pyxllib 0.3.120__tar.gz → 0.3.121__tar.gz

{pyxllib-0.3.120/pyxllib.egg-info → pyxllib-0.3.121}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: pyxllib
-Version: 0.3.120
+Version: 0.3.121
 Summary: 厦门理工模式识别团队通用python代码工具库
 Home-page: https://github.com/XLPRUtils/pyxllib
 Author: code4101

{pyxllib-0.3.120 → pyxllib-0.3.121}/pyxllib/data/sqlite.py

@@ -150,7 +150,6 @@ class SqlBuilder:
             sql.append(f"WHERE {' AND '.join(self._where)}")
         return '\n'.join(sql)
 
-
     def __3_build_中级命令(self):
         pass
 
@@ -171,6 +170,7 @@ class SqlBuilder:
         sql._where = self._where.copy()
         return sql.build_select()
 
+
 class SqlBase:
     """ Sql语法通用的功能 """
 

{pyxllib-0.3.120 → pyxllib-0.3.121}/pyxllib/file/xlsxlib.py

@@ -3063,8 +3063,8 @@ def extract_workbook_summary3b(file_path,
         reduce_summary(res)
 
     time_dict = {'load_time': human_readable_number(load_time),
-                     'summary2_time': human_readable_number(summary2_time),
-                     'summary3_time': human_readable_number(summary3_time)}
+                 'summary2_time': human_readable_number(summary2_time),
+                 'summary3_time': human_readable_number(summary3_time)}
 
     if return_mode == 1:
         return res, time_dict

{pyxllib-0.3.120 → pyxllib-0.3.121/pyxllib.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: pyxllib
-Version: 0.3.120
+Version: 0.3.121
 Summary: 厦门理工模式识别团队通用python代码工具库
 Home-page: https://github.com/XLPRUtils/pyxllib
 Author: code4101

{pyxllib-0.3.120 → pyxllib-0.3.121}/pyxllib.egg-info/SOURCES.txt

@@ -139,6 +139,11 @@ pyxlpr/data/icdar/icdar2013.py
 pyxlpr/data/icdar/iou.py
 pyxlpr/data/icdar/rrc_evaluation_funcs_1_1.py
 pyxlpr/eval/__init__.py
+pyxlpr/openai2/__init__.py
+pyxlpr/openai2/openai2.py
+pyxlpr/openai2/_core/chat.py
+pyxlpr/openai2/_core/chat_in_cmd.py
+pyxlpr/openai2/_core/group_chat.py
 pyxlpr/ppocr/__init__.py
 pyxlpr/ppocr/configs/rec/multi_language/generate_multi_language_configs.py
 pyxlpr/ppocr/data/__init__.py

pyxllib-0.3.121/pyxlpr/openai2/__init__.py

@@ -0,0 +1,3 @@
+from ._core.chat import Chat, AKPool, system_msg, user_msg, assistant_msg
+from ._core.group_chat import GroupChat
+from ._core.chat_in_cmd import chat_in_cmd, ParseCmd

