# Stop Burning AI Credits — OpenClaw + Python 24/7 Job Finder 🦞🐍

## Метаданные

- **Канал:** Python Simplified
- **YouTube:** https://www.youtube.com/watch?v=ehfTs0wdW5g
- **Источник:** https://ekstraktznaniy.ru/video/44531

## Транскрипт

### Segment 1 (00:00 - 05:00) []

What if you never had to search for a job again? Because AI did it for you. You define exactly what you're looking for. The role, the salary, your skills and experience. And then when something really cool pops up, you get notified immediately. Not a day later and 200 applicants later, but actively scanning the web for new listings as frequently as you'd like. But the thing is, usually when you do something like this, you get hit with a giant bill and you expose yourself to prompt injections and bot blockers and all kinds of adventures that it's best to avoid. So today we will do it properly. We will use a hardened, ultra-sandboxed version of OpenClaw, and we will run it in a virtual private server that we fully control. It will then connect to WhatsApp and will get it to run 24/7. As though we've hired a personal assistant that never sleeps. Now, to keep things as affordable as possible, we will do all the heavy lifting with Python, selecting job listings, filtering them, and scheduling on the system level. Saving our precious LLM credits only for quick analysis and very basic decision making. Everything else will run with bash on the operating system of our server, meaning we are not invoking OpenClaw for these type of stuff. So stop wasting credits if you don't actually have to. And if you're ready, let's roll. So first of all, let's set up open. Now if you're new to this, OpenClaw is an open source AI assistant platform that connects to WhatsApp, Telegram, Discord and so on. And it allows us to talk to AI agents through these platforms. So ChatGPT Gemini, Grok and many others. Now, in our case, since we want our agents to always be available, we will deploy them on a remote system just like deploying a website. And the idea is, if our website runs perfectly when our computer is off, then so should our AI bots. Now, to keep things extra safe, we will use the One-Click-Deploy version that automatically blocks any kind of shenanigans you will see shortly. How. And for this, we'll go ahead and navigate to hostinger. As we've done in previous private server tutorials by typing hostinger. com/python. And if you do so, Hostinger will know that you came from this video because this is my special link. Yay! The link is as usual in the description and in the pinned comment. Now we already have a recommended server size for our KVM or our kernel based virtual machine. And funny enough, it's the exact same server specs that we use for our portfolio project. So each of these components is explained there in great detail. But briefly we have two CPU cores with eight gigabytes of memory and a very big chunk of disk space. We have 100GB running on NVMe, which basically means very fast. So your bot will respond quickly across all the social media apps you connect. Now, once we hit deploy, you can then choose between different plans we have monthly, yearly and so on. Just make sure to use my code Python. When you go for the yearly or the bi yearly plans, it will give you 10% discount. On top of all these extra savings you get for the longer periods. Again, this is my special code and you can use it for any kind of server, not just OpenClaw. So on my end I'm going to go for the two year plan. And even though I'm already subscribed to a whole bunch of AI platforms, I am paying ChatGPT and Gemini and Grok. I will keep those Nexus AI credits. There is an offer right now that basically doubles the credits you buy for free. So if you get five, you get five for free, or if you get 20, you get 20 for free. And I'm kind of curious to see how long they last me. I'll make sure to give you an update in the pinned comment. And also, this is actually the safest and fastest way to get OpenClaw up and running. You don't need to copy API keys and everything will be ready automatically. Now the last detail here is picking a server location near you. In my case, since I live in Bulgaria, I'm just going to go for Germany and then let's click on continue. We will then log in of course and enter our credit card details off camera. Hello everyone. It's been 24 hours later and yesterday I've spent the whole day optimizing and experimenting and making little changes in my code, and I really wish I took the 40 credits. Okay, I took the ten instead, but when you're using OpenClaw to troubleshoot itself for many, many hours, I finished filming yesterday's video at around midnight. Okay, so that was a lot of troubleshooting involved. These credits. They're not going to last you for very long for these type of use cases also. Okay. You can of course use your existing subscriptions

