From 670a94b2fb9704ec78a8767fdb5cc2f83ac08cff Mon Sep 17 00:00:00 2001 From: zihanjian Date: Thu, 25 Sep 2025 15:56:48 +0800 Subject: [PATCH] =?UTF-8?q?=E5=8F=8C=E6=A8=A1=E5=BC=8F?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- config.yaml.template | 14 +++-- robot.py | 133 +++++++++++++++++++++++++++++++++++++------ 2 files changed, 125 insertions(+), 22 deletions(-) diff --git a/config.yaml.template b/config.yaml.template index 3f440b3..3426136 100644 --- a/config.yaml.template +++ b/config.yaml.template @@ -92,7 +92,8 @@ weather: # -----天气提醒配置这行不填----- chatgpt: # -----chatgpt配置这行不填----- key: # 填写你 ChatGPT 的 key api: https://api.openai.com/v1 # 如果你不知道这是干嘛的，就不要改 - model: gpt-3.5-turbo # 可选：gpt-3.5-turbo、gpt-4、gpt-4-turbo、gpt-4.1-mini、o4-mini + model_flash: gpt-3.5-turbo # 快速回复模型（可选） + model_reasoning: gpt-3.5-turbo # 深度思考模型（可选） proxy: # 如果你在国内，你可能需要魔法，大概长这样：http://域名或者IP地址:端口号 prompt: 你是智能聊天机器人，你叫 wcferry # 根据需要对角色进行设定 max_history_messages: 20 # <--- 添加这一行，设置 ChatGPT 最多回顾 20 条历史消息 @@ -101,7 +102,8 @@ deepseek: # -----deepseek配置这行不填----- #思维链相关功能默认关闭，开启后会增加响应时间和消耗更多的token key: # 填写你的 DeepSeek API Key API Key的格式为sk-xxxxxxxxxxxxxxx api: https://api.deepseek.com # DeepSeek API 地址 - model: deepseek-chat # 可选: deepseek-chat (DeepSeek-V3), deepseek-reasoner (DeepSeek-R1) + model_flash: deepseek-chat # 快速回复模型 + model_reasoning: deepseek-reasoner # 深度思考模型 prompt: 你是智能聊天机器人，你叫 DeepSeek 助手 # 根据需要对角色进行设定 enable_reasoning: false # 是否启用思维链功能，仅在使用 deepseek-reasoner 模型时有效 show_reasoning: false # 是否在回复中显示思维过程，仅在启用思维链功能时有效 @@ -109,7 +111,8 @@ deepseek: # -----deepseek配置这行不填----- gemini: # -----gemini配置----- api_key: "YOUR_GOOGLE_API_KEY" # 必须 - model_name: "gemini-1.5-pro-latest" # 可选, 默认是 "gemini-1.5-pro-latest" + model_flash: "gemini-1.5-pro-latest" # 快速回复模型（可选） + model_reasoning: "gemini-1.5-pro-latest" # 深度思考模型（可选） prompt: "你是一个AI助手，请用通俗易懂的语言回答用户的问题。" # 可选 proxy: "http://127.0.0.1:7890" # 可选, 代理地址 max_history_messages: 20 # 可选, 对话历史长度 @@ -137,7 +140,8 @@ perplexity: # -----perplexity配置这行不填----- key: # 填写你的Perplexity API Key api: https://api.perplexity.ai # API地址 proxy: # 如果你在国内，你可能需要魔法，大概长这样：http://域名或者IP地址:端口号 - model: mixtral-8x7b-instruct # 可选模型包括：sonar-small-chat, sonar-medium-chat, sonar-pro, mixtral-8x7b-instruct等 + model_flash: mixtral-8x7b-instruct # 快速回复模型（可选） + model_reasoning: mixtral-8x7b-instruct # 深度思考模型（可选） prompt: 你是Perplexity AI助手，请用专业、准确、有帮助的方式回答问题 # 角色设定 trigger_keyword: ask # 触发Perplexity服务的前置词 allow_all: false # 是否允许所有群聊和用户使用Perplexity，设为true时忽略下面的白名单配置 @@ -146,4 +150,4 @@ perplexity: # -----perplexity配置这行不填----- ai_router: # -----AI路由器配置----- enable: true # 是否启用AI路由功能 - allowed_groups: [] # 允许使用AI路由的群聊ID列表，例如：["123456789@chatroom", "123456789@chatroom"] \ No newline at end of file + allowed_groups: [] # 允许使用AI路由的群聊ID列表，例如：["123456789@chatroom", "123456789@chatroom"] diff --git a/robot.py b/robot.py index 655630c..e48ae22 100644 --- a/robot.py +++ b/robot.py @@ -9,6 +9,7 @@ from threading import Thread import os import random import shutil +import copy from image import AliyunImage, GeminiImage from image.img_manager import ImageGenerationManager @@ -65,69 +66,130 @@ class Robot(Job): self.xml_processor = XmlProcessor(self.LOG) self.chat_models = {} + self.reasoning_chat_models = {} self.LOG.info("开始初始化各种AI模型...") # 初始化ChatGPT if ChatGPT.value_check(self.config.CHATGPT): try: - # 传入 message_summary 和 wxid + chatgpt_flash_conf = copy.deepcopy(self.config.CHATGPT) + flash_model_name = chatgpt_flash_conf.get("model_flash", "gpt-3.5-turbo") + chatgpt_flash_conf["model"] = flash_model_name self.chat_models[ChatType.CHATGPT.value] = ChatGPT( - self.config.CHATGPT, + chatgpt_flash_conf, message_summary_instance=self.message_summary, bot_wxid=self.wxid ) - self.LOG.info(f"已加载 ChatGPT 模型") + self.LOG.info(f"已加载 ChatGPT 模型: {flash_model_name}") + + reasoning_model_name = self.config.CHATGPT.get("model_reasoning") + if reasoning_model_name and reasoning_model_name != flash_model_name: + chatgpt_reason_conf = copy.deepcopy(self.config.CHATGPT) + chatgpt_reason_conf["model"] = reasoning_model_name + self.reasoning_chat_models[ChatType.CHATGPT.value] = ChatGPT( + chatgpt_reason_conf, + message_summary_instance=self.message_summary, + bot_wxid=self.wxid + ) + self.LOG.info(f"已加载 ChatGPT 推理模型: {reasoning_model_name}") except Exception as e: self.LOG.error(f"初始化 ChatGPT 模型时出错: {str(e)}") # 初始化DeepSeek if DeepSeek.value_check(self.config.DEEPSEEK): try: - # 传入 message_summary 和 wxid - self.chat_models[ChatType.DEEPSEEK.value] = DeepSeek( - self.config.DEEPSEEK, - message_summary_instance=self.message_summary, - bot_wxid=self.wxid - ) - self.LOG.info(f"已加载 DeepSeek 模型") + deepseek_flash_conf = copy.deepcopy(self.config.DEEPSEEK) + flash_model_name = deepseek_flash_conf.get("model_flash", "deepseek-chat") + deepseek_flash_conf["model"] = flash_model_name + self.chat_models[ChatType.DEEPSEEK.value] = DeepSeek( + deepseek_flash_conf, + message_summary_instance=self.message_summary, + bot_wxid=self.wxid + ) + self.LOG.info(f"已加载 DeepSeek 模型: {flash_model_name}") + + reasoning_model_name = self.config.DEEPSEEK.get("model_reasoning") + if not reasoning_model_name and flash_model_name != "deepseek-reasoner": + reasoning_model_name = "deepseek-reasoner" + + if reasoning_model_name and reasoning_model_name != flash_model_name: + deepseek_reason_conf = copy.deepcopy(self.config.DEEPSEEK) + deepseek_reason_conf["model"] = reasoning_model_name + self.reasoning_chat_models[ChatType.DEEPSEEK.value] = DeepSeek( + deepseek_reason_conf, + message_summary_instance=self.message_summary, + bot_wxid=self.wxid + ) + self.LOG.info(f"已加载 DeepSeek 推理模型: {reasoning_model_name}") except Exception as e: - self.LOG.error(f"初始化 DeepSeek 模型时出错: {str(e)}") + self.LOG.error(f"初始化 DeepSeek 模型时出错: {str(e)}") # 初始化Gemini if Gemini.value_check(self.config.GEMINI): try: - # 传入 message_summary 和 wxid + gemini_flash_conf = copy.deepcopy(self.config.GEMINI) + flash_model_name = gemini_flash_conf.get("model_flash", Gemini.DEFAULT_MODEL) + gemini_flash_conf["model_name"] = flash_model_name self.chat_models[ChatType.GEMINI.value] = Gemini( - self.config.GEMINI, + gemini_flash_conf, message_summary_instance=self.message_summary, bot_wxid=self.wxid ) - self.LOG.info(f"已加载 Gemini 模型") + self.LOG.info(f"已加载 Gemini 模型: {flash_model_name}") + + reasoning_model_name = self.config.GEMINI.get("model_reasoning") + if reasoning_model_name and reasoning_model_name != flash_model_name: + gemini_reason_conf = copy.deepcopy(self.config.GEMINI) + gemini_reason_conf["model_name"] = reasoning_model_name + self.reasoning_chat_models[ChatType.GEMINI.value] = Gemini( + gemini_reason_conf, + message_summary_instance=self.message_summary, + bot_wxid=self.wxid + ) + self.LOG.info(f"已加载 Gemini 推理模型: {reasoning_model_name}") except Exception as e: self.LOG.error(f"初始化 Gemini 模型时出错: {str(e)}") # 初始化Perplexity if Perplexity.value_check(self.config.PERPLEXITY): - self.chat_models[ChatType.PERPLEXITY.value] = Perplexity(self.config.PERPLEXITY) - self.perplexity = self.chat_models[ChatType.PERPLEXITY.value] # 单独保存一个引用用于特殊处理 - self.LOG.info(f"已加载 Perplexity 模型") + try: + perplexity_flash_conf = copy.deepcopy(self.config.PERPLEXITY) + flash_model_name = perplexity_flash_conf.get("model_flash", "sonar") + perplexity_flash_conf["model"] = flash_model_name + self.chat_models[ChatType.PERPLEXITY.value] = Perplexity(perplexity_flash_conf) + self.perplexity = self.chat_models[ChatType.PERPLEXITY.value] # 单独保存一个引用用于特殊处理 + self.LOG.info(f"已加载 Perplexity 模型: {flash_model_name}") + + reasoning_model_name = self.config.PERPLEXITY.get("model_reasoning") + if reasoning_model_name