WEBVTT 00:00:00.000 --> 00:00:03.120 Ever feel like your digital life is just overwhelming, 00:00:03.379 --> 00:00:06.000 like you're constantly wading through this flood 00:00:06.000 --> 00:00:08.599 of tech news, maybe a dozen tabs open, hitting 00:00:08.599 --> 00:00:10.839 refresh, and you still feel like you're missing 00:00:10.839 --> 00:00:12.980 the important stuff? Yeah, that feeling is real. 00:00:13.320 --> 00:00:17.039 It's this like background anxiety of trying to 00:00:17.039 --> 00:00:20.699 stay current, manage your own projects, and maybe 00:00:20.699 --> 00:00:23.260 just have a minute to think. What if you could 00:00:23.260 --> 00:00:25.899 actually... change that? What if there was this 00:00:25.899 --> 00:00:29.579 open source tool that lets you build your own 00:00:29.579 --> 00:00:32.039 kind of digital assistant to cut through all 00:00:32.039 --> 00:00:35.189 that noise? Well, welcome back to the deep dive. 00:00:35.270 --> 00:00:37.409 That's exactly what we're digging into today. 00:00:37.630 --> 00:00:40.210 How you can actually build your very own personal 00:00:40.210 --> 00:00:43.170 intelligence agent. We've got this really comprehensive 00:00:43.170 --> 00:00:45.789 guide here covering how to build a pretty sophisticated 00:00:45.789 --> 00:00:49.210 automation system. And our mission today, it's 00:00:49.210 --> 00:00:50.990 simple, really. We want to show you how to take 00:00:50.990 --> 00:00:53.149 that information overload and turn it into your 00:00:53.149 --> 00:00:56.710 own personalized AI powered command center. We'll 00:00:56.710 --> 00:00:59.149 walk through setting up the tool, building an 00:00:59.149 --> 00:01:01.829 AI news aggregator, exploring direct command 00:01:01.829 --> 00:01:05.159 line stuff. setting up a YouTube monitor and 00:01:05.159 --> 00:01:08.000 even getting into AI agents that can actually 00:01:08.000 --> 00:01:10.480 think and act. You know, that feeling of drowning 00:01:10.480 --> 00:01:13.340 in digital stuff. It's so common for anyone who's 00:01:13.340 --> 00:01:16.159 really into tech. Your home lab, personal projects, 00:01:16.280 --> 00:01:19.079 just keeping up. It can be completely exhausting. 00:01:19.379 --> 00:01:21.719 Oh, absolutely. And that's exactly where NN, 00:01:21.879 --> 00:01:24.519 that's N8N, by the way, comes in. Think of it 00:01:24.519 --> 00:01:27.959 as this like open source automation powerhouse. 00:01:28.200 --> 00:01:30.439 It's ready to genuinely shake up your workflow. 00:01:30.760 --> 00:01:33.189 N8N. Got it. Yeah, it's got a bit of a nerdy 00:01:33.189 --> 00:01:35.010 name, maybe a pun in there, but honestly, the 00:01:35.010 --> 00:01:36.969 power is serious. It sounds like it. I've heard 00:01:36.969 --> 00:01:39.609 comparisons. Like if the usual automation tools 00:01:39.609 --> 00:01:43.510 are, say, a friendly family car, then N8N is 00:01:43.510 --> 00:01:46.310 more like a custom -built street -legal F1 car. 00:01:46.430 --> 00:01:49.209 It's sleeker, way smarter, and just has this 00:01:49.209 --> 00:01:51.670 monstrous engine. That's a great analogy, actually. 00:01:52.189 --> 00:01:56.140 And, crucially, it's free. Right. And it really 00:01:56.140 --> 00:01:58.319 respects your privacy. That feels like a big 00:01:58.319 --> 00:02:00.700 deal. It's not just about automating tasks. It's 00:02:00.700 --> 00:02:03.219 about taking back control, keeping your data 00:02:03.219 --> 00:02:06.040 yours and building something exactly for you. 00:02:06.159 --> 00:02:08.259 That's the core idea. You become the architect. 00:02:08.580 --> 00:02:12.939 Yeah. But fair warning here. N .A .N. It's incredibly 00:02:12.939 --> 00:02:16.060 addictive. Yeah. Seriously. You might start simple, 00:02:16.159 --> 00:02:17.900 maybe just thinking, I'll build a little news 00:02:17.900 --> 00:02:20.120 digest. Right. Next thing you know, you're deep 00:02:20.120 --> 00:02:22.560 down the rabbit hole building AI agents that 00:02:22.560 --> 00:02:24.560 are troubleshooting your servers for you. You've 00:02:24.560 --> 00:02:27.680 been warned. Okay. Noted. So for this deep dive, 00:02:27.819 --> 00:02:30.090 what's the plan? We're building something. Specific? 