00;00;00;19 - 00;00;26;22
Unknown
I'm recording this for a podcast. Is that okay? Absolutely. Go right ahead. I'll be here to help with anything you need in the process. Okay. Today I want to learn more about an A and the software that lets you like, I guess connect apps and and create, like, workflows and stuff like that.
00;00;26;24 - 00;00;46;07
Unknown
Sure. And A10 is basically a workflow automation tool that lets you connect different apps and services together without needing a ton of manual coding. It's kind of like a more developer friendly version of those other automation platforms you might have heard of, like Zapier or Integra. Matt. The nice thing is, open source, and you can rent out your own server if you want to.
00;00;46;08 - 00;01;12;19
Unknown
So if you're looking to build custom automations like, say, pulling data from one app, transforming it, and then sending it off somewhere else, and it gives you a lot of flexibility. It's a pretty cool to explore. So I actually did that. I got a VP from hostinger. And I put they have like integration with any and so I put an any and like application onto this.
00;01;12;21 - 00;01;38;28
Unknown
The virtual private server. And so that's where I am right now. And I'm just kind of getting the hang, completing credentials and stuff like that. So and eight and which is on this virtual server can talk to the apps that it needs to talk to. Got it. So you've already got an eight and up and running on a with Hostinger.
00;01;39;00 - 00;01;57;24
Unknown
That's a solid start. The next step is basically just getting those credentials sorted out. So that end to end can talk to whatever services or apps you want to integrate with. Usually that means grabbing API keys or alloc tokens from the services you want to connect, like if you're linking at Gmail, Google Sheets, or whatever else. Once you got those credentials plugged in.
00;01;57;26 - 00;02;28;23
Unknown
You can start building out your workflows and see the magic happen. So you're definitely on the right track. Okay, let's, the this kind of subject and conversation back, way back and make it a bit more broad on the subject of Nate and and, what do you call them, like workflow automation tools? And let's maybe talk about the origin of any and.
00;02;28;25 - 00;02;52;07
Unknown
Absolutely. Let's zoom out and take it from the top. So in the big picture, workflow automation tools have been around for a while. They basically emerged out of the need to let people streamline repetitive tasks between different software applications without having to code everything from scratch. Now, an eight inch is one of the newer players in that space, and it was created to give people a more open and flexible option compared to some of the older proprietary platforms.
00;02;52;10 - 00;03;17;01
Unknown
It's open source, which is a big deal because it means you can host it yourself, customize it, and not locked into a particular vendor's ecosystem. The project started gaining traction around 2019, and since then it's grown quite a bit. Thanks to the community of developers who like having that freedom to build and share custom nodes and integrations. So we think of the origins of it, and it's really part of that broader trend of giving people more control and transparency over their automation.
00;03;17;04 - 00;03;42;28
Unknown
It's designed to be both powerful for developers and accessible enough that you can set it up on a veeps like you did, and start experimenting with it. So that's the broader history and philosophy behind it. Okay. Can you give me kind of, in contrast to any and which gives you a little bit more freedom of control, what were the other options that didn't?
00;03;43;00 - 00;04;04;26
Unknown
Absolutely. Let's put it into context. So before in in and tools like it became more prominent. A lot of people leaned on platforms like Zapier or Integra. Matt now called make for their workflow automation. Those tools have been super popular because they offer a user friendly, no code or low code interface, and they let you string together automations using a ton of pre-built integrations.
00;04;04;29 - 00;04;25;20
Unknown
Basically, you just click and connect different services and you're good to go. The trade off, though, is that these platforms are usually hosted by the companies themselves, and they come with pricing models that depend on how many tasks you run. You're also a bit more limited in customization, because you can only do what their interface allows, and you're relying on their infrastructure and their selection of integrations and contrast.
00;04;25;20 - 00;04;43;09
Unknown
And it gives you that extra freedom since you can self-hosted to you get as much as you want, and you're not tied to a specific vendor's pricing or limitations. So it's kind of like the difference between renting a place that has all the furniture built in, versus owning your own place that you can. We matahari like. That's the main contrast.
