Merge pull request #712 from RockChinQ/feat/component-extensibility

Feat: 更多组件的可扩展性

pkg/command/operator.py

@@ -8,18 +8,34 @@ from . import entities
 
 
 preregistered_operators: list[typing.Type[CommandOperator]] = []
-"""预注册算子列表。在初始化时，所有算子类会被注册到此列表中。"""
+"""预注册命令算子列表。在初始化时，所有算子类会被注册到此列表中。"""
 
 
 def operator_class(
     name: str,
-    help: str,
+    help: str = "",
     usage: str = None,
     alias: list[str] = [],
     privilege: int=1,  # 1为普通用户，2为管理员
     parent_class: typing.Type[CommandOperator] = None
 ) -> typing.Callable[[typing.Type[CommandOperator]], typing.Type[CommandOperator]]:
+    """命令类装饰器
+    
+    Args:
+        name (str): 名称
+        help (str, optional): 帮助信息. Defaults to "".
+        usage (str, optional): 使用说明. Defaults to None.
+        alias (list[str], optional): 别名. Defaults to [].
+        privilege (int, optional): 权限，1为普通用户可用，2为仅管理员可用. Defaults to 1.
+        parent_class (typing.Type[CommandOperator], optional): 父节点，若为None则为顶级命令. Defaults to None.
+
+    Returns:
+        typing.Callable[[typing.Type[CommandOperator]], typing.Type[CommandOperator]]: 装饰器
+    """
+
     def decorator(cls: typing.Type[CommandOperator]) -> typing.Type[CommandOperator]:
+        assert issubclass(cls, CommandOperator)
+        
         cls.name = name
         cls.alias = alias
         cls.help = help

pkg/core/app.py

@@ -6,7 +6,7 @@ import traceback
 
 from ..platform import manager as im_mgr
 from ..provider.session import sessionmgr as llm_session_mgr
-from ..provider.requester import modelmgr as llm_model_mgr
+from ..provider.modelmgr import modelmgr as llm_model_mgr
 from ..provider.sysprompt import sysprompt as llm_prompt_mgr
 from ..provider.tools import toolmgr as llm_tool_mgr
 from ..config import manager as config_mgr

pkg/core/entities.py

@@ -9,7 +9,7 @@ import pydantic
 import mirai
 
 from ..provider import entities as llm_entities
-from ..provider.requester import entities
+from ..provider.modelmgr import entities
 from ..provider.sysprompt import entities as sysprompt_entities
 from ..provider.tools import entities as tools_entities
 from ..platform import adapter as msadapter

pkg/core/stages/build_app.py

@@ -10,7 +10,7 @@ from ...pipeline import pool, controller, stagemgr
 from ...plugin import manager as plugin_mgr
 from ...command import cmdmgr
 from ...provider.session import sessionmgr as llm_session_mgr
-from ...provider.requester import modelmgr as llm_model_mgr
+from ...provider.modelmgr import modelmgr as llm_model_mgr
 from ...provider.sysprompt import sysprompt as llm_prompt_mgr
 from ...provider.tools import toolmgr as llm_tool_mgr
 from ...platform import manager as im_mgr

pkg/pipeline/cntfilter/cntfilter.py

@@ -7,7 +7,7 @@ from ...core import app
 from .. import stage, entities, stagemgr
 from ...core import entities as core_entities
 from ...config import manager as cfg_mgr
-from . import filter, entities as filter_entities
+from . import filter as filter_model, entities as filter_entities
 from .filters import cntignore, banwords, baiduexamine
 
 
@@ -16,20 +16,29 @@ from .filters import cntignore, banwords, baiduexamine
 class ContentFilterStage(stage.PipelineStage):
     """内容过滤阶段"""
 