and reasoning_model_name != flash_model_name: + perplexity_reason_conf = copy.deepcopy(self.config.PERPLEXITY) + perplexity_reason_conf["model"] = reasoning_model_name + self.reasoning_chat_models[ChatType.PERPLEXITY.value] = Perplexity(perplexity_reason_conf) + self.LOG.info(f"已加载 Perplexity 推理模型: {reasoning_model_name}") + except Exception as e: + self.LOG.error(f"初始化 Perplexity 模型时出错: {str(e)}") # 根据chat_type参数选择默认模型 + self.current_model_id = None if chat_type > 0 and chat_type in self.chat_models: self.chat = self.chat_models[chat_type] self.default_model_id = chat_type + self.current_model_id = chat_type else: # 如果没有指定chat_type或指定的模型不可用，尝试使用配置文件中指定的默认模型 self.default_model_id = self.config.GROUP_MODELS.get('default', 0) if self.default_model_id in self.chat_models: self.chat = self.chat_models[self.default_model_id] + self.current_model_id = self.default_model_id elif self.chat_models: # 如果有任何可用模型，使用第一个 self.default_model_id = list(self.chat_models.keys())[0] self.chat = self.chat_models[self.default_model_id] + self.current_model_id = self.default_model_id else: self.LOG.warning("未配置任何可用的模型") self.chat = None self.default_model_id = 0 + self.current_model_id = None self.LOG.info(f"默认模型: {self.chat}，模型ID: {self.default_model_id}") @@ -201,6 +263,30 @@ class Robot(Job): setattr(ctx, 'chat', self.chat) setattr(ctx, 'specific_max_history', specific_limit) + reasoning_triggered = bool(ctx.text and "想想" in ctx.text) + if reasoning_triggered: + self.LOG.info("检测到推理模式触发词，跳过AI路由。") + ctx.send_text("正在深度思考，请稍候...") + + previous_ctx_chat = ctx.chat + reasoning_chat = self._get_reasoning_chat_model() + if reasoning_chat: + ctx.chat = reasoning_chat + self.LOG.debug(f"使用推理模型 {reasoning_chat} 处理消息") + else: + self.LOG.warning("当前模型未配置推理模型，使用默认模型处理深度思考请求") + + reasoning_handled = False + try: + reasoning_handled = handle_chitchat(ctx, None) + finally: + ctx.chat = previous_ctx_chat + + if not reasoning_handled: + self.LOG.warning("推理模式处理消息失败，向用户返回降级提示") + ctx.send_text("抱歉，深度思考暂时遇到问题，请稍后再试。") + return + handled = False # 5. 优先尝试使用AI路由器处理消息（仅限私聊或@机器人） @@ -448,6 +534,13 @@ class Robot(Job): return None + def _get_reasoning_chat_model(self): + """获取当前聊天模型对应的推理模型实例""" + model_id = getattr(self, 'current_model_id', None) + if model_id is None: + return None + return self.reasoning_chat_models.get(model_id) + def _select_model_for_message(self, msg: WxMsg) -> None: """根据消息来源选择对应的AI模型 @@ -464,6 +557,7 @@ class Robot(Job): # 没有配置，使用默认模型 if self.default_model_id in self.chat_models: self.chat = self.chat_models[self.default_model_id] + self.current_model_id = self.default_model_id return # 群聊消息处理 @@ -477,10 +571,12 @@ class Robot(Job): if self.chat != self.chat_models[model_id]: self.chat = self.chat_models[model_id] self.LOG.info(f"已为群 {source_id} 切换到模型: {self.chat.__class__.__name__}") + self.current_model_id = model_id else: self.LOG.warning(f"群 {source_id} 配置的模型ID {model_id} 不可用，使用默认模型") if self.default_model_id in self.chat_models: self.chat = self.chat_models[self.default_model_id] + self.current_model_id = self.default_model_id return # 私聊消息处理 else: @@ -493,15 +589,18 @@ class Robot(Job): if self.chat != self.chat_models[model_id]: self.chat = self.chat_models[model_id] self.LOG.info(f"已为私聊用户 {source_id} 切换到模型: {self.chat.__class__.__name__}") + self.current_model_id = model_id else: self.LOG.warning(f"私聊用户 {source_id} 配置的模型ID {model_id} 不可用，使用默认模型") if self.default_model_id in self.chat_models: self.chat = self.chat_models[self.default_model_id] + self.current_model_id = self.default_model_id return # 如果没有找到对应配置，使用默认模型 if self.default_model_id in self.chat_models: self.chat = self.chat_models[self.default_model_id] + self.current_model_id = self.default_model_id def _get_specific_history_limit(self, msg: WxMsg) -> int: """根据消息来源和配置，获取特定的历史消息数量限制