WEBVTT 00:00.120 --> 00:03.240 So we saw previously how cool agents are. 00:03.480 --> 00:09.960 They take as input text and they're able to select the correct tools and execute them, and then curate 00:10.000 --> 00:11.480 an elaborate answer. 00:12.040 --> 00:15.960 They equipped with llms with the capability to do anything. 00:16.080 --> 00:19.040 So up until now we treated them as a black box. 00:19.040 --> 00:23.720 We don't know how this tool selection happens and what's happening under the hood. 00:23.720 --> 00:25.480 And for us it was like magic. 00:25.680 --> 00:32.080 And in this section we're going to dive deep into the internals of the react agent framework. 00:32.080 --> 00:34.440 And we're going to see how this is implemented. 00:34.440 --> 00:38.080 And we're going to implement our own version of the react agent. 00:38.240 --> 00:44.040 And by doing this, we'll understand all the moving parts that are happening in react algorithm. 00:44.280 --> 00:48.920 And we'll be able to build complex technologies with this information. 00:49.080 --> 00:53.360 So in my opinion, this is the most beautiful part in LLM application development. 00:53.360 --> 00:55.120 And I hope you enjoy this section. 00:55.520 --> 01:02.160 So remember, in the earlier videos we defined an agent by using the Create React agent which received 01:02.160 --> 01:05.760 the list of tools Dlrm and the react prompt. 01:07.400 --> 01:13.240 We then invoke the agent with the input of what we want to process and the agent when we run. 01:13.240 --> 01:17.260 It was smart enough to select the correct tool that we had in our toolbox. 01:17.420 --> 01:21.300 And to execute that tool and from that to curate an answer. 01:21.620 --> 01:25.940 And in the next couple of videos, we're going to implement all of that from zero. 01:26.300 --> 01:32.900 Now I think it's super important to know how to implement this kind of logic from zero, from the ground 01:32.940 --> 01:33.260 up. 01:33.460 --> 01:37.900 And this is going to be the basis for every Agentic application. 01:37.900 --> 01:43.380 And this is the core concept that you need to know when developing generative AI agents. 01:43.420 --> 01:48.180 This section of the course is going to provide you with depth and really understanding. 01:48.340 --> 01:50.340 What are generative AI agents. 01:50.700 --> 01:54.380 Now, you might have heard about Landgraf and Lange. 01:54.380 --> 02:00.180 Jane is pushing Landgraf for developing generative AI agents, and the LinkedIn team is very bullish 02:00.180 --> 02:00.980 on Landgraf. 02:00.980 --> 02:03.340 And I agree, it's an amazing technology. 02:03.780 --> 02:10.180 Now, what we're going to do in this section is actually a prerequisite to understand why Landgraf is 02:10.180 --> 02:12.420 important and why do we need Landgraf. 02:12.660 --> 02:18.500 So regardless, if you're going to implement your agents within LinkedIn or within Landgraf or with 02:18.500 --> 02:24.300 any other framework, I really recommend you watching this section because it will really give you the 02:24.340 --> 02:27.180 depth of understanding what our agents. 02:27.220 --> 02:27.620 All right. 02:27.720 --> 02:33.240 So the implementation I'm going to show you is going to be based with the link chain framework. 02:33.480 --> 02:39.320 Now there is a version of the same implementation but with the graph framework. 02:39.640 --> 02:44.160 And this is the official recommendation from the link chain team to use graph. 02:44.560 --> 02:51.840 However, I have to say that in order to understand JNI, I think it's very important to understand 02:51.840 --> 02:56.440 its history and to understand how things evolved, what they are today. 02:56.560 --> 03:02.800 Because if I were to show you right now the graph implementation, then you wouldn't have gotten that 03:02.800 --> 03:06.720 much depth out of it, because a lot of abstraction is going to be made. 03:06.840 --> 03:13.920 And I think building things from the ground up really gives you the sense and idea of why things are 03:13.920 --> 03:15.760 implemented the way they are. 03:17.280 --> 03:22.000 And if you want to support me in the course, I would appreciate it very much if you could leave me 03:22.000 --> 03:23.200 a Udemy rating. 03:23.240 --> 03:28.880 This really motivates me to continue and create new videos for this course and reach the content, and 03:28.880 --> 03:33.680 it also helps future students to decide whether this course is a right fit for them or not. 03:33.800 --> 03:39.280 So if you don't mind, I'd appreciate if you can pause, go to the ratings section and leave me a Udemy 03:39.320 --> 03:39.800 review. 03:40.120 --> 03:42.040 Thank you so much and see you in the next video.