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

Merge branch 'master' of https://github.com/MaiBot-Plus/MaiMbot-Pro-Max

Browse Source
This commit is contained in:
雅诺狐
2025-08-25 16:36:59 +08:00
parent 0bcc0ba8b5 119633e3c3
commit 8054262827
8 changed files with 146 additions and 60 deletions
Download Patch File Download Diff File Expand all files Collapse all files

20
__main__.py Normal file
View File

@@ -0,0 +1,20 @@
#!/usr/bin/env python3
"""Bot项目的主入口点"""
if __name__ == "__main__":
# 设置Python路径并执行bot.py
import sys
import os
from pathlib import Path
# 添加当前目录到Python路径
current_dir = Path(__file__).parent
sys.path.insert(0, str(current_dir))
# 执行bot.py的代码
bot_file = current_dir / "bot.py"
with open(bot_file, 'r', encoding='utf-8') as f:
exec(f.read())
# 这个文件是为了适配一键包使用的，在一键包项目之外没有用

9
src/chat/chat_loop/wakeup_manager.py
View File

@@ -26,6 +26,8 @@ class WakeUpManager:
self.angry_start_time = 0.0 # 愤怒状态开始时间
self.last_decay_time = time.time() # 上次衰减时间
self._decay_task: Optional[asyncio.Task] = None
self.last_log_time = 0
self.log_interval = 30
# 从配置文件获取参数
wakeup_config = global_config.wakeup_system
@@ -123,7 +125,12 @@ class WakeUpManager:
# 群聊未被艾特，不增加唤醒度
return False
logger.info(f"{self.context.log_prefix} 唤醒度变化: {old_value:.1f} -> {self.wakeup_value:.1f} (阈值: {self.wakeup_threshold})")
current_time = time.time()
if current_time - self.last_log_time > self.log_interval:
logger.info(f"{self.context.log_prefix} 唤醒度变化: {old_value:.1f} -> {self.wakeup_value:.1f} (阈值: {self.wakeup_threshold})")
self.last_log_time = current_time
else:
logger.debug(f"{self.context.log_prefix} 唤醒度变化: {old_value:.1f} -> {self.wakeup_value:.1f} (阈值: {self.wakeup_threshold})")
# 检查是否达到唤醒阈值
if self.wakeup_value >= self.wakeup_threshold:

