WEBVTT 00:00.080 --> 00:07.400 And just to show you what is harder to do if we open this up and look for sheets and open up the Google 00:07.400 --> 00:11.240 Sheets tool and go to Select Credentials and add a new credential. 00:11.240 --> 00:17.440 You may remember that when we were on NHN cloud, it was relatively easy to connect to a sheet. 00:17.440 --> 00:21.800 We pressed a button here, it launched a pop up and we logged into Google and we were done. 00:22.040 --> 00:28.840 And that works because any N cloud has already had like a handshake with Google in the past. 00:28.880 --> 00:31.160 Google knows who cloud is. 00:31.200 --> 00:36.280 There's been some sort of OAuth two integration that's happened at that level, and as a result, it's 00:36.280 --> 00:38.520 relatively quick to add yourself in. 00:38.880 --> 00:43.680 But when we're running on our local box, Google Cloud has no idea who this is. 00:44.160 --> 00:48.680 And so there's more hoops to jump through to be able to set this credentials in place. 00:48.760 --> 00:51.560 It's a bit of a rigmarole, but it's perfectly doable. 00:51.560 --> 00:55.440 We will do some of this tomorrow, but when you're running locally, you have to do it a lot. 00:55.480 --> 01:01.100 Okay, let's come out of that close and back we go and we'll delete this new box. 01:01.420 --> 01:05.380 The final thing I want to show you, I'm going to show you very quickly because it's super optional, 01:05.380 --> 01:09.140 as I think I'd mentioned, and it is about running a models locally. 01:09.140 --> 01:13.820 I do want to give you a demo of this, and for people that know about it, you can you can do this. 01:13.860 --> 01:18.300 And for those that are that are new, you can just just get get some intuition for it. 01:18.300 --> 01:21.700 But I do want to add in Olama. 01:21.940 --> 01:30.660 So Olama Olama what is it is a product that lets you run models locally on your computer in a fast way, 01:30.860 --> 01:33.900 and you can install it by pressing the download button here. 01:33.940 --> 01:38.940 It will bring up a little chat box, and you can use that to chat with different models locally, and 01:38.940 --> 01:40.980 it runs on your local computer. 01:40.980 --> 01:43.340 Running those models, you're probably already familiar. 01:43.340 --> 01:45.060 If you're listening to this part, you probably already have it. 01:45.100 --> 01:50.020 If not, give it a shot, download it and run it and get get a sense for that. 01:50.020 --> 01:53.380 We're about to hook up to that running locally. 01:53.380 --> 01:58.660 Okay, once you've installed Olama just by pressing the download button and following the quick instructions 01:58.660 --> 02:01.290 there, it should come up and it will look like this. 02:01.330 --> 02:02.730 It's like a chat interface. 02:02.730 --> 02:04.570 It might be white background or black background. 02:04.570 --> 02:05.650 This is a llama. 02:05.890 --> 02:10.690 And over here in this dropdown you pick one of the open source models that you could talk to. 02:10.930 --> 02:15.690 And you can also find out more details about them by going to the models page on their website. 02:15.730 --> 02:20.930 The ones with a cloud next to them are ones and which which say colon cloud in the name are not the 02:20.930 --> 02:21.610 ones you want to pick. 02:21.610 --> 02:22.930 They're the ones that would be remote. 02:22.930 --> 02:23.970 We want the ones local. 02:24.010 --> 02:27.290 On your computer, pick one of the ones that would be local on your computer. 02:27.330 --> 02:33.090 This button here is how you download it, so be sure to download a few models and pick the smaller models 02:33.090 --> 02:34.690 if you've got a smaller computer. 02:34.690 --> 02:40.010 And again, look at the models page over here for more information about which models to pick. 02:40.210 --> 02:46.890 Once you've picked one that's like, I could pick a GPT OS right here, you can then send a message 02:46.890 --> 02:48.050 to it like, hi there! 02:48.290 --> 02:55.450 And it will then think GPT OS is a big model that needs you to have at least 16GB of Ram on your GPU, 02:55.490 --> 02:55.730 maybe. 02:55.770 --> 02:57.810 Maybe more, or in your CPU. 02:57.810 --> 03:01.950 It will just be even slower, but eventually this will reply. 03:02.190 --> 03:06.830 Uh, you see, my screen just froze there because my computer is really, really working very hard on 03:06.830 --> 03:07.230 this. 03:07.430 --> 03:08.110 And there we go. 03:08.150 --> 03:10.630 It finally it says, hello, how can I help you today? 03:10.910 --> 03:16.830 And I think I jittered there because of course my computer is being slammed, as would yours. 