WEBVTT 00:00.280 --> 00:01.680 Well, I hope you're excited for this. 00:01.720 --> 00:03.200 We're going back to lab four. 00:03.480 --> 00:05.120 We are coming here. 00:05.480 --> 00:07.800 So we've looked through traders Pi. 00:07.840 --> 00:13.360 I remind you again that there is that constant at the top that you need to change if you're using the 00:13.400 --> 00:18.240 paid market provider, or you may be deciding to use a different market provider. 00:18.240 --> 00:21.520 If you've looked for some good MCP tools, it might be worth doing that. 00:21.520 --> 00:24.200 And so now we import trader from traders. 00:24.200 --> 00:24.600 This is it. 00:24.600 --> 00:29.920 This is a running that it's happening now we're going to create an instance of trader. 00:29.920 --> 00:31.960 And we're going to call it Alice. 00:33.240 --> 00:35.280 That seems to work okay. 00:35.760 --> 00:40.480 Now we are going to call init agent, which of course we have to await because it's async. 00:40.880 --> 00:44.560 So this is going to do things like creating all of our MCP servers. 00:46.080 --> 00:47.160 So this takes a little bit of time. 00:47.160 --> 00:51.920 All our MCP servers are creating and getting called done. 00:52.120 --> 00:54.440 So now we have an agent name is Alice. 00:54.640 --> 00:56.720 There are Alice's instructions. 00:57.240 --> 01:00.120 And uh, you can, uh, look through all that. 01:00.320 --> 01:04.320 There's all of the tools that are in there, and it's time for us to kick this off. 01:04.920 --> 01:06.800 So off it goes. 01:07.640 --> 01:14.160 So Alice is now running and there could be a lot of calls involved as before, but this time we are 01:14.160 --> 01:20.920 going through the module now because we have that flag that goes true false true false. 01:20.920 --> 01:24.800 It's going to start by doing trading rather than rebalancing. 01:25.200 --> 01:27.640 But we will let Alice do her thing. 01:27.920 --> 01:34.520 I'm just going to check that we should we should add in here some markdown to look at the of course 01:34.520 --> 01:35.080 to look. 01:36.440 --> 01:37.880 And Alice is finished already. 01:42.640 --> 01:44.080 But we should go and look at the trace. 01:44.080 --> 01:47.400 But first let's look at the results so we can see I'm too eager. 01:47.680 --> 01:48.520 So let's see. 01:48.560 --> 01:49.560 Trades executed. 01:49.600 --> 01:51.240 Bought five shares of Tesla. 01:51.280 --> 01:53.080 Bought ten shares of Nvidia. 01:53.400 --> 01:57.360 Rationale high trading volume indicates potential for short term gains. 01:57.720 --> 02:00.950 Remaining balance is 2.99 and portfolio appraisal. 02:00.990 --> 02:06.630 My portfolio now includes shares which are poised for potential rebounds amidst a volatile market. 02:06.670 --> 02:08.390 This all sounds perfectly sensible. 02:09.150 --> 02:10.830 I'm excited to see the trace. 02:10.830 --> 02:15.750 So as before, we go to, uh, platform OpenAI. 02:17.950 --> 02:22.870 And even though this output looks very legit, we shouldn't get excited until we've done the work to 02:22.910 --> 02:28.470 go and look at the trace and understand what's happening and confirm that this is performing as we would 02:28.470 --> 02:28.830 like. 02:28.990 --> 02:35.550 So there was 11 different tools in Alice Trading, which was what we were hoping to see and what we 02:35.550 --> 02:36.550 can see here. 02:36.790 --> 02:40.270 Okay, so Alice started with, uh, listing a bunch of tools. 02:40.270 --> 02:46.030 It went to the researcher who just did one web search this time, latest stock market news. 02:46.190 --> 02:48.510 And that came back a lot of currents. 02:48.670 --> 02:48.830 Wow. 02:48.870 --> 02:59.230 Lots of lots of, uh, checking of the market prices for Tesla, for Nvidia, for Amazon, for E.M. 02:59.270 --> 03:02.310 and it says unable to fetch current price or no price found. 03:03.030 --> 03:11.550 And PG is that Procter and Gamble and uh okay and then buy shares. 03:11.550 --> 03:12.830 Buy shares for Tesla. 03:12.870 --> 03:14.110 Buy shares for Nvidia. 03:14.470 --> 03:15.950 Buy shares for Amazon. 03:15.950 --> 03:17.750 And that was the response. 03:17.750 --> 03:20.510 So it definitely did did its work. 03:20.510 --> 03:22.030 It got the current stock market price. 03:22.030 --> 03:23.110 It bought the shares. 03:23.310 --> 03:25.870 Uh that is all great to see okay. 03:25.910 --> 03:31.270 And I've now gone through and run run a bunch more times for trading and rebalancing. 03:31.270 --> 03:38.790 It switches each time so we can see a bit of activity going on occasionally I've had MacD tools error 03:38.790 --> 03:40.350 ring as it has done here. 03:40.350 --> 03:42.630 And so that might be a cause for concern. 03:42.670 --> 03:46.510 And we can go and have a look at why of course by looking at the trace. 03:46.510 --> 03:54.630 But let's first see the final portfolio after rebalancing a Netflix is ten shares Tesla at ten shares. 