00:02:30.189 --> 00:02:32.550 Yeah. We're constructing a smart news aggregator 00:02:32.550 --> 00:02:35.030 with AI summaries, an automated YouTube channel 00:02:35.030 --> 00:02:36.949 monitor, getting instant Discord notifications, 00:02:37.349 --> 00:02:40.550 powerful command line integrations, and finally 00:02:40.550 --> 00:02:43.990 building a thinking AI agent. Wow. Okay. That's 00:02:43.990 --> 00:02:45.969 quite a build. So what's the fundamental shift 00:02:45.969 --> 00:02:48.370 this tool offers compared to the more traditional 00:02:48.370 --> 00:02:50.849 automation platforms? It's really about moving 00:02:50.849 --> 00:02:54.530 away from those fixed kind of rigid rules towards 00:02:54.530 --> 00:02:57.490 having that adaptable, powerful F1 car we talked 00:02:57.490 --> 00:02:59.800 about. One that respects... your privacy and 00:02:59.800 --> 00:03:01.960 lets you build precisely what you need. That 00:03:01.960 --> 00:03:04.240 makes sense. Adaptability and control. Okay, 00:03:04.300 --> 00:03:07.039 so once you decide you want this power, the first 00:03:07.039 --> 00:03:10.319 big question is, where do you actually put it? 00:03:10.379 --> 00:03:13.139 Where does it live? Exactly. That's the foundational 00:03:13.139 --> 00:03:15.300 decision. Where does your NANN instance run? 00:03:15.960 --> 00:03:18.340 Cloud or on -premise? It's that classic fork 00:03:18.340 --> 00:03:20.580 in the road. Right. The cloud path sounds like 00:03:20.580 --> 00:03:23.120 the express lane, especially if you're just starting 00:03:23.120 --> 00:03:26.099 out. Using a cloud provider seems like it gets 00:03:26.099 --> 00:03:28.560 you automating faster. It absolutely does. Less 00:03:28.560 --> 00:03:30.560 complexity, you don't manage the underlying server, 00:03:30.659 --> 00:03:33.319 better connectivity usually. It's quick. Like 00:03:33.319 --> 00:03:35.300 how quick? We're talking maybe a five -minute 00:03:35.300 --> 00:03:39.259 launch sequence. Sign up, pick a plan like a 00:03:39.259 --> 00:03:41.819 virtual private server with NNN pre -installed, 00:03:41.819 --> 00:03:44.180 maybe two cores, four gigs of RAM, deploy it, 00:03:44.219 --> 00:03:46.460 create your admin account. Done. Five minutes. 00:03:46.620 --> 00:03:48.960 So then there's the on -prem path. This is for 00:03:48.960 --> 00:03:52.039 the, let's call them the home lab warriors. Ah, 00:03:52.039 --> 00:03:54.780 yes. I know a few of those. Right. Running it 00:03:54.780 --> 00:03:57.080 on your own hardware, you get absolute control, 00:03:57.319 --> 00:04:00.939 maximum data privacy, and honestly, that satisfaction 00:04:00.939 --> 00:04:03.259 of running it yourself. Definitely more control, 00:04:03.340 --> 00:04:05.759 more privacy, maybe more rewarding if you're 00:04:05.759 --> 00:04:08.219 into that stuff. For sure. Full control, data 00:04:08.219 --> 00:04:11.039 stays home, personal reward. So for someone just 00:04:11.039 --> 00:04:14.020 curious, maybe a beginner, what's the biggest 00:04:14.020 --> 00:04:17.050 plus of starting with that cloud path? Speed 00:04:17.050 --> 00:04:19.189 and simplicity, hands down. You're automating 00:04:19.189 --> 00:04:21.569 almost instantly, no fuss. Right, gets you going 00:04:21.569 --> 00:04:23.649 quickly. Okay. Okay, so we know where we could 00:04:23.649 --> 00:04:25.790 build it. Let's actually build something, something 00:04:25.790 --> 00:04:28.589 useful right away, like that AI news aggregation 00:04:28.589 --> 00:04:30.350 machine you mentioned. Let's do it. The core 00:04:30.350 --> 00:04:33.350 idea in NAN, the philosophy, it's pretty intuitive. 00:04:33.490 --> 00:04:36.579 Think of a workflow like a... A digital Rube 00:04:36.579 --> 00:04:39.220 Goldberg machine. Okay. Data goes in, flows through 00:04:39.220 --> 00:04:41.720 these different blocks called nodes, each doing 00:04:41.720 --> 00:04:44.139 one specific job, and then something useful comes 00:04:44.139 --> 00:04:46.360 out. You've got triggers, which start things. 00:04:46.639 --> 00:04:49.199 The on switch. Exactly. Data nodes gather stuff. 00:04:49.300 --> 00:04:52.220 Logic nodes are the brains doing filtering or 00:04:52.220 --> 00:04:54.560 changing things. And action nodes, well, they 00:04:54.560 --> 00:04:56.860 do something. Send a message, save a file, whatever. 00:04:57.120 --> 00:04:59.279 It's like stacking Lego blocks, but with data. 00:04:59.480 --> 00:05:02.660 Perfect analogy. So you start with a blank workflow. 00:05:02.860 --> 00:05:05.779 First thing, add a manual trigger. That's your 00:05:05.779 --> 00:05:08.399 test button. Lets you run it anytime. Crucial 00:05:08.399 --> 00:05:10.879 for building, right? Instant feedback. Absolutely. 00:05:11.019 --> 00:05:13.300 But for real automation, you need it to run on 00:05:13.300 --> 00:05:15.699 its own. So you add a second trigger, a schedule 00:05:15.699 --> 00:05:18.759 trigger, maybe every day at 7 a .m. So you have 00:05:18.759 --> 00:05:21.500 both a test button and a schedule. Yep. It's 00:05:21.500 --> 00:05:24.040 kind of like a pro setup, a dev mode and a prod 00:05:24.040 --> 00:05:26.319 mode trigger. You can turn the schedule off while 00:05:26.319 --> 00:05:29.459 you're tinkering. Smart. Okay, triggers are set. 00:05:29.769 --> 00:05:32.329 Now, fetching the news. Right. You need a data 00:05:32.329 --> 00:05:35.870 gatherer. For RSS feeds, there's an RSS read 00:05:35.870 --> 00:05:39.240 node. You just point it at a feed URL, like a 00:05:39.240 --> 00:05:41.660 bleeping computer's feed, for example. Just paste 00:05:41.660 --> 00:05:44.100 the URL in? Pretty much. Then you run just that 00:05:44.100 --> 00:05:46.699 node and bam, that's the aha moment. What happens? 