feat:采用直接输出法构建动作planner
This commit is contained in:
@@ -634,12 +634,14 @@ HFC_STYLE_CONFIG = {
|
||||
},
|
||||
"simple": {
|
||||
"console_format": (
|
||||
"<level>{time:MM-DD HH:mm}</level> | <light-green>专注聊天</light-green> | <light-green>{message}</light-green>"
|
||||
"<level>{time:MM-DD HH:mm}</level> | <light-green>专注聊天 | {message}</light-green>"
|
||||
),
|
||||
"file_format": "{time:YYYY-MM-DD HH:mm:ss} | {level: <8} | {extra[module]: <15} | 专注聊天 | {message}",
|
||||
},
|
||||
}
|
||||
|
||||
|
||||
|
||||
CONFIRM_STYLE_CONFIG = {
|
||||
"console_format": "<RED>{message}</RED>", # noqa: E501
|
||||
"file_format": "{time:YYYY-MM-DD HH:mm:ss} | {level: <8} | {extra[module]: <15} | EULA与PRIVACY确认 | {message}",
|
||||
|
||||
@@ -10,8 +10,11 @@ logger = get_logger("mai_state")
|
||||
|
||||
|
||||
# -- 状态相关的可配置参数 (可以从 glocal_config 加载) --
|
||||
# enable_unlimited_hfc_chat = True # 调试用:无限专注聊天
|
||||
enable_unlimited_hfc_chat = False
|
||||
# The line `enable_unlimited_hfc_chat = False` is setting a configuration parameter that controls
|
||||
# whether a specific debugging feature is enabled or not. When `enable_unlimited_hfc_chat` is set to
|
||||
# `False`, it means that the debugging feature for unlimited focused chatting is disabled.
|
||||
enable_unlimited_hfc_chat = True # 调试用:无限专注聊天
|
||||
# enable_unlimited_hfc_chat = False
|
||||
prevent_offline_state = True
|
||||
# 目前默认不启用OFFLINE状态
|
||||
|
||||
|
||||
@@ -18,7 +18,7 @@ from src.heart_flow.sub_mind import SubMind
|
||||
# 定义常量 (从 interest.py 移动过来)
|
||||
MAX_INTEREST = 15.0
|
||||
|
||||
logger = get_logger("subheartflow")
|
||||
logger = get_logger("sub_heartflow")
|
||||
|
||||
PROBABILITY_INCREASE_RATE_PER_SECOND = 0.1
|
||||
PROBABILITY_DECREASE_RATE_PER_SECOND = 0.1
|
||||
|
||||
@@ -2,6 +2,7 @@ import asyncio
|
||||
import time
|
||||
import traceback
|
||||
import random # <--- 添加导入
|
||||
import json # <--- 确保导入 json
|
||||
from typing import List, Optional, Dict, Any, Deque, Callable, Coroutine
|
||||
from collections import deque
|
||||
from src.plugins.chat.message import MessageRecv, BaseMessageInfo, MessageThinking, MessageSending
|
||||
@@ -37,7 +38,7 @@ EMOJI_SEND_PRO = 0.3 # 设置一个概率,比如 30% 才真的发
|
||||
CONSECUTIVE_NO_REPLY_THRESHOLD = 3 # 连续不回复的阈值
|
||||
|
||||
|
||||
logger = get_logger("HFC") # Logger Name Changed
|
||||
logger = get_logger("hfc") # Logger Name Changed
|
||||
|
||||
|
||||
# 默认动作定义
|
||||
@@ -119,35 +120,6 @@ class ActionManager:
|
||||
"""重置为默认动作集"""
|
||||
self._available_actions = DEFAULT_ACTIONS.copy()
|
||||
|
||||
def get_planner_tool_definition(self) -> List[Dict[str, Any]]:
|
||||
"""获取当前动作集对应的规划器工具定义"""
|
||||
return [
|
||||
{
|
||||
"type": "function",
|
||||
"function": {
|
||||
"name": "decide_reply_action",
|
||||
"description": "根据当前聊天内容和上下文,决定机器人是否应该回复以及如何回复。",
|
||||
"parameters": {
|
||||
"type": "object",
|
||||
"properties": {
|
||||
"action": {
|
||||
"type": "string",
|
||||
"enum": list(self._available_actions.keys()),
|
||||
