FlowFuse v2.30: Native Expert Flow Creation using AI #flowfuse #artificialintelligence #industry40

Added: 2026-05-17

[00:00:00]FlowFuse version 2.3.0 is out right now, and one of its standout features is about to change the way you develop your industrial applications forever. Hey there, my name is Christopher Sandoval.

[00:00:09]I'm a developer relations advocate here at FlowFuse. And today we're going to look at how FlowFuse Expert uses natural language processing to help you build your flows using FlowFuse's AI. All right, let's check it out.

[00:00:25]The FlowFuse Expert just had a massive upgrade. In version 2.3.0, you can now use FlowFuse Expert to create industrial applications natively with natural language processing.

[00:00:36]Let's take a look at how that works and what it looks like in practice.

[00:00:39]So here we are in a new FlowFuse instance.

[00:00:42]What I want to do is create a flow that calls a specific weather URL and then feed that information to all of my systems.

[00:00:48]I can simply ask the FlowFuse Expert to use the URL that I pasted inside of our chat to generate that exact flow.

[00:00:55]After a few seconds of waiting, you can see that the flow is being built for you automatically.

[00:01:00]Not only are the nodes being called, they're being wired up together.

[00:01:04]So with one prompt, you can now build an entire function.

[00:01:07]But what if I wanted to take this a step further and build a dashboard?

[00:01:11]Again, with one simple request, I can actually ask FlowFuse Expert to make it for me. Here I want to take the output of the get weather data node and render the text in a particular way. After I ask FlowFuse Expert to create the node, it'll wait for a few seconds.

[00:01:25]And as you can see, the FlowFuse Expert is quite intelligent. It's going to detect that I actually don't have the dashboard nodes installed and it's going to prompt me to install them by opening up the palette manager itself.

[00:01:35]I'll go ahead and click install, wait a couple seconds, and I'll continue my conversation with the FlowFuse Expert and say, "Hey, now it's installed. Go ahead and create the dashboard that I requested."

[00:01:45]Now the FlowFuse Expert can look back at our conversation and pull out the original request, and then just a few seconds, we have the dashboard that I originally asked for.

[00:01:55]Now let's go ahead and check it out.

[00:01:59]And oh, it looks good, but the only problem is that it's reporting temperature in Celsius, and I actually need it to report it in Fahrenheit.

[00:02:06]Well, good news is I can go right back to the FlowFuse Expert and tell it exactly that.

[00:02:11]The people that are going to be using this dashboard are located in the United States. So, even though the site itself is in Europe, I actually need it to report [music] Fahrenheit.

[00:02:19]Let's give it that request and see what it comes up with.

[00:02:27]Now, I can already see that there are some additional errors with this dashboard, like the unit in the actual label is wrong. But, there's a couple of other things that I want to add to the dashboard before I go through the rest of the build process. [music] So, before we go through the bug fixes, let's add some additional functions to make it a little bit more complicated for the FlowFuse Expert. First of all, I'm just going to ask it where in the world does this longitude and latitude even correspond to?

[00:02:49]Without any additional contextual information, the FlowFuse Expert can tell you exactly where that location is.

[00:02:55]So, we'll go ahead and add a text element so that other people can know that information as well.

[00:03:04]Let's make a few other small edits, like changing up the order of how these elements display, and we'll go ahead and deploy.

[00:03:11]Now, as that was processing, I had a couple of other things I wanted to add in, but putting them into practice, it looks like I actually have an error.

[00:03:17]We've just replicated the location label multiple times. What I actually wanted it to do was say the site name and then the manager name for that site. So, I'll tell FlowFuse Expert exactly what I was trying to do, and then it'll look at what I actually did and try and find where I messed up.

[00:03:32]So, the FlowFuse Expert will go through every single thing that I created and try and figure out exactly what the problem was. It looks like it's already found a couple things to fix, so we'll go ahead and click deploy and see what we came up with.

