WEBVTT 00:00.160 --> 00:03.520 And a very warm welcome to week four, day three. 00:03.560 --> 00:04.200 Here we are. 00:04.240 --> 00:09.280 This is going to be the time that Landgraf is going to start to pay dividends. 00:09.280 --> 00:13.560 The investment that we've made in understanding the new terminology is going to come together, and 00:13.560 --> 00:15.240 we're going to see some real value. 00:15.480 --> 00:18.960 But first, as always, I have to give you a quick recap. 00:19.000 --> 00:22.720 Again, I'm going to get bored of it, but it's better that you get bored of it and you understand it. 00:22.840 --> 00:29.880 So before you can call Graph Invoke, which is how you kick off your graph, you have to define it. 00:29.920 --> 00:33.240 And defining it is these five steps to find your state class. 00:33.240 --> 00:34.240 The graph Builder. 00:34.240 --> 00:37.240 Create a node edges and compile the graph. 00:37.440 --> 00:38.840 Just a quick refresher. 00:38.840 --> 00:41.080 I'm sure you've got this all committed to memory now. 00:41.400 --> 00:43.440 Okay, so what are we actually going to cover today? 00:43.560 --> 00:48.160 Well, there's going to be a few ways we're going to go deeper into the world of graph. 00:48.200 --> 00:52.920 First of all, we're going to look at Lang Smith and have a moment of understanding how information 00:52.920 --> 00:54.360 gets logged there. 00:54.600 --> 00:59.240 We're then going to look at tools tool calling something that we've done many times now using the out 00:59.280 --> 01:02.000 of the box tools that come with Lang graph. 01:02.160 --> 01:05.680 And then we're going to build a custom tool again as we've done many times. 01:05.680 --> 01:07.000 So we'll see that working. 01:07.200 --> 01:12.610 And we'll finally end with checkpointing which is a very important part of it. 01:12.650 --> 01:13.290 Indeed. 01:13.770 --> 01:18.170 And to to tee that up to motivate checkpointing. 01:18.170 --> 01:21.290 I want to talk for a moment about something called the super step. 01:21.410 --> 01:23.050 So what's a super step? 01:23.290 --> 01:24.650 Well, a super step. 01:24.690 --> 01:25.770 Start with a docs. 01:25.810 --> 01:26.370 A super step. 01:26.370 --> 01:30.290 They define as a single iteration over the graph nodes. 01:30.290 --> 01:33.290 Nodes that run in parallel a part of the same super step. 01:33.410 --> 01:36.690 Nodes that run sequentially belong to a separate super step. 01:37.090 --> 01:38.770 So what does that mean exactly? 01:38.810 --> 01:40.690 Now this is super, super important. 01:40.730 --> 01:42.370 Super steps are super important. 01:42.370 --> 01:44.410 And it's something which you have to get your head around. 01:44.410 --> 01:48.770 And it's it's a it's a definitely may not be what you're expecting. 01:49.570 --> 01:59.610 So a graph defines one set of interactions between like agents and their use of tools and perhaps delegating 01:59.650 --> 02:00.930 to other agents. 02:00.930 --> 02:05.810 If you think back to the handoff of, uh, OpenAI agents SDK. 02:06.330 --> 02:10.450 So one invocation of the graph is just one kind of step. 02:10.450 --> 02:17.370 It's like when the user says one message, putting that message to our LLM, that is one invocation 02:17.370 --> 02:23.250 of the entire graph all the way through from top to bottom, and if then it comes back to the user with 02:23.250 --> 02:25.770 a response and the user types in another message. 02:25.770 --> 02:28.210 That's another invocation of the whole graph. 02:28.210 --> 02:32.090 And each of these invocations is a superstep. 02:32.130 --> 02:35.210 Every time you invoke the graph, that is a superstep. 02:35.450 --> 02:39.970 And so yeah, it's important to get your head around it, because you might initially think that maybe 02:39.970 --> 02:45.170 you could imagine that a node is like a human and a node is a chatbot, and that a graph is human and 02:45.170 --> 02:46.890 chatbot going backwards and forwards. 