pyxllib-0.3.121/pyxlpr/openai2/_core/chat.py

@@ -0,0 +1,277 @@
+from json import dumps as jsonDumps
+from json import loads as jsonLoads
+from pathlib import Path
+from typing import List, Literal
+from openai import OpenAI, AsyncOpenAI
+
+
+class AKPool:
+    """ 轮询获取api_key """
+
+    def __init__(self, apikeys: list):
+        self._pool = self._POOL(apikeys)
+
+    def fetch_key(self):
+        return next(self._pool)
+
+    @classmethod
+    def _POOL(cls, apikeys: list):
+        while True:
+            for x in apikeys:
+                yield x
+
+
+class MsgBase:
+    role_name: str
+    text: str
+
+    def __init__(self, text: str):
+        self.text = text
+
+    def __str__(self):
+        return self.text
+
+    def __iter__(self):
+        yield "role", self.role_name
+        yield "content", self.text
+
+
+system_msg = type("system_msg", (MsgBase,), {"role_name": "system"})
+user_msg = type("user_msg", (MsgBase,), {"role_name": "user"})
+assistant_msg = type("assistant_msg", (MsgBase,), {"role_name": "assistant"})
+
+
+class Temque:
+    """ 一个先进先出, 可设置最大容量, 可固定元素的队列 """
+
+    def __init__(self, maxlen: int = None):
+        self.core: List[dict] = []
+        self.maxlen = maxlen or float("inf")
+
+    def _trim(self):
+        core = self.core
+        if len(core) > self.maxlen:
+            dc = len(core) - self.maxlen
+            indexes = []
+            for i, x in enumerate(core):
+                if not x["pin"]:
+                    indexes.append(i)
+                if len(indexes) == dc:
+                    break
+            for i in indexes[::-1]:
+                core.pop(i)
+
+    def add_many(self, *objs):
+        for x in objs:
+            self.core.append({"obj": x, "pin": False})
+        self._trim()
+
+    def __iter__(self):
+        for x in self.core:
+            yield x["obj"]
+
+    def pin(self, *indexes):
+        for i in indexes:
+            self.core[i]["pin"] = True
+
+    def unpin(self, *indexes):
+        for i in indexes:
+            self.core[i]["pin"] = False
+
+    def copy(self):
+        que = self.__class__(maxlen=self.maxlen)
+        que.core = self.core.copy()
+        return que
+
+    def deepcopy(self):
+        ...  # 创建这个方法是为了以代码提示的方式提醒用户: copy 方法是浅拷贝
+
+    def __add__(self, obj: 'list|Temque'):
+        que = self.copy()
+        if isinstance(obj, self.__class__):
+            que.core += obj.core
+            que._trim()
+        else:
+            que.add_many(*obj)
+        return que
+
+
+class Chat:
+    """
+    [文档](https://lcctoor.github.io/arts/arts/openai2)
+
+    获取api_key:
+    * [获取链接1](https://platform.openai.com/account/api-keys)
+    * [获取链接2](https://www.baidu.com/s?wd=%E8%8E%B7%E5%8F%96%20openai%20api_key)
+    """
+
+    recently_request_data: dict  # 最近一次请求所用的参数
+
+    def __init__(self,
+                 # kwargs
+                 api_key: str|AKPool,
+                 base_url: str = None,  # base_url 参数用于修改基础URL
+                 timeout=None,
+                 max_retries=None,
+                 http_client=None,
+                 # request_kwargs
+                 model: Literal["gpt-4-1106-preview", "gpt-4-vision-preview", "gpt-4", "gpt-4-0314", "gpt-4-0613",
+                                "gpt-4-32k", "gpt-4-32k-0314", "gpt-4-32k-0613", "gpt-3.5-turbo"] = "gpt-3.5-turbo",
+                 # Chat
+                 msg_max_count: int=None,
+                 # kwargs
+                 **kwargs,
+                 ):
+        api_base = kwargs.pop('api_base', None)
+        base_url = base_url or api_base
+        MsgMaxCount = kwargs.pop('MsgMaxCount', None)
+        msg_max_count = msg_max_count or MsgMaxCount
+
+        if base_url: kwargs["base_url"] = base_url
+        if timeout: kwargs["timeout"] = timeout
+        if max_retries: kwargs["max_retries"] = max_retries
+        if http_client: kwargs["http_client"] = http_client
+
+        self.reset_api_key(api_key)
+        self._kwargs = kwargs
+        self._request_kwargs = {'model':model}
+        self._messages = Temque(maxlen=msg_max_count)
+
+    def reset_api_key(self, api_key: str|AKPool):
+        if isinstance(api_key, AKPool):
+            self._akpool = api_key
+        else:
+            self._akpool = AKPool([api_key])
+
+    def request(self, text: str=None, **kwargs):
+        messages = [{"role": "user", "content": text}]
+        messages += (kwargs.pop('messages', None) or [])  # 兼容官方包[openai]用户, 使其代码可以无缝切换到[openai2]
+        assert messages
+        self.recently_request_data = {
+            'api_key': (api_key := self._akpool.fetch_key()),
+        }
+        completion = OpenAI(api_key=api_key, **self._kwargs).chat.completions.create(**{
