WEBVTT 00:00.080 --> 00:01.640 Look in this journey. 00:01.640 --> 00:06.880 Some days are more challenging days and some days are more satisfying days. 00:07.160 --> 00:07.840 Today. 00:08.480 --> 00:10.040 Today is both. 00:10.200 --> 00:11.600 Today is a bit of both. 00:11.600 --> 00:13.000 I've got something challenging. 00:13.000 --> 00:14.800 I've got something satisfying. 00:15.040 --> 00:18.680 Today we are integrating 11 labs with Nan. 00:18.840 --> 00:21.520 We're putting them both together and it's going to be fun. 00:21.560 --> 00:21.760 Okay. 00:21.800 --> 00:27.960 So obviously we're going to be using Nan and 11 labs together in order to have voice agents that have 00:27.960 --> 00:29.440 workflows in Nan. 00:29.760 --> 00:35.600 Now there are actually two different ways to do this, two completely different integration patterns. 00:35.760 --> 00:37.000 And we're going to do both. 00:37.280 --> 00:39.280 Uh, one of them is definitely the better. 00:39.280 --> 00:43.240 And I will explain which and why, but we're still going to do both because it's good practice and it's 00:43.240 --> 00:45.760 going to help you really understand what's going on. 00:45.920 --> 00:52.040 So the first of them is where N is the boss, Nan is running the show. 00:52.040 --> 00:54.200 It is orchestrating everything. 00:54.200 --> 00:58.560 The end to end workflow is described in Nan with all the different steps. 00:58.560 --> 01:07.900 And one of those steps is calling out to the 11 labs API to turn text into speech, to text to speech. 01:07.940 --> 01:09.740 That's when we want the thing to talk to us. 01:09.740 --> 01:12.700 And we'll use 11 labs for generating the audio. 01:12.980 --> 01:19.140 And a different step, probably earlier in the process is start speech to text. 01:19.180 --> 01:24.060 This is when we take some audio and turn it into text, which would go into an LLM. 01:24.420 --> 01:29.300 And so these are steps on the workflow which is where Nam would call that to 11 labs. 01:29.500 --> 01:32.980 Hopefully that makes complete sense to you that that's how it can come together. 01:33.460 --> 01:34.780 So what's the other way. 01:34.820 --> 01:38.820 So the other way, as you probably guessed, is where 11 labs is running the show. 01:38.860 --> 01:39.860 It's the boss. 01:39.900 --> 01:44.420 The orchestration is happening driven by 11 labs first and foremost. 01:44.460 --> 01:49.740 You basically build a voice agent in 11 labs, much as we've already done. 01:49.980 --> 01:52.700 And you give that voice agent a tool. 01:52.820 --> 01:59.060 And we we didn't actually use tools, but I showed it to you, and you use that tool to make a call 01:59.060 --> 02:02.780 into nen to carry out some sub workflow. 02:02.980 --> 02:08.360 And basically you have your sort of business logic, your business workflow, be part of an A10. 02:08.400 --> 02:11.520 So, so in this way, the two of them kind of collaborating together. 02:11.520 --> 02:14.760 But 11 labs is the overall end to end process. 02:14.760 --> 02:16.600 It sort of it sort of owns the show. 02:16.760 --> 02:19.000 So what are the reasons for doing it each way? 02:19.040 --> 02:25.800 Well, the benefit of orchestration on the left is obviously that it is simple. 02:26.000 --> 02:33.560 It's clear you get all of your workflow in one place, you've got one canvas, and on it you see everything 02:33.560 --> 02:35.600 that's going on and that's great. 02:35.960 --> 02:38.360 It also gives you a more switchable API. 02:38.400 --> 02:44.800 You've got you've got a link to 11 labs, and you could swap that out for any other API that offers 02:44.960 --> 02:47.680 text to speech and speech to text. 02:47.760 --> 02:52.600 You just have to replace those, those two APIs and bam, you're onto a different vendor. 02:52.600 --> 02:54.200 And that's super convenient. 02:54.200 --> 02:57.880 So those are the benefits of an orchestration. 02:57.920 --> 03:02.080 And with that, you're probably wondering what could possibly be the benefits of 11 labs orchestration 03:02.080 --> 03:04.040 because it's clearly more complex. 03:04.040 --> 03:04.760 It's not really. 03:04.800 --> 03:06.450 It's kind of both orchestration. 03:06.490 --> 03:10.490 You're you're first and foremost starting in 11 labs, but then you've got some workflow and then it 03:10.490 --> 03:10.690 ends. 03:10.690 --> 03:13.050 So it just sounds more complicated all in all. 03:13.050 --> 03:14.890 So why would you pick this ever? 03:15.090 --> 03:16.450 Well, look, here's the thing. 03:16.730 --> 03:17.730 Latency. 03:17.770 --> 03:19.410 It's about latency. 03:19.730 --> 03:24.050 If you're using Nand to orchestrate, then you first have to collect the audio. 