gardel/Mofox-Core
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

remove：移除willing系统，移除reply2，移除能量值,移除reply_to改为message

Browse Source
This commit is contained in:
SengokuCola
2025-08-11 13:18:17 +08:00
parent 4b59eda5c0
commit 1f91967d2d
17 changed files with 155 additions and 893 deletions
Download Patch File Download Diff File Expand all files Collapse all files

120
src/chat/replyer/default_generator.py
View File

@@ -121,40 +121,11 @@ class DefaultReplyer:
def __init__(
self,
chat_stream: ChatStream,
model_set_with_weight: Optional[List[Tuple[TaskConfig, float]]] = None,
request_type: str = "focus.replyer",
request_type: str = "replyer",
):
self.request_type = request_type
if model_set_with_weight:
# self.express_model_configs = model_configs
self.model_set: List[Tuple[TaskConfig, float]] = model_set_with_weight
else:
# 当未提供配置时，使用默认配置并赋予默认权重
# model_config_1 = global_config.model.replyer_1.copy()
# model_config_2 = global_config.model.replyer_2.copy()
prob_first = global_config.chat.replyer_random_probability
# model_config_1["weight"] = prob_first
# model_config_2["weight"] = 1.0 - prob_first
# self.express_model_configs = [model_config_1, model_config_2]
self.model_set = [
(model_config.model_task_config.replyer_1, prob_first),
(model_config.model_task_config.replyer_2, 1.0 - prob_first),
]
# if not self.express_model_configs:
# logger.warning("未找到有效的模型配置，回复生成可能会失败。")
# # 提供一个最终的回退，以防止在空列表上调用 random.choice
# fallback_config = global_config.model.replyer_1.copy()
# fallback_config.setdefault("weight", 1.0)
# self.express_model_configs = [fallback_config]
self.express_model = LLMRequest(model_set=model_config.model_task_config.replyer, request_type=request_type)
self.chat_stream = chat_stream
self.is_group_chat, self.chat_target_info = get_chat_type_and_target_info(self.chat_stream.stream_id)
self.heart_fc_sender = HeartFCSender()
self.memory_activator = MemoryActivator()
self.instant_memory = InstantMemory(chat_id=self.chat_stream.stream_id)
@@ -163,14 +134,6 @@ class DefaultReplyer:
self.tool_executor = ToolExecutor(chat_id=self.chat_stream.stream_id, enable_cache=True, cache_ttl=3)
def _select_weighted_models_config(self) -> Tuple[TaskConfig, float]:
"""使用加权随机选择来挑选一个模型配置"""
configs = self.model_set
# 提取权重，如果模型配置中没有'weight'键，则默认为1.0
weights = [weight for _, weight in configs]
return random.choices(population=configs, weights=weights, k=1)[0]
async def generate_reply_with_context(
self,
reply_to: str = "",
@@ -179,8 +142,8 @@ class DefaultReplyer:
enable_tool: bool = True,
from_plugin: bool = True,
stream_id: Optional[str] = None,
reply_message: Optional[Dict[str, Any]] = None,
) -> Tuple[bool, Optional[Dict[str, Any]], Optional[str]]:
# sourcery skip: merge-nested-ifs
"""
回复器 (Replier): 负责生成回复文本的核心逻辑。
@@ -205,6 +168,7 @@ class DefaultReplyer:
extra_info=extra_info,
available_actions=available_actions,
enable_tool=enable_tool,
reply_message=reply_message,
)
if not prompt:
@@ -302,16 +266,11 @@ class DefaultReplyer:
traceback.print_exc()
return False, None, prompt if return_prompt else None
async def build_relation_info(self, reply_to: str = ""):
async def build_relation_info(self, sender: str, target: str):
if not global_config.relationship.enable_relationship:
return ""
relationship_fetcher = relationship_fetcher_manager.get_fetcher(self.chat_stream.stream_id)
if not reply_to:
return ""
sender, text = self._parse_reply_target(reply_to)
if not sender or not text:
return ""
# 获取用户ID
person_info_manager = get_person_info_manager()
@@ -418,7 +377,7 @@ class DefaultReplyer:
return memory_str
async def build_tool_info(self, chat_history: str, reply_to: str = "", enable_tool: bool = True) -> str:
async def build_tool_info(self, chat_history: str, sender: str, target: str, enable_tool: bool = True) -> str:
"""构建工具信息块
Args:
@@ -433,18 +392,11 @@ class DefaultReplyer:
if not enable_tool:
return ""
if not reply_to:
return ""
sender, text = self._parse_reply_target(reply_to)
if not text:
return ""
try:
# 使用工具执行器获取信息
tool_results, _, _ = await self.tool_executor.execute_from_chat_message(
sender=sender, target_message=text, chat_history=chat_history, return_details=False
sender=sender, target_message=target, chat_history=chat_history, return_details=False
)
if tool_results:
@@ -672,7 +624,8 @@ class DefaultReplyer:
extra_info: str = "",
available_actions: Optional[Dict[str, ActionInfo]] = None,
enable_tool: bool = True,
) -> str: # sourcery skip: merge-else-if-into-elif, remove-redundant-if