+            **self._request_kwargs,  # 全局参数
+            **kwargs,  # 单次请求的参数覆盖全局参数
+            "messages": list(self._messages + messages),
+            "stream": False,
+        })
+        answer: str = completion.choices[0].message.content
+        self._messages.add_many(*messages, {"role": "assistant", "content": answer})
+        return answer
+
+    def stream_request(self, text: str=None, **kwargs):
+        messages = [{"role": "user", "content": text}]
+        messages += (kwargs.pop('messages', None) or [])  # 兼容官方包[openai]用户, 使其代码可以无缝切换到[openai2]
+        assert messages
+        self.recently_request_data = {
+            'api_key': (api_key := self._akpool.fetch_key()),
+        }
+        completion = OpenAI(api_key=api_key, **self._kwargs).chat.completions.create(**{
+            **self._request_kwargs,  # 全局参数
+            **kwargs,  # 单次请求的参数覆盖全局参数
+            "messages": list(self._messages + messages),
+            "stream": True,
+        })
+        answer: str = ""
+        for chunk in completion:
+            if chunk.choices and (content := chunk.choices[0].delta.content):
+                answer += content
+                yield content
+        self._messages.add_many(*messages, {"role": "assistant", "content": answer})
+
+    async def async_request(self, text: str=None, **kwargs):
+        messages = [{"role": "user", "content": text}]
+        messages += (kwargs.pop('messages', None) or [])  # 兼容官方包[openai]用户, 使其代码可以无缝切换到[openai2]
+        assert messages
+        self.recently_request_data = {
+            'api_key': (api_key := self._akpool.fetch_key()),
+        }
+        completion = await AsyncOpenAI(api_key=api_key, **self._kwargs).chat.completions.create(**{
+            **self._request_kwargs,  # 全局参数
+            **kwargs,  # 单次请求的参数覆盖全局参数
+            "messages": list(self._messages + messages),
+            "stream": False,
+        })
+        answer: str = completion.choices[0].message.content
+        self._messages.add_many(*messages, {"role": "assistant", "content": answer})
+        return answer
+
+    async def async_stream_request(self, text: str=None, **kwargs):
+        messages = [{"role": "user", "content": text}]
+        messages += (kwargs.pop('messages', None) or [])  # 兼容官方包[openai]用户, 使其代码可以无缝切换到[openai2]
+        assert messages
+        self.recently_request_data = {
+            'api_key': (api_key := self._akpool.fetch_key()),
+        }
+        completion = await AsyncOpenAI(api_key=api_key, **self._kwargs).chat.completions.create(**{
+            **self._request_kwargs,  # 全局参数
+            **kwargs,  # 单次请求的参数覆盖全局参数
+            "messages": list(self._messages + messages),
+            "stream": True,
+        })
+        answer: str = ""
+        async for chunk in completion:
+            if chunk.choices and (content := chunk.choices[0].delta.content):
+                answer += content
+                yield content
+        self._messages.add_many(*messages, {"role": "assistant", "content": answer})
+
+    def rollback(self, n=1):
+        '''
+        回滚对话
+        '''
+        self._messages.core[-2 * n :] = []
+        for x in self._messages.core[-2:]:
+            x = x["obj"]
+            print(f"[{x['role']}]:{x['content']}")
+
+    def pin_messages(self, *indexes):
+        '''
+        锁定历史消息
+        '''
+        self._messages.pin(*indexes)
+
+    def unpin_messages(self, *indexes):
+        '''
+        解锁历史消息
+        '''
+        self._messages.unpin(*indexes)
+
+    def fetch_messages(self):
+        return list(self._messages)
+
+    def add_dialogs(self, *ms: dict|system_msg|user_msg|assistant_msg):
+        '''
+        添加历史对话
+        '''
+        messages = [dict(x) for x in ms]
+        self._messages.add_many(*messages)
+
+    def __getattr__(self, name):
+        match name:  # 兼容旧代码
+            case 'asy_request':
+                return self.async_request
+            case 'forge':
+                return self.add_dialogs
+            case 'pin':
+                return self.pin_messages
+            case 'unpin':
+                return self.unpin_messages
+            case 'dump':
+                return self._dump
+            case 'load':
+                return self._load
+        raise AttributeError(name)
+
+    def _dump(self, fpath: str):
+        """ 存档 """
+        jt = jsonDumps(self.fetch_messages(), ensure_ascii=False)
+        Path(fpath).write_text(jt, encoding="utf8")
+        return True
+
+    def _load(self, fpath: str):
+        """ 载入存档 """
+        jt = Path(fpath).read_text(encoding="utf8")
+        self._messages.add_many(*jsonLoads(jt))
+        return True

