mirror of
https://github.com/xinnan-tech/xiaozhi-esp32-server.git
synced 2026-07-30 16:23:57 +08:00
feat: 统一LLM错误处理并添加系统错误回复配置
在多个LLM提供者中移除try-catch块，将错误处理统一到connection.py的流处理层 添加system_error_response配置项，支持自定义系统错误时的回复内容 在意图识别和流处理中捕获异常时返回配置的错误回复，避免硬编码错误信息 Fixes #2075
This commit is contained in:
huozaimengli
@@ -25,151 +25,141 @@ class LLMProvider(LLMProviderBase):
|
||||
self.is_qwen3 = self.model_name and self.model_name.lower().startswith("qwen3")
|
||||
|
||||
def response(self, session_id, dialogue, **kwargs):
|
||||
try:
|
||||
# 如果是qwen3模型，在用户最后一条消息中添加/no_think指令
|
||||
if self.is_qwen3:
|
||||
# 复制对话列表，避免修改原始对话
|
||||
dialogue_copy = dialogue.copy()
|
||||
# 如果是qwen3模型，在用户最后一条消息中添加/no_think指令
|
||||
if self.is_qwen3:
|
||||
# 复制对话列表，避免修改原始对话
|
||||
dialogue_copy = dialogue.copy()
|
||||
|
||||
# 找到最后一条用户消息
|
||||
for i in range(len(dialogue_copy) - 1, -1, -1):
|
||||
if dialogue_copy[i]["role"] == "user":
|
||||
# 在用户消息前添加/no_think指令
|
||||
dialogue_copy[i]["content"] = (
|
||||
"/no_think " + dialogue_copy[i]["content"]
|
||||
)
|
||||
logger.bind(tag=TAG).debug(f"为qwen3模型添加/no_think指令")
|
||||
break
|
||||
|
||||
# 使用修改后的对话
|
||||
dialogue = dialogue_copy
|
||||
|
||||
responses = self.client.chat.completions.create(
|
||||
model=self.model_name, messages=dialogue, stream=True
|
||||
)
|
||||
is_active = True
|
||||
# 用于处理跨chunk的标签
|
||||
buffer = ""
|
||||
|
||||
for chunk in responses:
|
||||
try:
|
||||
delta = (
|
||||
chunk.choices[0].delta
|
||||
if getattr(chunk, "choices", None)
|
||||
else None
|
||||
# 找到最后一条用户消息
|
||||
for i in range(len(dialogue_copy) - 1, -1, -1):
|
||||
if dialogue_copy[i]["role"] == "user":
|
||||
# 在用户消息前添加/no_think指令
|
||||
dialogue_copy[i]["content"] = (
|
||||
"/no_think " + dialogue_copy[i]["content"]
|
||||
)
|
||||
content = delta.content if hasattr(delta, "content") else ""
|
||||
logger.bind(tag=TAG).debug(f"为qwen3模型添加/no_think指令")
|
||||
break
|
||||
|
||||
if content:
|
||||
# 将内容添加到缓冲区
|
||||
buffer += content
|
||||
# 使用修改后的对话
|
||||
dialogue = dialogue_copy
|
||||
|
||||
# 处理缓冲区中的标签
|
||||
while "<think>" in buffer and "</think>" in buffer:
|
||||
# 找到完整的<think></think>标签并移除
|
||||
pre = buffer.split("<think>", 1)[0]
|
||||
post = buffer.split("</think>", 1)[1]
|
||||
buffer = pre + post
|
||||
responses = self.client.chat.completions.create(
|
||||
model=self.model_name, messages=dialogue, stream=True
|
||||
)
|
||||
is_active = True
|
||||
# 用于处理跨chunk的标签
|
||||
buffer = ""
|
||||
|
||||
# 处理只有开始标签的情况
|
||||
if "<think>" in buffer:
|
||||
is_active = False
|
||||
buffer = buffer.split("<think>", 1)[0]
|
||||
for chunk in responses:
|
||||
try:
|
||||
delta = (
|
||||
chunk.choices[0].delta
|
||||
if getattr(chunk, "choices", None)
|
||||
else None
|
||||
)
|