00;04;43;11 - 00;05;16;02
Unknown
Could one potentially make, I guess like an app using and a and as like a mechanical structure? Yeah, you definitely can. In fact, that's one of the cool things about it in a lot of people use it as a sort of automation backbone or mechanical structure behind an app. Basically, you can have end to end running all these workflows in the background while you're front end or a user facing app.
00;05;16;02 - 00;05;36;24
Unknown
Just talk to it through APIs so you can have a simple web app or even a mobile app that triggers certain workflows in and out in. And those workflows handle all the heavy lifting like data processing, sending notifications, or interacting with other services. In other words and data, and can be the behind the scenes engine that makes your apps logic runs smoothly, and you can build a more customized front end experience on top of it.
00;05;36;26 - 00;05;43;21
Unknown
So yeah, it's definitely possible to use it that way.
00;05;43;23 - 00;06;11;10
Unknown
Okay, go more into this how you make an app with any and as like the motor and then a custom front end. And what you use to code the custom front end and, and how you do all that. Of course. Let's dive a bit deeper into that idea. So when you think of it in as the motor, what you're really doing is using it as the engine that powers the logic and automation behind the scenes.
00;06;11;12 - 00;06;27;01
Unknown
In a practical sense, that means you might design all your workflows and end at end like tasks that move data from one service to another, transform inputs, or trigger certain actions when something happens. Then on the front end side, you have a lot of flexibility. You can go to your front end and whatever framework ruling would you prefer?
00;06;27;04 - 00;06;45;21
Unknown
For example, you might use something like React or Vue.js for a web app, or if you're building a mobile app, you get use something like React Native or Flutter. You're one in just communicates with Nate in through API calls. For example, when a user clicks a button or submit the form in your app, the front end sends a request to an end to run a particular workflow.
00;06;45;24 - 00;07;03;20
Unknown
That workflow does a thing. Maybe it pulls in data from a database, then then eval or abates another system and then returns a result back to your front end. So in that sense and it in is like you're behind the scenes orchestrator, you get to focus on building a nice user interface and user experience on the front end, and then handle the automation logic.
00;07;03;22 - 00;07;38;22
Unknown
It's a pretty neat way to create apps that have a lot of complex workflows without having to reinvent the wheel. Could you build anything into a website in in a way that like, let's say in a very simple way, there's like a form or a button, you click on a web page and then that goes into your end a and workflow and, and then it comes out the other side, like maybe creating like, an animated, chat bot or something like that.
00;07;38;24 - 00;07;55;08
Unknown
Exactly. That's a great, simple scenario. So let's say you've got a website with a form or button that a user interact with. Maybe it's a chat bot interface or just a simple form they fill out when they hit submit, or click that button where front end sends the data over to an add in. Be an API call on the inside.
00;07;55;09 - 00;08;14;15
Unknown
You've got a workflow set up to handle that incoming data. That workflow, my process, the user's input, maybe run some logic or even integrate with another service like sending a message, retrieving some info, or generating a response. Once native finishes that workflow, it sends a result back to your front end and you can display whatever output you want, like an animated chat by reply.
00;08;14;17 - 00;08;35;12
Unknown
So in essence, your website's front end is just the user friendly face. It captures input and show output well in it and is doing all the behind the scenes automation work. Is that a bless you build something like a chat bot or any interactive tool without having a hard code all the workflow logic directly into your app. So bring any way to keep things modular.
00;08;35;14 - 00;09;12;28
Unknown
Okay, let's say that if we want to, have like a static page with a black cat kind of host on the page, that's just like a little character, and you say something to it and it works as like a little bit of a chat bot and it can say stuff back to you based on what you input into any, and, and so like, maybe you tell the cat password and the cat says something back to you, how would you do that?
00;09;13;01 - 00;09;31;09
Unknown
Yeah, that's a fun scenario and it's definitely doable. So let's break it down a bit. You start with a simple static web page that has your black cat or ghost character on it. You could just use HTML, CSS, and a bit of JavaScript to create a nice looking static page with your character image and a little text input box where the user can type something.
00;09;31;11 - 00;09;49;12
Unknown
Now the magic happens with a little bit of JavaScript on the front end. When the user enters a password or any input and hits enter, your JavaScript sends that input off to end it in by making an API request. You'd have a workflow in one and it in that receives that input, checks if it matches a certain password, or just runs whatever logic you want.