03:16.870 --> 03:19.670 But you could also pick a much smaller model. 03:19.710 --> 03:21.270 I've got a really tiny one here. 03:21.310 --> 03:22.830 Gemma from Google. 03:22.990 --> 03:24.430 And if I say, uh. 03:24.670 --> 03:25.110 Hi. 03:25.390 --> 03:26.470 I'm great. 03:26.990 --> 03:28.750 How are you now? 03:28.750 --> 03:30.310 Going to Gemma three. 03:30.470 --> 03:36.470 It's going to be a bit of a random but quick answer that we'll get from a tiny model that will hopefully 03:36.470 --> 03:39.910 be somewhat coherent at the end of it, but don't expect too much. 03:39.910 --> 03:40.590 But it's fine. 03:40.590 --> 03:41.190 I'm doing well. 03:41.190 --> 03:42.110 Thanks for asking. 03:42.110 --> 03:45.070 All right, so download a few models, play with them. 03:45.070 --> 03:48.390 And then coming up next we're going to connect them to Nan. 03:48.430 --> 03:53.990 So the first thing I'm going to do is come back to the terminal which is running our n810 inside our 03:53.990 --> 03:54.870 container here. 03:54.870 --> 03:58.430 And I'm going to stop it and I'm going to clear my screen. 03:58.860 --> 04:05.380 and now I'm going to run the same command as before to start it up again, but with one tiny difference. 04:05.380 --> 04:08.220 If I show you this command, I've got it with the slashes again. 04:08.220 --> 04:11.100 But but you can, uh, take them out. 04:11.220 --> 04:14.100 This is the extra line I've put in there. 04:14.300 --> 04:20.820 Now, what this is doing for people new to Docker is it's saying I also want to to make it so that there 04:20.820 --> 04:28.100 is a new a web, a web address available inside my container called Host Internal. 04:28.220 --> 04:31.860 And that gives me access to my host computer. 04:32.060 --> 04:37.820 And that's going to mean that we're going to be able to run things within that can connect to the host, 04:37.820 --> 04:39.860 which is where Obama is running. 04:40.140 --> 04:44.380 So when I when I run this again, it's going to launch my container. 04:44.380 --> 04:45.100 It's much quicker. 04:45.100 --> 04:47.460 It's going to launch my my container based on the image. 04:47.460 --> 04:48.700 It's already downloaded. 04:48.700 --> 04:50.420 So it ran just like that. 04:50.420 --> 04:56.220 And we can now go back in to to localhost five, six, seven eight. 04:56.220 --> 05:02.320 And we will find that this web address is a web address that it will know about inside there. 05:02.320 --> 05:03.160 Let's go see. 05:03.200 --> 05:06.160 Okay, so I'm now back looking at my workflow. 05:06.160 --> 05:10.480 And I'm going to refresh this screen, leave the site, refresh the screen because I want to bring it 05:10.480 --> 05:12.200 back up again because we've stopped and started. 05:12.200 --> 05:14.120 And in the meantime here we are. 05:14.240 --> 05:14.760 Okay. 05:14.840 --> 05:18.280 First up I'm going to delete this open router chat model. 05:18.280 --> 05:20.080 Let's just just just bin that. 05:20.120 --> 05:22.360 And I'm going to add in a new chat model. 05:22.360 --> 05:28.440 The chat model that I'm picking it's going to be Olama Olama chat model okay. 05:29.240 --> 05:32.520 And now we need to select a credential to connect with. 05:32.800 --> 05:33.880 And that's kind of weird. 05:33.880 --> 05:35.160 We don't have a credential. 05:35.160 --> 05:35.560 Exactly. 05:35.560 --> 05:38.280 Because we're going to be just connecting to the local computer. 05:38.280 --> 05:40.880 It's not really a credential, it's the URL. 05:41.200 --> 05:45.920 Now people that use Olama will recognize this because localhost 11434. 05:45.960 --> 05:50.360 That is indeed where Olama runs and listens on your local computer. 05:50.360 --> 05:55.720 If you've if you've launched Olama and you're chatting with it, it's listening on that port 11434 it's 05:55.720 --> 05:56.720 ready for us. 05:57.100 --> 05:58.740 So what can we do here? 05:58.780 --> 06:06.380 Well, I'm going to paste in a different URL which is host internal and the same port 11434. 06:06.660 --> 06:11.100 And that of course that is where we have told the container. 06:11.100 --> 06:13.860 We've mapped that to our host computer. 06:14.060 --> 06:18.580 Now again, if this is this is this is a too much, too much technical detail for you. 06:18.620 --> 06:19.260 Don't worry about it. 06:19.260 --> 06:20.100 Just get the gist. 06:20.100 --> 06:21.780 I just want to show you this working. 06:21.940 --> 06:27.020 And now when I press save we will hopefully see green the connection tested successfully. 06:27.020 --> 06:30.660 This is hooked up to a llama running on my local machine. 