Notes:
1. GPT-4 web 25 req per 3 hours.
2. GPT-3.5 web 5 req per 10 seconds.
3. 当使用train-embedding-v2/api模式训练数据时,需要人工对长文本进行分段,每段控制在100-2000个token内,然后使用chat-gpt3.5-embedding/api模式提问。
Updates:
2023-7-12
1. Support for training data as embedding
2. Searching training store by cosine similarity to provide reference for answer
2023-7-4
1. Provides external connection capability which support Baidu and weather search
2. Refactoring code, uniform all exception handling
2023-6-20
1. Add System role and GPT-3.5-16k
2023-5-27
1. Add GPT-4 and more modes
2023-5-26
1. Add unofficial web mode, support long input and continue conversation
2023-5-19
1. Better error handling
2023-5-18
1. Change the algorithm for calculating token
2. Add Reset feature to terminate the conversation
2023-5-16
1. Support continuous conversation mode
2. Show chat history
2023-5-10
1. Add token statistics feature
2023-3-10
1. Start