diff --git a/pkg/config/migrations/m2_openai_config_migration.py b/pkg/config/migrations/m2_openai_config_migration.py new file mode 100644 index 00000000..2f2553ef --- /dev/null +++ b/pkg/config/migrations/m2_openai_config_migration.py @@ -0,0 +1,47 @@ +from __future__ import annotations + +from .. import migration + + +@migration.migration_class("openai-config-migration", 2) +class OpenAIConfigMigration(migration.Migration): + """OpenAI配置迁移 + """ + + async def need_migrate(self) -> bool: + """判断当前环境是否需要运行此迁移 + """ + return 'openai-config' in self.ap.provider_cfg.data + + async def run(self): + """执行迁移 + """ + old_openai_config = self.ap.provider_cfg.data['openai-config'].copy() + + if 'keys' not in self.ap.provider_cfg.data: + self.ap.provider_cfg.data['keys'] = {} + + if 'openai' not in self.ap.provider_cfg.data['keys']: + self.ap.provider_cfg.data['keys']['openai'] = [] + + self.ap.provider_cfg.data['keys']['openai'] = old_openai_config['api-keys'] + + self.ap.provider_cfg.data['model'] = old_openai_config['chat-completions-params']['model'] + + del old_openai_config['chat-completions-params']['model'] + + if 'requester' not in self.ap.provider_cfg.data: + self.ap.provider_cfg.data['requester'] = {} + + if 'openai-chat-completions' not in self.ap.provider_cfg.data['requester']: + self.ap.provider_cfg.data['requester']['openai-chat-completions'] = {} + + self.ap.provider_cfg.data['requester']['openai-chat-completions'] = { + 'base-url': old_openai_config['base_url'], + 'args': old_openai_config['chat-completions-params'], + 'timeout': old_openai_config['request-timeout'], + } + + del self.ap.provider_cfg.data['openai-config'] + + await self.ap.provider_cfg.dump_config() \ No newline at end of file diff --git a/pkg/core/app.py b/pkg/core/app.py index e1df54ec..18cbeffa 100644 --- a/pkg/core/app.py +++ b/pkg/core/app.py @@ -53,6 +53,8 @@ class Application: plugin_setting_meta: config_mgr.ConfigManager = None + llm_models_meta: config_mgr.ConfigManager = None + # ========================= ctr_mgr: center_mgr.V2CenterAPI = None diff --git a/pkg/core/stages/load_config.py b/pkg/core/stages/load_config.py index 7f83dbad..9e61c1cb 100644 --- a/pkg/core/stages/load_config.py +++ b/pkg/core/stages/load_config.py @@ -18,7 +18,6 @@ class LoadConfigStage(stage.BootingStage): ap.provider_cfg = await config.load_json_config("data/config/provider.json", "templates/provider.json") ap.system_cfg = await config.load_json_config("data/config/system.json", "templates/system.json") - ap.plugin_setting_meta = await config.load_json_config("plugins/plugins.json", "templates/plugin-settings.json") await ap.plugin_setting_meta.dump_config() @@ -27,3 +26,6 @@ class LoadConfigStage(stage.BootingStage): ap.adapter_qq_botpy_meta = await config.load_json_config("data/metadata/adapter-qq-botpy.json", "templates/metadata/adapter-qq-botpy.json") await ap.adapter_qq_botpy_meta.dump_config() + + ap.llm_models_meta = await config.load_json_config("data/metadata/llm-models.json", "templates/metadata/llm-models.json") + await ap.llm_models_meta.dump_config() diff --git a/pkg/core/stages/migrate.py b/pkg/core/stages/migrate.py index 5e84b182..edf3cba6 100644 --- a/pkg/core/stages/migrate.py +++ b/pkg/core/stages/migrate.py @@ -4,7 +4,7 @@ import importlib from .. import stage, app from ...config import migration -from ...config.migrations import m1_sensitive_word_migration +from ...config.migrations import m1_sensitive_word_migration, m2_openai_config_migration @stage.stage_class("MigrationStage") diff --git a/pkg/provider/modelmgr/apis/chatcmpl.py b/pkg/provider/modelmgr/apis/chatcmpl.py index 4965acf7..c8d1a812 100644 --- a/pkg/provider/modelmgr/apis/chatcmpl.py +++ b/pkg/provider/modelmgr/apis/chatcmpl.py @@ -17,8 +17,8 @@ from ... import entities as llm_entities from ...tools import entities as tools_entities -@api.requester_class("openai-chat-completion") -class OpenAIChatCompletion(api.LLMAPIRequester): +@api.requester_class("openai-chat-completions") +class OpenAIChatCompletions(api.LLMAPIRequester): """OpenAI ChatCompletion API 请求器""" client: openai.AsyncClient @@ -26,8 +26,8 @@ class OpenAIChatCompletion(api.LLMAPIRequester): async def initialize(self): self.client = openai.AsyncClient( api_key="", - base_url=self.ap.provider_cfg.data['openai-config']['base_url'], - timeout=self.ap.provider_cfg.data['openai-config']['request-timeout'], + base_url=self.ap.provider_cfg.data['requester']['openai-chat-completions']['base-url'], + timeout=self.ap.provider_cfg.data['requester']['openai-chat-completions']['timeout'], http_client=httpx.AsyncClient( proxies=self.ap.proxy_mgr.get_forward_proxies() ) @@ -58,7 +58,7 @@ class OpenAIChatCompletion(api.LLMAPIRequester): ) -> llm_entities.Message: self.client.api_key = use_model.token_mgr.get_token() - args = self.ap.provider_cfg.data['openai-config']['chat-completions-params'].copy() + args = self.ap.provider_cfg.data['requester']['openai-chat-completions']['args'].copy() args["model"] = use_model.name if use_model.model_name is None else use_model.model_name if use_model.tool_call_supported: diff --git a/pkg/provider/modelmgr/modelmgr.py b/pkg/provider/modelmgr/modelmgr.py index a91c3110..f5b130c7 100644 --- a/pkg/provider/modelmgr/modelmgr.py +++ b/pkg/provider/modelmgr/modelmgr.py @@ -1,11 +1,15 @@ from __future__ import annotations +import aiohttp + from . import entities from ...core import app -from . import token +from . import token, api from .apis import chatcmpl +FETCH_MODEL_LIST_URL = "https://api.qchatgpt.rockchin.top/api/v2/fetch/model_list" + class ModelManager: """模型管理器""" @@ -13,10 +17,16 @@ class ModelManager: ap: app.Application model_list: list[entities.LLMModelInfo] + + requesters: dict[str, api.LLMAPIRequester] + + token_mgrs: dict[str, token.TokenManager] def __init__(self, ap: app.Application): self.ap = ap self.model_list = [] + self.requesters = {} + self.token_mgrs = {} async def get_model_by_name(self, name: str) -> entities.LLMModelInfo: """通过名称获取模型 @@ -24,171 +34,72 @@ class ModelManager: for model in self.model_list: if model.name == name: return model - raise ValueError(f"不支持模型: {name} , 请检查配置文件") - + raise ValueError(f"无法确定模型 {name} 的信息，请在元数据中配置") + async def initialize(self): - openai_chat_completion = chatcmpl.OpenAIChatCompletion(self.ap) - await openai_chat_completion.initialize() - openai_token_mgr = token.TokenManager("openai", list(self.ap.provider_cfg.data['openai-config']['api-keys'])) + + # 初始化token_mgr, requester + for k, v in self.ap.provider_cfg.data['keys'].items(): + self.token_mgrs[k] = token.TokenManager(k, v) - model_list = [ - entities.LLMModelInfo( - name="gpt-3.5-turbo", - token_mgr=openai_token_mgr, - requester=openai_chat_completion, - tool_call_supported=True, - ), - entities.LLMModelInfo( - name="gpt-3.5-turbo-1106", - token_mgr=openai_token_mgr, - requester=openai_chat_completion, - tool_call_supported=True, - ), - entities.LLMModelInfo( - name="gpt-3.5-turbo-16k", - token_mgr=openai_token_mgr, - requester=openai_chat_completion, - tool_call_supported=True, - ), - entities.LLMModelInfo( - name="gpt-3.5-turbo-0613", - token_mgr=openai_token_mgr, - requester=openai_chat_completion, - tool_call_supported=True, - ), - entities.LLMModelInfo( - name="gpt-3.5-turbo-16k-0613", - token_mgr=openai_token_mgr, - requester=openai_chat_completion, - tool_call_supported=True, - ), - entities.LLMModelInfo( - name="gpt-3.5-turbo-0301", - token_mgr=openai_token_mgr, - requester=openai_chat_completion, - tool_call_supported=True, - ) - ] + for