140
src/chat/utils/utils_video.py
View File

@@ -130,6 +130,8 @@ class VideoAnalyzer:
# 新增的线程池配置
self.use_multiprocessing = getattr(config, 'use_multiprocessing', True)
self.max_workers = getattr(config, 'max_workers', 2)
self.frame_extraction_mode = getattr(config, 'frame_extraction_mode', 'fixed_number')
self.frame_interval_seconds = getattr(config, 'frame_interval_seconds', 2.0)
# 将配置文件中的模式映射到内部使用的模式名称
config_mode = config.analysis_mode
@@ -163,6 +165,8 @@ class VideoAnalyzer:
self.enable_frame_timing = True
self.use_multiprocessing = True # 默认启用线程池
self.max_workers = 2 # 默认最大2个线程
self.frame_extraction_mode = "fixed_number"
self.frame_interval_seconds = 2.0
self.batch_analysis_prompt = """请分析这个视频的内容。这些图片是从视频中按时间顺序提取的关键帧。
请提供详细的分析，包括：
@@ -314,6 +318,8 @@ class VideoAnalyzer:
async def _extract_frames_fallback(self, video_path: str) -> List[Tuple[str, float]]:
"""帧提取的降级方法 - 原始异步版本"""
frames = []
frame_count = 0
extracted_count = 0
cap = cv2.VideoCapture(video_path)
fps = cap.get(cv2.CAP_PROP_FPS)
total_frames = int(cap.get(cv2.CAP_PROP_FRAME_COUNT))
@@ -321,61 +327,97 @@ class VideoAnalyzer:
logger.info(f"视频信息: {total_frames}帧, {fps:.2f}FPS, {duration:.2f}秒")
# 使用numpy优化帧间隔计算
if duration > 0:
frame_interval = max(1, int(duration / self.max_frames * fps))
else:
frame_interval = 30 # 默认间隔
logger.info(f"计算得出帧间隔: {frame_interval} (将提取约{min(self.max_frames, total_frames // frame_interval + 1)}帧)")
# 使用numpy计算目标帧位置
target_frames = np.arange(0, min(self.max_frames, total_frames // frame_interval + 1)) * frame_interval
target_frames = target_frames[target_frames < total_frames].astype(int)
extracted_count = 0
for target_frame in target_frames:
# 跳转到目标帧
cap.set(cv2.CAP_PROP_POS_FRAMES, target_frame)
ret, frame = cap.read()
if not ret:
continue
# 使用numpy优化图像处理
frame_rgb = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)
if self.frame_extraction_mode == "time_interval":
# 新模式：按时间间隔抽帧
time_interval = self.frame_interval_seconds
next_frame_time = 0.0
# 转换为PIL图像并使用numpy进行尺寸计算
height, width = frame_rgb.shape[:2]
max_dim = max(height, width)
if max_dim > self.max_image_size:
# 使用numpy计算缩放比例
ratio = self.max_image_size / max_dim
new_width = int(width * ratio)
new_height = int(height * ratio)
while cap.isOpened():
ret, frame = cap.read()
if not ret:
break
# 使用opencv进行高效缩放
frame_resized = cv2.resize(frame_rgb, (new_width, new_height), interpolation=cv2.INTER_LANCZOS4)
pil_image = Image.fromarray(frame_resized)
current_time = cap.get(cv2.CAP_PROP_POS_MSEC) / 1000.0
if current_time >= next_frame_time:
# 转换为PIL图像并压缩
frame_rgb = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)
pil_image = Image.fromarray(frame_rgb)
# 调整图像大小
if max(pil_image.size) > self.max_image_size:
ratio = self.max_image_size / max(pil_image.size)
new_size = tuple(int(dim * ratio) for dim in pil_image.size)
pil_image = pil_image.resize(new_size, Image.Resampling.LANCZOS)
# 转换为base64
buffer = io.BytesIO()
pil_image.save(buffer, format='JPEG', quality=self.frame_quality)
frame_base64 = base64.b64encode(buffer.getvalue()).decode('utf-8')
frames.append((frame_base64, current_time))
extracted_count += 1
logger.debug(f"提取第{extracted_count}帧 (时间: {current_time:.2f}s)")
next_frame_time += time_interval
else:
# 使用numpy优化帧间隔计算
if duration > 0:
frame_interval = max(1, int(duration / self.max_frames * fps))
else:
pil_image = Image.fromarray(frame_rgb)
frame_interval = 30 # 默认间隔
logger.info(f"计算得出帧间隔: {frame_interval} (将提取约{min(self.max_frames, total_frames // frame_interval + 1)}帧)")
# 使用numpy计算目标帧位置
target_frames = np.arange(0, min(self.max_frames, total_frames // frame_interval + 1)) * frame_interval
target_frames = target_frames[target_frames < total_frames].astype(int)
# 转换为base64
buffer = io.BytesIO()
pil_image.save(buffer, format='JPEG', quality=self.frame_quality)
frame_base64 = base64.b64encode(buffer.getvalue()).decode('utf-8')
# 计算时间戳
timestamp = target_frame / fps if fps > 0 else 0
frames.append((frame_base64, timestamp))
extracted_count += 1
logger.debug(f"提取第{extracted_count}帧 (时间: {timestamp:.2f}s, 帧号: {target_frame})")
# 每提取一帧让步一次
await asyncio.sleep(0.001)
extracted_count = 0
for target_frame in target_frames:
# 跳转到目标帧
cap.set(cv2.CAP_PROP_POS_FRAMES, target_frame)
ret, frame = cap.read()
if not ret:
continue
# 使用numpy优化图像处理
frame_rgb = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)
# 转换为PIL图像并使用numpy进行尺寸计算
height, width = frame_rgb.shape[:2]
max_dim = max(height, width)
if max_dim > self.max_image_size:
# 使用numpy计算缩放比例
ratio = self.max_image_size / max_dim
new_width = int(width * ratio)
new_height = int(height * ratio)
# 使用opencv进行高效缩放
frame_resized = cv2.resize(frame_rgb, (new_width, new_height), interpolation=cv2.INTER_LANCZOS4)
pil_image = Image.fromarray(frame_resized)
else:
pil_image = Image.fromarray(frame_rgb)
# 转换为base64
buffer = io.BytesIO()
pil_image.save(buffer, format='JPEG', quality=self.frame_quality)
frame_base64 = base64.b64encode(buffer.getvalue()).decode('utf-8')
# 计算时间戳
timestamp = target_frame / fps if fps > 0 else 0
frames.append((frame_base64, timestamp))
extracted_count += 1
logger.debug(f"提取第{extracted_count}帧 (时间: {timestamp:.2f}s, 帧号: {target_frame})")
# 每提取一帧让步一次
await asyncio.sleep(0.001)
cap.release()
logger.info(f"✅ 成功提取{len(frames)}帧")
return frames