||||
content = delta.content if hasattr(delta, "content") else ""
|
||||
|
||||
# 处理只有结束标签的情况
|
||||
if "</think>" in buffer:
|
||||
is_active = True
|
||||
buffer = buffer.split("</think>", 1)[1]
|
||||
if content:
|
||||
# 将内容添加到缓冲区
|
||||
buffer += content
|
||||
|
||||
# 如果当前处于活动状态且缓冲区有内容，则输出
|
||||
if is_active and buffer:
|
||||
yield buffer
|
||||
buffer = "" # 清空缓冲区
|
||||
# 处理缓冲区中的标签
|
||||
while "<think>" in buffer and "</think>" in buffer:
|
||||
# 找到完整的<think></think>标签并移除
|
||||
pre = buffer.split("<think>", 1)[0]
|
||||
post = buffer.split("</think>", 1)[1]
|
||||
buffer = pre + post
|
||||
|
||||
except Exception as e:
|
||||
logger.bind(tag=TAG).error(f"Error processing chunk: {e}")
|
||||
# 处理只有开始标签的情况
|
||||
if "<think>" in buffer:
|
||||
is_active = False
|
||||
buffer = buffer.split("<think>", 1)[0]
|
||||
|
||||
except Exception as e:
|
||||
logger.bind(tag=TAG).error(f"Error in Ollama response generation: {e}")
|
||||
yield "【Ollama服务响应异常】"
|
||||
# 处理只有结束标签的情况
|
||||
if "</think>" in buffer:
|
||||
is_active = True
|
||||
buffer = buffer.split("</think>", 1)[1]
|
||||
|
||||
# 如果当前处于活动状态且缓冲区有内容，则输出
|
||||
if is_active and buffer:
|
||||
yield buffer
|
||||
buffer = "" # 清空缓冲区
|
||||
|
||||
except Exception as e:
|
||||
logger.bind(tag=TAG).error(f"Error processing chunk: {e}")
|
||||
|
||||
def response_with_functions(self, session_id, dialogue, functions=None):
|
||||
try:
|
||||
# 如果是qwen3模型，在用户最后一条消息中添加/no_think指令
|
||||
if self.is_qwen3:
|
||||
# 复制对话列表，避免修改原始对话
|
||||
dialogue_copy = dialogue.copy()
|
||||
# 如果是qwen3模型，在用户最后一条消息中添加/no_think指令
|
||||
if self.is_qwen3:
|
||||
# 复制对话列表，避免修改原始对话
|
||||
dialogue_copy = dialogue.copy()
|
||||
|
||||
# 找到最后一条用户消息
|
||||
for i in range(len(dialogue_copy) - 1, -1, -1):
|
||||
if dialogue_copy[i]["role"] == "user":
|
||||
# 在用户消息前添加/no_think指令
|
||||
dialogue_copy[i]["content"] = (
|
||||
"/no_think " + dialogue_copy[i]["content"]
|
||||
)
|
||||
logger.bind(tag=TAG).debug(f"为qwen3模型添加/no_think指令")
|
||||
break
|
||||
|
||||
# 使用修改后的对话
|
||||
dialogue = dialogue_copy
|
||||
|
||||
stream = self.client.chat.completions.create(
|
||||
model=self.model_name,
|
||||
messages=dialogue,
|
||||
stream=True,
|
||||
tools=functions,
|
||||
)
|
||||
|
||||
is_active = True
|
||||
buffer = ""
|
||||
|
||||
for chunk in stream:
|
||||
try:
|
||||
delta = (
|
||||
chunk.choices[0].delta
|
||||
if getattr(chunk, "choices", None)
|
||||
else None
|
||||
)
|
||||
content = delta.content if hasattr(delta, "content") else None
|
||||
tool_calls = (
|
||||
delta.tool_calls if hasattr(delta, "tool_calls") else None
|
||||
# 找到最后一条用户消息
|
||||
for i in range(len(dialogue_copy) - 1, -1, -1):
|
||||
if dialogue_copy[i]["role"] == "user":
|
||||
# 在用户消息前添加/no_think指令
|
||||
