主动发言的性能问题

[yaowan233]

现在每三分钟都要从数据库获取一天群内所有消息（
这部分能不能做成缓存，搞个缓存记录一下最近的消息，不然太吃cpu了

nonebot-plugin-learning-chat/nonebot_plugin_learning_chat/handler.py

Lines 348 to 350 in 1c9d2ed

	for group_id in groups:
	if messages := await ChatMessage.filter(group_id=group_id, time__gte=today_time):
	total_messages[group_id] = messages

[CMHopeSunshine]

还好吧，几万条数据也都不到0.5秒就获取完，如果要搞缓存，那就是内存换cpu了，我觉得暂时不必

[yaowan233]

但我这服务器如果开着主动发言的话cpu一直爆满（

[CMHopeSunshine]

是内存爆还是cpu爆啊，我的服务器2核4g，cpu占用长期不到15%，我觉得不是这个的问题，内存占用大倒是真的，因为用到了jieba库，本插件会有200m左右的内存占用

[yaowan233]

没 就是cpu 我服务器是单核2.5g的

[yaowan233]

我只要把定时任务关了就没问题了 肯定不是jieba库的问题

[CMHopeSunshine]

这样的话建议试试用原版牛牛吧，它是把消息存内存里定期做持久化的，我在复刻它的时候就是想去掉缓存以优化内存占用

[yaowan233]

好吧 找到问题了
问题是有一万多个context，但是在这个群里找不到一个answer，所以一直在进行数据库操作（

nonebot-plugin-learning-chat/nonebot_plugin_learning_chat/handler.py

Lines 409 to 450 in 1c9d2ed

	for context in contexts:
	if (not speak_list or random.random() < config.speak_continuously_probability) and len(
	speak_list) < config.speak_continuously_max_len:
	if answers := await ChatAnswer.filter(context=context,
	group_id=group_id,
	count__gte=config.answer_threshold):
	answer = random.choices(answers,
	weights=[
	answer.count + 1 if answer.time >= today_time else answer.count
	for answer in answers])[0]
	message = random.choice(answer.messages)
	if len(message) < 2:
	continue
	if message.startswith('&#91;') and message.endswith('&#93;'):
	continue
	if any(word in message for word in ban_words):
	continue
	speak_list.append(message)
	follow_answer = answer
	while random.random() < config.speak_continuously_probability and len(
	speak_list) < config.speak_continuously_max_len:
	if (follow_context := await ChatContext.filter(
	keywords=follow_answer.keywords).first()) and (
	follow_answers := await ChatAnswer.filter(
	group_id=group_id,
	context=follow_context,
	count__gte=config.answer_threshold)):
	follow_answer = random.choices(follow_answers,
	weights=[
	a.count + 1 if a.time >= today_time else a.count
	for a in follow_answers])[0]
	message = random.choice(follow_answer.messages)
	if len(message) < 2:
	continue
	if message.startswith('&#91;') and message.endswith('&#93;'):
	continue
	if all(word not in message for word in ban_words):
	speak_list.append(message)
	else:
	break
	else:
	break

[yaowan233]

就是在其他群里面学到了很多，但就是有个群里没有学到什么，这时候就一直在这个for里面循环了
建议可以直接查询answer表，这样可以省下很多时间