"description": "决定采取的行动:"
|
||||
+ ", ".join([f"'{k}'({v})" for k, v in self._available_actions.items()]),
|
||||
},
|
||||
"reasoning": {"type": "string", "description": "做出此决定的简要理由。"},
|
||||
"emoji_query": {
|
||||
"type": "string",
|
||||
"description": "如果行动是'emoji_reply',指定表情的主题或概念。如果行动是'text_reply'且希望在文本后追加表情,也在此指定表情主题。",
|
||||
},
|
||||
},
|
||||
"required": ["action", "reasoning"],
|
||||
},
|
||||
},
|
||||
}
|
||||
]
|
||||
|
||||
|
||||
# 在文件开头添加自定义异常类
|
||||
class HeartFCError(Exception):
|
||||
@@ -222,7 +194,6 @@ class HeartFChatting:
|
||||
max_tokens=256,
|
||||
request_type="response_heartflow",
|
||||
)
|
||||
self.tool_user = ToolUser()
|
||||
self.heart_fc_sender = HeartFCSender()
|
||||
|
||||
# LLM规划器配置
|
||||
@@ -784,225 +755,208 @@ class HeartFChatting:
|
||||
async def _planner(self, current_mind: str, cycle_timers: dict, is_re_planned: bool = False) -> Dict[str, Any]:
|
||||
"""
|
||||
规划器 (Planner): 使用LLM根据上下文决定是否和如何回复。
|
||||
重构为:让LLM返回结构化JSON文本,然后在代码中解析。
|
||||
|
||||
参数:
|
||||
current_mind: 子思维的当前思考结果
|
||||
cycle_timers: 计时器字典
|
||||
is_re_planned: 是否为重新规划
|
||||
is_re_planned: 是否为重新规划 (此重构中暂时简化,不处理 is_re_planned 的特殊逻辑)
|
||||
"""
|
||||
logger.info(f"{self.log_prefix}[Planner] 开始{'重新' if is_re_planned else ''}执行规划器")
|
||||
logger.info(f"{self.log_prefix}[Planner] 开始执行规划器 (JSON解析模式)")
|
||||
|
||||
# --- 新增:检查历史动作并调整可用动作 ---
|
||||
lian_xu_wen_ben_hui_fu = 0 # 连续文本回复次数
|
||||
actions_to_remove_temporarily = []
|
||||
probability_roll = random.random() # 在循环外掷骰子一次,用于概率判断
|
||||
|
||||
# 反向遍历最近的循环历史
|
||||
# --- 检查历史动作并决定临时移除动作 (逻辑保持不变) ---
|
||||
lian_xu_wen_ben_hui_fu = 0
|
||||
probability_roll = random.random()
|
||||
for cycle in reversed(self._cycle_history):
|
||||
# 只关心实际执行了动作的循环
|
||||
if cycle.action_taken:
|
||||
if cycle.action_type == "text_reply":
|
||||
lian_xu_wen_ben_hui_fu += 1
|
||||
else:
|
||||
break # 遇到非文本回复,中断计数
|
||||
# 检查最近的3个循环即可,避免检查过多历史 (如果历史很长)
|
||||
break
|
||||
if len(self._cycle_history) > 0 and cycle.cycle_id <= self._cycle_history[0].cycle_id + (
|
||||
len(self._cycle_history) - 4
|
||||
):
|
||||
break
|
||||
|
||||
logger.debug(f"{self.log_prefix}[Planner] 检测到连续文本回复次数: {lian_xu_wen_ben_hui_fu}")
|
||||
|
||||
# 根据连续次数决定临时移除哪些动作
|
||||
if lian_xu_wen_ben_hui_fu >= 3:
|
||||
logger.info(f"{self.log_prefix}[Planner] 连续回复 >= 3 次,强制移除 text_reply 和 emoji_reply")
|
||||
actions_to_remove_temporarily.extend(["text_reply", "emoji_reply"])
|
||||
elif lian_xu_wen_ben_hui_fu == 2:
|
||||
if probability_roll < 0.8: # 80% 概率
|
||||
if probability_roll < 0.8:
|
||||
logger.info(f"{self.log_prefix}[Planner] 连续回复 2 次,80% 概率移除 text_reply 和 emoji_reply (触发)")
|
||||
actions_to_remove_temporarily.extend(["text_reply", "emoji_reply"])
|
||||
else:
|
||||
logger.info(
|
||||
f"{self.log_prefix}[Planner] 连续回复 2 次,80% 概率移除 text_reply 和 emoji_reply (未触发)"
|
||||
)
|
||||
logger.info(f"{self.log_prefix}[Planner] 连续回复 2 次,80% 概率移除 text_reply 和 emoji_reply (未触发)")
|
||||
elif lian_xu_wen_ben_hui_fu == 1:
|
||||
