
 @register_tool()
 async def worker(goal, tools, work_dir, cache_messages=None):
@@ -64,24 +62,6 @@ async def worker(goal, tools, work_dir, cache_messages=None):
     work_agent = chatgpt(**work_agent_config)
     async def instruction_agent_task():
         while True:
-            # 获取工作agent的对话历史
-            # conversation_history = copy.deepcopy(work_agent.conversation["default"])
-            # conversation_history.pop(0)
-
-            # conversation_len = len(conversation_history) - 1
-            # message_index = 0
-            # while message_index < conversation_len:
-            #     if isinstance(conversation_history[message_index]["content"], str) and conversation_history[message_index]["content"].strip() == "":
-            #         conversation_history.pop(message_index)
-            #         conversation_len = conversation_len - 1
-            #     elif isinstance(conversation_history[message_index]["content"], list) and \
-            #     len(conversation_history[message_index]["content"]) > 0 and \
-            #     conversation_history[message_index]["content"][0].get("type") == "text" and \
-            #     conversation_history[message_index]["content"][0].get("text").strip() == "":
-            #         conversation_history.pop(message_index)
-            #         conversation_len = conversation_len - 1
-            #     else:
-            #         message_index = message_index + 1
 
             instruction_prompt = f"""
 </work_agent_conversation_end>
@@ -96,6 +76,8 @@ async def worker(goal, tools, work_dir, cache_messages=None):
             conversation_history = copy.deepcopy(work_agent.conversation["default"])
             conversation_history.pop(0)
             instruction_agent.conversation["default"][1:] = conversation_history
+            if "find_and_click_element" in str(tools_json):
+                instruction_prompt = await get_current_screen_image_message(instruction_prompt)
             next_instruction = await instruction_agent.ask_async(instruction_prompt)
             print("\n🤖 指令智能体生成的下一步指令:", next_instruction)
             if "fetch_gpt_response_stream HTTP Error', 'status_code': 404" in next_instruction:
@@ -107,6 +89,13 @@ async def worker(goal, tools, work_dir, cache_messages=None):
                 print("\n❌ 指令智能体生成的指令不符合要求，请重新生成。")
                 continue
             else:
+                if conversation_history == []:
+                    next_instruction = (
+                        "任务描述：\n"
+                        f"{goal}\n\n"
+                        "现在开始执行第一步：\n"
+                        f"{next_instruction}"
+                    )
                 break
         return next_instruction
 
@@ -120,7 +109,8 @@ async def worker(goal, tools, work_dir, cache_messages=None):
             if "任务已完成" in next_instruction:
                 print("\n✅ 任务已完成！")
                 break
-
+        if "find_and_click_element" in str(tools_json):
+            next_instruction = await get_current_screen_image_message(next_instruction)
         result = await work_agent.ask_async(next_instruction)
         if result.strip() == '' or result.strip() == '</content>\n</write_to_file>':
             print("\n❌ 工作智能体回复为空，请重新生成指令。")

beswarm/utils.py

@@ -8,4 +8,40 @@ def extract_xml_content(text, xml_tag):
         result = match.group(1)
     if not result:
         return ''
-    return result
+    return result
+
+import io
+import base64
+from .aient.src.aient.core.utils import get_image_message, get_text_message
+
+async def get_current_screen_image_message(prompt):
+    print("instruction agent 正在截取当前屏幕...")
+    try:
+        import pyautogui
+        # 使用 pyautogui 截取屏幕，返回 PIL Image 对象
+        screenshot = pyautogui.screenshot()
+        # img_width, img_height = screenshot.size # 获取截图尺寸
+        img_width, img_height = pyautogui.size()
+        print(f"截图成功，尺寸: {img_width}x{img_height}")
+
+        # 将 PIL Image 对象转换为 Base64 编码的 PNG 字符串
+        buffered = io.BytesIO()
+        screenshot.save(buffered, format="PNG")
+        base64_encoded_image = base64.b64encode(buffered.getvalue()).decode("utf-8")
+        IMAGE_MIME_TYPE = "image/png" # 截图格式为 PNG
+
+    except ImportError:
+        # Pillow 也是 pyautogui 的依赖，但以防万一单独处理
+        print("\n❌ 请安装所需库: pip install Pillow pyautogui")
+        return False
+    except Exception as e:
+        print(f"\n❌ 截取屏幕或处理图像时出错: {e}")
+        return False
+
+    engine_type = "gpt"
+    message_list = []
+    text_message = await get_text_message(prompt, engine_type)
+    image_message = await get_image_message(f"data:{IMAGE_MIME_TYPE};base64," + base64_encoded_image, engine_type)
+    message_list.append(text_message)
+    message_list.append(image_message)
+    return message_list

