✨ Implement long operation prompt

jimmidier commented 1 year ago

Cc: @duanxinhuan

[ ] All tasks are updated and closed
[ ] No high priority defects/bugs are on that user story
[ ] Code reviewed by one other team member (e.g. Checked by Bill)
[ ] All acceptance criteria have been met
[ ] Deployed to Dev
[ ] (If implemented) Pass automated UI testing
[ ] (If not implemented above) Do a manual test on all supported commands (at least one for each LUIS intent) against Dev
[ ] Complex code is documented (removing technical debt)
[ ] Deployed to Production
[ ] Add description about the changes for the PBI
[ ] Tick off the definition of done tasks in the PBI
[ ] (Optional) A Done Video, particularly for PBI with multiple stages

Thanks!

jimmidier commented 1 year ago

The solution from the reference doc is not suitable in this case for two reasons

It's specifically designed for activity processing that is supposed to be long-running, whereas in our case a long-running query is treated as irregular and only happens in rare case e.g. server error/network issue. This results in the actively queuing and prompting approach from the doc causing unnecessary user prompting and development overhead if applied to SophieBot.
The sample from the doc is based on Bot Framework SDK which uses different bot implementation, turn context and dialogs than the adaptive variations we're using, thus is not usable directly.

Alternative options:

(Recommended) Use the built-in ShowTyping feature. However, this built-in feature has a known issue see https://github.com/microsoft/botbuilder-dotnet/issues/3638
Implement a new IMiddleware to do "long operation monitoring" tasks. This implementation will see on each turn if the bot takes too long to respond. If so, let the user know that the bot is still processing.

jimmidier commented 1 year ago

For the recommended option above, demonstration is as below: 2023-09-06_19-31-57

SSWConsulting / SSW.SophieBot