WEBVTT

00:00.320 --> 00:06.560
And so to recap, we just experienced a crew I project, which is in fact a UV project under the hood.

00:06.920 --> 00:11.160
We created it by saying crew I create crew debate.

00:11.440 --> 00:14.960
It created this directory structure which now should land with you.

00:15.000 --> 00:20.160
You should be more familiar debate census than debate again then config.

00:20.360 --> 00:24.920
We set up our agents YAML to define each of the agents, including the model.

00:25.360 --> 00:30.160
We set up the tasks that included the expected output and the output file.

00:30.520 --> 00:35.800
Crew is where we had the various decorators, and we brought it all together with our functions for

00:35.800 --> 00:37.640
agents, tasks and the crew.

00:37.680 --> 00:39.800
And we said that the process was sequential.

00:40.080 --> 00:46.800
And then we typed query run within our directory and it kicked the whole thing off, generated the output.

00:46.920 --> 00:48.720
And it was successful.

00:48.720 --> 00:52.320
And apparently there should be stricter regulation around llms.

00:52.480 --> 00:58.480
I encourage you to try debating some more controversial points of your liking and choose different models.

00:58.760 --> 01:01.600
So yes, we'll of course be getting much deeper into crew.

01:01.600 --> 01:06.280
But in the meantime, the assignment for you is to now play around with this.

01:06.320 --> 01:13.560
You could have a separate agent for the the agent that is proposing and opposing the motion.

01:13.600 --> 01:18.720
Break that into two different agents and have the tasks going to them separately.

01:18.800 --> 01:22.000
And the reason to do that is that then it's easy to have a different model.

01:22.000 --> 01:28.360
So that you could have OpenAI debate with deep seek or something like that, and then switch who's opposing

01:28.360 --> 01:32.080
and who's proposing to see whether it changes the outcome.

01:32.080 --> 01:38.520
And that's an amusing and entertaining way to battle Llms together, to see which are better at forming

01:38.520 --> 01:45.000
coherent arguments, persuasive arguments that convince a different model that is being the judge.

01:45.160 --> 01:49.840
So that allows you to come up with your own little leaderboard based on debating skills.

01:50.160 --> 01:53.160
So please go away and enjoy yourself with that.

01:53.160 --> 01:58.920
Get a good handle, get comfortable with the framework around crew and the sort of minimal scaffolding

01:58.960 --> 01:59.400
there.

01:59.680 --> 02:05.840
And next time we'll get a little bit deeper into crew and start building some more crew projects.

02:05.840 --> 02:06.800
I will see you then.