02:46.890 --> 02:51.330 But no, every time that there is that kind of human interaction, you should think of that as a whole 02:51.330 --> 02:52.650 invocation of the graph. 02:52.650 --> 02:57.850 Or in some situations, you might be resuming the graph from one point if it was paused for for a human 02:57.850 --> 02:58.570 to respond. 02:58.650 --> 03:04.690 So there's various ways of doing it, but each of these interactions is considered an entire superstep 03:04.890 --> 03:06.290 within a superstep. 03:06.330 --> 03:13.250 One one invocation of the graph belongs to to activities which happen in parallel as part of that step. 03:14.130 --> 03:20.730 So this is a bit repetitive, but yeah, the graph describes one full superstep an interaction between 03:20.730 --> 03:25.490 agents and tools and potentially multiple agents to achieve an outcome. 03:26.250 --> 03:28.170 Every user interaction. 03:28.380 --> 03:30.700 It's a fresh invoke call. 03:30.700 --> 03:33.460 It's a fresh time that you're calling Graph Invoke. 03:33.500 --> 03:35.220 There's also a graph resume. 03:35.220 --> 03:36.380 But but the point remains. 03:36.380 --> 03:37.780 It's a fresh call. 03:38.420 --> 03:44.460 And the reducer that I talked about, the thing that is able to combine the state that comes out of 03:44.460 --> 03:50.460 a call with the original state that applies during carrying out a single super step. 03:50.460 --> 03:53.180 That is how state is managed across the graph. 03:53.220 --> 03:57.740 That's how if multiple nodes update the same state state, it gets combined at the end. 03:58.020 --> 04:00.020 That's what the reducer is handling. 04:00.020 --> 04:06.780 But the reducer doesn't handle, uh, the separate super steps a separate super step is an entirely 04:06.780 --> 04:09.740 fresh invocation of the graph. 04:10.740 --> 04:13.740 So that's going to be a bit confusing. 04:13.740 --> 04:16.620 So what does that mean to show that visually. 04:16.660 --> 04:18.540 Next I want to draw you a diagram. 04:18.540 --> 04:24.620 So with this diagram it's now all going to become crystal clear I'm confident it all begins with defining 04:24.620 --> 04:25.180 the graph. 04:25.180 --> 04:30.460 As I keep saying this is the five things you got to do, including defining the nodes, the edges, 04:30.460 --> 04:31.780 and compiling the graph. 04:31.780 --> 04:32.900 And then you're set. 04:33.460 --> 04:39.100 The next thing you do is perhaps the user has a question, and that question is what you then use to 04:39.140 --> 04:40.780 invoke the graph. 04:40.940 --> 04:41.500 There you go. 04:41.500 --> 04:43.220 And that is called a super step. 04:43.260 --> 04:48.900 And out pops some kind of an answer after the agents and tools have done their thing, and then the 04:48.900 --> 04:51.540 user says something, a follow up question, they have something else. 04:51.540 --> 04:53.500 And that would be another super step. 04:54.060 --> 04:58.460 And then that might happen again with another follow up or with another external activity. 04:58.620 --> 05:03.220 Each of these are super steps, complete invocations of the graph. 05:03.420 --> 05:07.940 So just to make that really obvious, I'm putting a little picture of a graph by each one. 05:07.940 --> 05:11.420 The whole graph is is invoked each time. 05:11.900 --> 05:14.300 And that is what it means to have a super step. 05:14.500 --> 05:16.020 And why am I going on about this? 05:16.020 --> 05:22.140 Because when it comes to memory, when it comes to preserving context between these different calls, 05:22.660 --> 05:28.220 we need to involve something called checkpointing, which is something that landgraaf makes available 05:28.220 --> 05:35.900 to us to be able to keep track, to sort of freeze a record of the state after each super step. 05:35.900 --> 05:42.020 So we've got that tracked, and then next time we call a super step, it can recall the state exactly 05:42.020 --> 05:43.260 as it was Checkpointed. 05:43.580 --> 05:47.500 And that is one of the things that we're going to be doing in the lab right now.