00:05:46.980 --> 00:05:49.319 The output panel just lights up. Suddenly you 00:05:49.319 --> 00:05:51.920 see all the structured data titles, links, authors, 00:05:52.240 --> 00:05:55.360 dates, the content snippets, all pulled neatly 00:05:55.360 --> 00:05:58.120 from the feed. Ah, okay. So you see the raw data 00:05:58.120 --> 00:06:00.620 grabbed. Exactly. And you realize, okay, this 00:06:00.620 --> 00:06:02.160 isn't just clicking buttons. This is actually 00:06:02.160 --> 00:06:05.360 commanding information. So what's the core concept 00:06:05.360 --> 00:06:07.779 that makes... these workflows so flexible then? 00:06:07.920 --> 00:06:10.959 I'd say it's about chaining those specific function 00:06:10.959 --> 00:06:14.019 nodes like Lego blocks to build complex processes 00:06:14.019 --> 00:06:16.500 step by step. Precisely. OK, so we have the news 00:06:16.500 --> 00:06:20.139 data, but data trapped inside NANN isn't very 00:06:20.139 --> 00:06:22.939 useful, is it? No, you need to get it out. And 00:06:22.939 --> 00:06:26.149 you mentioned Discord. Like a private channel 00:06:26.149 --> 00:06:28.410 as a command center. Yeah, exactly. Pipe that 00:06:28.410 --> 00:06:30.189 curated news right into a place you already check. 00:06:30.370 --> 00:06:33.050 So you add a Discord node after the RSS node, 00:06:33.230 --> 00:06:35.910 set its operation to send message. And for connection, 00:06:36.110 --> 00:06:38.290 you use something called a webhook. Webhook. 00:06:38.389 --> 00:06:40.550 Sounds technical. It's just a special private 00:06:40.550 --> 00:06:43.230 URL. Think of it like a dedicated phone number 00:06:43.230 --> 00:06:45.629 for automated messages to that specific Discord 00:06:45.629 --> 00:06:47.670 channel. You grab it from your Discord channel 00:06:47.670 --> 00:06:50.129 settings under integrations. Gotcha. Copy paste 00:06:50.129 --> 00:06:53.850 the URL. Then what? Then the fun part. crafting 00:06:53.850 --> 00:06:57.550 the message. Instead of just new post, you use 00:06:57.550 --> 00:07:00.170 NAN's expressions, little snippets of code that 00:07:00.170 --> 00:07:02.589 pull data dynamically. Ah, so you can say like 00:07:02.589 --> 00:07:05.170 title .title and it fills in the actual title. 00:07:05.269 --> 00:07:07.569 Exactly. Title, author, link, date, dot, format 00:07:07.569 --> 00:07:11.529 it nicely. But when you first test this. Let 00:07:11.529 --> 00:07:16.769 me guess. Spam. A dozen messages? Yep. Because 00:07:16.769 --> 00:07:19.550 N810 runs that Discord node for every single 00:07:19.550 --> 00:07:22.220 item it got from the RSS feed. Right. You're 00:07:22.220 --> 00:07:24.819 going to tame that beast. So you slot in a limit 00:07:24.819 --> 00:07:27.300 node right between the RSS reader and Discord. 00:07:27.560 --> 00:07:30.519 Tell it only let three articles through or maybe 00:07:30.519 --> 00:07:32.120 five. Okay, that makes sense. Limit the flow. 00:07:32.240 --> 00:07:34.379 Test again. Perfect. Just the top few articles. 00:07:34.740 --> 00:07:37.240 And here's that pro tip again. Pin your data. 00:07:37.860 --> 00:07:40.379 After you run the RSS node and get good data, 00:07:40.540 --> 00:07:42.879 click the little pin icon on its output. Pin 00:07:42.879 --> 00:07:46.639 the data. Why? It saves that specific test data. 00:07:46.779 --> 00:07:48.899 So when you're tweaking the Discord message or 00:07:48.899 --> 00:07:51.120 adding filters later, you don't have to rerun 00:07:51.120 --> 00:07:53.180 the RSS node and fetch the news all over again 00:07:53.180 --> 00:07:55.980 every single time. Saves time, saves API calls 00:07:55.980 --> 00:07:57.839 if the source was something else. It's huge. 00:07:58.139 --> 00:08:01.319 Ah, okay. So why is pinning data such a pro tip 00:08:01.319 --> 00:08:03.860 then for building these workflows? Simply put, 00:08:04.000 --> 00:08:06.699 it prevents refetching data constantly, speeding 00:08:06.699 --> 00:08:10.019 up your building process and saving costs. Sponsor. 00:08:10.410 --> 00:08:12.850 All right. Buckle up. Time to unlock some real 00:08:12.850 --> 00:08:16.350 power. This next step is kind of why NAN feels 00:08:16.350 --> 00:08:18.709 like it's in a different league. We're going 00:08:18.709 --> 00:08:22.709 to add command line execution directly into the 00:08:22.709 --> 00:08:25.930 workflow. Command line? Like Linux terminal commands? 