if probability_roll < 0.4: # 40% 概率
|
||||
if probability_roll < 0.4:
|
||||
logger.info(f"{self.log_prefix}[Planner] 连续回复 1 次,40% 概率移除 text_reply (触发)")
|
||||
actions_to_remove_temporarily.append("text_reply")
|
||||
else:
|
||||
logger.info(f"{self.log_prefix}[Planner] 连续回复 1 次,40% 概率移除 text_reply (未触发)")
|
||||
# 如果 lian_xu_wen_ben_hui_fu == 0,则不移除任何动作
|
||||
# --- 结束:检查历史动作 ---
|
||||
# --- 结束检查历史动作 ---
|
||||
|
||||
# 获取观察信息
|
||||
observation = self.observations[0]
|
||||
if is_re_planned:
|
||||
await observation.observe()
|
||||
# if is_re_planned: # 暂时简化,不处理重新规划
|
||||
# await observation.observe()
|
||||
observed_messages = observation.talking_message
|
||||
observed_messages_str = observation.talking_message_str_truncate
|
||||
|
||||
# --- 使用 LLM 进行决策 --- #
|
||||
reasoning = "默认决策或获取决策失败"
|
||||
llm_error = False # LLM错误标志
|
||||
arguments = None # 初始化参数变量
|
||||
emoji_query = "" # <--- 在这里初始化 emoji_query
|
||||
# --- 使用 LLM 进行决策 (JSON 输出模式) --- #
|
||||
action = "no_reply" # 默认动作
|
||||
reasoning = "规划器初始化默认"
|
||||
emoji_query = ""
|
||||
llm_error = False # LLM 请求或解析错误标志
|
||||
|
||||
# 获取我们将传递给 prompt 构建器和用于验证的当前可用动作
|
||||
current_available_actions = self.action_manager.get_available_actions()
|
||||
|
||||
try:
|
||||
# --- 新增:应用临时动作移除 ---
|
||||
# --- 应用临时动作移除 ---
|
||||
if actions_to_remove_temporarily:
|
||||
self.action_manager.temporarily_remove_actions(actions_to_remove_temporarily)
|
||||
# 更新 current_available_actions 以反映移除后的状态
|
||||
current_available_actions = self.action_manager.get_available_actions()
|
||||
logger.debug(
|
||||
f"{self.log_prefix}[Planner] 临时移除的动作: {actions_to_remove_temporarily}, 当前可用: {list(self.action_manager.get_available_actions().keys())}"
|
||||
f"{self.log_prefix}[Planner] 临时移除的动作: {actions_to_remove_temporarily}, 当前可用: {list(current_available_actions.keys())}"
|
||||
)
|
||||
|
||||
# --- 构建提示词 ---
|
||||
replan_prompt_str = ""
|
||||
if is_re_planned:
|
||||
replan_prompt_str = await self._build_replan_prompt(
|
||||
self._current_cycle.action_type, self._current_cycle.reasoning
|
||||
)
|
||||
# --- 构建提示词 (调用修改后的 _build_planner_prompt) ---
|
||||
# replan_prompt_str = "" # 暂时简化
|
||||
# if is_re_planned:
|
||||
# replan_prompt_str = await self._build_replan_prompt(
|
||||
# self._current_cycle.action_type, self._current_cycle.reasoning
|
||||
# )
|
||||
prompt = await self._build_planner_prompt(
|
||||
observed_messages_str, current_mind, self.sub_mind.structured_info, replan_prompt_str
|
||||
observed_messages_str,
|
||||
current_mind,
|
||||
self.sub_mind.structured_info,
|
||||
"", # replan_prompt_str,
|
||||
current_available_actions # <--- 传入当前可用动作
|
||||
)
|
||||
|
||||
# --- 调用 LLM ---
|
||||
# --- 调用 LLM (普通文本生成) ---
|
||||
llm_content = None
|
||||
try:
|
||||
planner_tools = self.action_manager.get_planner_tool_definition()
|
||||
logger.debug(f"{self.log_prefix}[Planner] 本次使用的工具定义: {planner_tools}") # 记录本次使用的工具
|
||||
_response_text, _reasoning_content, tool_calls = await self.planner_llm.generate_response_tool_async(
