WEBVTT 00:02.160 --> 00:07.800 In this tutorial, I'm going to show you how to build a WhatsApp AI agent that can plan everything for 00:07.800 --> 00:08.120 you. 00:08.640 --> 00:12.720 We'll be using Deep Sigrun as the brain for our AI planner agent. 00:13.280 --> 00:19.000 It's an amazing reasoning model that's great for planning, decision making and complex reasoning tasks. 00:19.440 --> 00:27.000 The planner agent will analyze your requests and craft a detailed plan to our second agent, which is 00:27.040 --> 00:28.080 a research agent. 00:29.360 --> 00:30.720 Now let's do a quick demo. 00:30.920 --> 00:37.560 Let's say I wanted to plan a budget friendly one month stay in Thailand as a digital nomad. 00:38.200 --> 00:44.040 I need a cheap place to stay, good wifi, coworking spaces and some great attractions to visit. 00:45.320 --> 00:46.560 Let's test the workflow. 00:46.600 --> 00:53.000 Normally you wouldn't do this if the setup is active, but I wanted to show you the flow. 00:53.640 --> 00:55.920 So I'm going to hit this workflow. 00:58.640 --> 01:00.160 And send the message. 01:00.600 --> 01:08.950 Our agent is now analyzing the request and preparing a plan, passing it to our second agent, which 01:08.950 --> 01:11.270 is research agent. 01:14.150 --> 01:17.150 The research agent uses Serp API to search the web. 01:20.150 --> 01:23.870 We should get a response in WhatsApp in a sec. 01:24.710 --> 01:26.350 So let's just wait a moment. 01:31.990 --> 01:32.390 All right. 01:32.390 --> 01:32.830 Great. 01:32.870 --> 01:37.230 Let's have a look for a budget friendly one month stay in Chiangmai. 01:38.030 --> 01:42.870 So as you can see, our agent recommends Chiang Mai as a city to stay. 01:45.750 --> 01:50.670 It has prepared affordable accommodation options with website links. 01:51.190 --> 01:55.670 It also suggests that negotiation can help lower the price, which is really good point. 01:55.710 --> 02:03.100 Actually, it remembers that we are looking for a place with good wifi, which is great. 02:12.140 --> 02:15.780 It also funds coworking spaces. 02:18.300 --> 02:23.220 And provides website links with addresses. 02:32.060 --> 02:33.180 Top attractions. 02:36.780 --> 02:37.740 Cooking classes. 02:39.780 --> 02:41.300 Transportation options. 02:41.340 --> 02:44.060 It recommends to rent a scooter. 02:47.740 --> 02:47.940 Or. 02:47.940 --> 02:48.900 I really like this. 02:50.220 --> 02:57.700 It even provided a total cost estimate so you know exactly what to expect, which is impressive. 03:00.500 --> 03:02.180 You also have additional tips. 03:07.340 --> 03:07.700 All right. 03:07.700 --> 03:08.020 Great. 03:08.020 --> 03:09.460 So this is just an example. 03:09.700 --> 03:14.340 You can use this agent to plan anything both for personal and business use cases. 03:14.380 --> 03:18.940 For example you can use it for market analysis or competitors research. 03:19.140 --> 03:21.980 You can plan whatever you want with this agent. 03:22.180 --> 03:30.660 So just to quickly show you, when we open our Typekit R1 planner agent, we can see all the reasoning 03:30.660 --> 03:31.300 steps. 03:32.500 --> 03:37.140 So okay, so I need to plan a budget friendly one or stay in Thailand as a digital nomad. 03:41.900 --> 03:43.260 And 3D days. 03:45.220 --> 03:49.580 So you can check exactly the thinking process of this agent. 03:52.100 --> 03:56.220 That I should update a local events or festivals happening during my stay. 03:57.940 --> 04:07.450 So this agent really helps to plan our trip to Thailand And it passed all this information to our research 04:07.450 --> 04:10.690 agent, which has performed the web search. 