From 0d5a8c7fa748555a6050004e9c9fd580c336fb7a Mon Sep 17 00:00:00 2001 From: SengokuCola <1026294844@qq.com> Date: Wed, 4 Jun 2025 13:42:30 +0800 Subject: [PATCH] =?UTF-8?q?better=EF=BC=9A=E4=BC=98=E5=8C=96planner?= =?UTF-8?q?=E7=9A=84=E6=A0=BC=E5=BC=8F?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- .../focus_chat/planners/planner_complex.py | 363 ------------------ .../focus_chat/planners/planner_factory.py | 2 - .../focus_chat/planners/planner_simple.py | 43 ++- 3 files changed, 22 insertions(+), 386 deletions(-) delete mode 100644 src/chat/focus_chat/planners/planner_complex.py diff --git a/src/chat/focus_chat/planners/planner_complex.py b/src/chat/focus_chat/planners/planner_complex.py deleted file mode 100644 index 7f4aed7f7..000000000 --- a/src/chat/focus_chat/planners/planner_complex.py +++ /dev/null @@ -1,363 +0,0 @@ -import json # <--- 确保导入 json -import traceback -from typing import List, Dict, Any, Optional -from rich.traceback import install -from src.llm_models.utils_model import LLMRequest -from src.config.config import global_config -from src.chat.focus_chat.info.info_base import InfoBase -from src.chat.focus_chat.info.obs_info import ObsInfo -from src.chat.focus_chat.info.cycle_info import CycleInfo -from src.chat.focus_chat.info.mind_info import MindInfo -from src.chat.focus_chat.info.action_info import ActionInfo -from src.chat.focus_chat.info.structured_info import StructuredInfo -from src.chat.focus_chat.info.self_info import SelfInfo -from src.common.logger_manager import get_logger -from src.chat.utils.prompt_builder import Prompt, global_prompt_manager -from src.individuality.individuality import individuality -from src.chat.focus_chat.planners.action_manager import ActionManager -from json_repair import repair_json -from src.chat.focus_chat.planners.base_planner import BasePlanner - -logger = get_logger("planner") - -install(extra_lines=3) - - -def init_prompt(): - Prompt( - """ -你的自我认知是： -{self_info_block} -{extra_info_block} -{memory_str} -注意，除了下面动作选项之外，你在群聊里不能做其他任何事情，这是你能力的边界，现在请你选择合适的action: - -{action_options_text} - -你必须从上面列出的可用action中选择一个，并说明原因。 -你的决策必须以严格的 JSON 格式输出，且仅包含 JSON 内容，不要有任何其他文字或解释。 - -{moderation_prompt} - -你需要基于以下信息决定如何参与对话 -这些信息可能会有冲突，请你整合这些信息，并选择一个最合适的action： -{chat_content_block} - -{mind_info_block} -{cycle_info_block} - -请综合分析聊天内容和你看到的新消息，参考聊天规划，选择合适的action: - -请你以下面格式输出你选择的action： -{{ - "action": "action_name", - "reasoning": "说明你做出该action的原因", - "参数1": "参数1的值", - "参数2": "参数2的值", - "参数3": "参数3的值", - ... -}} - -请输出你的决策 JSON：""", - "planner_prompt", - ) - - Prompt( - """ -action_name: {action_name} - 描述：{action_description} - 参数： -{action_parameters} - 动作要求： -{action_require}""", - "action_prompt", - ) - - -class ActionPlanner(BasePlanner): - def __init__(self, log_prefix: str, action_manager: ActionManager): - super().__init__(log_prefix, action_manager) - # LLM规划器配置 - self.planner_llm = LLMRequest( - model=global_config.model.planner, - max_tokens=1000, - request_type="focus.planner", # 用于动作规划 - ) - - async def plan(self, all_plan_info: List[InfoBase], running_memorys: List[Dict[str, Any]]) -> Dict[str, Any]: - """ - 规划器 (Planner): 使用LLM根据上下文决定做出什么动作。 - - 参数: - all_plan_info: 所有计划信息 - running_memorys: 回忆信息 - """ - - action = "no_reply" # 默认动作 - reasoning = "规划器初始化默认" - action_data = {} - - try: - # 获取观察信息 - extra_info: list[str] = [] - - # 设置默认值 - nickname_str = "" - for nicknames in global_config.bot.alias_names: - nickname_str += f"{nicknames}," - name_block = f"你的名字是{global_config.bot.nickname},你的昵称有{nickname_str}，有人也会用这些昵称称呼你。" - - personality_block = individuality.get_personality_prompt(x_person=2, level=2) - identity_block = individuality.get_identity_prompt(x_person=2, level=2) - - self_info = name_block + personality_block + identity_block - current_mind = "你思考了很久，没有想清晰要做什么" - - cycle_info = "" - structured_info = "" - extra_info = [] - current_mind = "" - observed_messages = [] - observed_messages_str = "" - chat_type = "group" - is_group_chat = True - for info in all_plan_info: - if isinstance(info, ObsInfo): - observed_messages = info.get_talking_message() - observed_messages_str = info.get_talking_message_str_truncate() - chat_type = info.get_chat_type() - is_group_chat = chat_type == "group" - elif isinstance(info, MindInfo): - current_mind = info.get_current_mind() - elif isinstance(info, CycleInfo): - cycle_info = info.get_observe_info() - elif isinstance(info, SelfInfo): - self_info = info.get_processed_info() - elif isinstance(info, StructuredInfo): - structured_info = info.get_processed_info() - # print(f"structured_info: {structured_info}") - # elif not isinstance(info, ActionInfo): # 跳过已处理的ActionInfo - # extra_info.append(info.get_processed_info()) - - # 获取当前可用的动作 - current_available_actions = self.action_manager.get_using_actions() - - # 如果没有可用动作或只有no_reply动作，直接返回no_reply - if not current_available_actions or ( - len(current_available_actions) == 1 and "no_reply" in current_available_actions - ): - action = "no_reply" - reasoning = "没有可用的动作" if not current_available_actions else "只有no_reply动作可用，跳过规划" - logger.info(f"{self.log_prefix}{reasoning}") - self.action_manager.restore_actions() - logger.debug( - f"{self.log_prefix}沉默后恢复到默认动作集, 当前可用: {list(self.action_manager.get_using_actions().keys())}" - ) - return { - "action_result": {"action_type": action, "action_data": action_data, "reasoning": reasoning}, - "current_mind": current_mind, - "observed_messages": observed_messages, - } - - # --- 构建提示词 (调用修改后的 PromptBuilder 方法) --- - prompt = await self.build_planner_prompt( - self_info_block=self_info, - is_group_chat=is_group_chat, # <-- Pass HFC state - chat_target_info=None, - observed_messages_str=observed_messages_str, # <-- Pass local variable - current_mind=current_mind, # <-- Pass argument - structured_info=structured_info, # <-- Pass SubMind info - current_available_actions=current_available_actions, # <-- Pass determined actions - cycle_info=cycle_info, # <-- Pass cycle info - extra_info=extra_info, - running_memorys=running_memorys, - ) - - # --- 调用 LLM (普通文本生成) --- - llm_content = None - try: - prompt = f"{prompt}" - print(len(prompt)) - llm_content, (reasoning_content, _) = await self.planner_llm.generate_response_async(prompt=prompt) - logger.debug(f"{self.log_prefix}[Planner] LLM 原始 JSON 响应 (预期): {llm_content}") - logger.debug(f"{self.log_prefix}[Planner] LLM 原始理由 响应 (预期): {reasoning_content}") - except Exception as req_e: - logger.error(f"{self.log_prefix}[Planner] LLM 请求执行失败: {req_e}") - reasoning = f"LLM 请求失败，你的模型出现问题: {req_e}" - action = "no_reply" - - if llm_content: - try: - fixed_json_string = repair_json(llm_content) - if isinstance(fixed_json_string, str): - try: - parsed_json = json.loads(fixed_json_string) - except json.JSONDecodeError as decode_error: - logger.error(f"JSON解析错误: {str(decode_error)}") - parsed_json = {} - else: - # 如果repair_json直接返回了字典对象，直接使用 - parsed_json = fixed_json_string - - # 提取决策，提供默认值 - extracted_action = parsed_json.get("action", "no_reply") - extracted_reasoning = parsed_json.get("reasoning", "LLM未提供理由") - - # 将所有其他属性添加到action_data - action_data = {} - for key, value in parsed_json.items(): - if key not in ["action", "reasoning"]: - action_data[key] = value - - # 对于reply动作不需要额外处理，因为相关字段已经在上面的循环中添加到action_data - - if extracted_action not in current_available_actions: - logger.warning( - f"{self.log_prefix}LLM 返回了当前不可用或无效的动作: '{extracted_action}' (可用: {list(current_available_actions.keys())})，将强制使用 'no_reply'" - ) - action = "no_reply" - reasoning = f"LLM 返回了当前不可用的动作 '{extracted_action}' (可用: {list(current_available_actions.keys())})。