|
||||
prompt=prompt,
|
||||
tools=planner_tools,
|
||||
)
|
||||
logger.debug(f"{self.log_prefix}[Planner] 原始人 LLM响应: {_response_text}")
|
||||
# 假设 LLMRequest 有 generate_response 方法返回 (content, reasoning, model_name)
|
||||
# 我们只需要 content
|
||||
# !! 注意:这里假设 self.planner_llm 有 generate_response 方法
|
||||
# !! 如果你的 LLMRequest 类使用的是其他方法名,请相应修改
|
||||
llm_content, _, _ = await self.planner_llm.generate_response(prompt=prompt)
|
||||
logger.debug(f"{self.log_prefix}[Planner] LLM 原始 JSON 响应 (预期): {llm_content}")
|
||||
except Exception as req_e:
|
||||
logger.error(f"{self.log_prefix}[Planner] LLM请求执行失败: {req_e}")
|
||||
action = "error"
|
||||
reasoning = f"LLM请求失败: {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: {req_e}"
|
||||
llm_error = True
|
||||
# 直接返回错误结果
|
||||
return {
|
||||
"action": action,
|
||||
"reasoning": reasoning,
|
||||
"emoji_query": "",
|
||||
"current_mind": current_mind,
|
||||
"observed_messages": observed_messages,
|
||||
"llm_error": llm_error,
|
||||
}
|
||||
# 直接使用默认动作返回错误结果
|
||||
action = "no_reply" # 明确设置为默认值
|
||||
emoji_query = "" # 明确设置为空
|
||||
# 不再立即返回,而是继续执行 finally 块以恢复动作
|
||||
# return { ... }
|
||||
|
||||
# 默认错误状态
|
||||
action = "error"
|
||||
reasoning = "处理工具调用时出错"
|
||||
llm_error = True
|
||||
# --- 解析 LLM 返回的 JSON (仅当 LLM 请求未出错时进行) ---
|
||||
if not llm_error and llm_content:
|
||||
try:
|
||||
# 尝试去除可能的 markdown 代码块标记
|
||||
cleaned_content = llm_content.strip().removeprefix("```json").removeprefix("```").removesuffix("```").strip()
|
||||
if not cleaned_content:
|
||||
raise json.JSONDecodeError("Cleaned content is empty", cleaned_content, 0)
|
||||
parsed_json = json.loads(cleaned_content)
|
||||
|
||||
# 1. 验证工具调用
|
||||
success, valid_tool_calls, error_msg = process_llm_tool_calls(
|
||||
tool_calls, log_prefix=f"{self.log_prefix}[Planner] "
|
||||
)
|
||||
# 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")
|
||||
extracted_emoji_query = parsed_json.get("emoji_query", "")
|
||||
|
||||
if success and valid_tool_calls:
|
||||
# 2. 提取第一个调用并获取参数
|
||||
first_tool_call = valid_tool_calls[0]
|
||||
tool_name = first_tool_call.get("function", {}).get("name")
|
||||
arguments = extract_tool_call_arguments(first_tool_call, None)
|
||||
|
||||
# 3. 检查名称和参数
|
||||
expected_tool_name = "decide_reply_action"
|
||||
if tool_name == expected_tool_name and arguments is not None:
|
||||
# 4. 成功,提取决策
|
||||
extracted_action = arguments.get("action", "no_reply")
|
||||
# 验证动作
|
||||
if extracted_action not in self.action_manager.get_available_actions():
|
||||
# 如果LLM返回了一个此时不该用的动作(因为被临时移除了)
|
||||
# 或者完全无效的动作
|
||||
# 验证动作是否在当前可用列表中
|
||||
# !! 使用调用 prompt 时实际可用的动作列表进行验证
|
||||
if extracted_action not in current_available_actions:
|
||||
logger.warning(
|
||||
f"{self.log_prefix}[Planner] LLM返回了当前不可用或无效的动作: {extracted_action},将强制使用 'no_reply'"
|
||||
f"{self.log_prefix}[Planner] 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}"
|
||||
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}"
