Greetings! I have previously posted this proposal here. But noticed that my suggestion is more suitable for the Discourse AI plugin.
Recently, I’ve been using the notion.so website (There is a similar open source project https://appflowy.io/). On this site, there’s an option to assign tasks to artificial intelligence to edit parts of the text it has written. For instance, to improve, expand, summarize, and so on. Here’s how it works:
I assign a task to the AI to write text.
The AI generates the text.
I highlight the necessary part of the generated text and select a command from the menu.
After generating new additional text, I decide whether to replace the old text with the new one or insert it below, and so on.
So, is it possible to implement something like this on discourse using your plugin? For example:
I am creating a topic/query
The AI generates a text response
I select a part of the text and select the desired scenario in the list of actions that opens
(in standard discourse, such an action opens a menu for quoting or sharing).
The bot edits its generated text only in the place where I indicated, without the need to create a new response.
These are merely one or two illustrations; upon scrutinizing the video with care, one can draw an abundance of ideas from it. I surmise that the Discourse AI plugin will be adept in executing this task.
I suggest that this helper also work outside of the composer. For example, the first - I sent a message to the AI (chatbot), the second - he answered me, the third I want him to make changes to his answer and for this I select a part of the text, after which, along with the “quote”, “share” buttons appears AI assistant. I click on the assistant button and select the command I need or enter my own prompt. Fourth, the chatbot corrects its response depending on my command.
Here is an example:
After I click on the AI button, the same menu will open as I showed above:
Why do I suggest “AI” button outside of composer.
In the composer, when I select a part of the text and give a command (for example, to translate into English), the Chatbot does not know the context of the entire text. In order for him to know the context, I must send him the entire text.
After I sent a request to the Chatbot and it answered me, in order to correct its text, I have to give a command to correct one or another part of the text by clicking on the “Reply” or “Quote” button. The bot only replies in the next message, and does not correct its existing post. After receiving the answer (according to the results of my request), in order to combine the received (corrected) text and the main text (the first chatbot response), I have to copy the very first and last Chatbot response, then paste it into a third-party text editor (for example, Microsoft Word) and there further edit all text. This is rather inconvenient, since you have to make extra gestures.
And if I select a part of the text in the response received from the chatbot and give a command using the “AI” button, then the chatbot will remember the context. Also, if the chatbot corrects its source text, instead of replying to the user at the bottom of the topic, the user will not feel cluttered by the abundance of messages following one another.
Of course, I am not against the dialogue method, where messages follow one another, but I think it would not be superfluous to allow the user to edit the text inside one message (response) of the chatbot. And if you add the ability to view the history of changes in the bot message, as it is in regular Discourse topics
, then it will be convenient for the user to view the history of changes, as it is implemented on the site chat.openaicom
Ps.I hope I was able to convey my thoughts. Used google translator.
I am firmly confident that this will provide an additional impetus to the progression of all Discourse forums worldwide.
I am elated; this is assuredly going to be a supremely beneficial feature.
Of course, I understand that such a feature would not be applicable to all posts, but only to those posts that the bot has personally written to the user. If everyone starts editing AIbot’s posts, instructing it to modify them in one way or another, it would become a pointless and even detrimental function. Therefore, I hope that a similar feature for editing the bot’s personal responses by the bot itself upon the user’s command will be available in the future