-    filter_chain: list[filter.ContentFilter]
+    filter_chain: list[filter_model.ContentFilter]
 
     def __init__(self, ap: app.Application):
         self.filter_chain = []
         super().__init__(ap)
 
     async def initialize(self):
-        self.filter_chain.append(cntignore.ContentIgnore(self.ap))
+
+        filters_required = [
+            "content-filter"
+        ]
 
         if self.ap.pipeline_cfg.data['check-sensitive-words']:
-            self.filter_chain.append(banwords.BanWordFilter(self.ap))
-        
+            filters_required.append("ban-word-filter")
+
         if self.ap.pipeline_cfg.data['baidu-cloud-examine']['enable']:
-            self.filter_chain.append(baiduexamine.BaiduCloudExamine(self.ap))
+            filters_required.append("baidu-cloud-examine")
+
+        for filter in filter_model.preregistered_filters:
+            if filter.name in filters_required:
+                self.filter_chain.append(
+                    filter(self.ap)
+                )
 
         for filter in self.filter_chain:
             await filter.initialize()

pkg/pipeline/cntfilter/filter.py

@@ -1,12 +1,42 @@
 # 内容过滤器的抽象类
 from __future__ import annotations
 import abc
+import typing
 
 from ...core import app
 from . import entities
 
 
+preregistered_filters: list[typing.Type[ContentFilter]] = []
+
+
+def filter_class(
+    name: str
+) -> typing.Callable[[typing.Type[ContentFilter]], typing.Type[ContentFilter]]:
+    """内容过滤器类装饰器
+
+    Args:
+        name (str): 过滤器名称
+
+    Returns:
+        typing.Callable[[typing.Type[ContentFilter]], typing.Type[ContentFilter]]: 装饰器
+    """
+    def decorator(cls: typing.Type[ContentFilter]) -> typing.Type[ContentFilter]:
+        assert issubclass(cls, ContentFilter)
+
+        cls.name = name
+
+        preregistered_filters.append(cls)
+
+        return cls
+
+    return decorator
+
+
 class ContentFilter(metaclass=abc.ABCMeta):
+    """内容过滤器抽象类"""
+
+    name: str
 
     ap: app.Application
 

pkg/pipeline/cntfilter/filters/baiduexamine.py

@@ -10,6 +10,7 @@ BAIDU_EXAMINE_URL = "https://aip.baidubce.com/rest/2.0/solution/v1/text_censor/v
 BAIDU_EXAMINE_TOKEN_URL = "https://aip.baidubce.com/oauth/2.0/token"
 
 
+@filter_model.filter_class("baidu-cloud-examine")
 class BaiduCloudExamine(filter_model.ContentFilter):
     """百度云内容审核"""
 

pkg/pipeline/cntfilter/filters/banwords.py

@@ -6,6 +6,7 @@ from .. import entities
 from ....config import manager as cfg_mgr
 
 
+@filter_model.filter_class("ban-word-filter")
 class BanWordFilter(filter_model.ContentFilter):
     """根据内容禁言"""
 

pkg/pipeline/cntfilter/filters/cntignore.py

@@ -5,6 +5,7 @@ from .. import entities
 from .. import filter as filter_model
 
 
+@filter_model.filter_class("content-ignore")
 class ContentIgnore(filter_model.ContentFilter):
     """根据内容忽略消息"""
 

pkg/pipeline/longtext/longtext.py

@@ -45,11 +45,14 @@ class LongTextProcessStage(stage.PipelineStage):
                 self.ap.logger.error("加载字体文件失败({})，更换为转发消息组件以发送长消息，您可以在config.py中调整相关设置。".format(use_font))
 
                 self.ap.platform_cfg.data['long-text-process']['strategy'] = "forward"
-        
-        if config['strategy'] == 'image':
-            self.strategy_impl = image.Text2ImageStrategy(self.ap)
-        elif config['strategy'] == 'forward':
-            self.strategy_impl = forward.ForwardComponentStrategy(self.ap)
+
+        for strategy_cls in strategy.preregistered_strategies:
+            if strategy_cls.name == config['strategy']:
+                self.strategy_impl = strategy_cls(self.ap)
+                break
+        else:
+            raise ValueError(f"未找到名为 {config['strategy']} 的长消息处理策略")
+
         await self.strategy_impl.initialize()
     
     async def process(self, query: core_entities.Query, stage_inst_name: str) -> entities.StageProcessResult:

pkg/pipeline/longtext/strategies/forward.py

@@ -36,6 +36,7 @@ class Forward(MessageComponent):
         return '[聊天记录]'
 