03:24.170 --> 03:28.530 And then you have to call out to get that audio turned into text. 03:28.690 --> 03:32.970 If you're using 11 labs, it is a voice agent platform. 03:32.970 --> 03:36.970 And that means it's all built to be real time from the get go. 03:37.170 --> 03:40.890 As you're speaking, it's automatically converting that into text. 03:40.890 --> 03:43.210 That's not it's not a linear process. 03:43.210 --> 03:47.170 And it's designed to be great at this voice agent interaction. 03:47.170 --> 03:48.330 That's what it does. 03:48.490 --> 03:53.130 And from that perspective, it's great to be using it as much as we can. 03:53.690 --> 03:58.370 And in many ways, using them together like this gives us the ability to have the best of both. 03:58.370 --> 04:04.450 We can use all of the features and functions that you get in 11 labs when it comes to voices, and then 04:04.450 --> 04:07.070 also take advantage of ten for the orchestration. 04:07.070 --> 04:11.790 So yes, it's more complex, but it does mean you can use both platforms to the max. 04:11.790 --> 04:17.670 And as part of that, it's worth mentioning that 11 labs has some advanced functions, like like it 04:17.670 --> 04:25.190 has features to to for example, use um, a telephone like have it connect to to a phone and be able 04:25.190 --> 04:27.470 to speak on the phone to one of your voice agents. 04:27.470 --> 04:29.230 And this way you get that. 04:29.230 --> 04:34.270 Otherwise you're you're using n810 to collect the audio, and you don't get the advantage of all of 04:34.270 --> 04:36.750 these pro features in 11 labs. 04:36.750 --> 04:40.350 So there's some pretty big benefits on the right hand side as well. 04:40.350 --> 04:43.430 So which do you think I'm going to recommend as being the right way? 04:43.910 --> 04:46.510 It's definitely the right hand side. 04:46.710 --> 04:49.110 I you know, I like to say simple is always better. 04:49.110 --> 04:50.190 That's one of my mantras. 04:50.190 --> 04:54.590 I'm always trying to push you to be to take things, make things simpler, not more complex. 04:54.590 --> 05:00.630 But this is a clear case where you get you get a lot for the complexity and it's worth it. 05:00.790 --> 05:06.590 Latency is everything when it comes to voice agents, having to wait is agonizing if you're talking 05:06.630 --> 05:07.910 to to an agent. 05:07.910 --> 05:12.130 And so, yes, you need the voice agent from 11 labs. 05:12.130 --> 05:14.490 You need to use the right hand pattern. 05:14.530 --> 05:17.410 We're going to do both so we can experiment with both. 05:17.450 --> 05:21.690 Often, often people start with the left hand pattern because it's the more obvious. 05:21.690 --> 05:27.210 It's the sort of go to uh, but very quickly we'll discover the difference between them and why it is 05:27.210 --> 05:29.250 that the right one is much more compelling. 05:29.410 --> 05:30.930 But but as I say, we'll try both. 05:30.970 --> 05:32.170 Okay, now, don't hate me. 05:32.170 --> 05:37.370 But before we go and build all of this and we've got some juicy building to do today, but before we 05:37.410 --> 05:44.730 do need to do a little bit of refresher, a couple of things, API stuff, but, uh, it's good stuff. 05:44.730 --> 05:45.650 It's good to get through this. 05:45.650 --> 05:50.330 So first of all, just to quickly remind you about this thing webhook that we encountered a couple of 05:50.330 --> 05:56.410 days ago, this is this idea that you could have a node and it can be calling out to an endpoint, making 05:56.450 --> 05:59.770 like a web call to some URL, some web address. 05:59.890 --> 06:04.170 And that's that's one way you can think about making an API request to get information. 06:04.530 --> 06:06.690 And and this is it's like request response. 06:06.690 --> 06:09.330 You're hitting an HTTP route as it's called. 06:09.370 --> 06:11.030 Like a, like a like a web address. 06:11.510 --> 06:16.950 Uh, and then there's this other idea that someone else could call you. 06:16.990 --> 06:22.750 You could make a web endpoint available and say, hey, if someone if someone calls this web endpoint, 06:22.950 --> 06:27.790 which is like saying, if someone asks me if I can give them a web page at this address, I'm going 06:27.830 --> 06:30.870 to treat that like that's them telling me an event. 06:30.870 --> 06:36.430 They're telling me something's happened so I can say, hey, if you if you hit me on this web link, 06:36.470 --> 06:38.670 I'm going to know that you want to give me information. 06:38.670 --> 06:39.990 You want to tell me something? 06:40.030 --> 06:41.150 You want to tell me that? 06:41.150 --> 06:42.270 That a message has arrived. 06:42.270 --> 06:45.030 An email has arrived, a slack message has arrived. 