dialogue_copy[i]["content"] = (
|
||||
"/no_think " + dialogue_copy[i]["content"]
|
||||
)
|
||||
logger.bind(tag=TAG).debug(f"为qwen3模型添加/no_think指令")
|
||||
break
|
||||
|
||||
# 如果是工具调用，直接传递
|
||||
if tool_calls:
|
||||
yield None, tool_calls
|
||||
continue
|
||||
# 使用修改后的对话
|
||||
dialogue = dialogue_copy
|
||||
|
||||
# 处理文本内容
|
||||
if content:
|
||||
# 将内容添加到缓冲区
|
||||
buffer += content
|
||||
stream = self.client.chat.completions.create(
|
||||
model=self.model_name,
|
||||
messages=dialogue,
|
||||
stream=True,
|
||||
tools=functions,
|
||||
)
|
||||
|
||||
# 处理缓冲区中的标签
|
||||
while "<think>" in buffer and "</think>" in buffer:
|
||||
# 找到完整的<think></think>标签并移除
|
||||
pre = buffer.split("<think>", 1)[0]
|
||||
post = buffer.split("</think>", 1)[1]
|
||||
buffer = pre + post
|
||||
is_active = True
|
||||
buffer = ""
|
||||
|
||||
# 处理只有开始标签的情况
|
||||
if "<think>" in buffer:
|
||||
is_active = False
|
||||
buffer = buffer.split("<think>", 1)[0]
|
||||
for chunk in stream:
|
||||
try:
|
||||
delta = (
|
||||
chunk.choices[0].delta
|
||||
if getattr(chunk, "choices", None)
|
||||
else None
|
||||
)
|
||||
content = delta.content if hasattr(delta, "content") else None
|
||||
tool_calls = (
|
||||
delta.tool_calls if hasattr(delta, "tool_calls") else None
|
||||
)
|
||||
|
||||
# 处理只有结束标签的情况
|
||||
if "</think>" in buffer:
|
||||
is_active = True
|
||||
buffer = buffer.split("</think>", 1)[1]
|
||||
|
||||
# 如果当前处于活动状态且缓冲区有内容，则输出
|
||||
if is_active and buffer:
|
||||
yield buffer, None
|
||||
buffer = "" # 清空缓冲区
|
||||
except Exception as e:
|
||||
logger.bind(tag=TAG).error(f"Error processing function chunk: {e}")
|
||||
# 如果是工具调用，直接传递
|
||||
if tool_calls:
|
||||
yield None, tool_calls
|
||||
continue
|
||||
|
||||
except Exception as e:
|
||||
logger.bind(tag=TAG).error(f"Error in Ollama function call: {e}")
|
||||
yield f"【Ollama服务响应异常: {str(e)}】", None
|
||||
# 处理文本内容
|
||||
if content:
|
||||
# 将内容添加到缓冲区
|
||||
buffer += content
|
||||
|
||||
# 处理缓冲区中的标签
|
||||
while "<think>" in buffer and "</think>" in buffer:
|
||||
# 找到完整的<think></think>标签并移除
|
||||
pre = buffer.split("<think>", 1)[0]
|
||||
post = buffer.split("</think>", 1)[1]
|
||||
buffer = pre + post
|
||||
|
||||
# 处理只有开始标签的情况
|
||||
if "<think>" in buffer:
|
||||
is_active = False
|
||||
buffer = buffer.split("<think>", 1)[0]
|
||||
|
||||
# 处理只有结束标签的情况
|
||||
if "</think>" in buffer:
|
||||
is_active = True
|
||||
buffer = buffer.split("</think>", 1)[1]
|
||||
|
||||
# 如果当前处于活动状态且缓冲区有内容，则输出
|
||||
if is_active and buffer:
|
||||
yield buffer, None
|
||||
buffer = "" # 清空缓冲区
|
||||
except Exception as e:
|
||||
logger.bind(tag=TAG).error(f"Error processing function chunk: {e}")
|
||||
continue
|
||||
|
||||
Reference in New Issue
Block a user