WEBVTT 00:00.040 --> 00:04.160 Now, one thing we're expecting is we're expecting it to have an error initially because we haven't 00:04.160 --> 00:07.560 yet properly configured that that that chat input. 00:07.720 --> 00:11.440 Uh, but still we should be ready pretty much to at least see that happening. 00:11.520 --> 00:13.240 So let's go back into N810. 00:13.280 --> 00:13.920 Here we are. 00:14.200 --> 00:20.960 And make sure we we come here and we press Execute workflow so that this is up and running and waiting 00:20.960 --> 00:21.360 for us. 00:21.360 --> 00:22.960 We expect this will throw an error. 00:23.160 --> 00:29.400 Now we'll come back into 11 labs and we're going to come back to our agent, which is now equipped with 00:29.400 --> 00:30.240 this tool. 00:30.560 --> 00:32.000 And we're going to give it a try. 00:32.280 --> 00:36.520 Uh, let's uh let's see, uh, we're going to press preview. 00:37.520 --> 00:40.680 Up it comes and we're now going to call our agent. 00:40.880 --> 00:41.640 Here we go. 00:43.360 --> 00:43.680 Hello. 00:43.680 --> 00:44.880 How can I help you today? 00:45.200 --> 00:45.680 Hello. 00:45.720 --> 00:49.920 I'd like to know how many, uh, shares of spy I have in my portfolio. 00:52.400 --> 00:53.240 I'm sorry. 00:53.240 --> 00:56.680 I was unable to retrieve the number of spy shares you hold. 00:57.200 --> 01:01.730 Is there anything else I can help you with regarding your equity portfolio. 01:03.410 --> 01:04.650 Okay, let's stop it there. 01:04.690 --> 01:07.610 Let's go over to to Nan and see what happened. 01:08.050 --> 01:09.250 There was indeed so. 01:09.290 --> 01:10.130 So first up. 01:10.490 --> 01:11.410 First up. 01:11.450 --> 01:12.010 Wow. 01:12.410 --> 01:13.370 It did call. 01:13.370 --> 01:14.570 We weren't even watching it. 01:14.570 --> 01:16.530 And it came in and it called the webhook. 01:16.730 --> 01:18.050 And then something went wrong. 01:18.050 --> 01:19.930 And we know exactly what went wrong. 01:19.930 --> 01:24.370 What went wrong is that we were thinking it was like a chat trigger node and it wasn't. 01:24.370 --> 01:25.970 It was this stuff here. 01:26.170 --> 01:30.010 So what we really want to do is get this question here. 01:30.050 --> 01:33.850 Body question how many shares of Spy do I have in my portfolio? 01:33.890 --> 01:38.290 Do you see the way it's filled in the field that we described in 11 labs? 01:38.330 --> 01:44.250 Remember we described that the body of the question needed needed the body of the of the post, of the 01:44.290 --> 01:48.410 of the webhook needed to contain a field called question with a question. 01:48.650 --> 01:51.010 So we need this input to point to there. 01:51.010 --> 01:55.370 So we simply select define below just like before. 01:55.530 --> 01:57.970 And now we can just drag and drop from here. 01:57.970 --> 02:00.170 You can probably now already fill in that. 02:00.170 --> 02:06.690 That is going to be the JSON is the expression is going to be JSON dot body dot question. 02:06.690 --> 02:10.170 We want the body of this and we want to collect the question. 02:10.170 --> 02:12.730 And there is the question okay. 02:13.090 --> 02:15.650 And now we go back here again. 02:15.970 --> 02:18.810 And now we press execute Workflow again. 02:19.050 --> 02:21.770 And we will try this one more time okay. 02:21.850 --> 02:23.090 Over here we go. 02:23.330 --> 02:25.370 And now we're going to press preview. 02:25.730 --> 02:27.010 And let's try this. 02:28.450 --> 02:28.810 Hello. 02:28.850 --> 02:29.890 How can I help you today. 02:30.170 --> 02:30.690 Hello. 02:30.730 --> 02:33.530 How many shares of spy do I have in my portfolio? 02:34.890 --> 02:36.650 Let me check your portfolio for you. 02:39.450 --> 02:42.330 You have two shares of Spy in your portfolio. 02:43.730 --> 02:44.450 That's great. 02:44.450 --> 02:45.010 Thank you. 02:46.770 --> 02:47.890 So there you have it. 02:47.890 --> 02:50.010 I feel the need to thank it. 02:50.170 --> 02:53.690 Uh, success wasn't that cool. 02:53.690 --> 02:56.130 And, of course, what you saw, even though there still is. 02:56.290 --> 03:01.850 The latency isn't perfect and there's more that can be done there, but it's pretty impressive and certainly 03:01.890 --> 03:04.250 compared to the prior approach. 03:04.410 --> 03:07.810 And whilst I called this more complicated, it was actually pretty simple. 