03:55.030 --> 04:01.780 Uh five were bought and that leaves a cash balance and net profit and loss is down $60 on the trading 04:01.780 --> 04:03.220 activity of now. 04:03.220 --> 04:04.380 So not not the best of times. 04:04.380 --> 04:06.860 But I think the markets are are turbulent. 04:07.180 --> 04:12.300 Um, and so, uh, there's some explanation here of why it did what it did. 04:12.500 --> 04:14.740 But now let's go look at the trace. 04:15.460 --> 04:20.180 So here is the trace and we'll see why, uh, tools were erroring. 04:20.180 --> 04:20.500 You'll see. 04:20.500 --> 04:23.380 I've been doing trading, rebalancing, trading, rebalancing. 04:23.660 --> 04:32.260 Uh, so, um, Alice did, uh, some get company news and that had an error with the get company news 04:32.460 --> 04:38.580 route that is currently that tool, which is one of the tools, the free tools offered by the paid, 04:38.820 --> 04:43.180 uh, financial services, uh, MCP tool that we're using is not working. 04:43.220 --> 04:48.780 And so Alice instead just picked current stock price, which is working and which gave back a good stock 04:48.780 --> 04:49.300 price. 04:49.300 --> 04:56.100 So Alice wasn't perturbed by some, uh, ill behavior from one one API and just switched to other, 04:56.460 --> 05:01.020 uh, areas and then used the researcher tool, which would be a better way to get company news. 05:01.020 --> 05:03.340 And the researcher tool went off and did some research. 05:03.620 --> 05:06.940 Um, and the research tool got some information about Netflix. 05:07.220 --> 05:10.900 Um, but on getting some asking for some information on Apple. 05:11.100 --> 05:15.940 It got an error that the rate limit was exceeded because I've been running this too much, which is 05:15.940 --> 05:18.140 another little problem to be aware of. 05:18.380 --> 05:23.260 Uh, and uh, unperturbed, Alice kept trying, or the researcher kept trying. 05:23.500 --> 05:26.300 Uh, and, uh, in the end, I got another response. 05:26.300 --> 05:29.820 After a while, after doing a big query on some things. 05:30.060 --> 05:35.220 Um, but this may indicate that I need to put in some sort of a pause to make sure that these things 05:35.220 --> 05:39.700 don't, uh, mean that we get rate limits from from the brave web search. 05:39.940 --> 05:44.020 Um, but anyway, after doing all of that, after managing the fact that there were errors, I love 05:44.020 --> 05:49.740 that the fact that the autonomous agent was able to to handle it was retrying itself. 05:49.780 --> 05:52.460 I didn't give any instruction to retry. 05:52.580 --> 05:57.180 And also, when the get company news failed, it flipped over to the researcher instead. 05:57.180 --> 06:03.580 So it's really interesting to see that, to see that the agent is able to handle unexpected outcomes. 06:04.020 --> 06:08.020 And as a result, there has been some trading, there has been some rebalancing. 06:08.020 --> 06:11.300 And this is the the final outcome at this point. 06:11.300 --> 06:18.900 And so that completes the first part of our two parter on our capstone project, on building an autonomous 06:18.900 --> 06:22.220 agentic platform for making trading decisions. 06:22.540 --> 06:24.340 And it was a lot of fun. 06:24.500 --> 06:30.140 The thing to really bear in mind, the main learning point is that these kinds of agent platforms can 06:30.140 --> 06:38.420 be very powerful because by the nature of their autonomy, they are able to do things like handle API 06:38.420 --> 06:43.780 calls that aren't working and just trying out different approaches for getting their information, which 06:43.780 --> 06:49.900 which is extremely powerful and something that we wouldn't have thought possible only a few years ago. 06:49.940 --> 06:53.740 And this is why people are so excited about these kinds of platforms. 06:53.980 --> 06:57.410 Um, and then we're also seeing the power of MCP. 06:57.450 --> 07:03.290 The fact that there are all these different tools for the agents to be using, and that they are experimenting 07:03.290 --> 07:09.290 with using different tools, even saw in the last one how it tried using the get market information 07:09.290 --> 07:14.770 from from one MCP server, and when that failed, it flipped to using its research tool, which ends 07:14.770 --> 07:17.010 up using brave searches. 07:17.170 --> 07:23.770 So it's amazing to see how bringing together these different tools from MCP servers, from different 07:23.770 --> 07:30.970 authors across the world has allowed us to give extra capabilities, integrate capabilities into our 07:30.970 --> 07:33.050 agents really easily. 07:33.250 --> 07:39.130 And so that completes day four of our final week. 07:39.330 --> 07:40.850 I do hope you're enjoying it. 07:40.850 --> 07:48.450 I hope you're taking as much pleasure in this financial services or this autonomous alpha seeking agents 07:48.490 --> 07:49.290 as I am. 07:49.570 --> 07:54.290 And prepare yourself for the grand finale in the last day. 07:54.330 --> 07:55.490 I will see you then.