api_cls in api.preregistered_requesters: + api_inst = api_cls(self.ap) + await api_inst.initialize() + self.requesters[api_inst.name] = api_inst - self.model_list.extend(model_list) + # 尝试从api获取最新的模型信息 + try: + async with aiohttp.ClientSession() as session: + async with session.request( + method="GET", + url=FETCH_MODEL_LIST_URL, + ) as resp: + model_list = (await resp.json())['data']['list'] - gpt4_model_list = [ - entities.LLMModelInfo( - name="gpt-4-0125-preview", - token_mgr=openai_token_mgr, - requester=openai_chat_completion, - tool_call_supported=True, - ), - entities.LLMModelInfo( - name="gpt-4-turbo-preview", - token_mgr=openai_token_mgr, - requester=openai_chat_completion, - tool_call_supported=True, - ), - entities.LLMModelInfo( - name="gpt-4-1106-preview", - token_mgr=openai_token_mgr, - requester=openai_chat_completion, - tool_call_supported=True, - ), - entities.LLMModelInfo( - name="gpt-4-vision-preview", - token_mgr=openai_token_mgr, - requester=openai_chat_completion, - tool_call_supported=True, - ), - entities.LLMModelInfo( - name="gpt-4", - token_mgr=openai_token_mgr, - requester=openai_chat_completion, - tool_call_supported=True, - ), - entities.LLMModelInfo( - name="gpt-4-0613", - token_mgr=openai_token_mgr, - requester=openai_chat_completion, - tool_call_supported=True, - ), - entities.LLMModelInfo( - name="gpt-4-32k", - token_mgr=openai_token_mgr, - requester=openai_chat_completion, - tool_call_supported=True, - ), - entities.LLMModelInfo( - name="gpt-4-32k-0613", - token_mgr=openai_token_mgr, - requester=openai_chat_completion, - tool_call_supported=True, - ) - ] + for model in model_list: - self.model_list.extend(gpt4_model_list) + for index, local_model in enumerate(self.ap.llm_models_meta.data['list']): + if model['name'] == local_model['name']: + self.ap.llm_models_meta.data['list'][index] = model + break + else: + self.ap.llm_models_meta.data['list'].append(model) - one_api_model_list = [ - entities.LLMModelInfo( - name="OneAPI/SparkDesk", - model_name='SparkDesk', - token_mgr=openai_token_mgr, - requester=openai_chat_completion, - tool_call_supported=False, - ), - entities.LLMModelInfo( - name="OneAPI/chatglm_pro", - model_name='chatglm_pro', - token_mgr=openai_token_mgr, - requester=openai_chat_completion, - tool_call_supported=False, - ), - entities.LLMModelInfo( - name="OneAPI/chatglm_std", - model_name='chatglm_std', - token_mgr=openai_token_mgr, - requester=openai_chat_completion, - tool_call_supported=False, - ), - entities.LLMModelInfo( - name="OneAPI/chatglm_lite", - model_name='chatglm_lite', - token_mgr=openai_token_mgr, - requester=openai_chat_completion, - tool_call_supported=False, - ), - entities.LLMModelInfo( - name="OneAPI/qwen-v1", - model_name='qwen-v1', - token_mgr=openai_token_mgr, - requester=openai_chat_completion, - tool_call_supported=False, - ), - entities.LLMModelInfo( - name="OneAPI/qwen-plus-v1", - model_name='qwen-plus-v1', - token_mgr=openai_token_mgr, - requester=openai_chat_completion, - tool_call_supported=False, - ), - entities.LLMModelInfo( - name="OneAPI/ERNIE-Bot", - model_name='ERNIE-Bot', - token_mgr=openai_token_mgr, - requester=openai_chat_completion, - tool_call_supported=False, - ), - entities.LLMModelInfo( - name="OneAPI/ERNIE-Bot-turbo", - model_name='ERNIE-Bot-turbo', - token_mgr=openai_token_mgr, - requester=openai_chat_completion, - tool_call_supported=False, - ), - entities.LLMModelInfo( - name="OneAPI/gemini-pro", - model_name='gemini-pro', - token_mgr=openai_token_mgr, - requester=openai_chat_completion, - tool_call_supported=False, - ), - ] + await