[00:03:43]This is looking really good, but I actually overlooked the label for each of those items as well. So, I'll go back to the FlowFuse Expert and tell it, "Hey, I messed up. These things should be called something else. FlowFuse Expert is ready to solve that problem for me as well.

[00:04:01]So, here we have our dashboard, but I want to go a little bit further with this. Dashboarding is one thing, but can we have the FlowFuse Expert create actual simulated data and actual complex machinery?

[00:04:12]Let's see what we can do.

[00:04:14]We'll ask it to create a flow simulating a factory machine called Hanover press.

[00:04:19]We'll have it simulate a bunch of different units. We'll have this set to a repeatable timer, say 1 hour to start.

[00:04:25]Let's ask it to generate this machine.

[00:04:27]Give it a couple seconds to process.

[00:04:30]And then go ahead and deploy the flow.

[00:04:33]>> [snorts] >> And let's change the trigger say to once every 5 seconds and see if that changes the logic at all.

[00:04:38]And here we have our machine generating all of the data points that we asked the system to create.

[00:04:42]Okay, one machine is pretty easy, but what if we had it make a bunch of machines? This time we're going to ask it to generate a factory simulator that simulates 50 machines. Each machine is going to create its own emulated data points for heat, kilowatt hours used, vibration, number of units produced, and the total uptime expressed in seconds.

[00:05:00]Awesome, we have our flow. But let's go a step further. What if we wanted to store all of this data in the flow context? Let's ask FlowFuse Expert to do exactly that. So, let's go ahead and ask it to enter all of the data for each machine in its own flow context. We'll have flow.machine1 representing the first machine data set, and then flow.machine2 representing the second, and so on all the way to flow.machine50.

[00:05:24]Now, this was a somewhat intentionally ambiguous request. We didn't tell it the last machine's name, so let's see how well it renders all of those data sets.

[00:05:34]And here we have the flow. Let's go ahead and click deploy and see what it looks like in practice.

[00:05:38]>> [music] >> So, with this deployed, let's look at our flow context.

[00:05:43]And here we can see everything is rendered perfectly.

[00:05:47]Not only are we generating all of these data points with flows created by the FlowFuse expert, we're also writing it to the flow context for further processing down the line.

[00:05:55]Now, let's give it an even more complicated task. Since we have all this data saved in our flow context, we should be able to give it to our NCP server as context for any additional insights.

[00:06:05]So, let's give it exactly what we want.

[00:06:07]We'll tell it that we want it to take the machines in the flow context. We'll give it a couple examples of what those flow context names look like.

[00:06:15]And then we'll tell it we want it to connect to NCP so that we can do any sort of insights down the road.

[00:06:20]With our NCP server selected, we'll go ahead and ask it, "Hey, please tell me all about my machines and use the NCP tool available."

[00:06:28]Here, it's told us about all 50 machines that are connected, giving us a general overview about exactly what they're doing.

[00:06:34]But, I want to know what my best-performing and my worst-performing machines in terms of heat are across the entire data set. So, I'm going to ask our NCP solution via FlowFuse expert that exact question.

[00:06:44]And almost immediately, we have our answer.

[00:06:48]So, as you can see, this is a huge jump in capability for the FlowFuse expert.

[00:06:51]But, that's not all you get with this update. We've rolled out a ton of awesome improvements, including smoother iteration with FlowFuse expert, more usable snapshot comparisons, and a bevy of other smaller updates and fixes, including device editor auto recovery, tooltip cleanup, Git integration feature flags, a cleaner way for the front-end chat requests to time out in FlowFuse expert, and more detailed stop events in your auto logs to help you discover where your problem actually lies. I definitely suggest you go check out the entire change log and see what all is in this update. But for now, we hope you enjoyed this video. If you did, please consider giving us a like, comment, or subscribe.

[00:07:24]>> [music] >> And if this video interests you in FlowFuse, head on over to flowfuse.com to start your free 14-day trial. This has been Christopher Sider of all.

[00:07:30]Thanks for watching, and I'll see you next time.

[00:07:44]>> [music]