|
||||
emoji_query = ""
|
||||
llm_error = False # 视为逻辑修正而非 LLM 错误
|
||||
# --- 检查 'no_reply' 是否也恰好被移除了 (极端情况) ---
|
||||
if "no_reply" not in self.action_manager.get_available_actions():
|
||||
logger.error(
|
||||
f"{self.log_prefix}[Planner] 严重错误:'no_reply' 动作也不可用!无法执行任何动作。"
|
||||
)
|
||||
action = "error" # 回退到错误状态
|
||||
reasoning = "无法执行任何有效动作,包括 no_reply"
|
||||
llm_error = True
|
||||
# 检查 no_reply 是否也恰好被移除了 (极端情况)
|
||||
if "no_reply" not in current_available_actions:
|
||||
logger.error(f"{self.log_prefix}[Planner] 严重错误:'no_reply' 动作也不可用!无法执行任何动作。")
|
||||
action = "error" # 回退到错误状态
|
||||
reasoning = "无法执行任何有效动作,包括 no_reply"
|
||||
llm_error = True # 标记为严重错误
|
||||
else:
|
||||
llm_error = False # 视为逻辑修正而非 LLM 错误
|
||||
else:
|
||||
# 动作有效且可用,使用提取的值
|
||||
# 动作有效且可用
|
||||
action = extracted_action
|
||||
reasoning = arguments.get("reasoning", "未提供理由")
|
||||
emoji_query = arguments.get("emoji_query", "")
|
||||
llm_error = False # 成功处理
|
||||
# 记录决策结果
|
||||
reasoning = extracted_reasoning
|
||||
emoji_query = extracted_emoji_query
|
||||
llm_error = False # 解析成功
|
||||
logger.debug(
|
||||
f"{self.log_prefix}[要做什么]\nPrompt:\n{prompt}\n\n决策结果: {action}, 理由: {reasoning}, 表情查询: '{emoji_query}'"
|
||||
f"{self.log_prefix}[要做什么]\nPrompt:\n{prompt}\n\n决策结果 (来自JSON): {action}, 理由: {reasoning}, 表情查询: '{emoji_query}'"
|
||||
)
|
||||
elif tool_name != expected_tool_name:
|
||||
reasoning = f"LLM返回了非预期的工具: {tool_name}"
|
||||
logger.warning(f"{self.log_prefix}[Planner] {reasoning}")
|
||||
else: # arguments is None
|
||||
reasoning = f"无法提取工具 {tool_name} 的参数"
|
||||
logger.warning(f"{self.log_prefix}[Planner] {reasoning}")
|
||||
elif not success:
|
||||
reasoning = f"验证工具调用失败: {error_msg}"
|
||||
logger.warning(f"{self.log_prefix}[Planner] {reasoning}")
|
||||
else: # not valid_tool_calls
|
||||
# 如果没有有效的工具调用,我们需要检查 'no_reply' 是否是当前唯一可用的动作
|
||||
available_actions = list(self.action_manager.get_available_actions().keys())
|
||||
if available_actions == ["no_reply"]:
|
||||
logger.info(
|
||||
f"{self.log_prefix}[Planner] LLM未返回工具调用,但当前唯一可用动作是 'no_reply',将执行 'no_reply'"
|
||||
)
|
||||
action = "no_reply"
|
||||
reasoning = "LLM未返回工具调用,且当前仅 'no_reply' 可用"
|
||||
emoji_query = ""
|
||||
llm_error = False # 视为逻辑选择而非错误
|
||||
else:
|
||||
reasoning = "LLM未返回有效的工具调用"
|
||||
logger.warning(f"{self.log_prefix}[Planner] {reasoning}")
|
||||
# llm_error 保持为 True
|
||||
# 如果 llm_error 仍然是 True,说明在处理过程中有错误发生
|
||||
|
||||
except Exception as llm_e:
|
||||
logger.error(f"{self.log_prefix}[Planner] Planner LLM处理过程中发生意外错误: {llm_e}")
|
||||
except json.JSONDecodeError as json_e:
|
||||
logger.warning(f"{self.log_prefix}[Planner] 解析LLM响应JSON失败: {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" # 解析失败则默认不回复
|
||||
emoji_query = ""
|
||||
llm_error = True # 标记解析错误
|
||||
except Exception as parse_e:
|
||||
logger.error(f"{self.log_prefix}[Planner] 处理LLM响应时发生意外错误: {parse_e}")
|
||||
reasoning = f"处理LLM响应时发生意外错误: {parse_e}. 将使用默认动作 'no_reply'."
