External issue Progressive Web App

[TorontoJoe]

Affected version

2.3.4

I think this is a bug but hoping just my being too green with XF.

I recently launched my XF forum. Users are very excited about the app. A few things however that I'm noticing that don't function the way I would expect.

I am on XF 2.3 and app installed on iPhone 13 pro max. IOS 18, Safari

Most issues seem to be related to how notifications are directed.

Examples:

Moments ago I received a DM. I got a banner notification on my phone. Clicking on it took me to the forum. Not the DM I was sent or even the DM list. Just the forum. I noticed the notification for new DM so I clicked on that and I went to his message.... to which I replied.

A short time later he again replied to me and I was notified. I had minimized the app so when I clicked on the notification banner it took me to the DM section. It appears it only takes you to the place you last visited and not to the event that prompted the notification. This seems the same behaviour when receiving notifications for posts on subscribed threads/posts. It takes you to the app..... but you don't really know why, and then need to hunt for the event for which you got notified.

In both scenarios something makes this even more of a challenge. In the case of say the DM, I had it previously open so it at least took me back to the message string with the other member.... however the only thing there was my last message and not his reply. So it does not refresh with the new information.

Just as I type this another banner notification came in indicating a user has just posted on a thread to which I am subscribed.... I hit the banner and what opened is the app, back at the message thread noted above.

----

Similarly - When the app is installed it of course appears on the home screen with number of unattended notifications. This number.... in my case 4, notifies relative to both subscribed post notifications and DM notifications. There is no way to determine which. So when touching the icon, one is left searching for whatever this notifications was for.

This may all be my own ignorance but I'm normally pretty on the ball with this sort of this sort of thing.

Thoughts?

 

[TorontoJoe]

This also seems to apply to like notifications. Click on the notification, and then no sign of the item that was liked.

 

[XF Bug Bot]

Automated triage: Browser issue

Thank you for reporting this issue. Each notification includes a link to the exact message or post, and the site asks the device to open it. On iOS home-screen web apps, Safari sometimes opens the app's start page or last page instead. That is a known Safari bug that we can't fix from our side. The app badge shows unread direct messages and alerts combined.

This is an automated assessment. A member of staff may review and update it.

 

[TorontoJoe]

Nearly two years and now, a bug bot replies!?!?

I don't think this is a positive trend. We're all boned..... sigh....

Bots are the antithesis of the community forum experience.... PERIOD. If we're introducing AI into this sacred, community space..... I suppose we may as well just give up and call it quits and admit that human interaction is dead.

Or, we can reject it, violently... and be the hub of HUMAN communication.

Choose wisely. If this continues I think I'll be out of the Xenforo sphere.

We shouldn't bow to what FB is doing because a few people say they want a more "modern" experience. That's what we're trying to avoid.

I feel so gross even having to post this.... like it needs to be said

 

[Chris D]

On the other hand, the bot got it correct. Cool.

 

[Chris D]

Nearly two years and now, a bug bot replies!?!?

I don't think this is a positive trend. We're all boned..... sigh....

Bots are the antithesis of the community forum experience.... PERIOD. If we're introducing AI into this sacred, community space..... I suppose we may as well just give up and call it quits and admit that human interaction is dead.

Or, we can reject it, violently... and be the hub of HUMAN communication.

Choose wisely. If this continues I think I'll be out of the Xenforo sphere.

We shouldn't bow to what FB is doing because a few people say they want a more "modern" experience. That's what we're trying to avoid.

I feel so gross even having to post this.... like it needs to be said

Listen, I understand AI is not to everyone's tastes but to be clear; we're not trying to be Facebook and I don't really understand the connection. We had a backlog of over 1,000 bug reports, many of which were genuinely stale, blocked due to unclear or uncertain reproduction cases, some of which had less attention than they should have done, and the workload was overwhelming.

Leaving them as they were was unsustainable, providing a poor experience to customers, and a problem that was ultimately only getting worse rather than better.

