feat:支持qwen3模型的enable_thinking参数和thinking_budget参数
This commit is contained in:
@@ -32,7 +32,7 @@ def init_prompt():
|
||||
{chat_observe_info}
|
||||
|
||||
现在请你根据现有的信息,思考自我认同:请严格遵守以下规则
|
||||
1. 请严格参考最上方的人设,适当参考记忆和当前聊天内容
|
||||
1. 请严格参考最上方的人设,适当参考记忆和当前聊天内容,不要被记忆和当前聊天内容中相反的内容误导
|
||||
2. 你是一个什么样的人,你和群里的人关系如何
|
||||
3. 你的形象是什么
|
||||
4. 思考有没有人提到你,或者图片与你有关
|
||||
|
||||
@@ -180,8 +180,9 @@ class ActionPlanner:
|
||||
# --- 调用 LLM (普通文本生成) ---
|
||||
llm_content = None
|
||||
try:
|
||||
llm_content, _, _ = await self.planner_llm.generate_response(prompt=prompt)
|
||||
llm_content, reasoning_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}")
|
||||
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}"
|
||||
|
||||
@@ -26,7 +26,7 @@ class StructureObservation:
|
||||
for structured_info in self.structured_info:
|
||||
if structured_info.get("ttl") > 0:
|
||||
structured_info["ttl"] -= 1
|
||||
observed_structured_infos.append(structured_info)
|
||||
logger.debug(f"观察到结构化信息仍旧在: {structured_info}")
|
||||
observed_structured_infos.append(structured_info)
|
||||
logger.debug(f"观察到结构化信息仍旧在: {structured_info}")
|
||||
|
||||
self.structured_info = observed_structured_infos
|
||||
|
||||
@@ -117,6 +117,9 @@ class LLMRequest:
|
||||
self.model_name: str = model["name"]
|
||||
self.params = kwargs
|
||||
|
||||
self.enable_thinking = model.get("enable_thinking", False)
|
||||
self.temp = model.get("temp", 0.7)
|
||||
self.thinking_budget = model.get("thinking_budget", 4096)
|
||||
self.stream = model.get("stream", False)
|
||||
self.pri_in = model.get("pri_in", 0)
|
||||
self.pri_out = model.get("pri_out", 0)
|
||||
@@ -601,8 +604,9 @@ class LLMRequest:
|
||||
new_params = dict(params)
|
||||
|
||||
if self.model_name.lower() in self.MODELS_NEEDING_TRANSFORMATION:
|
||||
# 删除 'temperature' 参数(如果存在)
|
||||
new_params.pop("temperature", None)
|
||||
# 删除 'temperature' 参数(如果存在),但避免删除我们在_build_payload中添加的自定义温度
|
||||
if "temperature" in new_params and new_params["temperature"] == 0.7:
|
||||
new_params.pop("temperature")
|
||||
# 如果存在 'max_tokens',则重命名为 'max_completion_tokens'
|
||||
if "max_tokens" in new_params:
|
||||
new_params["max_completion_tokens"] = new_params.pop("max_tokens")
|
||||
@@ -632,6 +636,18 @@ class LLMRequest:
|
||||
"messages": messages,
|
||||
**params_copy,
|
||||
}
|
||||
|
||||
# 添加temp参数(如果不是默认值0.7)
|
||||
if self.temp != 0.7:
|
||||
payload["temperature"] = self.temp
|
||||
|
||||
# 添加enable_thinking参数(如果不是默认值False)
|
||||
if not self.enable_thinking:
|
||||
payload["enable_thinking"] = False
|
||||
|
||||
if self.thinking_budget != 4096:
|
||||
payload["thinking_budget"] = self.thinking_budget
|
||||
|
||||
if "max_tokens" not in payload and "max_completion_tokens" not in payload:
|
||||
payload["max_tokens"] = global_config.model.model_max_output_length
|
||||
# 如果 payload 中依然存在 max_tokens 且需要转换,在这里进行再次检查
|
||||
|
||||
@@ -11,7 +11,7 @@ class MuteAction(PluginAction):
|
||||
|
||||
action_name = "mute_action"
|
||||
action_description = (
|
||||
"如果某人违反了公序良俗,或者别人戳你太多,或者某人刷屏,一定要禁言某人,如果你很生气,可以禁言某人,可以自选禁言时长,视严重程度而定"
|
||||
"如果某人违反了公序良俗,或者别人戳你太多,或者某人刷屏,一定要禁言某人,如果你很生气,可以禁言某人,可以自选禁言时长,视严重程度而定。"
|
||||
)
|
||||
action_parameters = {
|
||||
"target": "禁言对象,输入你要禁言的对象的名字,必填",
|
||||
|
||||
Reference in New Issue
Block a user