|
||||
action = "no_reply"
|
||||
emoji_query = ""
|
||||
llm_error = True
|
||||
elif not llm_error and not llm_content:
|
||||
# LLM 请求成功但返回空内容
|
||||
logger.warning(f"{self.log_prefix}[Planner] LLM 返回了空内容。")
|
||||
reasoning = "LLM 返回了空内容,使用默认动作 'no_reply'."
|
||||
action = "no_reply"
|
||||
emoji_query = ""
|
||||
llm_error = True # 标记为空响应错误
|
||||
|
||||
# 如果 llm_error 在此阶段为 True,意味着请求成功但解析失败或返回空
|
||||
# 如果 llm_error 在请求阶段就为 True,则跳过了此解析块
|
||||
|
||||
except Exception as outer_e:
|
||||
logger.error(f"{self.log_prefix}[Planner] Planner 处理过程中发生意外错误: {outer_e}")
|
||||
logger.error(traceback.format_exc())
|
||||
action = "error"
|
||||
reasoning = f"Planner内部处理错误: {llm_e}"
|
||||
action = "error" # 发生未知错误,标记为 error 动作
|
||||
reasoning = f"Planner 内部处理错误: {outer_e}"
|
||||
emoji_query = ""
|
||||
llm_error = True
|
||||
# --- 新增:确保动作恢复 ---
|
||||
finally:
|
||||
if actions_to_remove_temporarily: # 只有当确实移除了动作时才需要恢复
|
||||
# --- 确保动作恢复 ---
|
||||
# 检查 self._original_actions_backup 是否有值来判断是否需要恢复
|
||||
if self.action_manager._original_actions_backup is not None:
|
||||
self.action_manager.restore_actions()
|
||||
logger.debug(
|
||||
f"{self.log_prefix}[Planner] 恢复了原始动作集, 当前可用: {list(self.action_manager.get_available_actions().keys())}"
|
||||
)
|
||||
# --- 结束:确保动作恢复 ---
|
||||
|
||||
# --- 新增:概率性忽略文本回复附带的表情(正确的位置)---
|
||||
# --- 结束确保动作恢复 ---
|
||||
|
||||
# --- 概率性忽略文本回复附带的表情 (逻辑保持不变) ---
|
||||
if action == "text_reply" and emoji_query:
|
||||
logger.debug(f"{self.log_prefix}[Planner] 大模型想让麦麦发文字时带表情: '{emoji_query}'")
|
||||
# 掷骰子看看要不要听它的
|
||||
logger.debug(f"{self.log_prefix}[Planner] 大模型建议文字回复带表情: '{emoji_query}'")
|
||||
if random.random() > EMOJI_SEND_PRO:
|
||||
logger.info(
|
||||
f"{self.log_prefix}[Planner] 但是麦麦这次不想加表情 ({1 - EMOJI_SEND_PRO:.0%}),忽略表情 '{emoji_query}'"
|
||||
)
|
||||
emoji_query = "" # 把表情请求清空,就不发了
|
||||
emoji_query = "" # 清空表情请求
|
||||
else:
|
||||
logger.info(f"{self.log_prefix}[Planner] 好吧,加上表情 '{emoji_query}'")
|
||||
# --- 结束:概率性忽略 ---
|
||||
|
||||
# --- 结束 LLM 决策 --- #
|
||||
# --- 结束概率性忽略 ---
|
||||
|
||||
# 返回结果字典
|
||||
return {
|
||||
"action": action,
|
||||
"reasoning": reasoning,
|
||||
"emoji_query": emoji_query,
|
||||
"current_mind": current_mind,
|
||||
"observed_messages": observed_messages,
|
||||
"llm_error": llm_error,
|
||||
"llm_error": llm_error, # 返回错误状态
|
||||
}
|
||||
|
||||
async def _get_anchor_message(self) -> Optional[MessageRecv]:
|
||||
@@ -1146,8 +1100,9 @@ class HeartFChatting:
|
||||
current_mind: Optional[str],
|
||||
structured_info: Dict[str, Any],
|
||||
replan_prompt: str,
|
||||
current_available_actions: Dict[str, str],
|
||||
) -> str:
|
||||
"""构建 Planner LLM 的提示词"""
|
||||
"""构建 Planner LLM 的提示词 (获取模板并填充数据)"""
|
||||
try:
|
||||
# 准备结构化信息块
|
||||
structured_info_block = ""
|
||||
@@ -1163,12 +1118,13 @@ class HeartFChatting:
|
||||
else:
|
||||
chat_content_block = "当前没有观察到新的聊天内容。\n"