### Segment 2 (05:00 - 10:00) [5:00]

like Grok chat, GPT Gemini, whatever. But when you do so make sure you do so from the get go. Don't wait until your next credits run out, and only then try switching to your open AI API key or whatever, because there is a bunch of rewiring involved. And I didn't know that. Okay. And this, this 40 credits offer that I didn't take. You don't get it when you try to top up. You only get it when you register and it's like a one time thing. And when you're trying to top up will pay the full price. So another thing I didn't know. So I'm telling it to you right now, I thought I should film it. So from my observations 40 credits is plenty, ten credits is not. But again, I've been using it for very heavy and complex tasks. If you're just following along with this video, it's not going to be this intense. Okay, so yeah, that's all I wanted to say. And let's go back to the workflow. And yesterday. Now once our payment is processed we'll get our very own OpenClaw gateway token. So let's quickly copy it. And don't worry, by the time this video goes live, this token will no longer exist. Additionally, we will enter our super secret and private WhatsApp number and we will click on deploy. Now, once our server is ready, we'll get a live OpenClaw project managed by Docker Manager, meaning our bot is encapsulated in a Docker container for extra isolation. If you're not sure what it means, check out my Docker Quick Start guide. But you don't really have to. Okay, so how exactly do we talk to our bot? Well, we simply click on this open cloud port. We then paste our gateway token like so and then boom. We are in now. And for some unknown reason at this stage you're getting a disconnected for no reason error. Then all you do is replace the chat session main part of your URL with question mark token equals your token, and then once you press enter, this should solve it. Additionally, if you forgot to copy your token earlier, just go back to your project on Docker Manager and click on the Copy Gateway token button. Beautiful. Now back in our chat, let's just quickly check if our chat bot actually works. Let's ask it. Hey buddy, are you alive and perfect? He is indeed alive and well. Next we will connect our WhatsApp account by clicking on channels and then show QR code and we immediately click on wait for scan only. Then we open our WhatsApp on our phone and then we go ahead and scan this QR code. Okay, we do so in our Linked Devices section. Now once your device is linked on your phone and shows up as Google Chrome, Open claw or whichever browser you've used, then back in your open dashboard, you will navigate to the config tab. You will then click on update and then when you return to your channels tab, boom! Our WhatsApp is officially configured and it's good to go. Now you only do this long process once when you connect a new device, but whatever you do, okay, make sure that you don't miss this wait for scan button because you will get stuck here for hours and you will completely lose your mind. Now, once everything is connected, we will go back to the chat area in the browser and at the same time we will open a conversation with ourselves on WhatsApp and then we will ask OpenClaw yo, is anyone home? And boom, open clock will respond on both interfaces. Perfect. And usually when we add external resources to communicate with it means that we need to be extra careful. But this is how this One-Click-Deploy image solves it. So if we navigate to the skills tab and we open the built in skills, we can see that the vast majority of them is actually blocked on purpose. And anything that may impose a safety risk is simply not allowed. What we can do instead is navigate to the agents tab, where we can customize the model we're chatting with. By the way, you can pick from this list. And also we can view the list of tools that are available to us. So we can do basic read and write. We also have an execute tool that can run bash scripts, and we also see a web search and a web fetch tools, as well as a cron task killing tool. Now these are the only tools that we will need for our project. And actually we can even bypass those web tools. We can use their low level bash equivalents. Perfect. So let's move on with using these tools for our job market intelligence spot for.

### Segment 3 (10:00 - 15:00) [10:00]

So we'll need some starter files that you can of course clone from my GitHub. And the link is in the pinned comment. Now in those starter files we first have a resumé. We have two Python scripts and a bash script that we will review quickly, and we basically have a while loop that runs every seconds and executes both our Python scripts, one after the other. Additionally, we have a resumé where we set our preferences. Now, this is a Json file that you will need to adapt to your skill in your experience. It basically helps us understand who you are and what exactly you're looking for. So please go ahead and enter things like target roles, and you can enter your salary and important flags such as relocation and remote work. These will help us filter listings that we're not really interested to see. We are pulling these into our Python files. Also, make sure to include your education and of course your technical skills, which is another important filter that our Python files rely on. Next we have cool jobs. py. Now this file reads RSS job listings or structured documents that are meant for machine reading. And that's human reading. Very similar to APIs. These feeds, they expect computers to access them. So this is why our bot is not going to be blocked. Now we access these feeds through a bash command named C URL. So no playwright, no selenium and no pip installing anything. And then after we go through a bunch of location filters and skill filters, any listing that survived it is being saved to a CSV file named jobs dot CSV. Perfect. And then lastly we have a pool desk or pool description that gets the latest timestamp from our new CSV file, and then it pulls all these latest job listings into a Json file named description dot Json. And this Json file is the only file that open will actually access this file erases any old listings, and it only stores the most recent batch file. To get this up and running on the system level, make sure you set an SSH or secure shell password first. If you're not sure what it means, check out my cloud computing video, but only if you are completely lost. Otherwise, we will navigate to the overview section and we'll go ahead and copy this root password command like so. We will then paste it in our terminal. And we will enter our password. Beautiful. We are in. Then we will go ahead and navigate to the workspace of Open Claw. But how are we supposed to find it. Well let's just ask it. Let's go back to our browser. chat and let's ask it save this Python file as Marias test. py. And then in a set of back quotes, it will then print yo yo and you have all done so. Then back in our terminal it will go ahead and type find slash dash name in our case a string of Marias test. py and beautiful. We got the exact URL in response actually the exact path. Now keep in mind it will be a bit different on your end, so let's quickly copy it and let's navigate there with CD followed by the path. And let's see exactly what it stores with less and great, we see a whole bunch of existing files on the server, and if you see a similar collection on your end, then you're in the right spot. I'm so sorry to interrupt you guys, but if you enjoyed this video and if you'd like to support this channel, please consider clicking on the high button or the like button or please leave some kind of a comment to help with the algorithm a little bit. I always forget to tell it in videos and then I feel very silly. So sorry. Once again, I'm out. Next we will copy our local files into our remote server for. This will go ahead and open a new terminal in addition to our old one, and we'll go ahead and navigate to our project folder with CD. In my case, job intelligence or wherever you saved the files you clone for my GitHub. Next, we will actually make sure that these files are there and beautiful. They are indeed here. Next we will go ahead and paste this lovely command that you can copy from the description. And if you'd like to find out exactly what it does, check out my cloud computing video. It's a secure copy command. Just make sure you revise the address of my server