00:08:26.149 --> 00:08:28.529 Exactly. It's like being handed the admin password 00:08:28.529 --> 00:08:30.810 to the server NAN is running on. You're taking 00:08:30.810 --> 00:08:33.710 the safety locks off. Okay. This sounds potentially 00:08:33.710 --> 00:08:37.149 dangerous, but powerful. How do you do it? You 00:08:37.149 --> 00:08:39.149 just add an execute command node, maybe branch 00:08:39.149 --> 00:08:40.990 off your main flow. In the command field, you 00:08:40.990 --> 00:08:43.529 could start simple, type ping 1 .1 1 .1 money 00:08:43.529 --> 00:08:46.750 C3. Ping Cloudflare's DNS from inside the workflow. 00:08:47.029 --> 00:08:49.570 Yeah, run it, and boom, real terminal output 00:08:49.570 --> 00:08:52.129 appears in the results panel. Whoa. Mind blown, 00:08:52.230 --> 00:08:54.049 right? What this actually means is you just ran 00:08:54.049 --> 00:08:55.629 an arbitrary command on the server's operating 00:08:55.629 --> 00:08:57.350 system. Anything you could type in a terminal, 00:08:57.470 --> 00:09:00.970 LS, Doi Clea, Docker OA, custom scripts, anything. 00:09:01.309 --> 00:09:03.450 Whoa. Okay, wait. So you could have an automated 00:09:03.450 --> 00:09:06.259 check on your Docker containers? Docker WA runs 00:09:06.259 --> 00:09:10.200 every hour. Or check disk space with DFTT. Precisely. 00:09:10.320 --> 00:09:14.700 Monitor system health, free DSM, run your own 00:09:14.700 --> 00:09:17.879 Python or bash security scripts on a schedule. 00:09:18.139 --> 00:09:21.220 Maybe even carefully automate server updates 00:09:21.220 --> 00:09:23.960 with apt -get update. The possibilities there 00:09:23.960 --> 00:09:27.259 are staggering. Imagine an assistant checking 00:09:27.259 --> 00:09:30.120 server health automatically. Wow. And to bring 00:09:30.120 --> 00:09:32.059 it all together, you use a merge node. Put it 00:09:32.059 --> 00:09:33.879 right before your Discord node. Merge. Yeah. 00:09:33.940 --> 00:09:35.960 You connect the output from your news filtering 00:09:35.960 --> 00:09:38.419 to one input and the output from your command 00:09:38.419 --> 00:09:40.419 execution like the ping results to the other 00:09:40.419 --> 00:09:43.340 input. Set the mode to append. Append. So it 00:09:43.340 --> 00:09:45.480 just sticks them together. Exactly. Combines 00:09:45.480 --> 00:09:47.899 the two streams of data. So your Discord message 00:09:47.899 --> 00:09:50.480 now has both the news and the server status report. 00:09:51.120 --> 00:09:54.179 One single pane of glass. That level of direct 00:09:54.179 --> 00:09:57.379 system access fused with web data, that's pretty 00:09:57.379 --> 00:09:59.000 unique. Okay. So what's the biggest difference 00:09:59.000 --> 00:10:01.220 this direct command line access really offers 00:10:01.220 --> 00:10:03.659 them? Freedom. Pure freedom. You're not limited 00:10:03.659 --> 00:10:05.919 by pre -built integrations anymore. You can make 00:10:05.919 --> 00:10:08.559 your systems do anything. Freedom. Got it. Okay. 00:10:08.659 --> 00:10:10.419 We have data. We have system access. Now we need 00:10:10.419 --> 00:10:12.889 the intelligence part, right? The brain. Make 00:10:12.889 --> 00:10:16.110 it feel like. J -A -R -V -I -S. Exactly. Time 00:10:16.110 --> 00:10:19.269 to add the J -A -R -V -I -S layer. We do this 00:10:19.269 --> 00:10:22.730 with AI, specifically large language models or 00:10:22.730 --> 00:10:25.269 LLMs. Okay. How do we plug that in? You add a 00:10:25.269 --> 00:10:28.919 node called basic LLM chain. You place it, say, 00:10:29.080 --> 00:10:31.120 after you filter the news articles, but before 00:10:31.120 --> 00:10:33.559 they get merged for Discord? LLM chain. Okay. 00:10:33.659 --> 00:10:35.460 And so you connect it to your AI model of choice. 00:10:36.259 --> 00:10:39.980 OpenAI is GPT -4, Anthropix Cloud, maybe even 00:10:39.980 --> 00:10:42.200 a local Alama model if you're running one. Right. 00:10:42.259 --> 00:10:44.230 And then you tell it what to do. The prompt. 00:10:44.549 --> 00:10:46.870 Precisely. And the prompt is everything. You 00:10:46.870 --> 00:10:48.750 can't just say, summarize this. You need to be 00:10:48.750 --> 00:10:51.710 specific. Like what? You are a world -class intelligence 00:10:51.710 --> 00:10:55.250 analyst briefing a tech CEO. Summarize the following 00:10:55.250 --> 00:10:58.190 article in exactly two sentences. Your summary 00:10:58.190 --> 00:11:00.610 must be engaging and directly state why this 00:11:00.610 --> 00:11:02.429 news is important for a tech professional to 00:11:02.429 --> 00:11:05.029 know. Here's the article content. And then you 00:11:05.029 --> 00:11:07.190 dynamically insert the article content from the 00:11:07.190 --> 00:11:10.409 previous note. Ah, okay. Give it a role, a specific 00:11:10.409 --> 00:11:13.639 task, length