WEBVTT 00:00.160 --> 00:03.880 And a very warm welcome to day two of week three. 00:03.920 --> 00:11.360 Our second day playing with crew, this time continuing doing some building and exploring the crew framework. 00:11.760 --> 00:15.440 To quickly recap what we did last time we learned about an agent. 00:15.480 --> 00:18.040 The agent being the smallest autonomous unit. 00:18.080 --> 00:20.040 It has an LM associated with it. 00:20.080 --> 00:21.120 It doesn't actually need to. 00:21.160 --> 00:23.720 You can have an agent with that name, but they typically do. 00:23.880 --> 00:25.840 It has a role, a goal, a backstory. 00:25.840 --> 00:28.800 And it also has memory and tools. 00:28.840 --> 00:30.960 Not that we've looked at either of them just yet. 00:31.560 --> 00:33.320 And then a task. 00:33.520 --> 00:37.600 This is the concept which doesn't have an analog in OpenAI agents SDK. 00:37.880 --> 00:44.440 A task is an assignment to be carried out with a description, expected output, perhaps an output file. 00:44.720 --> 00:47.120 And it's assigned to an agent. 00:47.320 --> 00:53.320 And then a crew, which is a team of agents and tasks together assigned to those agents. 00:53.320 --> 01:00.080 And they can run sequentially or hierarchically, in which case you'd have to assign a manager LM to 01:00.200 --> 01:03.480 figure out which task is assigned to which agent. 01:04.360 --> 01:09.360 So that's the overall structure of crew, which now should be pretty familiar to you. 01:09.360 --> 01:15.520 And you'll remember that there are five steps that we went through when we set up our first crew project. 01:15.520 --> 01:17.160 First of all, we created a project. 01:17.160 --> 01:23.320 We did create Create Crew, my project, or we did debate, but it could be whatever we want and it 01:23.320 --> 01:26.960 will set up that whole file system structure for us. 01:27.280 --> 01:31.200 We then go into source and then the name of the project and then config. 01:31.200 --> 01:35.360 And that is where we find the YAML files, which we can call them whatever we want. 01:35.360 --> 01:39.840 But by by by default there are two there one for the agents, one for the tasks. 01:39.840 --> 01:42.040 And we fill them in with the details. 01:42.320 --> 01:51.760 We then go to the crew.py module in our source project folder, and we create the agents, the tasks 01:51.760 --> 01:57.240 and the crew using the functions that are already set up there for us and using the decorators. 01:57.240 --> 02:03.890 And we reference the YAML config, although we don't need to, we could actually manually create a pass 02:03.890 --> 02:06.170 in those fields when we create each of those objects. 02:06.170 --> 02:12.570 But the config file makes that easy for us and keeps some of this prompting text separate from our code, 02:12.570 --> 02:14.490 which is a nice separation. 02:15.130 --> 02:24.010 Then fourth, we update the Main.py module to set the config to set up the run parameters that we did 02:24.210 --> 02:25.810 when we specify the inputs. 02:25.850 --> 02:33.010 The fact in our case that the topic of debate the motion was, as it happens, about AI regulation, 02:33.010 --> 02:36.890 and then we can run and the way we run is we type crew. 02:36.930 --> 02:44.490 I run from within the project folder, which behind the scenes does a run, and then it's off to the 02:44.490 --> 02:46.330 races with our crew framework. 02:47.290 --> 02:47.730 Okay. 02:47.730 --> 02:49.970 So we're going to go a little bit deeper in two ways. 02:49.970 --> 02:52.210 And we're going to set up another project today. 02:52.250 --> 02:57.890 And one of those ways is with tools, something that we're very familiar with already equipping agents 02:57.890 --> 02:58.850 with capabilities. 02:58.850 --> 03:04.530 We will see how to do that crew AI style And then the other one is about context. 03:04.530 --> 03:10.170 And this is how you tell crew what information is to be passed from one task to another. 03:10.170 --> 03:15.170 So these are the two extra details we're going to get into today when we build our second crew. 03:15.490 --> 03:17.650 And let's go and do that right now. 03:18.330 --> 03:23.370 So just before we go into crew, there's another of these APIs that I need you to sign up for, but 03:23.370 --> 03:25.650 you'll be pleased to hear it's another free one. 03:25.650 --> 03:29.970 And this is going to be quite a contrast with our experience with OpenAI agents SDK. 03:30.130 --> 03:38.770 We're going to use something called Serpa, which is a fast way to run Google queries from code. 03:38.810 --> 03:46.170 It's an API to run lightning fast Google search at an unbeatable price, and indeed it is unbeatable. 03:46.370 --> 03:52.370 So you go to Serpa Dev and you sign up by pressing that button there, and I've already done so and 03:52.370 --> 03:54.850 so I can sign in with my account. 03:54.850 --> 03:56.530 And here I am. 03:56.850 --> 04:03.460 And the thing is that you get 2500 free credits, which is more than enough for this course, and this 04:03.460 --> 04:08.300 will allow us to do plenty of searches quite happily indeed. 04:08.500 --> 04:11.020 And you will go to API key, which I won't do. 04:11.020 --> 04:16.980 Now you'll get my API key, but you go there to create your API key and as usual, copy it into your 04:16.980 --> 04:21.460 clipboard because that is something that you will need for your env file. 04:21.740 --> 04:28.780 And in your env file, you should have entered that in as Serpa underscore API underscore key cursor 04:28.780 --> 04:31.260 will actually prompt you for that if you start typing it. 04:31.580 --> 04:33.180 Serpa API key. 04:33.620 --> 04:36.140 And if you're interested Serp where that comes from. 04:36.140 --> 04:41.860 Serp stands for Search Engine results page, which is the common name for these kinds of services. 04:41.860 --> 04:44.500 And there is another one that's called something similar to Serpa. 04:44.500 --> 04:47.420 So be sure that you use Serpa API key. 04:47.620 --> 04:51.180 Not I think there's one called Serp API or something which is different. 04:51.180 --> 04:54.060 So be sure to have the right key there. 04:54.340 --> 04:58.260 Okay, so with that we are we are still in the debate folder. 04:58.260 --> 04:59.460 We will close that. 04:59.500 --> 05:07.620 Here we are in our crew folder in cursor for week three, and we are going to go into bringing up a 05:07.620 --> 05:08.820 new terminal. 05:08.860 --> 05:10.180 It's still got what we had before. 05:10.220 --> 05:12.260 Let's let's clear this let's exit that. 05:12.420 --> 05:14.620 Let's start a fresh terminal. 05:14.700 --> 05:15.660 Here it is. 05:15.780 --> 05:19.620 We're going to go into the third folder crew. 05:19.860 --> 05:23.380 And it's time for us again to create a new project. 05:23.420 --> 05:25.260 Do you remember how to create a crew project? 05:25.540 --> 05:26.620 Well here you do it. 05:26.660 --> 05:27.380 It's crew. 05:27.380 --> 05:33.020 I create crew again, as opposed to a flow, which is different. 05:33.260 --> 05:39.700 And then we're going to make one that's called financial researcher, somewhat inspired by the default 05:39.700 --> 05:40.940 one that's already there. 05:40.980 --> 05:43.460 Let's have a financial researcher. 05:43.620 --> 05:45.860 Let's create that right now. 05:47.060 --> 05:52.460 So we as before, we just choose OpenAI and GPT four or many. 05:52.500 --> 05:54.300 And we skip setting up the key. 05:54.300 --> 06:01.540 And it has created the financial Researcher project and a bunch of scaffolding for us, which is great. 06:01.540 --> 06:03.820 And this is where we will get started.