原始理由: {extracted_reasoning}" - else: - # 动作有效且可用 - action = extracted_action - reasoning = extracted_reasoning - - except Exception as json_e: - logger.warning( - f"{self.log_prefix}解析LLM响应JSON失败，模型返回不标准: {json_e}. LLM原始输出: '{llm_content}'" - ) - reasoning = f"解析LLM响应JSON失败: {json_e}. 将使用默认动作 'no_reply'." - action = "no_reply" - - except Exception as outer_e: - logger.error(f"{self.log_prefix}Planner 处理过程中发生意外错误，规划失败，将执行 no_reply: {outer_e}") - traceback.print_exc() - action = "no_reply" - reasoning = f"Planner 内部处理错误: {outer_e}" - - logger.debug( - f"{self.log_prefix}规划器Prompt:\n{prompt}\n\n决策动作:{action},\n动作信息: '{action_data}'\n理由: {reasoning}" - ) - - # 恢复到默认动作集 - self.action_manager.restore_actions() - logger.debug( - f"{self.log_prefix}规划后恢复到默认动作集, 当前可用: {list(self.action_manager.get_using_actions().keys())}" - ) - - action_result = {"action_type": action, "action_data": action_data, "reasoning": reasoning} - - plan_result = { - "action_result": action_result, - "current_mind": current_mind, - "observed_messages": observed_messages, - "action_prompt": prompt, - } - - return plan_result - - async def build_planner_prompt( - self, - self_info_block: str, - is_group_chat: bool, # Now passed as argument - chat_target_info: Optional[dict], # Now passed as argument - observed_messages_str: str, - current_mind: Optional[str], - structured_info: Optional[str], - current_available_actions: Dict[str, ActionInfo], - cycle_info: Optional[str], - extra_info: list[str], - running_memorys: List[Dict[str, Any]], - ) -> str: - """构建 Planner LLM 的提示词 (获取模板并填充数据)""" - try: - memory_str = "" - if global_config.focus_chat.parallel_processing: - memory_str = "" - if running_memorys: - memory_str = "以下是当前在聊天中，你回忆起的记忆：\n" - for running_memory in running_memorys: - memory_str += f"{running_memory['topic']}: {running_memory['content']}\n" - - chat_context_description = "你现在正在一个群聊中" - chat_target_name = None # Only relevant for private - if not is_group_chat and chat_target_info: - chat_target_name = ( - chat_target_info.get("person_name") or chat_target_info.get("user_nickname") or "对方" - ) - chat_context_description = f"你正在和 {chat_target_name} 私聊" - - chat_content_block = "" - if observed_messages_str: - chat_content_block = f"聊天记录：\n{observed_messages_str}" - else: - chat_content_block = "你还未开始聊天" - - mind_info_block = "" - if current_mind: - mind_info_block = f"对聊天的规划：{current_mind}" - else: - mind_info_block = "你刚参与聊天" - - personality_block = individuality.get_prompt(x_person=2, level=2) - - action_options_block = "" - for using_actions_name, using_actions_info in current_available_actions.items(): - # print(using_actions_name) - # print(using_actions_info) - # print(using_actions_info["parameters"]) - # print(using_actions_info["require"]) - # print(using_actions_info["description"]) - - using_action_prompt = await global_prompt_manager.get_prompt_async("action_prompt") - - param_text = "" - for param_name, param_description in using_actions_info["parameters"].items(): - param_text += f" {param_name}: {param_description}\n" - - require_text = "" - for require_item in using_actions_info["require"]: - require_text += f" - {require_item}\n" - - using_action_prompt = using_action_prompt.format( - action_name=using_actions_name, - action_description=using_actions_info["description"], - action_parameters=param_text, - action_require=require_text, - ) - - action_options_block += using_action_prompt - - extra_info_block = "\n".join(extra_info) - extra_info_block += f"\n{structured_info}" - if extra_info or structured_info: - extra_info_block = f"以下是一些额外的信息，现在请你阅读以下内容，进行决策\n{extra_info_block}\n以上是一些额外的信息，现在请你阅读以下内容，进行决策" - else: - extra_info_block = "" - - moderation_prompt_block = "请不要输出违法违规内容，不要输出色情，暴力，政治相关内容，如有敏感内容，请规避。" - - planner_prompt_template = await global_prompt_manager.get_prompt_async("planner_prompt") - prompt = planner_prompt_template.format( - self_info_block=self_info_block, - memory_str=memory_str, - # bot_name=global_config.bot.nickname, - prompt_personality=personality_block, - chat_context_description=chat_context_description, - chat_content_block=chat_content_block, - mind_info_block=mind_info_block, - cycle_info_block=cycle_info, - action_options_text=action_options_block, - # action_available_block=action_available_block, - extra_info_block=extra_info_block, - moderation_prompt=moderation_prompt_block, - ) - return prompt - - except Exception as e: - logger.error(f"构建 Planner 提示词时出错: {e}") - logger.error(traceback.format_exc()) - return "构建 Planner Prompt 时出错" - - -init_prompt() diff --git a/src/chat/focus_chat/planners/planner_factory.py b/src/chat/focus_chat/planners/planner_factory.py index e0df9f762..c92168238 100644 --- a/src/chat/focus_chat/planners/planner_factory.py +++ b/src/chat/focus_chat/planners/planner_factory.py @@ -1,6 +1,5 @@ from typing import Dict, Type from src.chat.focus_chat.planners.base_planner import BasePlanner -from src.chat.focus_chat.planners.planner_complex import ActionPlanner as ComplexActionPlanner from src.chat.focus_chat.planners.planner_simple import ActionPlanner as SimpleActionPlanner from src.chat.focus_chat.planners.action_manager import ActionManager from src.config.config import global_config @@ -14,7 +13,6 @@ class PlannerFactory: # 注册所有可用的规划器类型 _planner_types: Dict[str, Type[BasePlanner]] = { - "complex": ComplexActionPlanner, "simple": SimpleActionPlanner, } diff --git a/src/chat/focus_chat/planners/planner_simple.py b/src/chat/focus_chat/planners/planner_simple.py index 1c564026e..5f757a1ba 100644 --- a/src/chat/focus_chat/planners/planner_simple.py +++ b/src/chat/focus_chat/planners/planner_simple.py @@ -46,18 +46,13 @@ def init_prompt(): {mind_info_block} {cycle_info_block} -注意，除了下面动作选项之外，你在群聊里不能做其他任何事情，这是你能力的边界，现在请你选择合适的action: + {moderation_prompt} +注意，除了下面动作选项之外，你在群聊里不能做其他任何事情，这是你能力的边界，现在请你选择合适的action: {action_options_text} -以严格的 JSON 格式输出，且仅包含 JSON 内容，不要有任何其他文字或解释。 -请你以下面格式输出： -{{ - "action": "action_name" - "参数": "参数的值"(可能有多个参数), -}} - +请以动作的输出要求，以严格的 JSON 格式输出，且仅包含 JSON 内容。 请输出你提取的JSON，不要有任何其他文字或解释： """, @@ -66,11 +61,15 @@ def init_prompt(): Prompt( """ -动作名称：{action_name} - 描述：{action_description} - {action_parameters} - 使用该动作的场景： -{action_require}""", +动作：{action_name} +该动作的描述：{action_description} +使用该动作的场景： +{action_require} +输出要求： +{{ + "action": "{action_name}",{action_parameters} +}} +""", "action_prompt", ) @@ -342,18 +341,20 @@ class ActionPlanner(BasePlanner): using_action_prompt = await global_prompt_manager.get_prompt_async("action_prompt") - param_text = "" - for param_name, param_description in using_actions_info["parameters"].items(): - param_text += f" {param_name}: {param_description}\n" + if using_actions_info["parameters"]: + param_text = "\n" + for param_name, param_description in using_actions_info["parameters"].items(): + param_text += f' "{param_name}":"{param_description}"\n' + param_text = param_text.rstrip('\n') + else: + param_text = "" + require_text = "" for require_item in using_actions_info["require"]: - require_text += f"{require_item}\n" + require_text += f"- {require_item}\n" + require_text = require_text.rstrip('\n') - if param_text: - param_text = f"参数：\n{param_text}" - else: - param_text = "无需参数" using_action_prompt = using_action_prompt.format( action_name=using_actions_name,