{beswarm-0.1.57.dist-info → beswarm-0.1.59.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: beswarm
-Version: 0.1.57
+Version: 0.1.59
 Summary: MAS
 Requires-Python: >=3.11
 Description-Content-Type: text/markdown

{beswarm-0.1.57.dist-info → beswarm-0.1.59.dist-info}/RECORD

@@ -1,14 +1,14 @@
 beswarm/__init__.py,sha256=HZjUOJtZR5QhMuDbq-wukQQn1VrBusNWai_ysGo-VVI,20
-beswarm/utils.py,sha256=AdDCcqAIIKQEMl7PfryVgeT9G5sHe7QNsZnrvmTGA8E,283
+beswarm/utils.py,sha256=Z2Kuus2BLp9EHUC2ZNL9iUsb6NWnPj-MTA7SYzGyg24,1755
 beswarm/aient/main.py,sha256=SiYAIgQlLJqYusnTVEJOx1WNkSJKMImhgn5aWjfroxg,3814
-beswarm/aient/setup.py,sha256=6ygixQucKMW9S7_9BHl6IsJJN8lqDYcuKC1wOPsrzp8,487
+beswarm/aient/setup.py,sha256=qROUPbImzYY1XRZJ-vg_r1jy-51i5axd7ROuKJKxUJ4,487
 beswarm/aient/src/aient/__init__.py,sha256=SRfF7oDVlOOAi6nGKiJIUK6B_arqYLO9iSMp-2IZZps,21
 beswarm/aient/src/aient/core/__init__.py,sha256=NxjebTlku35S4Dzr16rdSqSTWUvvwEeACe8KvHJnjPg,34
 beswarm/aient/src/aient/core/log_config.py,sha256=kz2_yJv1p-o3lUQOwA3qh-LSc3wMHv13iCQclw44W9c,274
 beswarm/aient/src/aient/core/models.py,sha256=kF-HLi1I2k_G5r153ZHuiGH8_NmpTlFMfK0_myB28YQ,7366
-beswarm/aient/src/aient/core/request.py,sha256=VItemXnWzqzS10W-RuLVrARki1w7MZMBZdyqyA5axw8,61943
-beswarm/aient/src/aient/core/response.py,sha256=BNHLazjfQT8mVg7LnPLzlX429aQM3S03pumPbOpczCI,31518
-beswarm/aient/src/aient/core/utils.py,sha256=n3dyaApN4rrSduI8cjZbeD0mv8_O5LPTTbwRkj1_v4w,26540
+beswarm/aient/src/aient/core/request.py,sha256=9GbzEg7jIH8s-jXeB1gsfoOsDbwg4C6LqXvRxVTnqEs,65263
+beswarm/aient/src/aient/core/response.py,sha256=Z0Bjl_QvpUguyky1LIcsVks4BKKqT0eYEpDmKa_cwpQ,31978
+beswarm/aient/src/aient/core/utils.py,sha256=-naFCv8V-qhnqvDUd8BNbW1HR9CVAPxISrXoAz464Qg,26580
 beswarm/aient/src/aient/core/test/test_base_api.py,sha256=pWnycRJbuPSXKKU9AQjWrMAX1wiLC_014Qc9hh5C2Pw,524
 beswarm/aient/src/aient/core/test/test_geminimask.py,sha256=HFX8jDbNg_FjjgPNxfYaR-0-roUrOO-ND-FVsuxSoiw,13254
 beswarm/aient/src/aient/core/test/test_image.py,sha256=_T4peNGdXKBHHxyQNx12u-NTyFE8TlYI6NvvagsG2LE,319
@@ -25,7 +25,7 @@ beswarm/aient/src/aient/models/vertex.py,sha256=qVD5l1Q538xXUPulxG4nmDjXE1VoV4yu
 beswarm/aient/src/aient/plugins/__init__.py,sha256=p3KO6Aa3Lupos4i2SjzLQw1hzQTigOAfEHngsldrsyk,986
 beswarm/aient/src/aient/plugins/arXiv.py,sha256=yHjb6PS3GUWazpOYRMKMzghKJlxnZ5TX8z9F6UtUVow,1461
 beswarm/aient/src/aient/plugins/config.py,sha256=Vp6CG9ocdC_FAlCMEGtKj45xamir76DFxdJVvURNtog,6539
-beswarm/aient/src/aient/plugins/excute_command.py,sha256=A3WmfZboEikU1EHvtMWhBv-xHxCyMxbDddQ982I_8wE,10482
+beswarm/aient/src/aient/plugins/excute_command.py,sha256=huQSbNbeImV8BUIsQKE13BIhCAMr7aYRyXO4saE1dTI,10534
 beswarm/aient/src/aient/plugins/get_time.py,sha256=Ih5XIW5SDAIhrZ9W4Qe5Hs1k4ieKPUc_LAd6ySNyqZk,654
 beswarm/aient/src/aient/plugins/image.py,sha256=ZElCIaZznE06TN9xW3DrSukS7U3A5_cjk1Jge4NzPxw,2072
 beswarm/aient/src/aient/plugins/list_directory.py,sha256=JZVuImecMSfEv6jLqii-0uQJ1UCsrpMNmYlwW3PEDg4,1374