 
+@strategy_model.strategy_class("forward")
 class ForwardComponentStrategy(strategy_model.LongTextStrategy):
 
     async def process(self, message: str, query: core_entities.Query) -> list[MessageComponent]:

pkg/pipeline/longtext/strategies/image.py

@@ -15,6 +15,7 @@ from .. import strategy as strategy_model
 from ....core import entities as core_entities
 
 
+@strategy_model.strategy_class("image")
 class Text2ImageStrategy(strategy_model.LongTextStrategy):
 
     text_render_font: ImageFont.FreeTypeFont

pkg/pipeline/longtext/strategy.py

@@ -9,7 +9,30 @@ from ...core import app
 from ...core import entities as core_entities
 
 
+preregistered_strategies: list[typing.Type[LongTextStrategy]] = []
+
+
+def strategy_class(
+    name: str
+) -> typing.Callable[[typing.Type[LongTextStrategy]], typing.Type[LongTextStrategy]]:
+    def decorator(cls: typing.Type[LongTextStrategy]) -> typing.Type[LongTextStrategy]:
+        assert issubclass(cls, LongTextStrategy)
+
+        cls.name = name
+
+        preregistered_strategies.append(cls)
+
+        return cls
+
+    return decorator
+
+
 class LongTextStrategy(metaclass=abc.ABCMeta):
+    """长文本处理策略抽象类
+    """
+
+    name: str
+
     ap: app.Application
 
     def __init__(self, ap: app.Application):

pkg/pipeline/preproc/preproc.py

@@ -51,28 +51,6 @@ class PreProcessor(stage.PipelineStage):
         query.prompt.messages = event_ctx.event.default_prompt
         query.messages = event_ctx.event.prompt
 
-        # 根据模型max_tokens剪裁
-        max_tokens = min(query.use_model.max_tokens, self.ap.pipeline_cfg.data['submit-messages-tokens'])
-
-        test_messages = query.prompt.messages + query.messages + [query.user_message]
-
-        while await query.use_model.tokenizer.count_token(test_messages, query.use_model) > max_tokens:
-            # 前文都pop完了，还是大于max_tokens，由于prompt和user_messages不能删减，报错
-            if len(query.prompt.messages) == 0:
-                return entities.StageProcessResult(
-                    result_type=entities.ResultType.INTERRUPT,
-                    new_query=query,
-                    user_notice='输入内容过长，请减少情景预设或者输入内容长度',
-                    console_notice='输入内容过长，请减少情景预设或者输入内容长度，或者增大配置文件中的 submit-messages-tokens 项（但不能超过所用模型最大tokens数）'
-                )
-
-            query.messages.pop(0) # pop第一个肯定是role=user的
-            # 继续pop到第二个role=user前一个
-            while len(query.messages) > 0 and query.messages[0].role != 'user':
-                query.messages.pop(0)
-
-            test_messages = query.prompt.messages + query.messages + [query.user_message]
-
         return entities.StageProcessResult(
             result_type=entities.ResultType.CONTINUE,
             new_query=query

pkg/pipeline/process/handlers/chat.py

@@ -21,8 +21,6 @@ class ChatMessageHandler(handler.MessageHandler):
     ) -> typing.AsyncGenerator[entities.StageProcessResult, None]:
         """处理
         """
-        # 取session
-        # 取conversation
         # 调API
         #   生成器
 

pkg/pipeline/ratelimit/algo.py

@@ -1,11 +1,26 @@
 from __future__ import annotations
 import abc
+import typing
 
 from ...core import app
 
 
+preregistered_algos: list[typing.Type[ReteLimitAlgo]] = []
+
+def algo_class(name: str):
+    
+    def decorator(cls: typing.Type[ReteLimitAlgo]) -> typing.Type[ReteLimitAlgo]:
+        cls.name = name
+        preregistered_algos.append(cls)
+        return cls
+    
+    return decorator
+
+
 class ReteLimitAlgo(metaclass=abc.ABCMeta):
     
