Просмотр исходного кода

decouple message processing process

master
lanvent 1 год назад
Родитель
Сommit
d6037422ac
3 измененных файлов: 167 добавлений и 136 удалений
  1. +32
    -13
      bot/chatgpt/chat_gpt_bot.py
  2. +5
    -0
      bridge/bridge.py
  3. +130
    -123
      channel/wechat/wechat_channel.py

+ 32
- 13
bot/chatgpt/chat_gpt_bot.py Просмотреть файл

@@ -19,19 +19,24 @@ class ChatGPTBot(Bot):

def reply(self, query, context=None):
# acquire reply content
if not context or not context.get('type') or context.get('type') == 'TEXT':
if context['type']=='TEXT':
logger.info("[OPEN_AI] query={}".format(query))
session_id = context.get('session_id') or context.get('from_user_id')
session_id = context['session_id']
reply=None
if query == '#清除记忆':
self.sessions.clear_session(session_id)
return '记忆已清除'
reply={'type':'INFO', 'content':'记忆已清除'}
elif query == '#清除所有':
self.sessions.clear_all_session()
return '所有人记忆已清除'
reply={'type':'INFO', 'content':'所有人记忆已清除'}
elif query == '#更新配置':
load_config()
return '配置已更新'

reply={'type':'INFO', 'content':'配置已更新'}
elif query == '#DEBUG':
logger.setLevel('DEBUG')
reply={'type':'INFO', 'content':'DEBUG模式已开启'}
if reply:
return reply
session = self.sessions.build_session_query(query, session_id)
logger.debug("[OPEN_AI] session query={}".format(session))

@@ -41,12 +46,26 @@ class ChatGPTBot(Bot):

reply_content = self.reply_text(session, session_id, 0)
logger.debug("[OPEN_AI] new_query={}, session_id={}, reply_cont={}".format(session, session_id, reply_content["content"]))
if reply_content["completion_tokens"] > 0:
if reply_content['completion_tokens']==0 and len(reply_content['content'])>0:
reply={'type':'ERROR', 'content':reply_content['content']}
elif reply_content["completion_tokens"] > 0:
self.sessions.save_session(reply_content["content"], session_id, reply_content["total_tokens"])
return reply_content["content"]
reply={'type':'TEXT', 'content':reply_content["content"]}
else:
reply={'type':'ERROR', 'content':reply_content['content']}
logger.debug("[OPEN_AI] reply {} used 0 tokens.".format(reply_content))
return reply

elif context.get('type', None) == 'IMAGE_CREATE':
return self.create_img(query, 0)
elif context['type'] == 'IMAGE_CREATE':
ok, retstring=self.create_img(query, 0)
reply=None
if ok:
reply = {'type':'IMAGE', 'content':retstring}
else:
reply = {'type':'ERROR', 'content':retstring}
return reply
else:
reply= {'type':'ERROR', 'content':'Bot不支持处理{}类型的消息'.format(context['type'])}

def reply_text(self, session, session_id, retry_count=0) ->dict:
'''
@@ -104,7 +123,7 @@ class ChatGPTBot(Bot):
)
image_url = response['data'][0]['url']
logger.info("[OPEN_AI] image_url={}".format(image_url))
return image_url
return True,image_url
except openai.error.RateLimitError as e:
logger.warn(e)
if retry_count < 1:
@@ -112,10 +131,10 @@ class ChatGPTBot(Bot):
logger.warn("[OPEN_AI] ImgCreate RateLimit exceed, 第{}次重试".format(retry_count+1))
return self.create_img(query, retry_count+1)
else:
return "提问太快啦,请休息一下再问我吧"
return False,"提问太快啦,请休息一下再问我吧"
except Exception as e:
logger.exception(e)
return None
return False,str(e)
class SessionManager(object):
def __init__(self):


+ 5
- 0
bridge/bridge.py Просмотреть файл

@@ -15,6 +15,11 @@ class Bridge(object):
except ModuleNotFoundError as e:
print(e)


# 以下所有函数需要得到一个reply字典,格式如下:
# reply["type"] = "ERROR" / "TEXT" / "VOICE" / ...
# reply["content"] = reply的内容

def fetch_reply_content(self, query, context):
return self.bots["chat"].reply(query, context)



+ 130
- 123
channel/wechat/wechat_channel.py Просмотреть файл

@@ -46,62 +46,55 @@ class WechatChannel(Channel):

