SmythOS’ Post

Transcript

Hello and welcome to this new session. In this new session we're going to cover the async and await components. Well both together because these components work together. And well, that's why we're covered them and only one session. Well, let's see. 1st the async component. This component allow us to run a workflow synchronously. Until now all the workflows that we saw they are synchronous workflows, for example this one. Because all the components in here. Are executed one at a time, 1 after the other, step by step. But this async component allow us to run workflows asynchronously. And how this works? Well, first let's see some details about this component. First. We don't have extra settings. We don't have the gear button in this section. As we can have for example in the alarm prompt component. We don't have the gear button so no extra settings. Another thing is that we don't have the fault inputs but we have one default output that is called job ID. And the creation of inputs behaves in a different way than other components. Let's try it if we create a new input. We can see that we have they created input but in the outputs we have a new output with the same name but with the async label. So every time that we create a new input. A new output is created with the same name and with the async label. So how this works? Well, the async component turned all the input information into async outputs. So all the information that can be provided to these inputs can come from a synchronous workflows and once the async component is executed, the same information is going to be available in the outputs, but these outputs are going to be part of the asynchronous workflow. And well. So with that said, maybe you wonder why the job ID this default output doesn't have the async label? Yeah, this is because the job ad is the only output that can. Work with synchronous workflows. All the new outputs are part of the asynchronous workflow. So they cannot be used in a synchronous workflow. OK, with that said, let's see also the await. This component. Is the one in charge of retrieve the results of the asynchronous workflows. Because every time that the async component is executed at job, ID is generated. And this job is generated because we. We need a way to to retrieve the results from every asynchronous workflow. So that's exactly what the weight component receive in this job default input. It can receive an array of job IDs. And after the await component is executed, we're going to retrieve the result of those asynchronous workflows in the default output called results. And we have the settings. In the await component, the minimum jobs count and the maximum weight time in seconds. These settings help us to define how many time or for how many jobs we're going to wait to be completed. Uh. When we execute the await component for example. Let's imagine that we have 10 asynchronous workflows that are being executed. And but we just want to wait until five of them are completed and then that's it. So we can set the minimum jobs count to 5. Then we have to set the another settings, the maximum weight time in seconds, how many time we're going to wait for A5 jobs to be completed, so you can say for example 10 seconds. So this means. That we're going to wait for five jobs to be completed in a period of time of 10 seconds. This 10 seconds is the is the whole time. But for example if five jobs are completed in five seconds where after 5 seconds wait component is going to be executed and it is going to retrieve the results. So this is like the the maximum time. So yeah, this help us a lot to. For example, there are some uh, situation where. I don't know the execution of our agent well, don't have to surpass some amount of time. So this is a voice component can help us? To. To define a time of less than that time out so we don't trigger errors or I don't know, some other situation can happen if we run our agents maybe in other platforms where they have some defined timeout that well we cannot control, so our endpoints cannot take too long being executed. So well. We just see the basics of these components now we can. See an example of the execution. In here we have a very simple workflow. Let's see what we have. We have an LLM prompt component. This is just using the echo option, so it's just going to print what we have in the prompt. We have an array with three prompts. One generates 30 random characters and then generate A10 poems. Then give 10 random countries. This array is going to be provided to the for each component. And each prompt is going to be executed in this in the loop branch. So each element of the. Of the array is going to be delivered to this async component. So. As you can see, all the elements that we bind to this async outputs and they have a label here in the corner of the component that says async. That means that this component is part of the async workflow. The same is going to happen if I connect any kind of UH component in this branch. You can see now we have the same label, the same async label. So what happened here? Well. In the loop branch, we're going to receive each prompt of the this array. This array well this prompt is going to be delivered in this async output. And is going to be delivered to this LLM prompt that the only thing that is going to do is execute that prompt. As you can see. So the prompt is executed. And after all the elements of the array is executed, we're going to have it in the result output. And these results are going to be delivered to the await. Why in the await? Well, one thing that you have to know is that. Since this. 2 components. Are are part of an asynchronous workflow? Well they are. We are not going to have the results. The only thing that we are going to have is the job IDs as you as I said. Every time that the async component is executed, we generate a job ID. Because the asynchronous branch is executed independently, they have. They are executed while. In a separate in a different way from the synchronous workflow. So those jobs ideas are going to come to the away component. We have here that we're we're just going to wait for one job to be completed and. At times that we're going to wait is 10 seconds. And we're we're just showing that results with an API output. So let's run this. If I run this, I have the three prompts. And. Let's run it. Each prompt. Well, one thing that you can notice here is that. The execution was really fast. Because normally while the LLM prompt takes some time to be executed because it's a, it's a model. But in using the async component, well we are only generating a job ID. The LDL prompt component is executed independently. So we don't have to wait for those workflows to be completed. And we can check the result of the for each component. And we have only IDs. These are the three job IDs generated for each iteration of the for each component. These are delivered to the await component. And when they await component is executed, we retrieve the results of these jobs. As we can see here. Perhaps the job ID? And we have the information. Well, there are results. 1st we have. That well, they prompt about random characters as well. We have the result. We have any status of done. And we have these other two jobs. That only has the status of pending. So that means that, uh, well, based on the settings of the await component. We just wait for one job to be completed. So once we had one job completed. Well, they await component retrieves automatically the results of the three job IDs. That's why the other two jobs are pending, because they didn't have enough time to be completed. We just wait one second. Well, we just wait, no, for one job to be completed. So, uh. Another thing that we can, uh, notice. Is well as I said in the session about the for each component. The output format we have the option array of results that and I said that this specific option. Is here because. We use it for the async and await components. That's why we are using this option. To retrieve the job IDs. Because we just want an array with nothing else than the job ideas, we don't need the extra information. Uh, I don't know that this is that information that we can see when we use the full and minimal options. We don't want that. We just want as simple array with the job. It is in this case. That's why when you use the async and await components. And the for each component. Yeah well you have to set this output format to array of results. OK, so. As I said, these async outputs cannot be used in synchronous and normal workflows. So what happen if I want to bind this to?