constraints, the audience. Very 00:11:13.639 --> 00:11:16.980 detailed. Super detailed. Pro tips, be specific, 00:11:17.340 --> 00:11:20.600 two sentences, give context for a tech professional, 00:11:20.799 --> 00:11:23.639 and definitely test different models. GPT -4 00:11:23.639 --> 00:11:26.600 often gives much better, more nuanced summaries, 00:11:26.600 --> 00:11:29.659 usually worth the cost. Makes sense. And you 00:11:29.659 --> 00:11:32.139 mentioned AI for the ping results too. Yeah. 00:11:32.570 --> 00:11:35.129 You could add another LLM node after the execute 00:11:35.129 --> 00:11:37.070 command node. Give it a different personality 00:11:37.070 --> 00:11:39.710 in the prompt. You are a laid back network admin 00:11:39.710 --> 00:11:42.690 with a sense of humor. Analyze this ping output. 00:11:43.009 --> 00:11:45.289 If it's all good, tell me in a funny, reassuring 00:11:45.289 --> 00:11:47.590 way the internet is fine. If there are errors 00:11:47.590 --> 00:11:50.570 or packet loss, be serious and flag it. Huh. 00:11:51.080 --> 00:11:53.159 That's brilliant. So you get AI checking your 00:11:53.159 --> 00:11:55.659 network status with personality. Exactly. And 00:11:55.659 --> 00:11:57.980 sometimes getting the prompt just right. I still 00:11:57.980 --> 00:12:00.000 wrestle with prompt drift myself, you know. Yeah. 00:12:00.259 --> 00:12:02.840 Trying to get the AI to follow instructions exactly 00:12:02.840 --> 00:12:05.720 every time. But when it works, the power is undeniable. 00:12:05.940 --> 00:12:07.940 Yeah, I can imagine. So the daily briefing goes 00:12:07.940 --> 00:12:10.919 from just raw data to actual insight, AI summaries, 00:12:10.940 --> 00:12:13.539 AI analysis of system checks. Yeah. So what's 00:12:13.539 --> 00:12:15.460 the secret to getting those really useful AI 00:12:15.460 --> 00:12:18.799 summaries? It really boils down to this. Precise, 00:12:18.799 --> 00:12:21.379 contextualized prompts make. make all the difference 00:12:21.379 --> 00:12:25.419 got it precision and the prompt okay that news 00:12:25.419 --> 00:12:27.620 aggregator is already pretty powerful yeah but 00:12:27.620 --> 00:12:30.279 you mentioned monitoring multiple sources like 00:12:30.279 --> 00:12:32.340 YouTube channels that sounds like a step up in 00:12:32.340 --> 00:12:35.159 complexity it is but it's totally doable we've 00:12:35.159 --> 00:12:37.860 handled one stream of data now we build a system 00:12:37.860 --> 00:12:39.899 that monitors multiple sources loops through 00:12:39.899 --> 00:12:42.340 them and applies smart filtering the mission 00:12:42.340 --> 00:12:45.799 build that personal Scout for your favorite YouTube 00:12:45.799 --> 00:12:48.370 tech channels Okay, how do you even get YouTube 00:12:48.370 --> 00:12:50.690 updates without opening YouTube? Is there an 00:12:50.690 --> 00:12:53.590 API? There is, but there's a simpler way. The 00:12:53.590 --> 00:12:57.529 secret YouTube RSS hack. RSS. For YouTube. Yep. 00:12:57.649 --> 00:12:59.850 Every single YouTube channel has a hidden RSS 00:12:59.850 --> 00:13:02.230 feed. It just lists their latest videos. It's 00:13:02.230 --> 00:13:05.309 like a secret backdoor. No API keys needed. Wow, 00:13:05.429 --> 00:13:07.830 okay. That's handy. So how do you build the monitor 00:13:07.830 --> 00:13:09.389 with that? It's kind of a three -step process. 00:13:09.669 --> 00:13:12.129 First, create your watch list. You use a node 00:13:12.129 --> 00:13:14.889 called set field to just create a list, an array 00:13:14.889 --> 00:13:17.429 of the YouTube channel IDs you care about. Okay. 00:13:17.470 --> 00:13:20.309 A list of IDs. Then you need the one -by -one 00:13:20.309 --> 00:13:22.929 machine. There's a node called split out. It 00:13:22.929 --> 00:13:25.350 takes that list and processes each channel ID 00:13:25.350 --> 00:13:27.889 individually. It basically loops through your 00:13:27.889 --> 00:13:30.269 list. So it handles the looping for you. Clever. 00:13:30.330 --> 00:13:34.049 Exactly. And third, you use the RSS read node 00:13:34.049 --> 00:13:37.250 again, but this time the URL isn't fixed. It 00:13:37.250 --> 00:13:39.570 dynamically uses the channel ID that the split 00:13:39.570 --> 00:13:41.730 out node is currently processing. So it fetches 00:13:41.730 --> 00:13:44.409 the feed for each channel in your list, one after 00:13:44.409 --> 00:13:45.850 the other. Okay, fetch feed for channel one, 00:13:45.889 --> 00:13:47.850 then channel two, etc. But wouldn't that pull 00:13:47.850 --> 00:13:50.970 old videos too? It would. By default, it grabs 00:13:50.970 --> 00:13:53.710 the last 10 or 15 videos, old or new, which is 00:13:53.710 --> 00:13:56.590 noisy. So we add a filter node right after the 00:13:56.590 --> 00:13:59.879 RSS read. Filter based on what? Date. Precisely. 