reply_message: Optional[Dict[str, Any]] = None,
) -> str:
"""
构建回复器上下文
@@ -682,7 +635,7 @@ class DefaultReplyer:
available_actions: 可用动作
enable_timeout: 是否启用超时处理
enable_tool: 是否启用工具调用
reply_message: 回复的原始消息
Returns:
str: 构建好的上下文
"""
@@ -698,8 +651,21 @@ class DefaultReplyer:
mood_prompt = chat_mood.mood_state
else:
mood_prompt = ""
sender, target = self._parse_reply_target(reply_to)
if reply_to:
#兼容旧的reply_to
sender, target = self._parse_reply_target(reply_to)
else:
# 获取 platform，如果不存在则从 chat_stream 获取，如果还是 None 则使用默认值
platform = reply_message.get("chat_info_platform")
person_id = person_info_manager.get_person_id(
platform, # type: ignore
reply_message.get("user_id"), # type: ignore
)
person_name = await person_info_manager.get_value(person_id, "person_name")
sender = person_name
target = reply_message.get('processed_plain_text')
person_info_manager = get_person_info_manager()
person_id = person_info_manager.get_person_id_by_person_name(sender)
user_id = person_info_manager.get_value_sync(person_id, "user_id")
@@ -744,12 +710,12 @@ class DefaultReplyer:
self._time_and_run_task(
self.build_expression_habits(chat_talking_prompt_short, target), "expression_habits"
),
self._time_and_run_task(self.build_relation_info(reply_to), "relation_info"),
self._time_and_run_task(self.build_relation_info(sender, target), "relation_info"),
self._time_and_run_task(self.build_memory_block(chat_talking_prompt_short, target), "memory_block"),
self._time_and_run_task(
self.build_tool_info(chat_talking_prompt_short, reply_to, enable_tool=enable_tool), "tool_info"
self.build_tool_info(chat_talking_prompt_short, sender, target, enable_tool=enable_tool), "tool_info"
),
self._time_and_run_task(self.get_prompt_info(chat_talking_prompt_short, reply_to), "prompt_info"),
self._time_and_run_task(self.get_prompt_info(chat_talking_prompt_short, sender, target), "prompt_info"),
)
# 任务名称中英文映射
@@ -899,12 +865,17 @@ class DefaultReplyer:
raw_reply: str,
reason: str,
reply_to: str,
reply_message: Optional[Dict[str, Any]] = None,
) -> str: # sourcery skip: merge-else-if-into-elif, remove-redundant-if
chat_stream = self.chat_stream
chat_id = chat_stream.stream_id
is_group_chat = bool(chat_stream.group_info)
sender, target = self._parse_reply_target(reply_to)
if reply_message:
sender = reply_message.get("sender")
target = reply_message.get("target")
else:
sender, target = self._parse_reply_target(reply_to)
# 添加情绪状态获取
if global_config.mood.enable_mood:
@@ -930,7 +901,7 @@ class DefaultReplyer:
# 并行执行2个构建任务
expression_habits_block, relation_info = await asyncio.gather(
self.build_expression_habits(chat_talking_prompt_half, target),
self.build_relation_info(reply_to),
self.build_relation_info(sender, target),
)
keywords_reaction_prompt = await self.build_keywords_reaction_prompt(target)
@@ -1035,34 +1006,25 @@ class DefaultReplyer:
async def llm_generate_content(self, prompt: str):
with Timer("LLM生成", {}): # 内部计时器，可选保留
# 加权随机选择一个模型配置
selected_model_config, weight = self._select_weighted_models_config()
logger.info(f"使用模型集生成回复: {selected_model_config} (选中概率: {weight})")
express_model = LLMRequest(model_set=selected_model_config, request_type=self.request_type)
# 直接使用已初始化的模型实例
logger.info(f"使用模型集生成回复: {self.express_model.model_for_task}")
if global_config.debug.show_prompt:
logger.info(f"\n{prompt}\n")
else:
logger.debug(f"\n{prompt}\n")
content, (reasoning_content, model_name, tool_calls) = await express_model.generate_response_async(prompt)
content, (reasoning_content, model_name, tool_calls) = await self.express_model.generate_response_async(prompt)
logger.debug(f"replyer生成内容: {content}")
return content, reasoning_content, model_name, tool_calls
async def get_prompt_info(self, message: str, reply_to: str):
async def get_prompt_info(self, message: str, sender: str, target: str):
related_info = ""
start_time = time.time()
from src.plugins.built_in.knowledge.lpmm_get_knowledge import SearchKnowledgeFromLPMMTool
if not reply_to:
logger.debug("没有回复对象，跳过获取知识库内容")
return ""
sender, content = self._parse_reply_target(reply_to)
if not content:
logger.debug("回复对象内容为空，跳过获取知识库内容")
return ""
logger.debug(f"获取知识库内容，元消息：{message[:30]}...，消息长度: {len(message)}")
# 从LPMM知识库获取知识
try:
@@ -1080,7 +1042,7 @@ class DefaultReplyer:
time_now=time_now,
chat_history=message,
sender=sender,
target_message=content,
target_message=target,
)
_, _, _, _, tool_calls = await llm_api.generate_with_model_with_tools(
prompt,