# start message listener
itchat.run()
# handle_* 系列函数处理收到的消息后构造context,然后调用handle函数处理context
# context是一个字典,包含了消息的所有信息,包括以下key
# type: 消息类型,包括TEXT、VOICE、CMD_IMAGE_CREATE
# content: 消息内容,如果是TEXT类型,content就是文本内容,如果是VOICE类型,content就是语音文件名,如果是CMD_IMAGE_CREATE类型,content就是图片生成命令
# session_id: 会话id
# isgroup: 是否是群聊
# msg: 原始消息对象
# receiver: 需要回复的对象

def handle_voice(self, msg):
if conf().get('speech_recognition') != True :
return
logger.debug("[WX]receive voice msg: " + msg['FileName'])
thread_pool.submit(self._do_handle_voice, msg)

def _do_handle_voice(self, msg):
from_user_id = msg['FromUserName']
other_user_id = msg['User']['UserName']
if from_user_id == other_user_id:
file_name = TmpDir().path() + msg['FileName']
msg.download(file_name)
query = super().build_voice_to_text(file_name)
if conf().get('voice_reply_voice'):
self._do_send_voice(query, from_user_id)
else:
self._do_send_text(query, from_user_id)
context = { 'isgroup': False, 'msg': msg, 'receiver': other_user_id}
context['type']='VOICE'
context['session_id']=other_user_id
thread_pool.submit(self.handle, context)


def handle_text(self, msg):
logger.debug("[WX]receive text msg: " + json.dumps(msg, ensure_ascii=False))
content = msg['Text']
self._handle_single_msg(msg, content)

def _handle_single_msg(self, msg, content):
from_user_id = msg['FromUserName']
to_user_id = msg['ToUserName'] # 接收人id
other_user_id = msg['User']['UserName'] # 对手方id
match_prefix = self.check_prefix(content, conf().get('single_chat_prefix'))
match_prefix = check_prefix(content, conf().get('single_chat_prefix'))
if "」\n- - - - - - - - - - - - - - -" in content:
logger.debug("[WX]reference query skipped")
return
if from_user_id == other_user_id and match_prefix is not None:
# 好友向自己发送消息
if match_prefix != '':
str_list = content.split(match_prefix, 1)
if len(str_list) == 2:
content = str_list[1].strip()

img_match_prefix = self.check_prefix(content, conf().get('image_create_prefix'))
if img_match_prefix:
content = content.split(img_match_prefix, 1)[1].strip()
thread_pool.submit(self._do_send_img, content, from_user_id)
else :
thread_pool.submit(self._do_send_text, content, from_user_id)
elif to_user_id == other_user_id and match_prefix:
# 自己给好友发送消息
str_list = content.split(match_prefix, 1)
if len(str_list) == 2:
content = str_list[1].strip()
img_match_prefix = self.check_prefix(content, conf().get('image_create_prefix'))
if img_match_prefix:
content = content.split(img_match_prefix, 1)[1].strip()
thread_pool.submit(self._do_send_img, content, to_user_id)
else:
thread_pool.submit(self._do_send_text, content, to_user_id)
if match_prefix:
content=content.replace(match_prefix,'',1).strip()
else:
return
context = { 'isgroup': False, 'msg': msg, 'receiver': other_user_id}
context['session_id']=other_user_id
img_match_prefix = check_prefix(content, conf().get('image_create_prefix'))
if img_match_prefix:
content=content.replace(img_match_prefix,'',1).strip()
context['type']='CMD_IMAGE_CREATE'
else:
context['type']='TEXT'
context['content']=content
thread_pool.submit(self.handle, context)


def handle_group(self, msg):
@@ -122,100 +115,114 @@ class WechatChannel(Channel):
logger.debug("[WX]reference query skipped")
return ""
config = conf()
match_prefix = (msg['IsAt'] and not config.get("group_at_off", False)) or self.check_prefix(origin_content, config.get('group_chat_prefix')) \
or self.check_contain(origin_content, config.get('group_chat_keyword'))
if ('ALL_GROUP' in config.get('group_name_white_list') or group_name in config.get('group_name_white_list') or self.check_contain(group_name, config.get('group_name_keyword_white_list'))) and match_prefix:
img_match_prefix = self.check_prefix(content, conf().get('image_create_prefix'))
match_prefix = (msg['IsAt'] and not config.get("group_at_off", False)) or check_prefix(origin_content, config.get('group_chat_prefix')) \
or check_contain(origin_content, config.get('group_chat_keyword'))
if ('ALL_GROUP' in config.get('group_name_white_list') or group_name in config.get('group_name_white_list') or check_contain(group_name, config.get('group_name_keyword_white_list'))) and match_prefix:
context = { 'isgroup': True, 'msg': msg, 'receiver': group_id}
img_match_prefix = check_prefix(content, conf().get('image_create_prefix'))
if img_match_prefix:
content = content.split(img_match_prefix, 1)[1].strip()
thread_pool.submit(self._do_send_img, content, group_id)
content=content.replace(img_match_prefix,'',1).strip()
context['type']='CMD_IMAGE_CREATE'
else:
thread_pool.submit(self._do_send_group, content, msg)

def send(self, msg, receiver):
itchat.send(msg, toUserName=receiver)
logger.info('[WX] sendMsg={}, receiver={}'.format(msg, receiver))