04:11.810 --> 04:16.570 Okay, now let's jump into A10 and build these agents from scratch. 04:19.930 --> 04:22.770 Now let's start building our agents. 04:23.610 --> 04:27.890 So usually the first thing we need to do is add our trigger node. 04:28.450 --> 04:30.610 In this setup we'll be using WhatsApp. 04:31.170 --> 04:38.890 And some time ago I recorded a video on how to easily connect WhatsApp to your agents in an A10 in just 04:38.890 --> 04:39.690 six minutes. 04:40.050 --> 04:41.610 So I'll play that later. 04:41.810 --> 04:48.610 For now, let's focus on building our agents first, and then we'll integrate WhatsApp as an input and 04:48.610 --> 04:52.570 output for the first and last nodes in our workflow. 04:54.170 --> 04:56.690 So let's search for agent node. 04:58.570 --> 05:00.170 Click on add node button. 05:02.490 --> 05:05.200 Search for Agent. 05:09.440 --> 05:10.640 Let's close it for now. 05:12.320 --> 05:16.360 So as I said, we trigger our workflow, uh, in WhatsApp. 05:16.360 --> 05:18.600 So we need to remove this node. 05:22.320 --> 05:26.640 Now we need to add the chat model. 05:28.040 --> 05:36.120 So as the brain for our planner agent we'll use R1 through grok the host distilled version of this model 05:36.120 --> 05:37.120 which we can use. 05:38.160 --> 05:42.560 So add chat model search for grok. 05:48.120 --> 05:54.160 Now to get started with grok go to console.com. 05:56.840 --> 05:58.520 And you will land on this page. 05:59.600 --> 06:05.870 So it's a super fast AI hosting platform around ten times faster than most services. 06:06.390 --> 06:08.390 They host a lot of models themselves. 06:10.430 --> 06:15.270 So click on the console to access the playground. 06:16.510 --> 06:17.230 Sign up. 06:18.950 --> 06:20.870 Head over to API keys. 06:24.030 --> 06:25.510 Create a new API key. 06:27.230 --> 06:28.070 Give it a name. 06:34.190 --> 06:34.950 Copy it. 06:38.270 --> 06:43.790 Now if you want to learn more about these services and what models they offer. 06:44.550 --> 06:48.590 Under documentation click on models. 06:50.670 --> 06:53.630 And you can see all the models they they host. 06:56.470 --> 06:58.070 So we are interested in this one. 07:02.390 --> 07:07.710 You can even use their API for free for some time on the free plan, which is great. 07:08.550 --> 07:10.030 Now go back to N810. 07:15.110 --> 07:20.350 Create new credential and paste your API key right here. 07:22.150 --> 07:22.790 Hit save. 07:32.750 --> 07:35.110 Now you will have access to all the models. 07:36.110 --> 07:42.990 So just select Dopesick or one distiller llama seven TB. 07:45.430 --> 07:46.550 And you are all set. 07:47.150 --> 07:52.670 We want the planner agents output to be retained and passed to the research agent. 07:53.430 --> 07:56.670 So we need to add some kind of memory. 07:57.710 --> 08:00.230 So let's add the window buffer memory node. 08:00.790 --> 08:02.990 So click on this small plus button. 08:05.670 --> 08:10.860 Select window buffer memory, which is the easiest way to give our agents a memory. 08:14.260 --> 08:23.100 Let's keep the context window length to five, which means our agent will be able to remember the five 08:23.140 --> 08:24.300 last interactions. 08:25.420 --> 08:26.300 Let's close it. 08:32.780 --> 08:35.820 Now let's set up our research agent. 08:37.060 --> 08:40.140 So click this plus button to add nodes. 08:41.420 --> 08:43.740 Search for agent again. 08:46.180 --> 08:47.380 Let's rename it to. 08:50.220 --> 08:51.260 Research agent. 08:59.660 --> 09:00.420 Confirm. 