00:13:59.899 --> 00:14:02.200 We filter based on date logic. The filter's job 00:14:02.200 --> 00:14:04.679 is simple. Check each video's publication date 00:14:04.679 --> 00:14:06.379 and only let it pass if it meets our condition. 00:14:06.600 --> 00:14:09.279 Like, is this video new? And that involves those 00:14:09.279 --> 00:14:11.879 scary date expressions you mentioned. It can 00:14:11.879 --> 00:14:13.720 look intimidating, yeah. You'll see things like 00:14:13.720 --> 00:14:17.580 jcsun .isodate .towermat compared to something 00:14:17.580 --> 00:14:21.440 like nasaday .minus .format uiyy -mndd. Okay, 00:14:21.559 --> 00:14:23.399 yeah, that looks a bit much. But break it down. 00:14:23.940 --> 00:14:26.639 The first part just grabs the video's date. make 00:14:26.639 --> 00:14:29.159 sure it's treated as a date and formats it consistently 00:14:29.159 --> 00:14:32.820 the second part gets today's date now subtracts 00:14:32.820 --> 00:14:35.399 three days and formats that the filter just checks 00:14:35.399 --> 00:14:37.899 if the video date is after three days ago so 00:14:37.899 --> 00:14:40.080 it's just saying only show me videos from the 00:14:40.080 --> 00:14:42.620 last three days exactly much simpler when you 00:14:42.620 --> 00:14:45.639 translate it and honestly Don't feel you need 00:14:45.639 --> 00:14:48.379 to memorize this stuff. Use ChatGPT. Totally. 00:14:48.480 --> 00:14:51.259 Ask your favorite LLM to be your AI code translator. 00:14:51.500 --> 00:14:53.860 Describe what you want. I need to filter for 00:14:53.860 --> 00:14:55.840 videos newer than three days, and it'll often 00:14:55.840 --> 00:14:58.779 spit out the exact expression you need. Works 00:14:58.779 --> 00:15:01.659 surprisingly well. Nice tip. So why is that split 00:15:01.659 --> 00:15:03.779 -out node, the looping one, so essential for 00:15:03.779 --> 00:15:05.960 this kind of monitor? Because it lets you process 00:15:05.960 --> 00:15:08.539 a list of items individually, making the whole 00:15:08.539 --> 00:15:11.159 thing scalable to many channels. Okay, everything 00:15:11.159 --> 00:15:13.779 we've built so far. Powerful automation, yes. 00:15:14.039 --> 00:15:16.679 Scripting on steroids. But this next part, it's 00:15:16.679 --> 00:15:18.320 different. We're giving the machine a brain. 00:15:18.639 --> 00:15:21.779 Welcome to AI agents. AI agents. How are they 00:15:21.779 --> 00:15:23.919 different from just an AI node in a workflow? 00:15:24.299 --> 00:15:26.559 A regular workflow follows a fixed script, right? 00:15:26.879 --> 00:15:30.080 Trigger, fetch data, filter, summarize, notify, 00:15:30.519 --> 00:15:35.059 done. An AI agent, it can think, it can plan, 00:15:35.120 --> 00:15:37.679 it can use tools to achieve a goal you give it. 00:15:37.870 --> 00:15:40.909 Think dot plan. Use tools. Okay, that sounds 00:15:40.909 --> 00:15:42.929 like a big leap. It is. Think of the difference 00:15:42.929 --> 00:15:45.169 between an old video game character stuck on 00:15:45.169 --> 00:15:48.769 rails and a modern NPC in a game like Red Dead 00:15:48.769 --> 00:15:52.230 Redemption 2. That MTC has goals, perceives the 00:15:52.230 --> 00:15:54.690 world, makes decisions. That's kind of the leap 00:15:54.690 --> 00:15:57.269 from a workflow to an agent. Script versus mind. 00:15:57.549 --> 00:15:59.389 Okay, I get the concept. How do you build one 00:15:59.389 --> 00:16:02.250 in NAN, the J -A -R -V -I -S blueprint? It's 00:16:02.250 --> 00:16:04.210 surprisingly straightforward. The core is the 00:16:04.210 --> 00:16:07.490 AI agent node itself. That's the brain. a chat 00:16:07.490 --> 00:16:09.509 trigger node so you can talk to it. And you hook 00:16:09.509 --> 00:16:11.889 it up to a powerful AI model, GPT -4, Claw 3 00:16:11.889 --> 00:16:14.190 Opus, something good at reasoning. Right. Brain 00:16:14.190 --> 00:16:16.549 chat interface, powerful model. But you said 00:16:16.549 --> 00:16:18.929 tools. Yes. An agent needs superpowers, right? 00:16:19.090 --> 00:16:22.129 Tools. And in ADN, tools are just other nodes 00:16:22.129 --> 00:16:23.870 or even entire workflows you've already built. 00:16:24.070 --> 00:16:26.289 Oh. So the ping command we made earlier could 00:16:26.289 --> 00:16:29.389 be a tool. Exactly. You define it as a tool for 00:16:29.389 --> 00:16:32.669 the agent. You give it the command, ping 1 .1 00:16:32.669 --> 00:16:37.659 .1 .1 IC4. And crucially... A description. This 00:16:37.659 --> 00:16:39.840 is vital. Why is the description so important? 00:16:40.120 --> 00:16:42.059 Because that's how the AI brain knows when to 00:16:42.059 --> 00:16:44.179 use the tool. For the ping tool, the description 00:16:44.179 --> 00:16:47.379 might be. Use this to check if a public website 00:16:47.379 --> 00:16:49.820 is up and responding. This is for testing the 00:16:49.820 --> 00:16:52.120 general internet connection. Clear and specific. 