06:45.030 --> 06:48.790 And when you do it that way, it's described as as having a webhook. 06:48.990 --> 06:54.150 And the webhook is being called by the external system, like slack in this case. 06:54.430 --> 06:57.990 So that was the the idea behind a webhook that we met before. 06:58.030 --> 06:58.350 Okay. 06:58.390 --> 07:01.070 And now now I'm gonna before with next. 07:01.070 --> 07:02.550 After this we're going to go and build stuff. 07:02.550 --> 07:09.390 But I want to give you some terminology, some, some technical stuff, which I guess what the usual 07:09.390 --> 07:12.850 story for some of you, this is like yeah, yeah, yeah, come on I know all this, but for some of 07:12.850 --> 07:14.970 you, this is going to be hairy. 07:15.130 --> 07:18.730 And I want to just say you don't need to remember anything I'm about to say. 07:18.730 --> 07:21.850 If it doesn't mean anything to you, just just just listen. 07:21.890 --> 07:23.210 Like, get the gist. 07:23.210 --> 07:27.330 I'm going to repeat this one a few times, you know, so you let it sink in over time. 07:27.450 --> 07:32.090 But just to get you more comfortable with the terminology, let's dig in. 07:32.090 --> 07:35.210 So I've used the expression calling an API a fair bit. 07:35.210 --> 07:40.650 And I'm talking about calling a web API the most standard kind of APIs. 07:40.770 --> 07:45.890 And yeah, the way it's achieved is you're making a web request, you're using this, this protocol, 07:45.890 --> 07:50.770 this, this standard called HTTP, which is basically the way that web pages are exchanged. 07:50.770 --> 07:56.330 So you're using this HTTP standard to make a request over the internet to a web address. 07:56.490 --> 07:58.970 And that's what we're calling calling an API. 07:59.210 --> 08:05.770 And the URL that you use, the web address that you're going to request or hit is called an endpoint. 08:05.770 --> 08:07.370 As you know, you've heard this. 08:07.370 --> 08:12.410 And so the language that you might you might hear if you're saying it properly is I'm calling an API 08:12.530 --> 08:19.820 by making an HTTP request to an endpoint, calling an API by making an HTTP request to an endpoint. 08:19.820 --> 08:21.540 And that should just connect with you. 08:21.580 --> 08:24.020 You should be like, okay, I hear that language. 08:24.020 --> 08:25.140 I know what it means. 08:25.500 --> 08:31.540 If you're on the receiving end of this, if you want someone to call a URL to tell you something, to 08:31.620 --> 08:35.900 give you some information, like an event has happened, I need to trigger something. 08:36.100 --> 08:39.620 Then that endpoint you would describe as a webhook, that's that's what it means. 08:39.620 --> 08:45.500 So you might say, hey, I've set up a webhook make you can make an HTTP request to that webhook to 08:45.540 --> 08:51.860 tell me about X like you've, you've received an email or something like that or the other way I could 08:51.860 --> 08:57.260 say, hey, can you please tell me your webhook and I will call your webhook to to inform you that an 08:57.260 --> 08:59.380 event has happened that you need to know about. 08:59.380 --> 09:04.740 So that's the kind of language you might use the terminology to talk about webhooks. 09:04.740 --> 09:11.660 And then just to give you one more detail, there are in fact different types of these HTTP calls, 09:11.780 --> 09:13.020 different flavors of them. 09:13.020 --> 09:14.760 It's actually known as the method. 09:14.760 --> 09:22.040 Different methods of an HTTP call, sometimes called the verb, are the most common are get and post. 09:22.200 --> 09:27.280 So you sometimes hear people say an http get and an HTTP post. 09:27.320 --> 09:33.040 Two different types of HTTP request and usually a get. 09:33.040 --> 09:36.800 An HTTP get is used when you're trying to retrieve information. 09:36.800 --> 09:42.000 You want to find something you want to get information about it, you're doing an HTTP get, and the 09:42.000 --> 09:48.160 information will come back in the form of JSON as the response to your HTTP get. 09:48.640 --> 09:55.960 And an HTTP post is usually used to send information to the third party, and you send that information 09:55.960 --> 09:57.840 in the form of JSON. 09:58.000 --> 10:01.680 So that is an http get and an http post. 10:01.680 --> 10:06.120 And I say this because we will be selecting get and post in a couple of places. 10:06.120 --> 10:08.840 So this is your first time of hearing about it. 10:08.840 --> 10:10.720 Or maybe you know this stuff back to front. 10:10.760 --> 10:13.560 Anyways that that is the terminology. 10:13.600 --> 10:14.800 We're done with it. 10:14.840 --> 10:16.880 It's time to do some building.