self.ap.llm_models_meta.dump_config() - self.model_list.extend(one_api_model_list) + except Exception as e: + self.ap.logger.debug(f'获取最新模型列表失败: {e}') + + default_model_info: entities.LLMModelInfo = None + + for model in self.ap.llm_models_meta.data['list']: + if model['name'] == 'default': + default_model_info = entities.LLMModelInfo( + name=model['name'], + model_name=None, + token_mgr=self.token_mgrs[model['token_mgr']], + requester=self.requesters[model['requester']], + tool_call_supported=model['tool_call_supported'] + ) + break + + for model in self.ap.llm_models_meta.data['list']: + + try: + + model_name = model.get('model_name', default_model_info.model_name) + token_mgr = self.token_mgrs[model['token_mgr']] if 'token_mgr' in model else default_model_info.token_mgr + requester = self.requesters[model['requester']] if 'requester' in model else default_model_info.requester + tool_call_supported = model.get('tool_call_supported', default_model_info.tool_call_supported) + + model_info = entities.LLMModelInfo( + name=model['name'], + model_name=model_name, + token_mgr=token_mgr, + requester=requester, + tool_call_supported=tool_call_supported + ) + self.model_list.append(model_info) + + except Exception as e: + self.ap.logger.error(f"初始化模型 {model['name']} 失败: {e} ,请检查配置文件") diff --git a/pkg/provider/session/sessionmgr.py b/pkg/provider/session/sessionmgr.py index 4ffec020..9130cbe2 100644 --- a/pkg/provider/session/sessionmgr.py +++ b/pkg/provider/session/sessionmgr.py @@ -50,7 +50,7 @@ class SessionManager: conversation = core_entities.Conversation( prompt=await self.ap.prompt_mgr.get_prompt(session.use_prompt_name), messages=[], - use_model=await self.ap.model_mgr.get_model_by_name(self.ap.provider_cfg.data['openai-config']['chat-completions-params']['model']), + use_model=await self.ap.model_mgr.get_model_by_name(self.ap.provider_cfg.data['model']), use_funcs=await self.ap.tool_mgr.get_all_functions(), ) session.conversations.append(conversation) diff --git a/templates/metadata/llm-models.json b/templates/metadata/llm-models.json new file mode 100644 index 00000000..8775e665 --- /dev/null +++ b/templates/metadata/llm-models.json @@ -0,0 +1,34 @@ +{ + "list": [ + { + "name": "default", + "requester": "openai-chat-completions", + "token_mgr": "openai", + "tool_call_supported": false + }, + { + "name": "gpt-3.5-turbo", + "tool_call_supported": true + }, + { + "name": "gpt-4", + "tool_call_supported": true + }, + { + "name": "gpt-4-turbo-preview", + "tool_call_supported": true + }, + { + "name": "gpt-4-32k", + "tool_call_supported": true + }, + { + "model_name": "SparkDesk", + "name": "OneAPI/SparkDesk" + }, + { + "model_name": "gemini-pro", + "name": "OneAPI/gemini-pro" + } + ] +} \ No newline at end of file diff --git a/templates/provider.json b/templates/provider.json index 23360277..d1a68d77 100644 --- a/templates/provider.json +++ b/templates/provider.json @@ -1,17 +1,20 @@ { "enable-chat": true, - "openai-config": { - "api-keys": [ + "keys": { + "openai": [ "sk-1234567890" - ], - "base_url": "https://api.openai.com/v1", - "chat-completions-params": { - "model": "gpt-3.5-turbo" - }, - "request-timeout": 120 + ] }, + "requester": { + "openai-chat-completions": { + "base-url": "https://api.openai.com/v1", + "args": {}, + "timeout": 120 + } + }, + "model": "gpt-3.5-turbo", "prompt-mode": "normal", "prompt": { - "default": "如果用户之后想获取帮助，请你说”输入!help获取帮助“。" + "default": "" } } \ No newline at end of file