Bot accessed from a channel using @911bot should only post emergency info into a direct chat from the caller

[jchesshir]

Once emergency information has been requested from a channel and delivered there, tracking who might be misusing that information could become practically impossible. 911 Bot's emergency information should at the very least refuse to give the information over the channel and instead tell the channel to ask for the information from a direct chat.

A nice to have would be for 911 Bot to detect who requested the information from the channel, initiate a direct connection to that user if it has not already been established, and then continue with the interaction from there.

[jchesshir]

To hopefully clarify the @sdmackay comment in #16, while one should be able to issue the command "@911bot emergency" in a channel, the 911 Bot should respond by opening up a direct channel to the user who issues the command to verify emergency and deliver the requested information. So we should probably consider the nice-to-have in the above description to be a part of the specs for this enhancement.

[SkyeHoefling]

Doesn't Slackbot send private messages to us within the channel? I think having the UX be integrated into the channel is nice, but make it so only you can see the message in the channel. When an emergency is happening people tend to panic and the easier it is to use the bot the better.

[sohail-surge]

Could be fixed by #23

[jchesshir]

I've contemplated the idea suggested in #23. However, I've not seen how it will help us. I've tried sending a slash command through the reply to redirect the message directly to a user. However, Slack did not pick up the slash command coming from the reply like it does from the GUI.

The best way I can see to implement this is to change the bot token to a user token to allow 911bot to initiate a direct message, rather than just respond to one. However, I don't know how to make that happen with the slackbot interface. The best way I can see right now is to refactor out the components of the healthcheck module needed to initiate and send a single dm. Once that functionality is available, ironing out the other functionality should be trivial.

[sohail-surge]

As mentioned in the chat, if you look here you see that this is already being done so it's a matter of changing the emergency information to be sent in a private message:

https://github.com/surgeforward/911bot/blob/master/bot/plugins.py#L86-L89

However, the user interface should still be considered. What happens if they are in a channel and don't see that they should check the direct message? Take a look at design.org, specifically the "Retrieving emergency contact for user" section and how that would change before you make any code changes.

[jchesshir]

Agreed. So I think this issue should track changes to how 911Bot responds to a request in a channel via the @ notation, while #23 should be used to track new functionality of 911Bot responding to the / notation.

For the current issue, if the reply can be sent directly to the channel but be kept visible only to the requesting user, that would be optimal. Otherwise, a direct chat should be sent to the requesting user, along with a reply in the channel instructing the user to look for the direct chat .