long querys

[mccoysc]

If there are too many characters in the query, it can result in token limit issues. For example, submitting a webpage link where the text exceeds the token limit. In such cases, it is recommended to split the query and provide additional prompts to ChatGPT, such as informing ChatGPT that a long article will be split into several short paragraphs and submitted in separate queries until completion.

[mccoysc]

and,i got...until now,you don't have conversation mode......so,the feature will not be ok until conversation mode is ok

[mccoysc]

if no conversation mode, spliting query to pices result chatGPT do not know the previous pieces

[Leask]

[mccoysc]

[mccoysc]

ChatGPT works based on reading comprehension and uses the comprehension results to answer user queries. When given an article (system prompt), it first comprehends it and then uses the results to answer the user's questions. The ChatGPT API interface does not cache any user queries (with the exception of the web version of ChatGPT). Therefore, when using the API mode, if the program implementer does not implement session mode, ChatGPT will not know the user's previous queries.

In the screenshot provided, ChatGPT knows the user's previous query either because the user implemented session mode or used the web version of the ChatGPT API interface.

[Leask]

對話session的功能當然是已經實現了的。

On Wed, May 10, 2023 at 1:51 AM mccoysc @.***> wrote:

ChatGPT works based on reading comprehension and uses the comprehension results to answer user queries. When given an article (system prompt), it first comprehends it and then uses the results to answer the user's questions. The ChatGPT API interface does not cache any user queries (with the exception of the web version of ChatGPT). Therefore, when using the API mode, if the program implementer does not implement session mode, ChatGPT will not know the user's previous queries.

— Reply to this email directly, view it on GitHub https://github.com/Leask/halbot/issues/34#issuecomment-1541390179, or unsubscribe https://github.com/notifications/unsubscribe-auth/AABY4PQ7LSWFV4O7TJW5DU3XFMUGJANCNFSM6AAAAAAX4GZDOI . You are receiving this because you commented.Message ID: @.***>

--

Sincerely,

Sixia "Leask" Huang https://leaskh.com

[Leask]

睡下了，明天我有很多會，有空給你發截圖說明這個問題。你徹底誤解了。

[mccoysc]

對話session的功能當然是已經實現了的。 On Wed, May 10, 2023 at 1:51 AM mccoysc @.> wrote: ChatGPT works based on reading comprehension and uses the comprehension results to answer user queries. When given an article (system prompt), it first comprehends it and then uses the results to answer the user's questions. The ChatGPT API interface does not cache any user queries (with the exception of the web version of ChatGPT). Therefore, when using the API mode, if the program implementer does not implement session mode, ChatGPT will not know the user's previous queries. — Reply to this email directly, view it on GitHub <#34 (comment)>, or unsubscribe https://github.com/notifications/unsubscribe-auth/AABY4PQ7LSWFV4O7TJW5DU3XFMUGJANCNFSM6AAAAAAX4GZDOI . You are receiving this because you commented.Message ID: @.> -- Sincerely, Sixia "Leask" Huang https://leaskh.com

1、从我部署的看，没有实现。 2、从代码看，你没有把会话里的所有对话内容给到chaGPT，所以也不可能有会话模式

[Leask]

認真讀代碼，實現在chat gpt api庫裡面。時差問題，我先睡了。

[mccoysc]

認真讀代碼，實現在chat gpt api庫裡面。時差問題，我先睡了。

如果是实现了会话，那就需要基于会话模式解决下长问题的拆分，不然像网页文章很长时，它就无法处理

[mccoysc]

看起来确实没有记录会话