pyxllib-0.3.121/pyxlpr/openai2/_core/chat_in_cmd.py

@@ -0,0 +1,67 @@
+import sys
+from .chat import Chat, AKPool
+from arts.cooltypes import moduledb
+
+
+mdb = moduledb.DB('https://pypi.org/project/openai2', depth=3)['chat_in_cmd']
+apikeys: moduledb.File = mdb['apikeys']['data_1']
+apikeys.setdefault('list', [])
+record: moduledb.File = mdb['records']['chat_1']
+record.setdefault('messages', [])
+
+
+def chat_in_cmd(apikeys:list, newchat=False, model='gpt-4-1106-preview', msg_max_count=30):
+    print(f"\n\033[0m你已进入命令行聊天模式, 当前使用'{model}'模型, 请确保你的apikey支持该模型.", end='')
+    gpt = Chat(api_key=AKPool(apikeys), model=model, msg_max_count=msg_max_count)
+    if not newchat:
+        gpt.add_dialogs(*record['messages'])
+    while True:
+        user = input('\n\n\033[32;1m:')
+        print('\033[0m')
+        for x in gpt.stream_request(user):
+            print(x, end='', flush=True)
+        record['messages'] = gpt.fetch_messages()
+
+
+命令提示 = '''指令集:
+openai2 add_apikey <apikey> | 添加1个apikey              | 如需添加多个, 可执行多次
+openai2 read_apikey         | 查看所有apikey             |
+openai2 clear_apikey        | 清除所有apikey             |
+openai2 chat                | 继续上次的对话              |
+openai2 newchat             | 清空对话记录, 然后开始新对话 |
+'''
+
+
+def ParseCmd():
+    kws = sys.argv[1:]
+    if kws:
+        kw = kws[0].lower()
+        
+        # 添加apikey
+        if kw == 'add_apikey' and len(kws) > 1:
+            apikeys['list'].append( kws[1] )
+            apikeys.save()
+            print('添加 apikey 成功.')
+
+        # 继续上次的对话
+        elif kw == 'chat':
+            chat_in_cmd(apikeys['list'], newchat=False)
+        
+        # 清空对话记录, 然后开始新对话
+        elif kw == 'newchat':
+            chat_in_cmd(apikeys['list'], newchat=True)
+        
+        # 查看所有apikey
+        elif kw == 'read_apikey':
+            for x in apikeys['list']:
+                print(x)
+        
+        # 清除所有apikey
+        elif kw == 'clear_apikey':
+            apikeys['list'] = []
+            print('已清除所有 apikey .')
+
+        else:
+            print(命令提示)
+    else:
+        print(命令提示)

