WEBVTT 00:00.200 --> 00:06.720 And of course, the key point for us to have focused on here is that this second agent that did the 00:06.760 --> 00:13.880 summary of the research report on Tesla was taking advantage of the output from the first agent, because 00:13.880 --> 00:15.960 that was included in its context. 00:16.200 --> 00:19.320 And that's how it was able to give what it gave. 00:19.640 --> 00:24.760 And you'll note, if we look into this, that it's as of October 2023. 00:25.080 --> 00:31.160 And that's a bit disappointing because that's clearly not right now a key financial metrics Q3 2023. 00:31.600 --> 00:34.440 And yeah, we're not we're not super happy about that. 00:34.440 --> 00:40.560 And that is, of course, because we're relying on the context on the on the knowledge cut off from 00:40.560 --> 00:46.720 deep seq that did our research and deep seq didn't have more, was last trained back in 2023 and doesn't 00:46.720 --> 00:51.600 have more recent information, which is unfortunate, but something that we can now fix. 00:51.720 --> 00:56.800 So the way we're going to fix it is by adding in a tool, which is the big plan for for this week. 00:56.800 --> 00:58.600 And it's going to be quite easy. 00:58.840 --> 01:00.680 So we do that by going back. 01:00.720 --> 01:08.640 Let's close this and go back to the other crew module which is where we define our crew. 01:08.880 --> 01:11.920 And we are going to add another import here. 01:12.000 --> 01:14.720 We're going to import from Cru AI tools. 01:14.960 --> 01:21.680 And yeah it's there's not going to be much of a job for us in, in the not not too distant future. 01:21.960 --> 01:23.240 So yes indeed. 01:23.280 --> 01:29.280 Clever old cursor knows that we want the Serpa dev tool here, which is indeed what we want, which 01:29.280 --> 01:36.920 is a tool from crew AI that's able to do Google lookups using our Serpa dev account. 01:36.960 --> 01:42.400 And so you need to put your Serpa API key into the EMV file. 01:42.400 --> 01:51.120 But now the now is the challenging, the difficult task of giving the researcher the ability to use 01:51.240 --> 01:52.280 that tool. 01:52.480 --> 01:53.720 That's what we want to do. 01:53.760 --> 01:59.920 And the way we're going to do that is we're going to say that we only want our researcher tool to have 01:59.920 --> 02:00.080 it. 02:00.080 --> 02:01.720 So it's not difficult at all. 02:01.720 --> 02:03.320 It is really rather easy. 02:03.360 --> 02:11.280 We simply create an instance of the Serpa dev tool, and we pass that in in this tools list right here. 02:11.440 --> 02:12.600 That's all it will take. 02:12.720 --> 02:16.200 I save that and we should be good to go. 02:16.560 --> 02:21.600 Maybe we'll go back to models, though, and pick a different model. 02:21.800 --> 02:22.720 Let me see. 02:23.320 --> 02:25.480 We're going to go back to agents. 02:25.800 --> 02:32.880 And why don't we just use OpenAI, GPT four or mini so that we have a faster time of it? 02:33.560 --> 02:35.240 Okay, let's give this a shot. 02:35.280 --> 02:37.760 Let's bring up our terminal. 02:37.960 --> 02:39.640 Let's exit that one actually. 02:39.680 --> 02:40.920 Let's just start again. 02:40.960 --> 02:42.040 I could have typed clear. 02:42.320 --> 02:44.960 We go into the third directory into crew. 02:45.000 --> 02:49.480 We go into our financial researcher directory and now we type crew. 02:49.520 --> 02:55.160 I run and we're hoping now that it's going to go and look up using that tool. 02:55.320 --> 02:59.440 It's going to look up Google about Tesla. 02:59.600 --> 03:01.400 And we're going to get some information. 03:02.000 --> 03:04.160 Financial researcher in progress. 03:04.200 --> 03:05.720 There's something happening here. 03:06.080 --> 03:07.000 There's lots happening. 03:07.080 --> 03:08.760 Search the internet with Serpa. 03:09.160 --> 03:10.920 You can see it's doing its stuff. 03:11.040 --> 03:12.760 Plenty of search is happening. 03:13.160 --> 03:14.640 Tesla latest news today. 03:15.080 --> 03:17.520 I see 2025 appearing in the search results. 03:17.560 --> 03:18.560 This is promising. 03:18.800 --> 03:20.920 This is good news, but we'll soon see. 03:21.480 --> 03:23.800 So the researcher is still working. 03:23.800 --> 03:25.720 It's now on the report writer. 03:25.720 --> 03:27.050 We're now talking to grok. 03:27.290 --> 03:28.810 And here is the summary. 03:28.850 --> 03:29.850 Here is the report. 03:29.850 --> 03:34.130 And it's great to see we do indeed have a report as of early 2025. 03:34.250 --> 03:42.010 And we've got the most recent news from Tesla hasn't done as brilliantly as usual in the last month 03:42.010 --> 03:44.490 or so in the last few weeks, including. 03:44.490 --> 03:47.930 It's definitely got relevant recent information in this report. 03:47.930 --> 03:50.970 And it is, of course, a clear and accurate report. 03:51.090 --> 03:54.050 And, you know, it's fantastic to see this stuff. 03:54.250 --> 03:57.810 And I guess the the usual things for you to appreciate. 03:57.850 --> 04:01.130 One is that this is actually a high quality output. 04:01.130 --> 04:06.890 This is something that would genuinely, legitimately be useful if you spent about ten minutes, 15 04:06.890 --> 04:12.410 minutes yourself searching the internet, doing some searches, gathering some information, synthesizing 04:12.410 --> 04:17.010 it into an email like this, you probably would come up with something quite similar. 04:17.010 --> 04:21.130 It's it's it's proper work that's been done with a good output. 04:21.330 --> 04:27.290 And the other part of this is how easy it was to build this whole infrastructure. 04:27.330 --> 04:33.610 Thanks to crew, thanks to just a couple of of commands, building this and then writing in plain English 04:33.610 --> 04:39.850 in these YAML files, Our objectives for our couple of agents and tasks, and then just giving it the 04:39.850 --> 04:42.650 tool, it was able to handle all of this. 04:42.650 --> 04:51.570 And unlike OpenAI's costs of 2.5 cents per lookup, these costs have been coming out of our free credits. 04:51.570 --> 04:53.930 And so this has cost us nothing. 04:54.170 --> 04:54.730 Okay. 04:54.970 --> 05:00.090 Well, I hope you enjoyed that and I hope you've done it yourself and had a similar outcome. 05:00.250 --> 05:03.050 And I will now see you for the wrap up. 05:03.530 --> 05:04.810 Well, I hope you enjoyed that. 05:04.810 --> 05:07.930 And you've tried it yourself and experimented with some different models. 05:07.930 --> 05:14.210 If you're using a llama, experiment with some different free open source models, try messing around 05:14.210 --> 05:19.850 with the instructions and the backstories and see what it what it takes, what sorts of differences 05:19.850 --> 05:26.330 you can make, and try other kinds of quick agent projects that involve using searches. 05:26.490 --> 05:28.890 And you're getting more and more familiar with Crewe. 05:28.890 --> 05:33.410 And I imagine, like me, that you quite enjoy using Crewe AI as well. 05:33.410 --> 05:35.370 It has a lot to love. 05:35.570 --> 05:41.290 All right, next time we're going to take this project a step further, and we're going to experiment 05:41.290 --> 05:43.650 with some of the more advanced features of Crewe. 05:43.690 --> 05:44.490 See you then.