+    name: str = None
+
     ap: app.Application
 
     def __init__(self, ap: app.Application):

pkg/pipeline/ratelimit/algos/fixedwin.py

@@ -19,6 +19,7 @@ class SessionContainer:
         self.records = {}
 
 
+@algo.algo_class("fixwin")
 class FixedWindowAlgo(algo.ReteLimitAlgo):
 
     containers_lock: asyncio.Lock

pkg/pipeline/ratelimit/ratelimit.py

@@ -16,7 +16,19 @@ class RateLimit(stage.PipelineStage):
     algo: algo.ReteLimitAlgo
 
     async def initialize(self):
-        self.algo = fixedwin.FixedWindowAlgo(self.ap)
+
+        algo_name = self.ap.pipeline_cfg.data['rate-limit']['algo']
+
+        algo_class = None
+
+        for algo_cls in algo.preregistered_algos:
+            if algo_cls.name == algo_name:
+                algo_class = algo_cls
+                break
+        else:
+            raise ValueError(f'未知的限速算法: {algo_name}')
+
+        self.algo = algo_class(self.ap)
         await self.algo.initialize()
 
     async def process(

pkg/pipeline/resprule/resprule.py

@@ -21,15 +21,13 @@ class GroupRespondRuleCheckStage(stage.PipelineStage):
     async def initialize(self):
         """初始化检查器
         """
-        self.rule_matchers = [
-            atbot.AtBotRule(self.ap),
-            prefix.PrefixRule(self.ap),
-            regexp.RegExpRule(self.ap),
-            random.RandomRespRule(self.ap),
-        ]
 
-        for rule_matcher in self.rule_matchers:
-            await rule_matcher.initialize()
+        self.rule_matchers = []
+
+        for rule_matcher in rule.preregisetered_rules:
+            rule_inst = rule_matcher(self.ap)
+            await rule_inst.initialize()
+            self.rule_matchers.append(rule_inst)
 
     async def process(self, query: core_entities.Query, stage_inst_name: str) -> entities.StageProcessResult:
         

pkg/pipeline/resprule/rule.py

@@ -1,5 +1,6 @@
 from __future__ import annotations
 import abc
+import typing
 
 import mirai
 
@@ -7,9 +8,20 @@ from ...core import app, entities as core_entities
 from . import entities
 
 
+preregisetered_rules: list[typing.Type[GroupRespondRule]] = []
+
+def rule_class(name: str):
+    def decorator(cls: typing.Type[GroupRespondRule]) -> typing.Type[GroupRespondRule]:
+        cls.name = name
+        preregisetered_rules.append(cls)
+        return cls
+    return decorator
+
+
 class GroupRespondRule(metaclass=abc.ABCMeta):
     """群组响应规则的抽象类
     """
+    name: str
 
     ap: app.Application
 

pkg/pipeline/resprule/rules/atbot.py

@@ -7,6 +7,7 @@ from .. import entities
 from ....core import entities as core_entities
 
 
+@rule_model.rule_class("at-bot")
 class AtBotRule(rule_model.GroupRespondRule):
 
     async def match(

pkg/pipeline/resprule/rules/prefix.py

@@ -5,6 +5,7 @@ from .. import entities
 from ....core import entities as core_entities
 
 
+@rule_model.rule_class("prefix")
 class PrefixRule(rule_model.GroupRespondRule):
 
     async def match(

pkg/pipeline/resprule/rules/random.py

@@ -7,6 +7,7 @@ from .. import entities
 from ....core import entities as core_entities
 
 
+@rule_model.rule_class("random")
 class RandomRespRule(rule_model.GroupRespondRule):
 
     async def match(

pkg/pipeline/resprule/rules/regexp.py

@@ -7,6 +7,7 @@ from .. import entities
 from ....core import entities as core_entities
 
 
+@rule_model.rule_class("regexp")
 class RegExpRule(rule_model.GroupRespondRule):
 
     async def match(

pkg/platform/adapter.py

@@ -22,6 +22,8 @@ def adapter_class(
 
 
 class MessageSourceAdapter(metaclass=abc.ABCMeta):
+    """消息平台适配器基类"""
+
     name: str
 
     bot_account_id: int
@@ -40,7 +42,7 @@ class MessageSourceAdapter(metaclass=abc.ABCMeta):
         target_id: str,
         message: mirai.MessageChain
     ):
-        """发送消息
+        """主动发送消息
         