09:02.380 --> 09:03.620 And let's close it for now. 09:08.330 --> 09:11.810 Let's get rid of this friend chat. 09:11.850 --> 09:12.770 Message received. 09:12.810 --> 09:13.650 Trigger note. 09:23.250 --> 09:29.810 Now we also need to decide which large language model we want to use for the brain of our research agent. 09:37.250 --> 09:41.330 I want to use OpenAI's GPT four. 09:45.450 --> 09:50.770 For this tutorial, we'll use GPT four mini because it's fast and cost effective. 09:51.370 --> 09:52.290 Let's close it. 09:56.290 --> 10:01.890 Now we need to give our research agent access to some tools. 10:01.890 --> 10:06.810 In that case, we need to connect it to CEP API. 10:07.720 --> 10:09.400 So click on this plus button. 10:10.920 --> 10:12.800 Search for Serp API. 10:14.520 --> 10:20.200 Serp API basically allows the agent to search the web. 10:23.000 --> 10:24.280 So select this node. 10:26.920 --> 10:34.680 Now we need to create an account and provide credentials to connect our account to this node. 10:35.760 --> 10:39.360 So click on this dropdown Create new credential. 10:40.320 --> 10:42.920 As you can see we need API key. 10:44.040 --> 10:45.840 So go to Serp API. 10:47.960 --> 10:50.000 And you will land on this page. 10:51.520 --> 10:58.360 So this tool allows to scrape search engines using the API service. 10:59.560 --> 11:00.680 Create an account. 11:01.640 --> 11:03.080 Go to API key. 11:07.640 --> 11:12.510 Create your API key, copy it and paste it into an A-10. 11:14.270 --> 11:16.030 So we need to create a credential. 11:19.030 --> 11:20.310 And paste it right here. 11:21.870 --> 11:22.910 Then click save. 11:23.590 --> 11:25.910 And it should show as successful. 11:26.230 --> 11:32.710 Now that we have set up the planner agent and research agent, let's complete the workflow by integrating 11:32.710 --> 11:35.790 WhatsApp so you can communicate with your planner agent. 11:36.310 --> 11:40.110 So now I'm going to play a video on how to do it in a few minutes. 11:40.670 --> 11:46.950 Before we jump into configuration, just quickly to give you some insights, why using WhatsApp as a 11:46.950 --> 11:52.110 communication channel is so important, and why almost every business should use it. 11:52.310 --> 11:56.870 Did you know that WhatsApp is one of the fastest growing communication channels for businesses? 11:57.430 --> 12:04.230 Over 50 million businesses are already using WhatsApp to connect with their customers, and the app 12:04.230 --> 12:06.470 has over 3 billion users globally. 12:07.150 --> 12:18.590 WhatsApp messages have a 98% open rate, compared to just 21% for email, and 68% of users say WhatsApp 12:18.590 --> 12:20.510 is the easiest way to reach a business. 12:21.070 --> 12:26.710 It's less common in the US but is very popular in Europe, South America and Asia. 12:27.270 --> 12:29.270 Now let's move on to the integration part. 12:29.870 --> 12:33.110 First, we need a trigger to activate our workflow. 12:33.630 --> 12:36.830 So we'll start with the WhatsApp business cloud node. 12:37.510 --> 12:39.430 So search for WhatsApp. 12:41.510 --> 12:43.150 Choose WhatsApp Business Cloud. 12:45.230 --> 12:46.750 And choose on messages. 12:48.550 --> 12:56.870 Basically this node acts as a trigger it listens to for incoming messages on your WhatsApp account. 12:57.950 --> 13:02.430 Now we have to set up credentials to connect to our WhatsApp business account. 13:05.710 --> 13:13.700 So we will need our client ID and client Client's secret, so we need to get it from Facebook business 13:13.700 --> 13:14.260 account. 13:14.500 --> 13:17.100 Since WhatsApp business is managed under meta. 13:17.420 --> 13:22.660 So go to business.facebook.com and log in with your Facebook account. 