03:07.810 --> 03:11.010 We've got this thing running fast and it was super cool. 03:11.010 --> 03:14.650 And if you go over here you can see everything ran lots of ticks. 03:14.650 --> 03:21.130 Of course, I hardly need to tell you that this webhook was called by 11 labs and everything ran successfully. 03:21.130 --> 03:26.690 So let's talk through the round trip of what actually happened as a way to kind of revise what this 03:26.690 --> 03:32.050 is all about, because it was quite simple to build this, but there's a lot of moving parts behind 03:32.050 --> 03:32.450 the scenes. 03:32.450 --> 03:34.930 The complexity is in what actually happened. 03:35.210 --> 03:38.690 We start our journey in 11 labs where we set up an agent. 03:38.690 --> 03:42.370 Just saying, you're a helpful assistant and using Russell as our agent. 03:42.370 --> 03:46.370 And and then we equipped Russell with a tool. 03:46.370 --> 03:51.490 And that tool was there's some tools that we could have used like an end conversation tool, a skip 03:51.530 --> 03:52.450 turn tool. 03:52.450 --> 03:53.330 We didn't use that. 03:53.420 --> 03:58.980 We set up a new tool, which was a webhook tool, which means this is a tool that's able to call out 03:58.980 --> 04:01.580 to the internet, to a webhook that you give it. 04:01.580 --> 04:02.940 And here's how we set it up. 04:03.140 --> 04:04.260 We gave it a webhook. 04:04.260 --> 04:06.820 We gave it one that we'd taken from N810. 04:06.980 --> 04:07.900 We gave it a webhook. 04:07.900 --> 04:14.460 That means it can submit a request to N810 to do some processing, and we use a webhook. 04:14.500 --> 04:20.860 We use the method of post, which means that we can post a body of JSON to to this endpoint. 04:21.100 --> 04:24.420 Um, and look, I didn't save after I changed this, I don't think. 04:24.420 --> 04:25.340 Let me just move. 04:25.340 --> 04:28.420 This should have been a bit longer or we risk some sort of timeout. 04:28.460 --> 04:30.220 Always remember to press the save button here. 04:30.220 --> 04:31.460 It's easy to miss that. 04:31.660 --> 04:34.060 Uh, you probably saw me doing that and were asking me. 04:34.260 --> 04:37.380 Uh, okay, so we do that, we call this post. 04:37.380 --> 04:43.420 And in the description we can specify to the LM, we need you to provide some information and that will 04:43.420 --> 04:46.500 be sent in the post message in the body of it. 04:46.500 --> 04:49.180 And we said, hey, we want you to fill in a field called question. 04:49.220 --> 04:55.430 We want it to be the specific question about the users equity portfolio that the specialist should answer. 04:55.510 --> 04:57.790 And so our LLM did that. 04:57.830 --> 05:01.790 It populated that question and that got sent to Nan. 05:01.830 --> 05:06.190 And that's where we pick up the tail in and end the webhook gets called. 05:06.190 --> 05:11.070 This is the same URL that gets called that then passes it to the AI agent. 05:11.110 --> 05:13.590 It doesn't look for some chat information. 05:13.590 --> 05:19.470 No, it looks for JSON, which is exactly where the question comes in right there. 05:19.630 --> 05:22.030 That then goes to Gemini in the usual way. 05:22.270 --> 05:23.950 It also uses a tool. 05:24.230 --> 05:27.030 The tool it uses is one to look up in Google Sheets. 05:27.230 --> 05:33.270 And then that goes into a respond to webhook, which just takes all of this JSON and sends it back. 05:33.430 --> 05:38.350 Back to Russell, who is waiting for us over here and comes back as this tool. 05:38.350 --> 05:38.870 And Russell. 05:38.910 --> 05:42.710 The agent then responds, wow, take a breath. 05:42.830 --> 05:49.150 There's a lot of moving parts, even though it looks so simple and the result is a lower latency, faster, 05:49.150 --> 05:56.430 more responsive approach, which has the combination of the best functionality from both 11 labs and 05:56.430 --> 05:57.230 from Nln. 05:57.230 --> 06:01.350 And if you try this yourself, you probably noticed that if you try to ask a follow on question, then 06:01.390 --> 06:02.750 you fail miserably. 06:02.750 --> 06:08.310 And that's because we're using the test version, which only executes workflows once at a time. 06:08.350 --> 06:11.190 Surely it's time for us to publish the production version. 06:11.190 --> 06:14.950 To do that, we want to grab this production webhook URL. 06:15.230 --> 06:16.070 We come here. 06:16.110 --> 06:18.750 We'll save it and we'll press the publish button. 06:18.750 --> 06:20.550 So this gets published out there. 