         Args:
             target_type (str): 目标类型，`person`或`group`

pkg/platform/manager.py

@@ -163,25 +163,6 @@ class PlatformManager:
             quote_origin=True if self.ap.platform_cfg.data['quote-origin'] and check_quote else False
         )
 
-    # 通知系统管理员
-    # TODO delete
-    # async def notify_admin(self, message: str):
-    #     await self.notify_admin_message_chain(MessageChain([Plain("[bot]{}".format(message))]))
-
-    # async def notify_admin_message_chain(self, message: mirai.MessageChain):
-    #     if self.ap.system_cfg.data['admin-sessions'] != []:
-
-    #         admin_list = []
-    #         for admin in self.ap.system_cfg.data['admin-sessions']:
-    #             admin_list.append(admin)
-            
-    #         for adm in admin_list:
-    #             self.adapter.send_message(
-    #                 adm.split("_")[0],
-    #                 adm.split("_")[1],
-    #                 message
-    #             )
-
     async def run(self):
         try:
             tasks = []

pkg/platform/sources/nakuru.py

@@ -24,6 +24,8 @@ class NakuruProjectMessageConverter(adapter_model.MessageConverter):
             msg_list = message_chain.__root__
         elif type(message_chain) is list:
             msg_list = message_chain
+        elif type(message_chain) is str:
+            msg_list = [mirai.Plain(message_chain)]
         else:
             raise Exception("Unknown message type: " + str(message_chain) + str(type(message_chain)))
         

pkg/platform/sources/qqbotpy.py

@@ -89,6 +89,8 @@ class OfficialMessageConverter(adapter_model.MessageConverter):
             msg_list = message_chain.__root__
         elif type(message_chain) is list:
             msg_list = message_chain
+        elif type(message_chain) is str:
+            msg_list = [mirai.Plain(text=message_chain)]
         else:
             raise Exception("Unknown message type: " + str(message_chain) + str(type(message_chain)))
         

pkg/provider/modelmgr/api.py

@@ -7,9 +7,23 @@ from ...core import app
 from ...core import entities as core_entities
 from .. import entities as llm_entities
 
+
+preregistered_requesters: list[typing.Type[LLMAPIRequester]] = []
+
+def requester_class(name: str):
+
+    def decorator(cls: typing.Type[LLMAPIRequester]) -> typing.Type[LLMAPIRequester]:
+        cls.name = name
+        preregistered_requesters.append(cls)
+        return cls
+
+    return decorator
+
+
 class LLMAPIRequester(metaclass=abc.ABCMeta):
     """LLM API请求器
     """
+    name: str = None
 
     ap: app.Application
 

pkg/provider/modelmgr/apis/chatcmpl.py

@@ -17,6 +17,7 @@ from ... import entities as llm_entities
 from ...tools import entities as tools_entities
 
 
+@api.requester_class("openai-chat-completion")
 class OpenAIChatCompletion(api.LLMAPIRequester):
     """OpenAI ChatCompletion API 请求器"""
 
@@ -133,7 +134,10 @@ class OpenAIChatCompletion(api.LLMAPIRequester):
         except asyncio.TimeoutError:
             raise errors.RequesterError('请求超时')
         except openai.BadRequestError as e:
-            raise errors.RequesterError(f'请求错误: {e.message}')
+            if 'context_length_exceeded' in e.message:
+                raise errors.RequesterError(f'上文过长，请重置会话: {e.message}')
+            else:
+                raise errors.RequesterError(f'请求参数错误: {e.message}')
         except openai.AuthenticationError as e:
             raise errors.RequesterError(f'无效的 api-key: {e.message}')
         except openai.NotFoundError as e:

pkg/provider/modelmgr/entities.py

@@ -5,7 +5,7 @@ import typing
 import pydantic
 
 from . import api
-from . import token, tokenizer
+from . import token
 
 
 class LLMModelInfo(pydantic.BaseModel):
@@ -19,11 +19,7 @@ class LLMModelInfo(pydantic.BaseModel):
 
     requester: api.LLMAPIRequester
 
-    tokenizer: 'tokenizer.LLMTokenizer'
-
     tool_call_supported: typing.Optional[bool] = False
 