@@ -39,7 +39,7 @@ beswarm/aient/src/aient/prompt/__init__.py,sha256=GBtn6-JDT8KHFCcuPpfSNE_aGddg5p
 beswarm/aient/src/aient/prompt/agent.py,sha256=y2GETN6ScC5yQVs75VFfzm4YUWzblbqLYz0Sy6JnPRw,24950
 beswarm/aient/src/aient/utils/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 beswarm/aient/src/aient/utils/prompt.py,sha256=UcSzKkFE4-h_1b6NofI6xgk3GoleqALRKY8VBaXLjmI,11311
-beswarm/aient/src/aient/utils/scripts.py,sha256=wutPtgbs-WXo5AACLpnCJaRQBOSKXWNnsf2grbYDzyQ,29098
+beswarm/aient/src/aient/utils/scripts.py,sha256=LD8adnfuRrJoY2tWKseXOPJXaxbrUmz4czsnUvHswNY,29096
 beswarm/aient/test/chatgpt.py,sha256=Hvl7FuDt1c74N5TVBmhErOPvJbJJzA7FNp5VoZM4u30,4957
 beswarm/aient/test/claude.py,sha256=IyB4qI1eJLwlSfDNSnt2FhbQWYyBighHUjJxEXc3osQ,1095
 beswarm/aient/test/test.py,sha256=rldnoLQdtRR8IKFSIzTti7eIK2MpPMoi9gL5qD8_K44,29
@@ -119,17 +119,16 @@ beswarm/queries/tree-sitter-languages/ruby-tags.scm,sha256=vIidsCeE2A0vdFN18yXKq
 beswarm/queries/tree-sitter-languages/rust-tags.scm,sha256=9ljM1nzhfPs_ZTRw7cr2P9ToOyhGcKkCoN4_HPXSWi4,1451
 beswarm/queries/tree-sitter-languages/scala-tags.scm,sha256=UxQjz80JIrrJ7Pm56uUnQyThfmQNvwk7aQzPNypB-Ao,1761
 beswarm/queries/tree-sitter-languages/typescript-tags.scm,sha256=OMdCeedPiA24ky82DpgTMKXK_l2ySTuF2zrQ2fJAi9E,1253
-beswarm/tools/UIworker.py,sha256=1sEC76VGFwo48lSx6KOvhJwhgBj7UWAHAAH9BG_lp-M,6439
-beswarm/tools/__init__.py,sha256=jOfYY4EYkwmz-FTJGrI1CyaIYkGWsmGzZBGsoupeX9M,1088
+beswarm/tools/__init__.py,sha256=oDsCE7Coy3TXM0pTRS_4mWTEyPnsKVK7Vco1idSVxJk,1041
 beswarm/tools/click.py,sha256=TygaekCXTmU3fIu6Uom7ZcyzEgYMlCC_GX-5SmWHuLI,20762
 beswarm/tools/edit_file.py,sha256=xlAD0HB_xM0yZYc0eJwLE-9mAkywXa2UQPNHzG1OaW4,7664
 beswarm/tools/planner.py,sha256=lguBCS6kpwNPoXQvqH-WySabVubT82iyWOkJnjt6dXw,1265
-beswarm/tools/repomap.py,sha256=CwvwoN5Swr42EzrORTTeV8MMb7mPviy4a4b0fxBu50k,40828
+beswarm/tools/repomap.py,sha256=N09K0UgwjCN7Zjg_5TYlVsulp3n2fztYlS8twalChU8,45003
 beswarm/tools/search_arxiv.py,sha256=9slwBemXjEqrd7-YgVmyMijPXlkhZCybEDRVhWVQ9B0,7937
 beswarm/tools/search_web.py,sha256=B24amOnGHnmdV_6S8bw8O2PdhZRRIDtJjg-wXcfP7dQ,11859
 beswarm/tools/think.py,sha256=WLw-7jNIsnS6n8MMSYUin_f-BGLENFmnKM2LISEp0co,1760
-beswarm/tools/worker.py,sha256=b-FvSEP27-zMYNcqaQeVBoWxaSf2cX_7_1p1GAF6h04,6191
-beswarm-0.1.57.dist-info/METADATA,sha256=FdjwXDGhJw1_43unl9YjSFbzl8jB5Pu-g0KX_Hn_TNA,3553
-beswarm-0.1.57.dist-info/WHEEL,sha256=zaaOINJESkSfm_4HQVc5ssNzHCPXhJm0kEUakpsEHaU,91
-beswarm-0.1.57.dist-info/top_level.txt,sha256=pJw4O87wvt5882smuSO6DfByJz7FJ8SxxT8h9fHCmpo,8
-beswarm-0.1.57.dist-info/RECORD,,
+beswarm/tools/worker.py,sha256=PRsAa856KYNPbi2L8rDdWEKx3u-c_WzxoI1x8AFkmtc,5613
+beswarm-0.1.59.dist-info/METADATA,sha256=LrUP7SrPGRG87cgSLMuE7liHKGGMIe4bQUdOhbfRifk,3553
+beswarm-0.1.59.dist-info/WHEEL,sha256=zaaOINJESkSfm_4HQVc5ssNzHCPXhJm0kEUakpsEHaU,91
+beswarm-0.1.59.dist-info/top_level.txt,sha256=pJw4O87wvt5882smuSO6DfByJz7FJ8SxxT8h9fHCmpo,8
+beswarm-0.1.59.dist-info/RECORD,,