def _do_send_voice(self, query, reply_user_id):
try:
if not query:
return
context = dict()
context['from_user_id'] = reply_user_id
reply_text = super().build_reply_content(query, context)
if reply_text:
replyFile = super().build_text_to_voice(reply_text)
itchat.send_file(replyFile, toUserName=reply_user_id)
logger.info('[WX] sendFile={}, receiver={}'.format(replyFile, reply_user_id))
except Exception as e:
logger.exception(e)

def _do_send_text(self, query, reply_user_id):
try:
if not query:
return
context = dict()
context['session_id'] = reply_user_id
reply_text = super().build_reply_content(query, context)
if reply_text:
self.send(conf().get("single_chat_reply_prefix") + reply_text, reply_user_id)
except Exception as e:
logger.exception(e)

def _do_send_img(self, query, reply_user_id):
try:
if not query:
return
context = dict()
context['type'] = 'IMAGE_CREATE'
img_url = super().build_reply_content(query, context)
if not img_url:
return

# 图片下载
context['type']='TEXT'
context['content']=content

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)
# 统一的发送函数,根据reply的type字段发送不同类型的消息

def send(self, reply, receiver):
if reply['type']=='TEXT':
itchat.send(reply['content'], toUserName=receiver)
logger.info('[WX] sendMsg={}, receiver={}'.format(reply, receiver))
elif reply['type']=='ERROR' or reply['type']=='INFO':
itchat.send(reply['content'], toUserName=receiver)
logger.info('[WX] sendMsg={}, receiver={}'.format(reply, receiver))
elif reply['type']=='VOICE':
itchat.send_file(reply['content'], toUserName=receiver)
logger.info('[WX] sendFile={}, receiver={}'.format(reply['content'], receiver))
elif reply['type']=='IMAGE_URL': # 从网络下载图片
img_url = reply['content']
pic_res = requests.get(img_url, stream=True)
image_storage = io.BytesIO()
for block in pic_res.iter_content(1024):
image_storage.write(block)
image_storage.seek(0)

# 图片发送
itchat.send_image(image_storage, reply_user_id)
logger.info('[WX] sendImage, receiver={}'.format(reply_user_id))
except Exception as e:
logger.exception(e)

def _do_send_group(self, query, msg):
if not query:
return
context = dict()
group_name = msg['User']['NickName']
group_id = msg['User']['UserName']
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 \
self.check_contain(group_name, group_chat_in_one_session)):
context['session_id'] = group_id
itchat.send_image(image_storage, toUserName=receiver)
logger.info('[WX] sendImage url=, receiver={}'.format(img_url,receiver))
elif reply['type']=='IMAGE': # 从文件读取图片
image_storage = reply['content']
image_storage.seek(0)
itchat.send_image(image_storage, toUserName=receiver)
logger.info('[WX] sendImage, receiver={}'.format(receiver))
# 处理消息
def handle(self, context):
content=context['content']
reply=None

logger.debug('[WX] ready to handle context: {}'.format(context))
# reply的构建步骤
if context['type']=='TEXT' or context['type']=='CMD_IMAGE_CREATE':
reply = super().build_reply_content(content,context)
elif context['type']=='VOICE':
msg=context['msg']
file_name = TmpDir().path() + msg['FileName']
msg.download(file_name)
reply = super().build_voice_to_text(file_name)
if reply['type'] != 'ERROR' and reply['type'] != 'INFO':
reply = super().build_reply_content(reply['content'],context)
if reply['type']=='TEXT':
if conf().get('voice_reply_voice'):
reply = super().build_text_to_voice(reply['content'])
else:
context['session_id'] = msg['ActualUserName']
reply_text = super().build_reply_content(query, context)
if reply_text:
reply_text = '@' + msg['ActualNickName'] + ' ' + reply_text.strip()
self.send(conf().get("group_chat_reply_prefix", "") + reply_text, group_id)


def check_prefix(self, content, prefix_list):
for prefix in prefix_list:
if content.startswith(prefix):
return prefix
return None
logger.error('[WX] unknown context type: {}'.format(context['type']))
return
logger.debug('[WX] ready to decorate reply: {}'.format(reply))
# reply的包装步骤
if reply:
if reply['type']=='TEXT':
reply_text=reply['content']
if context['isgroup']:
reply_text = '@' + context['msg']['ActualNickName'] + ' ' + reply_text.strip()
reply_text=conf().get("group_chat_reply_prefix","")+reply_text
else:
reply_text=conf().get("single_chat_reply_prefix","")+reply_text
reply['content']=reply_text
elif reply['type']=='ERROR' or reply['type']=='INFO':
reply['content']=reply['type']+": "+ reply['content']
elif reply['type']=='IMAGE_URL' or reply['type']=='VOICE':
pass
else:
logger.error('[WX] unknown reply type: {}'.format(reply['type']))
return
if reply:
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(self, content, keyword_list):
if not keyword_list:
return None
for ky in keyword_list:
if content.find(ky) != -1:
return True
def check_contain(content, keyword_list):
if not keyword_list:
return None
for ky in keyword_list:
if content.find(ky) != -1:
return True
return None


Загрузка…
Отмена
Сохранить