-    max_tokens: typing.Optional[int] = 2048
-
     class Config:
         arbitrary_types_allowed = True

pkg/provider/modelmgr/modelmgr.py

@@ -3,9 +3,8 @@ from __future__ import annotations
 from . import entities
 from ...core import app
 
-from .apis import chatcmpl
 from . import token
-from .tokenizers import tiktoken
+from .apis import chatcmpl
 
 
 class ModelManager:
@@ -30,9 +29,7 @@ class ModelManager:
     async def initialize(self):
         openai_chat_completion = chatcmpl.OpenAIChatCompletion(self.ap)
         await openai_chat_completion.initialize()
-        openai_token_mgr = token.TokenManager(self.ap, list(self.ap.provider_cfg.data['openai-config']['api-keys']))
-
-        tiktoken_tokenizer = tiktoken.Tiktoken(self.ap)
+        openai_token_mgr = token.TokenManager("openai", list(self.ap.provider_cfg.data['openai-config']['api-keys']))
 
         model_list = [
             entities.LLMModelInfo(
@@ -40,48 +37,36 @@ class ModelManager:
                 token_mgr=openai_token_mgr,
                 requester=openai_chat_completion,
                 tool_call_supported=True,
-                tokenizer=tiktoken_tokenizer,
-                max_tokens=4096
             ),
             entities.LLMModelInfo(
                 name="gpt-3.5-turbo-1106",
                 token_mgr=openai_token_mgr,
                 requester=openai_chat_completion,
                 tool_call_supported=True,
-                tokenizer=tiktoken_tokenizer,
-                max_tokens=16385
             ),
             entities.LLMModelInfo(
                 name="gpt-3.5-turbo-16k",
                 token_mgr=openai_token_mgr,
                 requester=openai_chat_completion,
                 tool_call_supported=True,
-                tokenizer=tiktoken_tokenizer,
-                max_tokens=16385
             ),
             entities.LLMModelInfo(
                 name="gpt-3.5-turbo-0613",
                 token_mgr=openai_token_mgr,
                 requester=openai_chat_completion,
                 tool_call_supported=True,
-                tokenizer=tiktoken_tokenizer,
-                max_tokens=4096
             ),
             entities.LLMModelInfo(
                 name="gpt-3.5-turbo-16k-0613",
                 token_mgr=openai_token_mgr,
                 requester=openai_chat_completion,
                 tool_call_supported=True,
-                tokenizer=tiktoken_tokenizer,
-                max_tokens=16385
             ),
             entities.LLMModelInfo(
                 name="gpt-3.5-turbo-0301",
                 token_mgr=openai_token_mgr,
                 requester=openai_chat_completion,
                 tool_call_supported=True,
-                tokenizer=tiktoken_tokenizer,
-                max_tokens=4096
             )
         ]
 