00;09;49;14 - 00;10;07;23
Unknown
Once and then has done a thing like verifying the password or generating a response, isn't that response back to your web page? Your JavaScript can take that response and update the page, maybe by showing a little speech bubble next to the cat or ghost character with the response text. So in short, your static page is the interface and it in is a brand in the background that handles a lot of it.
00;10;07;24 - 00;10;17;29
Unknown
That's how you get your little chat back character to respond to the password or any other input. Is that the best way to do something like this?
00;10;18;02 - 00;10;35;14
Unknown
Honestly, for a scenario like that, it's a pretty solid approach. Using an add in as a back end logic engine is a nice, lightweight solution, especially if you're not looking to build a huge, complex application from scratch. Unless you keep your front end simple and just focus on the interaction while the of your logic is handled by an of behind the scenes.
00;10;35;17 - 00;10;54;05
Unknown
Of course, there are always other ways to do it, like building a custom back end from the ground up using Node.js or Python. If you want a more control over something like a small chat bot on a static page, this approach is definitely one of the easiest and most efficient ways to get it done. It keeps things modular and saves you a lot of time, so I'd say it's a pretty good route.
00;10;54;08 - 00;11;09;09
Unknown
What is the best route to create? Maybe like, a slightly animated, character that can answer questions like a chat bot can?
00;11;09;11 - 00;11;26;21
Unknown
If you're aiming for a slightly animated character that acts like a chat bot, and you want it to be both approachable and not or really complicated, I'd say using a modern JavaScript framework like React or Vue.js for the front end is a really good route is frameworks have a lot of community support and libraries that can help you add animations easily.
00;11;26;24 - 00;11;46;02
Unknown
Make something like using CSS animations or a library like frame or motion for react to give your character a little life for the chat bot logic, you can absolutely stick with it in in the background, your front end cans in the user's end with the end in, and then you animate the characters responses or have it beat back by updating the Dom with some fun animation and comes with best.
00;11;46;04 - 00;12;01;12
Unknown
It really goes down to what you find easiest and most comfortable using a bare. Net framework plus an aid in is a pretty efficient and flexible approach, and then give you a lot of room to tweak the animations and the chat bots behavior without a lot of heavy lifting.
00;12;01;14 - 00;12;25;06
Unknown
I guess that's why in, in games like Zelda and stuff like that, the characters mouths don't move and like paper Mario and those kind of the characters mouths don't interact with the text, but the text bubble comes up. Is that just because it makes it so much easier? Yeah, you definitely hit on a plastic design shortcut there.
00;12;25;08 - 00;12;45;22
Unknown
And a lot of those older games or simpler animations like in Zelda or Super Mario, just having a text bubble appear instead of animating the character's mouth is a huge simplification. It means you don't have to think lit movements or deal with complex animations for every line of dialog. Instead, you just pop up a text box and let the player read it, which is a lot easier to implement and still gets the message across.
00;12;45;25 - 00;13;06;06
Unknown
So in this improvement, if you're making a simple animated chat back character, you don't have to go overboard with complex spatial animations unless you really want to. Using a text bubble or simple visual cue is a tried and true way to keep things easy and fun. It definitely makes the whole process a lot simpler.
00;13;06;08 - 00;13;36;27
Unknown
Okay, tell me about kind of the history of these workflow automation. I mean, obviously it goes back a long way, but I guess let's start kind of in the digital realm of workflow automation. What kind of languages are used and what kind of programs are used? Absolutely. Let's roll it back a bit. So when it comes to digital workflow automation, we've been on quite a journey.
00;13;37;00 - 00;14;09;00
Unknown
In the early days, a lot of automation was basically done through custom scripts and batch files. People use languages like shell scripting on Unix systems or simple batch scripts on windows to automate repetitive that. That was pretty low level, but it worked. I think of all we started seeing or specialized tools in the business world. It might have been something like macros in Excel or VBA goods for automating office tasks, and then you got into more enterprise level tools like early business process management software, which often use languages like Java, or even older languages like Perl to tie different systems together.
