1.itchat添加群组语音回复文本功能;2.itchat添加群组语音回复语音功能;3.更新redeme

This commit is contained in:
Chiaki
2023-03-27 16:46:53 +08:00
parent 7c4037147c
commit 9b389ffc33
2 changed files with 42 additions and 9 deletions

2
app.py
View File

@@ -24,4 +24,4 @@ def run():
logger.exception(e)
if __name__ == '__main__':
run()
run()

View File

@@ -33,18 +33,20 @@ def handler_single_msg(msg):
WechatChannel().handle_text(msg)
return None
@itchat.msg_register(TEXT, isGroupChat=True)
def handler_group_msg(msg):
WechatChannel().handle_group(msg)
return None
@itchat.msg_register(VOICE)
def handler_single_voice(msg):
WechatChannel().handle_voice(msg)
return None
@itchat.msg_register(VOICE, isGroupChat=True)
def handler_group_voice(msg):
WechatChannel().handle_group_voice(msg)
return None
class WechatChannel(Channel):
def __init__(self):
@@ -89,7 +91,6 @@ class WechatChannel(Channel):
context.kwargs = {'isgroup': False, 'msg': msg, 'receiver': other_user_id, 'session_id': other_user_id}
thread_pool.submit(self.handle, context).add_done_callback(thread_pool_callback)
@time_checker
def handle_text(self, msg):
logger.debug("[WX]receive text msg: " + json.dumps(msg, ensure_ascii=False))
@@ -169,6 +170,33 @@ class WechatChannel(Channel):
thread_pool.submit(self.handle, context).add_done_callback(thread_pool_callback)
def handle_group_voice(self, msg):
if conf().get('group_speech_recognition') != True:
return
logger.debug("[WX]receive voice for group msg: " + msg['FileName'])
group_name = msg['User'].get('NickName', None)
group_id = msg['User'].get('UserName', None)
create_time = msg['CreateTime'] # 消息时间
if conf().get('hot_reload') == True and int(create_time) < int(time.time()) - 60: #跳过1分钟前的历史消息
logger.debug("[WX]history group message skipped")
return
# 验证群名
if not group_name:
return ""
if ('ALL_GROUP' in conf().get('group_name_white_list') or group_name in conf().get('group_name_white_list') or check_contain(group_name, conf().get('group_name_keyword_white_list'))):
context = Context(ContextType.VOICE,msg['FileName'])
context.kwargs = {'isgroup': True, 'msg': msg, 'receiver': group_id}
group_chat_in_one_session = conf().get('group_chat_in_one_session', [])
if ('ALL_GROUP' in group_chat_in_one_session or
group_name in group_chat_in_one_session or
check_contain(group_name, group_chat_in_one_session)):
context['session_id'] = group_id
else:
context['session_id'] = msg['ActualUserName']
thread_pool.submit(self.handle, context).add_done_callback(thread_pool_callback)
# 统一的发送函数每个Channel自行实现根据reply的type字段发送不同类型的消息
def send(self, reply : Reply, receiver):
if reply.type == ReplyType.TEXT:
@@ -206,19 +234,26 @@ class WechatChannel(Channel):
reply = e_context['reply']
if not e_context.is_pass():
logger.debug('[WX] ready to handle context: type={}, content={}'.format(context.type, context.content))
if context.type == ContextType.TEXT or context.type == ContextType.IMAGE_CREATE:
if context.type == ContextType.TEXT or context.type == ContextType.IMAGE_CREATE: # 文字和图片消息
reply = super().build_reply_content(context.content, context)
elif context.type == ContextType.VOICE:
elif context.type == ContextType.VOICE: # 语音消息
msg = context['msg']
file_name = TmpDir().path() + context.content
msg.download(file_name)
reply = super().build_voice_to_text(file_name)
if reply.type != ReplyType.ERROR and reply.type != ReplyType.INFO:
context.content = reply.content # 语音转文字后将文字内容作为新的context
# 如果是群消息,判断是否触发关键字
if context['isgroup']:
match_prefix = check_prefix(context.content, conf().get('group_chat_prefix')) or check_contain(context.content, conf().get('group_chat_keyword'))
if match_prefix != True:
return
context.type = ContextType.TEXT
reply = super().build_reply_content(context.content, context)
if reply.type == ReplyType.TEXT:
if conf().get('voice_reply_voice'):
if context['isgroup']:
reply.content = '@' + context['msg']['ActualNickName'] + ' ' + reply.content
reply = super().build_text_to_voice(reply.content)
else:
logger.error('[WX] unknown context type: {}'.format(context.type))
@@ -255,14 +290,12 @@ class WechatChannel(Channel):
logger.debug('[WX] ready to send reply: {} to {}'.format(reply, context['receiver']))
self.send(reply, context['receiver'])
def check_prefix(content, prefix_list):
for prefix in prefix_list:
if content.startswith(prefix):
return prefix
return None
def check_contain(content, keyword_list):
if not keyword_list:
return None