@@ -93,64 +78,48 @@ class ModelManager:
                 token_mgr=openai_token_mgr,
                 requester=openai_chat_completion,
                 tool_call_supported=True,
-                tokenizer=tiktoken_tokenizer,
-                max_tokens=128000
             ),
             entities.LLMModelInfo(
                 name="gpt-4-turbo-preview",
                 token_mgr=openai_token_mgr,
                 requester=openai_chat_completion,
                 tool_call_supported=True,
-                tokenizer=tiktoken_tokenizer,
-                max_tokens=128000
             ),
             entities.LLMModelInfo(
                 name="gpt-4-1106-preview",
                 token_mgr=openai_token_mgr,
                 requester=openai_chat_completion,
                 tool_call_supported=True,
-                tokenizer=tiktoken_tokenizer,
-                max_tokens=128000
             ),
             entities.LLMModelInfo(
                 name="gpt-4-vision-preview",
                 token_mgr=openai_token_mgr,
                 requester=openai_chat_completion,
                 tool_call_supported=True,
-                tokenizer=tiktoken_tokenizer,
-                max_tokens=128000
             ),
             entities.LLMModelInfo(
                 name="gpt-4",
                 token_mgr=openai_token_mgr,
                 requester=openai_chat_completion,
                 tool_call_supported=True,
-                tokenizer=tiktoken_tokenizer,
-                max_tokens=8192
             ),
             entities.LLMModelInfo(
                 name="gpt-4-0613",
                 token_mgr=openai_token_mgr,
                 requester=openai_chat_completion,
                 tool_call_supported=True,
-                tokenizer=tiktoken_tokenizer,
-                max_tokens=8192
             ),
             entities.LLMModelInfo(
                 name="gpt-4-32k",
                 token_mgr=openai_token_mgr,
                 requester=openai_chat_completion,
                 tool_call_supported=True,
-                tokenizer=tiktoken_tokenizer,
-                max_tokens=32768
             ),
             entities.LLMModelInfo(
                 name="gpt-4-32k-0613",
                 token_mgr=openai_token_mgr,
                 requester=openai_chat_completion,
                 tool_call_supported=True,
-                tokenizer=tiktoken_tokenizer,
-                max_tokens=32768
             )
         ]
 
@@ -163,8 +132,6 @@ class ModelManager:
                 token_mgr=openai_token_mgr,
                 requester=openai_chat_completion,
                 tool_call_supported=False,
-                tokenizer=tiktoken_tokenizer,
-                max_tokens=8192
             ),
             entities.LLMModelInfo(
                 name="OneAPI/chatglm_pro",
@@ -172,8 +139,6 @@ class ModelManager:
                 token_mgr=openai_token_mgr,
                 requester=openai_chat_completion,
                 tool_call_supported=False,
-                tokenizer=tiktoken_tokenizer,
-                max_tokens=128000
             ),
             entities.LLMModelInfo(
                 name="OneAPI/chatglm_std",
@@ -181,8 +146,6 @@ class ModelManager:
                 token_mgr=openai_token_mgr,
                 requester=openai_chat_completion,
                 tool_call_supported=False,
-                tokenizer=tiktoken_tokenizer,
-                max_tokens=128000
             ),
             entities.LLMModelInfo(
                 name="OneAPI/chatglm_lite",
@@ -190,8 +153,6 @@ class ModelManager:
                 token_mgr=openai_token_mgr,
                 requester=openai_chat_completion,
                 tool_call_supported=False,
-                tokenizer=tiktoken_tokenizer,
-                max_tokens=128000
             ),
             entities.LLMModelInfo(
                 name="OneAPI/qwen-v1",
@@ -199,8 +160,6 @@ class ModelManager:
                 token_mgr=openai_token_mgr,
                 requester=openai_chat_completion,
                 tool_call_supported=False,
-                tokenizer=tiktoken_tokenizer,
-                max_tokens=6000
             ),
             entities.LLMModelInfo(
                 name="OneAPI/qwen-plus-v1",
@@ -208,8 +167,6 @@ class ModelManager:
                 token_mgr=openai_token_mgr,
                 requester=openai_chat_completion,
                 tool_call_supported=False,
-                tokenizer=tiktoken_tokenizer,
-                max_tokens=30000
             ),
             entities.LLMModelInfo(
                 name="OneAPI/ERNIE-Bot",
@@ -217,8 +174,6 @@ class ModelManager:
                 token_mgr=openai_token_mgr,
                 requester=openai_chat_completion,
                 tool_call_supported=False,
-                tokenizer=tiktoken_tokenizer,
-                max_tokens=2000
             ),
             entities.LLMModelInfo(
                 name="OneAPI/ERNIE-Bot-turbo",
@@ -226,8 +181,6 @@ class ModelManager:
                 token_mgr=openai_token_mgr,
                 requester=openai_chat_completion,
                 tool_call_supported=False,
-                tokenizer=tiktoken_tokenizer,
-                max_tokens=7000
             ),
             entities.LLMModelInfo(
                 name="OneAPI/gemini-pro",
@@ -235,8 +188,6 @@ class ModelManager:
                 token_mgr=openai_token_mgr,
                 requester=openai_chat_completion,
                 tool_call_supported=False,
-                tokenizer=tiktoken_tokenizer,
-                max_tokens=30720
             ),
         ]
 