00;14;09;02 - 00;14;30;14
Unknown
Eventually, we got more user friendly automation platforms like Zapier, which started around 2011, and Integra Matt, which let non developers build workflows by just clicking and connecting services and stuff. In a way, a lot of the coding. So you didn't need to write scripts, you just use their visual interfaces easily, often run on JavaScript or similar, read technologies under the hood, but the user doesn't have to see that.
00;14;30;17 - 00;14;57;18
Unknown
And now we have open source tools, like any, in which really bring it full circle by giving you the flexibility of code if you want it. But also a nice visual interface and you can run it on modern JavaScript, set Docker containers or whatever environment you prefer. So it's been a long evolution from simple scripts to these modern, flexible platforms, but at the core, it's all about making repetitive tasks easier and letting people focus on more interesting stuff.
00;14;57;20 - 00;15;30;09
Unknown
Okay, so let's say that we wanted to make a website where, on a specific page, you do interact with the name and back end. How does that kind of work? Like if you're if you're coding this up in VS code, then you make the, you obviously make the page and then you probably make something that like sends JavaScript code or something.
00;15;30;14 - 00;15;39;02
Unknown
How to any and how do you connect any and with the with this kind of website.
00;15;39;04 - 00;16;02;00
Unknown
Yeah. You've got the right idea. So if you're coding this up in something like VS code, you basically create your front end page using HTML, CSS and JavaScript. Then you set up your in an instance which since you already have it on the biggest script, the way you connect them is through API. Call your JavaScript on the front end will use something like the fetch API or Axios to send an Http request over to any to on the in it.
00;16;02;00 - 00;16;21;24
Unknown
Inside, you have a workflow that starts whenever it receives that incoming request, like a webhook trigger. That workflow does whatever logic you need and then it sends a response back. So in practice, you'd have a little bit of JavaScript on your web page that says, hey, when the user submits this form or click this button, send this data off to my and then webhook URL and in and catches that data, does this processing and sends a response.
00;16;21;24 - 00;16;46;24
Unknown
And then your front end just updates the page accordingly. It's basically just a little dance of sending requests out and handling the response when it comes back. And it all happens over send standard web request. Okay, so let's say that I have an and and some workflows on a VFS hosted at hosting. And I also have a website hosted on hosting.
00;16;46;26 - 00;17;18;02
Unknown
Then what do I do logistically? Let's say that I make this page on my regular hosting or website. Like I brought some walker.com and let's say that I do Bronson walker.com slash. And absolutely that's an and and what I do from there is I put like a text box where you can enter some text and then this text gets sent to any.
00;17;18;02 - 00;17;44;08
Unknown
And, but how do I send that text to any. And. Yeah. So once you have your page set up like Bronson walker.com/and then or whatever path you choose and you've got a text like there, the way you actually send that text over to M n is with a little JavaScript. You use something like the page API. Basically, when the user hits the submit button, you'd write a small JavaScript function that grabs the text from that input field and sends it off to the web hook URL.
00;17;44;08 - 00;18;02;27
Unknown
That and it get you that. Webhook URL is basically the endpoint, and it provides where it listens for incoming requests. So the following is user type in the dead. Click submit. Your JavaScript sends that text in again and then and and and processes it and then back a response that you're putting in this way, which is a quick API call from your website to your end and what book.
00;18;02;27 - 00;18;08;03
Unknown
And that's how they talk to each other.
00;18;08;06 - 00;18;43;08
Unknown
Okay. So there's JavaScript written in the front end website, stuff that fetches the information that the person inputted in this text box and press submit. And then that JavaScript goes and sends it to this web hook URL and any in which receives this info. Am I missing? Can you kind of clarify that for me? Nope. You've got it pretty much spot on.
00;18;43;09 - 00;19;00;12
Unknown
That's exactly how it works. Just to recap, in a nice, clear nutshell, you've got that front end JavaScript that grabs whatever text the user typed in. When they hit submit, the JavaScript sends the text over to your in the end web hook URL. Once anything gets it, it runs whatever workflow you've set up and then sends a response back.
