feat(reply): 引入统一格式过滤器，优化回复内容清理逻辑
This commit is contained in:
Windpicker-owo
@@ -1812,8 +1812,9 @@ class DefaultReplyer:
|
||||
# 移除 [SPLIT] 标记，防止消息被分割
|
||||
content = content.replace("[SPLIT]", "")
|
||||
|
||||
# 应用增强的格式过滤器
|
||||
content = self._apply_system_format_filter(content)
|
||||
# 应用统一的格式过滤器
|
||||
from src.chat.utils.utils import filter_system_format_content
|
||||
content = filter_system_format_content(content)
|
||||
|
||||
logger.debug(f"replyer生成内容: {content}")
|
||||
return content, reasoning_content, model_name, tool_calls
|
||||
@@ -2100,72 +2101,7 @@ class DefaultReplyer:
|
||||
except Exception as e:
|
||||
logger.error(f"存储聊天记忆失败: {e}")
|
||||
|
||||
def _apply_system_format_filter(self, content: str) -> str:
|
||||
"""
|
||||
应用增强的系统格式过滤器，移除各种系统格式化文本
|
||||
|
||||
此方法过滤以下类型的系统格式化内容：
|
||||
1. 回复格式：[回复xxx]，说：xxx
|
||||
2. 表情包格式：[表情包：xxx]
|
||||
3. 图片格式：[图片:xxx]
|
||||
4. @格式：@<xxx>
|
||||
5. 原有的[回复开头格式
|
||||
|
||||
Args:
|
||||
content: 原始内容
|
||||
|
||||
Returns:
|
||||
过滤后的内容
|
||||
"""
|
||||
import re
|
||||
|
||||
if not content:
|
||||
return content
|
||||
|
||||
original_content = content
|
||||
cleaned_content = content.strip()
|
||||
|
||||
# 1. 移除回复格式：[回复xxx]，说：xxx（各种变体）
|
||||
# 匹配所有包含"]，说："格式的回复
|
||||
cleaned_content = re.sub(r"\[回复[^\]]*\]，说：\s*", "", cleaned_content)
|
||||
# 匹配 [回复<xxx:数字>]，说：xxx 格式
|
||||
cleaned_content = re.sub(r"\[回复<[^>]*>\]，说：\s*", "", cleaned_content)
|
||||
|
||||
# 2. 处理原有的[回复开头格式（保持向后兼容）
|
||||
# 注意：这步要在上面处理完成后再执行，避免冲突
|
||||
if cleaned_content.startswith("[回复"):
|
||||
last_bracket_index = cleaned_content.rfind("]")
|
||||
if last_bracket_index != -1:
|
||||
cleaned_content = cleaned_content[last_bracket_index + 1 :].strip()
|
||||
|
||||
# 3. 移除表情包格式：[表情包：xxx]
|
||||
cleaned_content = re.sub(r"\[表情包：[^\]]*\]", "", cleaned_content)
|
||||
|
||||
# 4. 移除图片格式：[图片:xxx]
|
||||
cleaned_content = re.sub(r"\[图片:[^\]]*\]", "", cleaned_content)
|
||||
|
||||
# 5. 移除@格式：@<xxx>
|
||||
cleaned_content = re.sub(r"@<[^>]*>", "", cleaned_content)
|
||||
|
||||
# 6. 移除其他可能的系统格式
|
||||
# [表情包(描述生成失败)] 等错误格式
|
||||
cleaned_content = re.sub(r"\[表情包\([^)]*\)\]", "", cleaned_content)
|
||||
# [图片(描述生成失败)] 等错误格式
|
||||
cleaned_content = re.sub(r"\[图片\([^)]*\)\]", "", cleaned_content)
|
||||
|
||||
# 清理多余空格
|
||||
cleaned_content = re.sub(r"\s+", " ", cleaned_content).strip()
|
||||
|
||||
# 记录过滤操作
|
||||
if cleaned_content != original_content.strip():
|
||||
logger.info(
|
||||
f"[格式过滤器] 检测到并清理了系统格式化文本。"
|
||||
f"原始内容: '{original_content}', "
|
||||
f"清理后: '{cleaned_content}'"
|
||||
)
|
||||
|
||||
return cleaned_content
|
||||
|
||||
|
||||
|
||||
def weighted_sample_no_replacement(items, weights, k) -> list:
|
||||
"""
|
||||
|
||||
@@ -923,3 +923,69 @@ def assign_message_ids_flexible(
|
||||
# # 增强版本 - 使用时间戳
|
||||
# result3 = assign_message_ids_flexible(messages, prefix="ts", use_timestamp=True)
|
||||
# # 结果: [{'id': 'ts123a1b', 'message': 'Hello'}, {'id': 'ts123c2d', 'message': 'World'}, {'id': 'ts123e3f', 'message': 'Test message'}]
|
||||
|
||||
|
||||
def filter_system_format_content(content: str | None) -> str:
|
||||
"""
|
||||
过滤系统格式化内容，移除回复、@、图片、表情包等系统生成的格式文本
|
||||
|
||||
此方法过滤以下类型的系统格式化内容：
|
||||
1. 回复格式：[回复xxx]，说：xxx
|
||||
2. 表情包格式：[表情包：xxx]
|
||||
3. 图片格式：[图片:xxx]
|
||||
4. @格式：@<xxx>
|
||||
5. 错误格式：[表情包(...)]、[图片(...)]