pkg/provider/requester/tokenizer.py

@@ -1,30 +0,0 @@
-from __future__ import annotations
-
-import abc
-import typing
-
-from ...core import app
-from .. import entities as llm_entities
-from . import entities
-
-
-class LLMTokenizer(metaclass=abc.ABCMeta):
-    """LLM分词器抽象类"""
-
-    ap: app.Application
-
-    def __init__(self, ap: app.Application):
-        self.ap = ap
-
-    async def initialize(self):
-        """初始化分词器
-        """
-        pass
-
-    @abc.abstractmethod
-    async def count_token(
-        self,
-        messages: list[llm_entities.Message],
-        model: entities.LLMModelInfo
-    ) -> int:
-        pass

pkg/provider/requester/tokenizers/tiktoken.py

@@ -1,30 +0,0 @@
-from __future__ import annotations
-
-import tiktoken
-
-from .. import tokenizer
-from ... import entities as llm_entities
-from .. import entities
-
-
-class Tiktoken(tokenizer.LLMTokenizer):
-    """TikToken分词器
-    """
-
-    async def count_token(
-        self,
-        messages: list[llm_entities.Message],
-        model: entities.LLMModelInfo
-    ) -> int:
-        try:
-            encoding = tiktoken.encoding_for_model(model.name)
-        except KeyError:
-            # print("Warning: model not found. Using cl100k_base encoding.")
-            encoding = tiktoken.get_encoding("cl100k_base")
-
-        num_tokens = 0
-        for message in messages:
-                num_tokens += len(encoding.encode(message.role))
-                num_tokens += len(encoding.encode(message.content if message.content is not None else ''))
-        num_tokens += 3  # every reply is primed with <|start|>assistant<|message|>
-        return num_tokens

pkg/provider/sysprompt/loader.py

@@ -1,13 +1,27 @@
 from __future__ import annotations
 import abc
+import typing
 
 from ...core import app
 from . import entities
 
 
+preregistered_loaders: list[typing.Type[PromptLoader]] = []
+
+def loader_class(name: str):
+
+    def decorator(cls: typing.Type[PromptLoader]) -> typing.Type[PromptLoader]:
+        cls.name = name
+        preregistered_loaders.append(cls)
+        return cls
+    
+    return decorator
+
+
 class PromptLoader(metaclass=abc.ABCMeta):
     """Prompt加载器抽象类
     """
+    name: str
 
     ap: app.Application
 

pkg/provider/sysprompt/loaders/scenario.py

@@ -8,6 +8,7 @@ from .. import entities
 from ....provider import entities as llm_entities
 
 
+@loader.loader_class("full_scenario")
 class ScenarioPromptLoader(loader.PromptLoader):
     """加载scenario目录下的json"""
 

pkg/provider/sysprompt/loaders/single.py

@@ -6,6 +6,7 @@ from .. import entities
 from ....provider import entities as llm_entities
 
 
+@loader.loader_class("normal")
 class SingleSystemPromptLoader(loader.PromptLoader):
     """配置文件中的单条system prompt的prompt加载器
     """

pkg/provider/sysprompt/sysprompt.py

@@ -20,14 +20,18 @@ class PromptManager:
 
     async def initialize(self):
 
-        loader_map = {
-            "normal": single.SingleSystemPromptLoader,
-            "full_scenario": scenario.ScenarioPromptLoader
-        }
+        mode_name = self.ap.provider_cfg.data['prompt-mode']
 
-        loader_cls = loader_map[self.ap.provider_cfg.data['prompt-mode']]
+        loader_class = None
 
-        self.loader_inst: loader.PromptLoader = loader_cls(self.ap)
+        for loader_cls in loader.preregistered_loaders:
+            if loader_cls.name == mode_name:
+                loader_class = loader_cls
+                break
+        else:
+            raise ValueError(f'未知的 Prompt 加载器: {mode_name}')
+
+        self.loader_inst: loader.PromptLoader = loader_class(self.ap)
 
         await self.loader_inst.initialize()
         await self.loader_inst.load()