|
||||
|
||||
# 准备当前思维块
|
||||
# 准备当前思维块 (修改以匹配模板)
|
||||
current_mind_block = ""
|
||||
if current_mind:
|
||||
current_mind_block = f"{current_mind}"
|
||||
# 模板中占位符是 {current_mind_block},它期望包含"你的内心想法:"的前缀
|
||||
current_mind_block = f"你的内心想法:\n{current_mind}"
|
||||
else:
|
||||
current_mind_block = "[没有特别的想法]"
|
||||
current_mind_block = "你的内心想法:\n[没有特别的想法]"
|
||||
|
||||
# 准备循环信息块 (分析最近的活动循环)
|
||||
recent_active_cycles = []
|
||||
@@ -1208,23 +1164,40 @@ class HeartFChatting:
|
||||
|
||||
# 包装提示块,增加可读性,即使没有连续回复也给个标记
|
||||
if cycle_info_block:
|
||||
# 模板中占位符是 {cycle_info_block},它期望包含"【近期回复历史】"的前缀
|
||||
cycle_info_block = f"\n【近期回复历史】\n{cycle_info_block}\n"
|
||||
else:
|
||||
# 如果最近的活动循环不是文本回复,或者没有活动循环
|
||||
cycle_info_block = "\n【近期回复历史】\n(最近没有连续文本回复)\n"
|
||||
|
||||
individuality = Individuality.get_instance()
|
||||
# 模板中占位符是 {prompt_personality}
|
||||
prompt_personality = individuality.get_prompt(x_person=2, level=2)
|
||||
|
||||
# 获取提示词模板并填充数据
|
||||
prompt = (await global_prompt_manager.get_prompt_async("planner_prompt")).format(
|
||||
# --- 构建可用动作描述 (用于填充模板中的 {action_options_text}) ---
|
||||
action_options_text = "当前你可以选择的行动有:\n"
|
||||
action_keys = list(current_available_actions.keys())
|
||||
for name in action_keys:
|
||||
desc = current_available_actions[name]
|
||||
action_options_text += f"- '{name}': {desc}\n"
|
||||
|
||||
# --- 选择一个示例动作键 (用于填充模板中的 {example_action}) ---
|
||||
example_action_key = action_keys[0] if action_keys else "no_reply"
|
||||
|
||||
# --- 获取提示词模板 ---
|
||||
planner_prompt_template = await global_prompt_manager.get_prompt_async("planner_prompt")
|
||||
|
||||
# --- 填充模板 ---
|
||||
prompt = planner_prompt_template.format(
|
||||
bot_name=global_config.BOT_NICKNAME,
|
||||
prompt_personality=prompt_personality,
|
||||
structured_info_block=structured_info_block,
|
||||
chat_content_block=chat_content_block,
|
||||
current_mind_block=current_mind_block,
|
||||
replan=replan_prompt,
|
||||
replan="", # 暂时留空 replan 信息
|
||||
cycle_info_block=cycle_info_block,
|
||||
action_options_text=action_options_text, # 传入可用动作描述
|
||||
example_action=example_action_key # 传入示例动作键
|
||||
)
|
||||
|
||||
return prompt
|
||||
@@ -1232,7 +1205,7 @@ class HeartFChatting:
|
||||
except Exception as e:
|
||||
logger.error(f"{self.log_prefix}[Planner] 构建提示词时出错: {e}")
|
||||
logger.error(traceback.format_exc())
|
||||
return ""
|
||||
return "[构建 Planner Prompt 时出错]" # 返回错误提示,避免空字符串
|
||||
|
||||
# --- 回复器 (Replier) 的定义 --- #
|
||||
async def _replier_work(
|
||||
@@ -1273,7 +1246,7 @@ class HeartFChatting:
|
||||
try:
|
||||
with Timer("LLM生成", {}): # 内部计时器,可选保留
|
||||
content, reasoning_content, model_name = await self.model_normal.generate_response(prompt)
|
||||
logger.info(f"{self.log_prefix}[Replier-{thinking_id}]\\nPrompt:\\n{prompt}\\n生成回复: {content}\\n")
|
||||
# logger.info(f"{self.log_prefix}[Replier-{thinking_id}]\\nPrompt:\\n{prompt}\\n生成回复: {content}\\n")
|