### Segment 4 (15:00 - 20:00) [15:00]

to the address of your server, as well as the name of my project to the name of your project. Now let's give it a quick run. And beautiful. Our files were copied, so now when we go back to our original SSH terminal and we go ahead and check the content of our folder with LS, our files were officially added to the server. Yay! Now, before we can go ahead and execute our loop, we will need to give ourselves permissions. If we don't do this, we will be denied. So we will type C, h, mode 700 and we will give ourselves permissions for the exec loop as h file. Next we will go ahead and launch our loop using no hub as in no hang up. And this will ensure that our bot keeps hunting 24/7 even after we close our laptop or disconnect from our secure shell. For this, you would have to bear with me and type the following. So no hub followed by the name of the file we are calling it on in our case dot slash exec loop dot as h, followed by a greater than symbol, followed by a name of a file that doesn't exist yet. Output dot log, which is a file to which we will save all these echoes from our shell script, as well as all the print statements from our Python files. It's a log file. Then we will go ahead and add two greater than and symbol one, which means print errors and regular output. And then finally we will add a space and another and symbol. And this will make sure that everything runs behind the scenes. Linux. What can I say. Now let's quickly give it a run. And okay let's verify that it actually works. Let's go ahead and review the content of our brand new log file, which get output dot log. And great. We found 28 job listings. So let's make sure they're actually there. Let's quickly check the content of our Json file with cat desk dot Json and boom! We have a whole bunch of Json objects here and we are officially good to proceed. Great. Our heavy lifting engine is running, and now you may be tempted to type your orders in the open cloud chat, just like I did several hours ago, but it doesn't always work because sometimes open forgets he can do things now. What do you currently see on the screen is the result of my natural language instructions. It did find me a job match and it did send me an alert on WhatsApp. Okay, you can see it here, but it's been more than five minutes and I this process just didn't repeat itself. So let me show you what works every single time. So let's quickly navigate to the cron jobs tab and we will manually set up this task okay. So we will call it Job alerts. And we will set it to run every five minutes. Also we will copy these very specific instructions from the description okay. Don't change anything. I've spent a lot of time perfecting them with open clock okay. And then once we schedule this task, those job alerts will pretty much start coming in like clockwork. Now, while we wait for this clock, let's quickly check our logs and okay, we see this one job listing that I received earlier, as well as the 28 listings from even earlier that I just didn't catch on time. And this is actually great news because it means that our system works perfectly, it ignores old listings, and it only focuses on the new ones. So what are we going to do? You know, for the sake of the demonstration, we will basically erase our data files and we will start collecting everything from scratch. For this we will remove with our M, our jobs dot CSV and our desk dot Json. Then it no longer exists. Now this will force Python to start everything from scratch. And we still have a few more minutes to spare. Now what if we have less than one minute left in our countdown? We'll pull out our phone and we will wait for some alerts to come in. It should happen any moment now. Okay, so let's just take a look and boom, we get a whole bunch of job matches. Wow. And this should be the 28 listings that we missed from earlier okay. Now obviously we didn't get 28 job alerts because OpenClaw has done some additional filtering, much more strict filtering on its own. But we got about like, what is it, six of them, which is awesome. Now I'm going to keep my eye on it and I will update in a few more minutes. Hopefully we'll get more of them now fast forward to about 30 minutes later.

### Segment 5 (20:00 - 21:00) [20:00]

We got two more job matches, one of them at 1209 and another one at 1204, and then anywhere in between 1144 and 1204. Nothing interesting happened, so we got no alerts for it. Okay. Now hopefully you're seeing similar results on your end, but if you're having any. I just got another one at 1214. Amazing. Now if you're having any issues okay. On your end please leave me a comment below and I will help you. I think I've explored this OpenClaw so deeply I may have become an expert. So best of luck on your job search and hopefully this video help you. And thank you so much for watching. If you found this video helpful, please share it with the world and don't forget to leave it a huge thumbs up and all kinds of comments! Now, if you'd like to see more videos of this kind, you can always subscribe to my channel and turn on the notification bell. I'll see you soon in an awesome tutorial! This one. So in the meantime that I.