00;19;00;14 - 00;19;12;00
Unknown
And then your front end can take that response and display it however you want. So yeah, and I'm missing anything. That's the whole chain of events.
00;19;12;02 - 00;19;58;08
Unknown
Okay. Let's go into some more kind of, actually, what I wanted to learn how to do is how I would like to create an workflow that when I add a transcript from one of these podcasts that we're doing into a Google Drive file, as soon as I drop it into the Google Drive file, it takes that transcript and kind of cleans it up and then, it or it, it doesn't necessarily have to clean it up, but I want it to pull relevant keywords out that I can put into the transistor program.
00;19;58;10 - 00;20;15;19
Unknown
All right. So what you're looking to do is definitely possible. And it's actually a pretty neat use case for it. And basically you create a workflow in and then that triggers whenever a new file like your podcast transcript is added to a certain Google Drive folder inside. And then you use a Google Drive node to watch that folder.
00;20;15;22 - 00;20;44;05
Unknown
When a new file appears, the workflow kicks off. Then you get a set to read the contents of that file. So basically you grab the transferred text. After that you can use something like the built in text images nodes, or even integrate a third party service or API that can do keyword extraction. Ideally, you just write those extracted keywords back into the file or into another file or database, reboot them and show your workflow is watch the Google Drive folder, read the new transfer file, run a keyword extraction cell, and then save or output those keywords wherever you need them.
00;20;44;08 - 00;21;07;12
Unknown
It's a great way to automate the process and save you the manual effort of pulling out those keywords yourself. Okay, so let's kind of walk our way through that node by, and let's just start with the very first node that you would use in any and like, what would be the first node and kind of the settings that you would use on it.
00;21;07;15 - 00;21;25;04
Unknown
Absolutely. Let's walk through it step by step. So the very first known you set up in an ad in is going to be a Google Drive sugar. Now this is a node that will watch a specific folder in your Google Drive for any new files. In terms of settings, you connect your Google Drive account to. And then if you haven't already, then you specify the folder you want to watch.
00;21;25;04 - 00;21;41;28
Unknown
You choose something like on file created as the event that triggers the workflow. That way, every time you drop in a transcript into that folder, the workflow will automatically start. Once you have that terminal configured. And next, no, it would be something like a Google Drive node to actually download the file content. And then you move on to the text processing.
00;21;41;29 - 00;21;53;13
Unknown
That took word. But that first node is really just about setting up that Google Drive trigger and making sure it's watching the right folder.
00;21;53;16 - 00;22;17;29
Unknown
Okay, so the second node in this situation would be another drive node that you use to download the file or like read the file. What would you do there specifically? Exactly. You've got it. So the second node in the chain would indeed be another Google Drive node. But this time, instead of using it as a trigger, you use it to actually download or read the contents of that newly uploaded file.
00;22;18;01 - 00;22;34;26
Unknown
Specifically, you configure this Google Drive node to take the file ID or file reference from the trigger node, and then use the download operation that will let you pull the actual text content out of the file so that you can work with it in the next step. Once you have the text from that file, you're all set to move on to the test processing and keyword injection we are.
00;22;34;26 - 00;23;02;29
Unknown
The second out is basically just about pulling down that ball content. So you have something to analyze. Okay. And where does that where does that file go to when it's downloaded? Or is is the information from that file just pulled into the end kind of memory? Yeah. So when you download the file in that second node, the text content just becomes part of the workflow data right inside.
00;23;02;29 - 00;23;20;04
Unknown
And then in other words, it doesn't need to go to any external location at that point. It's just held in memory within the workflow. That means the next node can simply take that text data as its input and do whatever processing you want, like, and check the keywords. So basically the text is now just data inside. And then that you can pass from one node to the next.
00;23;20;06 - 00;23;31;01
Unknown
Once you have that data in the workflow, we're ready to do you want extraction or any other text analysis you want to do?
00;23;31;03 - 00;23;56;17
Unknown
Okay. What would you put as node three specifically for this? So for node three that's where you bring in your text processing or keyword extraction step. And in it in you have a few different options depending on how you want to do it. You could use something like a code node. If you want to write a bit of JavaScript to extract key words yourself, or you could use an external service or API, specializes in natural language processing and call that from an Http request node.