pyxllib-0.3.121/pyxlpr/openai2/_core/group_chat.py

@@ -0,0 +1,120 @@
+from json import dumps as jsonDumps
+from typing import List, Tuple, Union, Literal
+from openai import OpenAI
+from .chat import system_msg, user_msg, assistant_msg, Chat, AKPool
+
+
+class GroupChat(Chat):
+
+    roles = {
+        "苏轼":'宋朝诗人，他的词风格独特，既有儒家的教诲，又有生活的乐趣。',
+        "李清照":'宋代著名的女词人，其词句优美，情感真挚。',
+        "杜甫":'唐朝著名诗人。',
+        "小许":"一个聪明的程序员",
+        "小郑":"一个帅气的男人",
+        "小张":"一个漂亮的女人"
+    }
+
+    _top_messages: List[dict] = None
+
+    def __init__(self, *vs, **kvs):
+        Chat.__init__(self, *vs, **kvs)
+        self.roles = self.roles.copy()
+        self._new_dialogs = []
+        if not self.__class__._top_messages:
+            self.__class__._top_messages = self._render_top_messages()
+    
+    # 欺骗编辑器进行代码提示
+    if len(__file__) < 0:
+        def __init__(self,
+                    # kwargs
+                    api_key: Union[str, AKPool],
+                    base_url: str = None,  # base_url 参数用于修改基础URL
+                    timeout=None,
+                    max_retries=None,
+                    http_client=None,
+                    # request_kwargs
+                    model: Literal["gpt-4-1106-preview", "gpt-4-vision-preview", "gpt-4", "gpt-4-0314", "gpt-4-0613",
+                                    "gpt-4-32k", "gpt-4-32k-0314", "gpt-4-32k-0613", "gpt-3.5-turbo"] = "gpt-3.5-turbo",
+                    # Chat
+                    msg_max_count=None,
+                    # kwargs
+                    **kwargs,
+                    ):
+            ...
+    
+    def add_dialog(self, speaker:str, audiences:list, remark:str):
+        '''
+        add_dialog('李白', ['杜甫', '小许'], '你们好呀')
+        add_dialog('杜甫', ['李白'], '你好, 你今天写诗了吗?')
+        '''
+        self.roles.setdefault(speaker, '')
+        for x in audiences:
+            self.roles.setdefault(x, '')
+        self._new_dialogs.append( f"【{speaker}】对【{'、'.join(audiences)}】说：{remark}".replace('\n', ' ') )
+    
+    def _get_user_text(self, roles: List[Tuple[str, List[str]]]):
+        # 补充人物数据
+        for speaker, audiences in roles:
+            self.roles.setdefault(speaker, '')
+            for x in audiences:
+                self.roles.setdefault(x, '')
+        # 新增的对话记录
+        _new_dialogs = [f"{i+1}、{x}" for i, x in enumerate(self._new_dialogs)]
+        _new_dialogs = '\n'.join(_new_dialogs)
+        # 需要模拟的对话
+        roles = [{'speaker':speaker, 'audiences':'、'.join(audiences), 'remark':''} for speaker, audiences in roles]
+        roles = jsonDumps(roles, ensure_ascii=False)
+        # 用户的发言
+        text = f'''以下是新增的对话记录：\n\n{_new_dialogs}\n\n请根据以下JSON文档的内容，模拟角色的视角进行发言，并把发言填充到'remark'字段中：\n\n```json\n{roles}\n```\n\n请把填充完整后的JSON文档发给我，只返回JSON文档即可，勿包含任何其它信息，否则会干扰我的解析。'''
+        return text
+    
+    def request(self, roles: List[Tuple[str, List[str]]], **kwargs):
+        '''
+        request([
+            ('苏轼', ['李清照', '杜甫']),
+            ('小明', ['小东']),
+        ])
+        '''
+        messages = [{"role": "user", "content": self._get_user_text(roles)}]
+        self.recently_request_data = {