06:30.940 --> 06:33.300 So now I can I can exit out of here. 06:33.500 --> 06:39.500 We now have this set up and when I go here, I'll see all of the models that I have downloaded locally 06:39.500 --> 06:43.940 in llama and from within the UI you can you can download more models. 06:44.260 --> 06:48.580 So we could pick GPT OSS, the DB version. 06:48.580 --> 06:50.060 I want to shake things up. 06:50.060 --> 06:53.500 I'm going to pick uh minstrel three. 06:53.620 --> 06:58.610 This is the latest model for me from the French AI startup mistral. 06:58.650 --> 06:59.770 That's very popular. 07:00.210 --> 07:02.810 So minstrel three to have a bit of a different take. 07:03.010 --> 07:09.050 Let's use that open source model running locally on my computer right here. 07:09.410 --> 07:13.010 Uh, and, uh, so X escape out of here. 07:13.130 --> 07:13.890 Here we are. 07:13.930 --> 07:14.490 Back. 07:14.770 --> 07:15.370 Ready? 07:15.410 --> 07:16.450 Let's chat. 07:16.490 --> 07:22.170 We're going to be chatting with Nan running locally to a model running locally on O llama. 07:22.210 --> 07:23.210 Let's give it a try. 07:23.250 --> 07:23.890 Okay. 07:23.930 --> 07:24.970 Open chat. 07:25.010 --> 07:25.810 Hi there. 07:28.330 --> 07:32.650 It's not going to be fast because, uh, it's going to be chugging away on my computer. 07:32.650 --> 07:35.890 I can see my GPU is, uh, is, uh, hammering up. 07:36.530 --> 07:37.050 Hello. 07:37.090 --> 07:38.370 How can I assist you today? 07:38.530 --> 07:39.210 Uh, there we go. 07:39.250 --> 07:40.090 It worked. 07:40.130 --> 07:40.770 Okay. 07:40.810 --> 07:42.210 Big moment for us. 07:42.410 --> 07:47.210 What is the share price of Apple? 07:48.530 --> 07:50.090 All right, off it goes. 07:50.290 --> 07:52.050 I can see the tool just got called. 07:52.490 --> 07:55.010 I can see my GPUs being used again. 07:55.050 --> 07:55.450 There we go. 07:55.490 --> 07:56.510 It already responded. 07:56.510 --> 07:59.830 Current closing share price is $270. 08:00.270 --> 08:01.470 There we have it. 08:01.470 --> 08:02.710 That's a success. 08:02.990 --> 08:04.950 Don't worry if you if you didn't do this part of it. 08:04.950 --> 08:05.630 That doesn't matter. 08:05.630 --> 08:09.830 If you're into open source models running on your computer, then for sure you should do this. 08:09.830 --> 08:16.390 Run a llama, do this, give it a try, and, uh, hopefully you will see that working nicely. 08:16.710 --> 08:22.750 As I say, this is this is a bit more of a technical point, but I think it's really cool to see everything 08:22.750 --> 08:26.950 running locally on a machine and and running locally and the model running locally too. 08:26.990 --> 08:29.310 And with that we can save this. 08:29.350 --> 08:31.150 We can go back to the overview. 08:31.190 --> 08:33.630 You'll notice a very familiar screen here. 08:33.630 --> 08:35.070 And indeed you've got workflows. 08:35.070 --> 08:40.510 The credentials of course shows us the different credentials that we've set up so far and executions 08:40.510 --> 08:41.590 the same as before. 08:41.710 --> 08:46.310 This is where we can see all the different executions that have happened for all of the workflows in 08:46.350 --> 08:47.470 this account. 08:47.590 --> 08:52.270 And that wraps up the tour of Nh10 running self-hosted. 08:52.430 --> 08:53.790 I will see you for the wrap up. 08:53.830 --> 08:54.550 All right. 08:54.550 --> 08:59.180 Well, I hope that the technical people feel satisfied with the bit of technical detail that we got 08:59.180 --> 08:59.860 to there. 08:59.900 --> 09:01.340 The commercial people know that. 09:01.340 --> 09:03.540 They've seen it got some perspective of what it's like. 09:03.700 --> 09:07.420 And for the rest of it, we're going to be focusing on the commercial side. 09:07.460 --> 09:12.540 But tech people, you can feel free to keep going in the self-hosted version and keep working with open 09:12.540 --> 09:16.220 source models potentially running locally on your machine. 09:16.620 --> 09:20.300 But that then, is a wrap today. 09:20.340 --> 09:24.180 One of week three self-hosted and local. 09:25.420 --> 09:26.300 We did all that. 09:26.340 --> 09:29.060 We ran it in locally and it looked pretty much the same. 09:29.180 --> 09:34.620 And we even briefly at the end there connected to Olama, which you can do two if you wish. 09:34.740 --> 09:41.780 From tomorrow we go back to the cloud version to polish off the week with advanced integrations with 09:41.860 --> 09:48.660 MCP, and then with lots of stuff leading up to the capstone project all about amplifying your business. 09:49.020 --> 09:54.300 With that, you are 73% of the way through to being a pro. 09:54.660 --> 09:55.500 See you tomorrow.