WEBVTT 00:00.440 --> 00:03.640 Well, have I got a treat in store for you? 00:04.080 --> 00:12.400 Uh, so welcome to day five of week five, the wrap up project for Autogen week in the form of the agent 00:12.400 --> 00:13.320 creator. 00:13.760 --> 00:17.440 And, well, let me tell you about this project. 00:17.440 --> 00:20.120 It's going to be quite quick, but it's going to be fun. 00:20.560 --> 00:22.520 So look, it's this is the big idea. 00:22.520 --> 00:25.840 It's one of these things that has pluses and minuses. 00:25.840 --> 00:28.360 It's a it's a bit of an out there project. 00:28.400 --> 00:31.920 Look, first of all this is designed to be primarily educational. 00:31.920 --> 00:37.920 So this is here to teach some of the innards of what you can do with this kind of framework and use 00:37.920 --> 00:39.760 Autogen in a different way. 00:39.800 --> 00:42.960 So something that's different, unlike anything else that we've done. 00:43.320 --> 00:44.800 Secondly, it's entertaining. 00:44.800 --> 00:47.680 This is something that should be intellectually entertaining. 00:47.800 --> 00:50.160 This is quite, quite out there. 00:50.320 --> 00:55.320 And on that front it is it's an edgy idea and thing to try with. 00:55.360 --> 01:00.640 And it's very in vogue in terms of thinking about autonomy, agency and so on. 01:00.640 --> 01:02.810 One, but those are the three pluses. 01:02.810 --> 01:04.850 There are also three minuses. 01:05.130 --> 01:12.450 Uh, it's not particularly commercial, so I do like to focus this whole course on commercial benefits. 01:12.450 --> 01:17.450 There is a twist that there is something a bit commercial about it, as you'll see, but it's a bit 01:17.450 --> 01:20.330 of a of a side benefit of it. 01:20.570 --> 01:22.370 Um, it's unreliable. 01:22.370 --> 01:28.570 The project that we're about to do, by its very nature is something that may sometimes fail, and also 01:28.570 --> 01:29.690 it's unsafe. 01:29.810 --> 01:33.930 So we are going to be building something that's going to be creating and running Python code, and it's 01:33.930 --> 01:39.410 going to be doing so in a, in a, generally without guardrails on, which means you have to run this 01:39.410 --> 01:42.610 at your own risk, take whatever precautions you wish. 01:42.890 --> 01:47.330 Um, I may in time put this into a Docker container or something, but that would be quite a bore because 01:47.330 --> 01:51.130 you'd need to install all of, uh, autogen in the Docker container, so it wouldn't it wouldn't be 01:51.130 --> 01:51.530 easy. 01:51.730 --> 01:55.130 Um, but as it is, take it for what it is. 01:55.130 --> 01:56.410 You could just watch me run it. 01:56.410 --> 02:00.770 If you don't feel confident that you know what's going on and that you're comfortable taking the risk 02:00.770 --> 02:06.210 of letting an agent write Python code and then executing that Python code. 02:06.250 --> 02:10.290 Natively on your box, which is not for the faint of heart. 02:10.490 --> 02:12.050 So eyes wide open. 02:12.090 --> 02:14.530 You don't need to execute any code if you don't want to. 02:14.650 --> 02:16.210 And please do do. 02:16.250 --> 02:17.570 Take that to heart. 02:18.010 --> 02:20.930 Run the code yourself only if you're comfortable with it. 02:20.970 --> 02:22.650 Okay, well, hopefully I've intrigued you somewhat. 02:22.650 --> 02:24.290 Let me tell you what this idea is. 02:24.570 --> 02:30.410 So we're going to explore the flexible, dynamic aspect of Autogen. 02:30.610 --> 02:37.210 We are going to make a creator agent an agent, as I say that can write a Python module. 02:37.250 --> 02:41.250 It will it will use an existing Python module as its kind of template. 02:41.770 --> 02:45.530 And that Python module is going to be an agent. 02:45.530 --> 02:51.250 It's going to be an autogen agent chat agent that's running in Autogen core. 02:51.650 --> 02:57.250 So we're going to build an agent that can write a Python module of an agent. 02:57.250 --> 03:02.700 It's going to create an agent, and it's going to change it so that the Python module it makes is something 03:02.700 --> 03:10.860 new, a different agent that doesn't exist, and then the creator agent is actually going to register 03:10.900 --> 03:15.300 its creation with the distributed runtime with autogen core. 03:16.020 --> 03:21.020 In other words, dare I say it, it's going to kind of give, give, give. 03:21.060 --> 03:22.020 Do I say give birth? 03:22.020 --> 03:22.540 It's going to. 03:22.580 --> 03:24.420 It's going to instantiate this agent. 03:24.460 --> 03:25.060 It's going to. 03:25.100 --> 03:25.940 Let's get technical. 03:25.940 --> 03:31.060 It's going to instantiate the agent that's going to be a running agent that was created by the creator 03:31.060 --> 03:31.660 agent. 03:31.860 --> 03:37.340 And it's then it's also as part of writing this agent based on the template it will use. 03:37.740 --> 03:41.060 Its agents are going to be able to message each other. 03:41.060 --> 03:46.060 So it's going to be possible for agents to message each other and and interact. 03:46.060 --> 03:48.740 These agents that have been created by the agent. 03:50.500 --> 03:57.180 And overall, they're going to have an objective of coming up with business ideas, commercial ideas 03:57.180 --> 04:01.940 for putting agent I, agent I into practice to make money. 04:01.980 --> 04:09.060 So there is a commercial angle subtly, because the overall objective of our of our team of created 04:09.060 --> 04:12.260 agents is to be to try and make you money by coming up with ideas. 04:12.260 --> 04:14.580 And of course, you can shift the the overall objective. 04:14.620 --> 04:19.060 You can make it something different or, you know, maybe not money with a genetic eye, but how to 04:19.060 --> 04:21.340 make a quick buck in some other way. 04:21.380 --> 04:27.540 You can assign it some other task and let it think, but the idea is that you can spawn this army of 04:27.580 --> 04:34.100 agents that can think about something and interact with each other, and you can help put the plant 04:34.220 --> 04:37.020 the initial seed and then see what happens. 04:37.620 --> 04:43.380 And as part of the educational aspect of this, we're going to be heavily using asynchronous Python. 04:43.380 --> 04:48.780 It would be quite a drag if this thing had to happen in a serial way, with each agent being created 04:48.780 --> 04:51.180 and then messaging the others one by one. 04:51.180 --> 04:57.620 That would take quite a long time, but now we're going to use Asyncio async IO to make sure that things 04:57.620 --> 04:58.220 fly. 04:59.460 --> 05:02.980 And so with that introduction, let's get to the code.