+            'api_key': (api_key := self._akpool.fetch_key()),
+        }
+        completion = OpenAI(api_key=api_key, **self._kwargs).chat.completions.create(**{
+            **self._request_kwargs,  # 全局参数
+            **kwargs,  # 单次请求的参数覆盖全局参数
+            "messages": self._top_messages + list(self._messages) + messages,
+            "stream": False,
+        })
+        answer: str = completion.choices[0].message.content
+        self._messages.add_many(*messages, {"role": "assistant", "content": answer})
+        self._new_dialogs = []
+        return answer
+    
+    def _render_top_messages(self):
+        messages = []
+        # 任务提示
+        messages.append(system_msg(f'''以下的JSON文档描述了一些人物的角色信息：\n\n```json\n{jsonDumps(self.roles, ensure_ascii=False)}\n```\n\n你需要了解这些人物的信息。user每次都会收集这些人物的对话记录发送给你，并让你站在其中的某些角色的视角上对其它角色进行模拟发言，你每次返回JSON文档即可，勿包含任何其它信息，否则会干扰user的解析。你在发言时可以编造，比如在回答年龄时，可以随意编一个年龄，但要注意上下文逻辑闭环。'''))
+        # 第一轮教学
+        self.add_dialog('李白', ['杜甫', '小许'], '你们好呀')
+        self.add_dialog('杜甫', ['李白'], '你好, 你今天写诗了吗?')
+        self.add_dialog('小许', ['李白'], '你好, 你吃饭了吗?')
+        messages.append(user_msg(self._get_user_text([('李白',['小许']), ('李白',['杜甫']), ('李白', ['杜甫','小许'])])))
+        messages.append(assistant_msg('''[{"speaker": "李白", "audiences": "小许", "remark": "我今天写诗了"}, {"speaker": "李白", "audiences": "杜甫", "remark": "我吃饭了"}, {"speaker": "李白", "audiences": "杜甫、小许", "remark": "你们有什么有趣的事情分享吗?"}]'''))
+        self._new_dialogs = []
+        # 第二轮教学
+        self.add_dialog('小郑', ['小张'], '你是谁?')
+        self.add_dialog('小张', ['小郑'], '我叫小张,今年13岁')
+        self.add_dialog('小许', ['小郑', '小张'], '你们是哪里人?')
+        messages.append(user_msg(self._get_user_text([('小郑',['小张']), ('小郑',['小许']), ('小张', ['小许'])])))
+        messages.append(assistant_msg('''[{"speaker": "小郑", "audiences": "小张", "remark": "哦哦, 我比你大1岁"}, {"speaker": "小郑", "audiences": "小许", "remark": "我是河南人"}, {"speaker": "小张", "audiences": "小许", "remark": "不告诉你"}]'''))
+        self._new_dialogs = []
+        # 第三轮教学
+        self.add_dialog('小许', ['小张'], '呵呵,这么神秘呀?')
+        self.add_dialog('小许', ['小郑'], '你知道小张是哪里人吗?')
+        messages.append(user_msg(self._get_user_text([('小郑',['小许']), ('小张',['小许'])])))
+        messages.append(assistant_msg('''[{"speaker": "小郑", "audiences": "小许", "remark": "哈哈, 我也不知道"}, {"speaker": "小张", "audiences": "小许", "remark": "哈哈, 开个玩笑啦, 我是湖北人"}]'''))
+        self._new_dialogs = []
+        # 导出
+        messages = [dict(x) for x in messages]
+        return messages

pyxllib-0.3.121/pyxlpr/openai2/openai2.py

@@ -0,0 +1,3 @@
+from arts.openai2._core.chat import Chat, AKPool, system_msg, user_msg, assistant_msg
+from arts.openai2._core.group_chat import GroupChat
+from arts.openai2._core.chat_in_cmd import chat_in_cmd, ParseCmd

{pyxllib-0.3.120 → pyxllib-0.3.121}/setup.py

@@ -45,7 +45,7 @@ _dir = Path(__file__).parent
 
 setup(
     name='pyxllib',  # pip 安装时用的名字
-    version='0.3.120',  # 当前版本，每次更新上传到pypi都需要修改; 第4位版本号一般是修紧急bug
+    version='0.3.121',  # 当前版本，每次更新上传到pypi都需要修改; 第4位版本号一般是修紧急bug
     author='code4101',
     author_email='877362867@qq.com',
     url='https://github.com/XLPRUtils/pyxllib',