- Use logging instead of printing.
- Fixed bugs during the refactoring.
- Improve the pytest support.
- Refactored the codes by @GiggleLiu to obey the DRY principle. Thanks for his contribution!
- Initial pytest support added.
- Improved academic prompts added to the bot, type
/helpto access them. - Context feature now disabled when using academic prompts.
- Fixed issue with conversation trimming when single prompt exceeding token limit.
- Add
/endto end the current conversation, which is shorter and fits Zulip syntexes.
- Upgraded the model from
text-davinci-003togpt-3.5-turboby default. It can be configured inconfig.ini. - Added contextual support, allowing the bot to answer questions based on the context of the conversation. Type
停止会话orend the conversationto end the current conversation. - The output will also show numbers of tokens used in the conversation.
- Implemented conversation history trimming to ensure it stays within OpenAI's maximum token limit.
- Fixed a bug causing the bot to crash during long conversations.
- Resolved an issue where the bot would reply privately when mentioned in a group conversation.