00:16:52.440 --> 00:16:55.379 Okay. So you could have another tool. Maybe pinging 00:16:55.379 --> 00:16:57.940 a local server. Yep. A server health tool. Command 00:16:57.940 --> 00:17:04.140 ping 192 .168 .1 .1SE3. Description. Use this 00:17:04.140 --> 00:17:06.599 to check if the main local file server is responding. 00:17:06.880 --> 00:17:09.880 The server's host name is mainframe. Got it. 00:17:09.940 --> 00:17:12.059 Tools are nodes with good descriptions. And then? 00:17:12.140 --> 00:17:14.500 Then the magic moment. You open the chat interface 00:17:14.500 --> 00:17:16.619 connected to your agent, and you just talk to 00:17:16.619 --> 00:17:18.819 it. Like, is the internet working? Exactly. And 00:17:18.819 --> 00:17:21.180 the agent receives that. Its AI brain thinks, 00:17:21.240 --> 00:17:23.059 okay, user wants to know about internet status. 00:17:23.700 --> 00:17:25.940 Aha, I have a tool for checking public websites. 00:17:26.019 --> 00:17:27.579 It runs the website ping tool automatically. 00:17:27.980 --> 00:17:29.539 Wow. Then it gets the result and responds to 00:17:29.539 --> 00:17:31.200 you, yes, the internet connection seems stable 00:17:31.200 --> 00:17:33.900 based on the ping test. And if I ask, check if 00:17:33.900 --> 00:17:36.240 the mainframe is up. It uses the other tool, 00:17:36.359 --> 00:17:38.759 the server health tool. And if you ask something 00:17:38.759 --> 00:17:41.619 complex, like give me a full network status report. 00:17:41.960 --> 00:17:45.410 It might use. Both tools. It might. If it reasons, 00:17:45.549 --> 00:17:47.529 that's the best way to answer. It could run both 00:17:47.529 --> 00:17:50.509 pings, then synthesize the results into one response 00:17:50.509 --> 00:17:53.450 for you. That's genuinely powerful, having a 00:17:53.450 --> 00:17:55.289 conversation with your infrastructure. And this 00:17:55.289 --> 00:17:57.869 is just scratching the surface. Imagine giving 00:17:57.869 --> 00:18:00.769 it a server medic tool belt tools to check disk 00:18:00.769 --> 00:18:03.369 space, find large files, maybe even clear logs. 00:18:03.690 --> 00:18:06.369 You could ask servers slow, find out why. Or 00:18:06.369 --> 00:18:08.789 a cybersecurity analyst agent, checking firewall 00:18:08.789 --> 00:18:11.480 rules. running network scans. The possibilities 00:18:11.480 --> 00:18:13.619 just explode once you start thinking in terms 00:18:13.619 --> 00:18:16.480 of agents and tools. It's a completely different 00:18:16.480 --> 00:18:18.819 way of interacting with your systems. Okay, so 00:18:18.819 --> 00:18:21.140 what's the single most important element when 00:18:21.140 --> 00:18:24.400 giving an AI agent one of these superpower tools? 00:18:24.720 --> 00:18:27.980 Without a doubt, the tool's clear, precise description. 00:18:28.670 --> 00:18:30.730 That's what guides the AI's decision -making 00:18:30.730 --> 00:18:33.450 process. Now, as you build more of these, whether 00:18:33.450 --> 00:18:36.349 simple workflows or complex agents, you'll want 00:18:36.349 --> 00:18:39.349 some best practices and automators code, if you 00:18:39.349 --> 00:18:42.009 will. Okay, lay it on me. First, save early, 00:18:42.130 --> 00:18:46.089 save often. Just Shreetrol plus S, CMD plus S 00:18:46.089 --> 00:18:49.250 all the time. Seriously, workflows can get complex 00:18:49.250 --> 00:18:52.660 fast. Good habit. What else? Pin your test data. 00:18:52.799 --> 00:18:55.200 We talked about this. Saves so much time, saves 00:18:55.200 --> 00:18:57.619 API costs, especially when tuning AI prompts. 00:18:57.819 --> 00:19:00.200 Just pin the input data to the AI node and you 00:19:00.200 --> 00:19:02.779 can tweak the prompt over and over without new 00:19:02.779 --> 00:19:05.720 API calls. Right. The pinning tip. Crucial. Also, 00:19:05.819 --> 00:19:08.400 duplicate for experimentation. Before you try 00:19:08.400 --> 00:19:10.480 something major, just duplicate the whole workflow. 00:19:10.599 --> 00:19:12.880 It's your safe sandbox, like saving your game 00:19:12.880 --> 00:19:14.460 before a boss fight. Makes sense. Don't break 00:19:14.460 --> 00:19:16.779 the working version. And related to AI, watch 00:19:16.779 --> 00:19:19.500 your AI tokens. Refining pumps with pin data 00:19:19.500 --> 00:19:21.920 helps, but be mindful of how much you're asking 00:19:21.920 --> 00:19:24.220 the AI to process, especially with expensive 00:19:24.220 --> 00:19:29.099 models. Got it. Save, pin, duplicate, watch tokens. 00:19:29.480 --> 00:19:32.680 So once you master these basics, what's the next 00:19:32.680 --> 00:19:34.900 quest? What other cool things could you build? 