beswarm/tools/UIworker.py

@@ -1,147 +0,0 @@
-import os
-import io
-import copy
-import base64
-import platform
-from datetime import datetime
-from ..aient.src.aient.plugins import register_tool, get_function_call_list
-
-from ..aient.src.aient.models import chatgpt
-from ..aient.src.aient.prompt import system_prompt, instruction_system_prompt
-from ..aient.src.aient.core.utils import get_image_message, get_text_message
-
-from ..utils import extract_xml_content
-
-async def get_current_screen_image_message(prompt):
-    print("instruction agent 正在截取当前屏幕...")
-    try:
-        import pyautogui
-        # 使用 pyautogui 截取屏幕，返回 PIL Image 对象
-        screenshot = pyautogui.screenshot()
-        # img_width, img_height = screenshot.size # 获取截图尺寸
-        img_width, img_height = pyautogui.size()
-        print(f"截图成功，尺寸: {img_width}x{img_height}")
-
-        # 将 PIL Image 对象转换为 Base64 编码的 PNG 字符串
-        buffered = io.BytesIO()
-        screenshot.save(buffered, format="PNG")
-        base64_encoded_image = base64.b64encode(buffered.getvalue()).decode("utf-8")
-        IMAGE_MIME_TYPE = "image/png" # 截图格式为 PNG
-
-    except ImportError:
-        # Pillow 也是 pyautogui 的依赖，但以防万一单独处理
-        print("\n❌ 请安装所需库: pip install Pillow pyautogui")
-        return False
-    except Exception as e:
-        print(f"\n❌ 截取屏幕或处理图像时出错: {e}")
-        return False
-
-    engine_type = "gpt"
-    message_list = []
-    text_message = await get_text_message(prompt, engine_type)
-    image_message = await get_image_message(f"data:{IMAGE_MIME_TYPE};base64," + base64_encoded_image, engine_type)
-    message_list.append(text_message)
-    message_list.append(image_message)
-    return message_list
-
-@register_tool()
-async def UIworker(goal, tools, work_dir, cache_messages=None):
-    """
-    启动一个 **工作智能体 (Worker Agent)** 来自动完成指定的任务目标 (`goal`)。
-
-    这个工作智能体接收一个清晰的任务描述、一组可供调用的工具 (`tools`)，以及一个工作目录 (`work_dir`)。
-    它会利用语言模型的能力，结合可用的工具，自主规划并逐步执行必要的操作，直到最终完成指定的任务目标。
-    核心功能是根据输入的目标，驱动整个任务执行流程。
-
-    Args:
-        goal (str): 需要完成的具体任务目标描述。工作智能体将围绕此目标进行工作。必须清晰、具体。
-        tools (list[str]): 一个包含可用工具函数对象的列表。工作智能体在执行任务时可能会调用这些工具来与环境交互（例如读写文件、执行命令等）。
-        work_dir (str): 工作目录的绝对路径。工作智能体将在此目录上下文中执行操作。
-
-    Returns:
-        str: 当任务成功完成时，返回字符串 "任务已完成"。
-    """
-
-    tools_json = [value for _, value in get_function_call_list(tools).items()]
-    work_agent_system_prompt = system_prompt.format(
-        os_version=platform.platform(),
-        workspace_path=work_dir,
-        shell=os.getenv('SHELL', 'Unknown'),
-        tools_list=tools_json
-    )
-
-    work_agent_config = {
-        "api_key": os.getenv("API_KEY"),
-        "api_url": os.getenv("BASE_URL"),
-        "engine": os.getenv("MODEL"),
-        "system_prompt": work_agent_system_prompt,