LLMs are good at making decisions, understanding context and code, and making reasonable judgement calls. The backlog is now reduced by half and we now have a strong basis for humans to feasibly work through what's remaining and allow us to do maintenance releases with clearer boundaries, and a clearer understanding of what's truly important to fix whenever we're working towards a new release. The triage pass has also enabled us to see more clearly which bugs require a much more detailed bug fix, the kinds of bug fixes we can only do for larger releases like 2.4 and 3.0, so rather than being completely ignored for years, we're now more focused on where they belong on the roadmap.

We have no intention or desire to remove human communication. Quite the opposite. While some of the decisions aren't popular, or may seem surprising, every single thread that the bot has replied to and actioned is now subscribed to (watched) by a member of staff, so if a member has a query about a decision, disagrees with it, or otherwise confirms their issue is resolved, that gets a human to come and talk to you, as has happened here.

In any case, some form of action trumps any form of inaction, in my opinion. I apologise it's not to your tastes, but we're solving a problem here and ultimately that will be a more positive step forwards for most customers even if you, personally, do not like the way it has been done.

 

[Andy.N]

I agree with Chris. Let's AI take care of the mundane task, the low value cleaning up and let us focus on the high value issues. If any issues that AI bot flags incorrectly and receive the proper attention or follow up by other customers, the XF team can take care of. My guess is that very few original reporter will follow up on these years-old reports which is the whole idea of using AI to clean up.

 

[Mendalla]

If any issues that AI bot flags incorrectly and receive the proper attention or follow up by other customers, the XF team can take care of.

And I have seen Chris override the bot's call at least a couple times so that is happening. This is actually one of the things that AI with proper training can handle fairly well. I imagine we will eventually enable the agent on the help desk at work to help with triaging.

 

[Andy.N]

And I have seen Chris override the bot's call at least a couple times so that is happening. This is actually one of the things that AI with proper training can handle fairly well. I imagine we will eventually enable the agent on the help desk at work to help with triaging.

This is the way. The amount of data that AI can shift through is order of magnitude compared to humans. I think we are losing the race to compete with their effectiveness so just let them do what they do best: supervised, well-trained jobs.
I'm actually excited to see how widespread AI is adopted so we can accomplish more things these days.

 

[TorontoJoe]

We have no intention or desire to remove human communication.

That’s exactly what happened. It was the bot and not a human that replied. It wasn’t until I called that out, that a human interjected.

Quite the opposite. While some of the decisions aren't popular, or may seem surprising, every single thread that the bot has replied to and actioned is now subscribed to (watched) by a member of staff, so if a member has a query about a decision, disagrees with it, or otherwise confirms their issue is resolved, that gets a human to come and talk to you, as has happened here.

There’s a big difference between LLM‘s being used as a background tool by developers … and the use as an alternative to an interaction with customers/members

The appropriateness of what tools you use in the background is perhaps debatable… But when a member or customer is faced with a bot, the immediate sentiment is one of being dismissed.

I have to believe they everyone here can see the distinction

 

[Chris D]

Almost nobody cares. They just want to see their bugs fixed and we have an efficient development process that allows us to balance maintenance and new development.

I’m sorry it isn’t to your tastes. You’ll see the bot a lot more. You’ll also see more frequent releases and more stable software as a result. That isn’t a bad thing and we won’t be changing anything about it because a few people aren’t in favour of the method of bug classification and triage.

That’s exactly what happened. It was the bot and not a human that replied. It wasn’t until I called that out, that a human interjected.

This is the important thing. This is the best of both worlds. You get a bot initially, but a relevant reply from a member gets a staff response when needed.

No one loses out here and far from it. It’s okay to disagree with the bot’s decision. It isn’t perfect, especially as the system we’ve designed is new and we’re ironing out a few things, and it won’t ever be perfect, frankly, but where we become aware of incorrect decisions we will address that and tune it for better results in the future.

 

[Itworx4me]

You’ll also see more frequent releases and more stable software as a result.

Careful what you say here. Your critics will hold you too this....

 

[TorontoJoe]

"Almost nobody cares"

This is comforting. In the end it feels much better to be completely dismissed by a human rather than the bot