00:19:35.160 --> 00:19:38.019 Oh, man, where to start? You could build AI that 00:19:38.019 --> 00:19:41.079 rates news articles from 110 based on your specific 00:19:41.079 --> 00:19:45.000 interests or an AI that summarizes YouTube transcripts 00:19:45.000 --> 00:19:46.880 to tell you if a video is actually worth watching. 00:19:47.400 --> 00:19:49.460 Transcript summary. I like that. For the home 00:19:49.460 --> 00:19:52.079 lab folks, build that master health check agent 00:19:52.079 --> 00:19:54.380 we talked about or a chat bot. You can just ask 00:19:54.380 --> 00:19:56.200 questions about your infrastructure, integrate 00:19:56.200 --> 00:19:59.359 with security cameras, maybe use AI object detection, 00:19:59.559 --> 00:20:02.000 automate social media posting. OK, lots of ideas. 00:20:02.140 --> 00:20:04.880 What about when things go wrong? Common problems. 00:20:05.140 --> 00:20:07.599 Good question. Quick troubleshooter. Connection 00:20:07.599 --> 00:20:09.980 failed. Almost always check your credentials 00:20:09.980 --> 00:20:12.579 or maybe an IP block on the server side. Workflow 00:20:12.579 --> 00:20:15.000 is not triggering. Double check the active toggle 00:20:15.000 --> 00:20:17.599 is on and look at the executions log for errors. 00:20:17.819 --> 00:20:21.559 What if the AI responses are just... Weird. 99 00:20:21.559 --> 00:20:24.359 % of the time, it's a prompt problem. Go back, 00:20:24.400 --> 00:20:26.680 refine your instructions, check the raw JSON 00:20:26.680 --> 00:20:29.720 view of the output. Sometimes the AI adds weird 00:20:29.720 --> 00:20:31.700 formatting. And getting too many notifications. 00:20:32.099 --> 00:20:36.420 More limit and filter nodes. Be aggressive. Automation 00:20:36.420 --> 00:20:39.819 should reduce noise, not add to it. Right. Okay, 00:20:39.900 --> 00:20:41.720 so beyond all the technical stuff, what's the 00:20:41.720 --> 00:20:44.279 mindset shift someone needs to really embrace 00:20:44.279 --> 00:20:47.289 this kind of automation power? I think it's starting 00:20:47.289 --> 00:20:50.829 to see the digital world less as a series of 00:20:50.829 --> 00:20:54.089 websites and apps and more like a giant Lego 00:20:54.089 --> 00:20:56.609 set. Everything's potentially connectable, ready 00:20:56.609 --> 00:20:59.170 for your creative connections. A giant Lego set. 00:20:59.329 --> 00:21:01.569 I like that. So thinking back on what we covered, 00:21:01.710 --> 00:21:04.680 it feels like a significant journey. You start 00:21:04.680 --> 00:21:07.940 with maybe feeling overwhelmed. And you end up 00:21:07.940 --> 00:21:09.940 deploying your own private automation engine, 00:21:10.019 --> 00:21:12.440 not just using a service. You have control. You 00:21:12.440 --> 00:21:14.920 built an AI intelligence agent analyzing the 00:21:14.920 --> 00:21:17.259 world for you. You gave your workflow direct 00:21:17.259 --> 00:21:20.740 OS access, maybe most profound. You gave an AI 00:21:20.740 --> 00:21:23.000 a brain and tools to think. It really is like 00:21:23.000 --> 00:21:25.220 that moment in The Matrix, you know, where Neo 00:21:25.220 --> 00:21:27.279 stops seeing the bullets. And starts seeing the 00:21:27.279 --> 00:21:31.400 code. Exactly. APIs, servers. Data streams, they 00:21:31.400 --> 00:21:33.960 stop looking like chores and start looking like 00:21:33.960 --> 00:21:37.180 interconnected systems, a vast play where that's 00:21:37.180 --> 00:21:39.259 your new superpower, understanding the code, 00:21:39.420 --> 00:21:41.259 making the connection. And it feels like the 00:21:41.259 --> 00:21:44.740 potential is just huge. You mentioned over 400 00:21:44.740 --> 00:21:48.160 pre -built integrations, plus building your own. 00:21:48.599 --> 00:21:50.720 It's not just endless possibilities, it feels 00:21:50.720 --> 00:21:53.640 like. Yeah. Like exploring this vast archipelago, 00:21:53.779 --> 00:21:56.119 maybe Ha Long Bay with all its islands. Each 00:21:56.119 --> 00:21:58.420 island is a new idea, a new connection waiting. 00:21:59.099 --> 00:22:00.940 That's a beautiful way to put it. And it leads 00:22:00.940 --> 00:22:02.880 to the real question, doesn't it? It's no longer 00:22:02.880 --> 00:22:05.059 if you can automate something. Right. That's 00:22:05.059 --> 00:22:06.960 simply what you will choose to automate first. 00:22:07.140 --> 00:22:09.900 A powerful question indeed. Well, thank you for 00:22:09.900 --> 00:22:11.779 joining us on this deep dive today. We encourage 00:22:11.779 --> 00:22:14.400 everyone listening. Keep exploring. Keep building. 00:22:14.619 --> 00:22:16.619 Absolutely. And always embrace that curiosity 00:22:16.619 --> 00:22:19.279 that drives this whole journey forward. Outro 00:22:19.279 --> 00:22:19.500 music.