-        "print_log": True,
-        # "max_tokens": 8000,
-        "temperature": 0.5,
-        "function_call_max_loop": 100,
-    }
-    if cache_messages:
-        work_agent_config["cache_messages"] = cache_messages
-
-    instruction_agent_config = {
-        "api_key": os.getenv("API_KEY"),
-        "api_url": os.getenv("BASE_URL"),
-        "engine": os.getenv("MODEL"),
-        "system_prompt": instruction_system_prompt.format(os_version=platform.platform(), tools_list=tools_json, workspace_path=work_dir, current_time=datetime.now().strftime("%Y-%m-%d %H:%M:%S")),
-        "print_log": False,
-        # "max_tokens": 4000,
-        "temperature": 0.7,
-        "use_plugins": False,
-    }
-
-    # 工作agent初始化
-    work_agent = chatgpt(**work_agent_config)
-    async def instruction_agent_task():
-        while True:
-            instruction_prompt = f"""
-任务目标: {goal}
-
-以上对话都是工作智能体的对话历史。
-
-根据以上对话历史和目标，请生成下一步指令。如果任务已完成，请回复"任务已完成"。
-            """
-            # 让指令agent分析对话历史并生成新指令
-            instruction_agent = chatgpt(**instruction_agent_config)
-            conversation_history = copy.deepcopy(work_agent.conversation["default"])
-            conversation_history.pop(0)
-            instruction_agent.conversation["default"][1:] = conversation_history
-            new_prompt = await get_current_screen_image_message(instruction_prompt)
-            next_instruction = await instruction_agent.ask_async(new_prompt)
-            print("\n🤖 指令智能体生成的下一步指令:", next_instruction)
-            if "fetch_gpt_response_stream HTTP Error', 'status_code': 404" in next_instruction:
-                raise Exception(f"Model: {instruction_agent_config['engine']} not found!")
-            if "'status_code': 413" in next_instruction:
-                raise Exception(f"The request body is too long, please try again.")
-            next_instruction = extract_xml_content(next_instruction, "instructions")
-            if not next_instruction:
-                print("\n❌ 指令智能体生成的指令不符合要求，请重新生成。")
-                continue
-            else:
-                break
-        return next_instruction
-
-    need_instruction = True
-    while True:
-        next_instruction = ''
-        if need_instruction:
-            next_instruction = await instruction_agent_task()
-
-            # 检查任务是否完成
-            if "任务已完成" in next_instruction:
-                print("\n✅ 任务已完成！")
-                break
-        new_prompt = await get_current_screen_image_message(next_instruction)
-        result = await work_agent.ask_async(new_prompt)
-        if result.strip() == '':
-            print("\n❌ 工作智能体回复为空，请重新生成指令。")
-            need_instruction = False
-            continue
-        print("✅ 工作智能体回复:", result)
-        need_instruction = True
-
-    return "任务已完成"