13:23.300 --> 13:28.180 If you are already logged in, it will take you to your Business Manager dashboard. 13:28.860 --> 13:34.100 If you have never used Facebook Business Manager, you will need to create a business portfolio. 13:34.620 --> 13:43.580 So go to business facebook.com slash settings and provide all the required details like your business 13:43.580 --> 13:48.820 name, email and any other information requested and click Create Business. 13:49.620 --> 13:52.700 Now choose the business portfolio you have created. 13:59.220 --> 14:05.580 And in your Business manager, look for the account section on the left hand menu and click apps. 14:09.460 --> 14:13.170 Now click add and create a new app ID. 14:17.050 --> 14:19.210 Here you have to provide an app name. 14:19.610 --> 14:23.370 So let's say an app contact email. 14:28.130 --> 14:28.810 So. 14:31.210 --> 14:32.410 Then click on next. 14:33.250 --> 14:34.130 Scroll down. 14:36.170 --> 14:36.570 Choose. 14:36.610 --> 14:37.050 Other. 14:38.490 --> 14:39.130 Next. 14:45.730 --> 14:46.410 Business. 14:46.610 --> 14:47.170 Next. 14:49.410 --> 14:50.850 And create the app. 14:52.130 --> 14:53.970 Now click on the app you created. 14:55.570 --> 14:56.490 App settings. 14:58.770 --> 14:59.530 Basic. 15:06.650 --> 15:09.970 If you remember, we are looking for an app ID and app Secret. 15:10.810 --> 15:12.440 So this is our app ID. 15:14.160 --> 15:14.840 Copy it. 15:17.120 --> 15:18.240 Go back to Nan. 15:20.800 --> 15:21.560 Paste it here. 15:23.920 --> 15:27.880 Show up secret then copy up secret. 15:30.280 --> 15:33.640 Paste it here and test connection. 15:34.360 --> 15:39.480 So now let's set up WhatsApp node to send replies back to the user. 15:40.520 --> 15:46.160 So this will allow your agent to respond to messages automatically. 15:48.360 --> 15:49.880 Search for WhatsApp. 15:53.360 --> 15:56.320 And choose send message node. 16:00.600 --> 16:04.440 Now you have to provide access token and business account ID. 16:07.720 --> 16:11.480 So let's go back to our Business Manager account. 16:12.440 --> 16:14.200 go through API setup. 16:25.240 --> 16:34.400 And here you have to generate access token and then copy this access token to place it here. 16:35.760 --> 16:38.200 And you can find business account ID. 16:40.920 --> 16:45.520 Right here just copy it and paste it here. 16:50.480 --> 16:53.800 Then hit Test Connection. 16:54.680 --> 16:55.680 Just the last step. 16:55.680 --> 16:59.360 You need to choose your business mobile number. 17:01.920 --> 17:04.880 Or you can use the test number to test the connection. 17:06.600 --> 17:13.790 Now a key part of building AI agents is giving them a clear, well-designed prompt to guide what they 17:13.790 --> 17:15.190 do and how they respond. 17:15.550 --> 17:21.590 I've put together prompts for both agents, and I will share them with a template so you can easily 17:21.590 --> 17:22.830 copy and paste them. 17:23.790 --> 17:26.150 Let me just quickly show you what it looks like. 17:26.750 --> 17:28.870 So for our planner agent. 17:31.870 --> 17:40.350 You might say something like you are an AI planner agent responsible for generating structured, actionable 17:40.390 --> 17:41.230 task plans. 17:41.550 --> 17:48.270 Your goal is to break down objectives into clear steps that guide execution and instructions. 17:50.030 --> 17:51.790 And for our research agent. 17:55.270 --> 17:56.990 The prompt is more complex. 17:58.910 --> 18:05.590 So the role instructions use API to search for relevant results. 18:10.910 --> 18:13.550 Avoid special characters, etc..