2
src/config/official_configs.py
View File

@@ -619,6 +619,8 @@ class VideoAnalysisConfig(ValidatedConfigBase):
enable: bool = Field(default=True, description="启用")
analysis_mode: str = Field(default="batch_frames", description="分析模式")
frame_extraction_mode: str = Field(default="fixed_number", description="抽帧模式")
frame_interval_seconds: float = Field(default=2.0, description="抽帧时间间隔")
max_frames: int = Field(default=8, description="最大帧数")
frame_quality: int = Field(default=85, description="帧质量")
max_image_size: int = Field(default=800, description="最大图像大小")

22
src/llm_models/utils_model.py
View File

@@ -340,16 +340,22 @@ class LLMRequest:
is_truncated = True
logger.warning("未检测到 [done] 标记，判定为截断")
if (is_empty_reply or is_truncated) and empty_retry_count < max_empty_retry:
empty_retry_count += 1
reason = "空回复" if is_empty_reply else "截断"
logger.warning(f"检测到{reason}，正在进行第 {empty_retry_count}/{max_empty_retry} 次重新生成")
if is_empty_reply or is_truncated:
if empty_retry_count < max_empty_retry:
empty_retry_count += 1
reason = "空回复" if is_empty_reply else "截断"
logger.warning(f"检测到{reason}，正在进行第 {empty_retry_count}/{max_empty_retry} 次重新生成")
if empty_retry_interval > 0:
await asyncio.sleep(empty_retry_interval)
if empty_retry_interval > 0:
await asyncio.sleep(empty_retry_interval)
model_info, api_provider, client = self._select_model()
continue
model_info, api_provider, client = self._select_model()
continue
else:
# 已达到最大重试次数，但仍然是空回复或截断
reason = "空回复" if is_empty_reply else "截断"
# 抛出异常，由外层重试逻辑或最终的异常处理器捕获
raise RuntimeError(f"经过 {max_empty_retry + 1} 次尝试后仍然是{reason}的回复")
# 记录使用情况
if usage := response.usage:

7
src/manager/schedule_manager.py
View File

@@ -418,7 +418,12 @@ class ScheduleManager:
if is_in_time_range:
# 检查是否被唤醒
if wakeup_manager and wakeup_manager.is_in_angry_state():
logger.info(f"在休眠活动 '{activity}' 期间，但已被唤醒。")
current_timestamp = datetime.now().timestamp()
if current_timestamp - self.last_sleep_log_time > self.sleep_log_interval:
logger.info(f"在休眠活动 '{activity}' 期间，但已被唤醒。")
self.last_sleep_log_time = current_timestamp
else:
logger.debug(f"在休眠活动 '{activity}' 期间，但已被唤醒。")
return False
current_timestamp = datetime.now().timestamp()

2
src/plugins/built_in/tts_plugin/plugin.py
View File

@@ -34,6 +34,8 @@ class TTSAction(BaseAction):
# 动作使用场景
action_require = [
"当需要发送语音信息时使用",
"当用户要求你说话时使用",
"当用户要求听你声音时使用",
"当用户明确要求使用语音功能时使用",
"当表达内容更适合用语音而不是文字传达时使用",
"当用户想听到语音回答而非阅读文本时使用",

4
template/bot_config_template.toml
View File

@@ -381,7 +381,9 @@ enable_friend_chat = false # 是否启用好友聊天
[video_analysis] # 视频分析配置
enable = true # 是否启用视频分析功能
analysis_mode = "batch_frames" # 分析模式："frame_by_frame"（逐帧分析，非常慢 "建议frames大于8时不要使用这个" ...但是详细）、"batch_frames"（批量分析，快但可能略简单 -其实效果也差不多）或 "auto"（自动选择）
max_frames = 16 # 最大分析帧数
frame_extraction_mode = "fixed_number" # 抽帧模式: "fixed_number" (固定总帧数) 或 "time_interval" (按时间间隔)
frame_interval_seconds = 2.0 # 按时间间隔抽帧的秒数（仅在 mode = "time_interval" 时生效）
max_frames = 16 # 最大分析帧数（仅在 mode = "fixed_number" 时生效）
frame_quality = 80 # 帧图像JPEG质量 (1-100)
max_image_size = 800 # 单帧最大图像尺寸(像素)
enable_frame_timing = true # 是否在分析中包含帧的时间信息