|
||||
6. [回复开头的格式
|
||||
|
||||
Args:
|
||||
content: 原始内容
|
||||
|
||||
Returns:
|
||||
过滤后的纯文本内容
|
||||
"""
|
||||
if not content:
|
||||
return ""
|
||||
|
||||
original_content = content
|
||||
cleaned_content = content.strip()
|
||||
|
||||
# 1. 移除回复格式：[回复xxx]，说：xxx（各种变体）
|
||||
# 匹配所有包含"]，说："格式的回复
|
||||
cleaned_content = re.sub(r"\[回复[^\]]*\]，说：\s*", "", cleaned_content)
|
||||
# 匹配 [回复<xxx:数字>]，说：xxx 格式
|
||||
cleaned_content = re.sub(r"\[回复<[^>]*>\]，说：\s*", "", cleaned_content)
|
||||
|
||||
# 2. 处理原有的[回复开头格式（保持向后兼容）
|
||||
# 注意：这步要在上面处理完成后再执行，避免冲突
|
||||
if cleaned_content.startswith("[回复"):
|
||||
last_bracket_index = cleaned_content.rfind("]")
|
||||
if last_bracket_index != -1:
|
||||
cleaned_content = cleaned_content[last_bracket_index + 1 :].strip()
|
||||
|
||||
# 3. 移除表情包格式：[表情包：xxx]
|
||||
cleaned_content = re.sub(r"\[表情包：[^\]]*\]", "", cleaned_content)
|
||||
|
||||
# 4. 移除图片格式：[图片:xxx]
|
||||
cleaned_content = re.sub(r"\[图片:[^\]]*\]", "", cleaned_content)
|
||||
|
||||
# 5. 移除@格式：@<xxx>
|
||||
cleaned_content = re.sub(r"@<[^>]*>", "", cleaned_content)
|
||||
|
||||
# 6. 移除其他可能的系统格式
|
||||
# [表情包(描述生成失败)] 等错误格式
|
||||
cleaned_content = re.sub(r"\[表情包\([^)]*\)\]", "", cleaned_content)
|
||||
# [图片(描述生成失败)] 等错误格式
|
||||
cleaned_content = re.sub(r"\[图片\([^)]*\)\]", "", cleaned_content)
|
||||
|
||||
# 清理多余空格
|
||||
cleaned_content = re.sub(r"\s+", " ", cleaned_content).strip()
|
||||
|
||||
# 记录过滤操作
|
||||
if cleaned_content != original_content.strip():
|
||||
logger.info(
|
||||
f"[系统格式过滤器] 检测到并清理了系统格式化文本。"
|
||||
f"原始内容: '{original_content}', "
|
||||
f"清理后: '{cleaned_content}'"
|
||||
)
|
||||
|
||||
return cleaned_content
|
||||
|
||||
@@ -412,7 +412,15 @@ class ProactiveThinkingPlanner:
|
||||
return None
|
||||
|
||||
logger.info(f"生成回复成功: {response[:50]}...")
|
||||
return response.strip()
|
||||
|
||||
# 应用格式过滤器，确保回复内容不包含系统格式化文本
|
||||
from src.chat.utils.utils import filter_system_format_content
|
||||
filtered_response = filter_system_format_content(response.strip())
|
||||
|
||||
if filtered_response != response.strip():
|
||||
logger.debug(f"主动思考回复已过滤系统格式: '{response.strip()}' -> '{filtered_response}'")
|
||||
|
||||
return filtered_response
|
||||
|
||||
except Exception as e:
|
||||
logger.error(f"生成回复失败: {e}", exc_info=True)
|
||||
|
||||
Reference in New Issue
Block a user