[CW-1931] Draft through API both for cross device support but also for automated drafting of conversation

[VSinghhh]

Is your enhancement request related to a problem? Please describe.
With a high volume of conversations and a fairly high complexity of some of our questions we want our bots to 'draft' replies rather than actually directly sending the replies.

Chatwoot has suport for drafts but only through local storage:
https://www.chatwoot.com/blog/v2-10-0/
https://www.youtube.com/watch?v=NAtxmt2lKQY
#4876

Describe the solution you'd like
We want the following flow

1. Our platform analyses the email - and while we have a high ability to confidently identify intents of the incoming emails, it is not 100% right. For this reason we would like the incoming email to be analyzed by our backend and assign a label to the conversation. We understand that the Chatwoot API works for this
2. We would like our backend to instruct the reply (draft?) to be written- but that our agent is the one who actually clicks "send" to send it. Essentially an auto-canned /drafted reply. I didnt see any API for drafting a message, and @pranavrajs in the YouTube clip says that drafts dont work between devices (the draft is in the browser local store rather than saved in the BE/system) at the time of launch of the draft feature - Jan 2023; but that it was coming.

The purpose is that our platform based on the analysis of incoming email does most of the work. If it is wrong in identifying the intent, the agent can relabel the email. (over time we will retrain our platform to identify the incoming email correctly in future (based on new label))

Td;lr: We want our platform to 'suggest'/autofil a reply - and if it is correct, the agent just clicks "send"

Describe alternatives you've considered
None available as far as we've seen

Additional context
Add any other context or screenshots about the enhancement request here.

_CW-1931

[pranavrajs]

@VSinghhh This seems similar to the reply suggestions feature that we are working on. I will add more details here later.

[VSinghhh]

Hi,
Just want to check in how its going with this?
A difference could be that the 'reply suggestions' that you guys might be building (LLM or similar?) might be different from our own BE-logic where we have our own setup which we can handle using drafts:

Should we built API/BE for drafts ourselves or wait to test your reply suggestions?

[sojan-official]

partial work done in #8018