06:20.590 --> 06:21.390 How exciting. 06:21.430 --> 06:22.550 Workflow is published. 06:22.590 --> 06:23.230 Got it. 06:23.590 --> 06:26.230 And now we want to come back to 11 labs. 06:26.230 --> 06:28.110 We want to come into our tool. 06:28.110 --> 06:29.870 We want to come into this tool. 06:30.070 --> 06:34.350 We need to go to this URL and change it for the production URL. 06:34.550 --> 06:37.110 Don't forget to press save and you always forget to press save. 06:37.150 --> 06:37.710 There we go. 06:37.750 --> 06:42.670 I didn't, and now it's time for us to test our production agent. 06:42.670 --> 06:45.750 And of course we should start by publishing it here as well. 06:45.750 --> 06:53.440 We also want to publish Live our agent on on this side to uh, so that is, uh, and we'll do a copy 06:53.480 --> 06:54.400 shareable link. 06:54.600 --> 07:00.680 Uh, so we have now published both we've published and and live and we've published 11 labs live. 07:00.840 --> 07:02.560 And it's time for us to try this out. 07:02.560 --> 07:05.920 I can open a new tab, I can paste in the shareable link. 07:05.920 --> 07:11.080 This is my published AI agent Russell on the internet. 07:11.120 --> 07:12.200 Let's give it a try. 07:12.520 --> 07:12.880 Hello. 07:12.920 --> 07:14.080 How can I help you today? 07:14.120 --> 07:14.560 Hello. 07:14.600 --> 07:18.240 I'm interested in how many shares of spy I have in my portfolio. 07:19.600 --> 07:23.400 Let me quickly check your portfolio for the number of spy shares you hold. 07:24.400 --> 07:27.800 Do you currently have two shares of Spy in your portfolio? 07:28.080 --> 07:31.240 Is there anything else I can help you with regarding your investments? 07:31.560 --> 07:32.000 Yeah. 07:32.000 --> 07:35.920 What what other stocks do I have in my portfolio apart from Spy? 07:37.360 --> 07:40.200 Let me check what other stocks you have in your portfolio. 07:42.960 --> 07:49.760 Besides Spy, your portfolio includes B and D, BMD, which is the Vanguard Total Bond Market ETF, 07:50.360 --> 07:56.480 the iShares Core Allocation ETF, and the iShares Core Aggressive Allocation ETF. 07:56.520 --> 07:59.680 Do you have any questions about these holdings? 08:00.800 --> 08:01.560 No thanks. 08:04.120 --> 08:05.200 There we have it. 08:05.440 --> 08:07.560 Well, everything ran nicely. 08:07.680 --> 08:14.360 That was us running in production with multiple questions between 11 labs and nt.m. 08:14.400 --> 08:15.240 The end to end flow. 08:15.280 --> 08:22.160 The second approach working perfectly with both 11 labs and deployed to production. 08:22.160 --> 08:25.880 Well, you remember I told you that today was going to be both challenging and satisfying? 08:25.920 --> 08:27.280 I think I delivered on both. 08:27.280 --> 08:27.960 It was a bit. 08:28.000 --> 08:30.560 It was a little bit fiddly from time to time. 08:30.760 --> 08:33.720 Uh, especially when we were configuring that, that request. 08:33.880 --> 08:38.400 Uh, again, don't worry if you if that was too much, we're going to do more of that. 08:38.400 --> 08:42.320 And also remember ChatGPT is so good with that kind of thing. 08:42.360 --> 08:44.200 This this is boilerplate stuff. 08:44.200 --> 08:49.200 So if you want help configuring a webhook, a tool like that, you can just describe it. 08:49.200 --> 08:51.720 You'll get step by step instructions so easily. 08:51.760 --> 08:54.160 Most important is that you get the gist of what we're doing. 08:54.200 --> 08:55.920 The basic building blocks. 08:56.040 --> 09:02.200 And essentially we've discovered that the best way to integrate 11 labs with N810 in order to build 09:02.200 --> 09:05.600 an interactive voice agent with relatively low latency. 09:05.640 --> 09:08.360 It's still not perfect, but it's much better. 09:08.400 --> 09:11.480 The right way to do it is to have 11 labs be the boss. 09:11.600 --> 09:17.840 Use the voice agents in 11 labs, give it a tool, and let it use that tool to call into our business 09:17.840 --> 09:22.240 logic running in Nan, which is what Nan is so, so great at. 09:22.640 --> 09:23.240 Okay. 09:23.360 --> 09:25.680 And that concludes week two. 09:25.800 --> 09:29.320 Day two integrating 11 labs and N810. 09:29.880 --> 09:31.640 That's that's a wrap. 09:31.640 --> 09:37.920 And that means that you are now at the 47% point about to cross the halfway mark. 09:38.120 --> 09:40.960 Tomorrow is going to be a whole new topic. 09:41.280 --> 09:42.760 Can't wait to see you then.