||||
# 捕捉 LLM 输出信息
|
||||
info_catcher.catch_after_llm_generated(
|
||||
prompt=prompt, response=content, reasoning_content=reasoning_content, model_name=model_name
|
||||
|
||||
@@ -7,13 +7,14 @@ from src.plugins.utils.chat_message_builder import build_readable_messages, get_
|
||||
from src.plugins.person_info.relationship_manager import relationship_manager
|
||||
from src.plugins.chat.utils import get_embedding
|
||||
import time
|
||||
from typing import Union, Optional
|
||||
from typing import Union, Optional, Dict, Any
|
||||
from ...common.database import db
|
||||
from ..chat.utils import get_recent_group_speaker
|
||||
from ..moods.moods import MoodManager
|
||||
from ..memory_system.Hippocampus import HippocampusManager
|
||||
from ..schedule.schedule_generator import bot_schedule
|
||||
from ..knowledge.knowledge_lib import qa_manager
|
||||
import traceback
|
||||
|
||||
logger = get_logger("prompt")
|
||||
|
||||
@@ -47,17 +48,15 @@ def init_prompt():
|
||||
"info_from_tools",
|
||||
)
|
||||
|
||||
# Planner提示词 - 优化版
|
||||
# Planner提示词 - 修改为要求 JSON 输出
|
||||
Prompt(
|
||||
"""你的名字是{bot_name},{prompt_personality},你现在正在一个群聊中。需要基于以下信息决定如何参与对话:
|
||||
'''你的名字是{bot_name},{prompt_personality},你现在正在一个群聊中。需要基于以下信息决定如何参与对话:
|
||||
{structured_info_block}
|
||||
{chat_content_block}
|
||||
你的内心想法:
|
||||
{current_mind_block}
|
||||
{replan}
|
||||
{cycle_info_block}
|
||||
|
||||
请综合分析聊天内容和你看到的新消息,参考内心想法,使用'decide_reply_action'工具做出决策。决策时请注意:
|
||||
请综合分析聊天内容和你看到的新消息,参考内心想法,并根据以下原则和可用动作做出决策。
|
||||
|
||||
【回复原则】
|
||||
1. 不回复(no_reply)适用:
|
||||
@@ -81,14 +80,34 @@ def init_prompt():
|
||||
- 避免重复或评价自己的发言
|
||||
- 不要和自己聊天
|
||||
|
||||
【必须遵守】
|
||||
- 遵守回复原则
|
||||
- 必须调用工具并包含action和reasoning
|
||||
- 你可以选择文字回复(text_reply),纯表情回复(emoji_reply),不回复(no_reply)
|
||||
- 并不是所有选择都可用
|
||||
- 选择text_reply或emoji_reply时必须提供emoji_query
|
||||
- 保持回复自然,符合日常聊天习惯""",
|
||||
"planner_prompt",
|
||||
【决策任务】
|
||||
{action_options_text}
|
||||
|
||||
你必须从上面列出的可用行动中选择一个,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。
|
||||
JSON 结构如下,包含三个字段 "action", "reasoning", "emoji_query":
|
||||
{{
|
||||
"action": "string", // 必须是上面提供的可用行动之一 (例如: '{example_action}')
|
||||
"reasoning": "string", // 做出此决定的详细理由和思考过程,说明你如何应用了回复原则
|
||||
"emoji_query": "string" // 可选。如果行动是 'emoji_reply',必须提供表情主题;如果行动是 'text_reply' 且你想附带表情,也在此提供表情主题,否则留空字符串 ""。遵循回复原则,不要滥用。
|
||||
}}
|
||||
|
||||
例如:
|
||||
{{
|
||||
"action": "text_reply",
|
||||
"reasoning": "用户提到了我,且问题比较具体,适合用文本回复。考虑到内容,可以带上一个微笑表情。",
|
||||
"emoji_query": "微笑"
|
||||
}}
|
||||
或
|
||||
{{
|
||||
"action": "no_reply",
|
||||
"reasoning": "我已经连续回复了两次,而且这个话题我不太感兴趣,根据回复原则,选择不回复,等待其他人发言。",
|
||||
"emoji_query": ""
|
||||
}}
|
||||
|
||||
请输出你的决策 JSON:
|
||||
''', # 使用三引号避免内部引号问题
|
||||
"planner_prompt", # 保持名称不变,替换内容
|
||||
)
|
||||
|
||||
Prompt(
|
||||
|
||||
Reference in New Issue
Block a user