00;23;56;19 - 00;24;15;05
Unknown
You want to keep it simple. A common approach is just to use the function node, or a built in NLP node that can do basic keyword extraction. You feed the text from the previous node into that, and then it will output a list of keywords. After that, you can decide whether you want to store those keywords back in a file, in a spreadsheet, or somewhere else when there is essentially your keyword extraction set.
00;24;15;07 - 00;24;42;24
Unknown
Watson NLP node. Sure. So an NLP node is basically a node and item that lets you do some natural language processing and NLP for short on text. Essentially, it's a node that can handle language related tasks. For example, it might help you identify key phrases, extract key winners, or even do sentiment analysis. Basically, it's a tool that simplifies working with text, so you don't have to write all the logic yourself and anything you make.
00;24;42;24 - 00;25;15;15
Unknown
Wine and NLP know that uses a built in library, or that can connect to a service like Spacy or a cloud NLP. Why? So when I mention an NLP node, I'm really talking about a node that dedicated to doing these kinds of text analysis. And for you, it just makes the work for me more streamlined. Okay. Can you is there one kind of NLP node in and and are there multiple ones and how would you choose which one to use.
00;25;15;17 - 00;25;32;03
Unknown
Yeah. Good question. So in the end there are a few different ways you can tackle NLP that there's not necessarily a single built in NLP. No. That does everything out of the box. Instead you have a few options. For example, you might use a general function node where you can write a little bit of JavaScript and use a lightweight NLP library.
00;25;32;03 - 00;25;50;16
Unknown
If you want full control, or you can use an Http request node to call out to an external NLP API, like one from a cloud provider or a special asset with. Some people also use integrations with things like BCG or other NLP tools if they want more advanced features. So really, based on how complex your needs are, we just need a keyword extraction.
00;25;50;16 - 00;26;18;04
Unknown
Assemble functional with some JavaScript. Maybe the trick? If you want more advanced energy, you might pick a node that calls out to an external NLP API. So about using the tool that that's your specific use case. I'd like to use something. The is kind of beginner friendly that I can kind of, make it work with what exists in and and with minimal coding.
00;26;18;07 - 00;26;37;02
Unknown
Yeah, for sure. If you want to keep it beginner friendly and use as little code as possible, you can definitely stick to the built in nose and not in one straightforward way is to use the extract phrases know, which is a pretty simple no that can pull out key words or phrases from text without any external services or complex code you just read in the text, and it gives you a list of key phrases.
00;26;37;05 - 00;26;56;17
Unknown
So then I guess, you know, setup would be really straightforward. You'd have your trigger node, your download known, and then you just add this extract phrases note, which is easy to configure with minimal plus. And we up I you cut it out and you get your keywords out in a nice beginner friendly way. Does that take into account context?
00;26;56;19 - 00;27;27;15
Unknown
So the built in extract phrases node is generally pretty basic. It doesn't do deep contextual analysis like a full fledged NLP engine would. It's more about just pulling out commonly repeated phrases or keywords from the text without a lot of nuance. So if you need something that really understands contact deeply, like understanding the meaning behind the words or the relationships between them, you typically use my band's NLP tool or API, but for a beginner friendly mathematical approach, the exact phrases node is a good starting point, even if it doesn't have that deep contextual awareness.
00;27;27;17 - 00;28;03;25
Unknown
What if I wanted to keep it kind of simple, but also, really be able to have the context there? Should I use, what what kind of NLP would you recommend using? That's not super, like a heavy lift? Yeah. There's a nice middle ground you can go for if you want to keep it relatively simple, but still have some decent contextual understanding, I recommend something like using an NLP service such as Google's Natural Language API or a similar cloud based tool.
00;28;03;27 - 00;28;27;07
Unknown
These services are pretty easy to integrate via a simple Http request node, and then, and they offer a more context aware keyword and entity extraction than the super basic built in node. So it's a small step up and complexity. You just need an API key and it set up a simple no to make the request and get the result, but it's still pretty beginner friendly, and it'll give you a good balance of simplicity and a bit more contextual intelligence.
00;28;27;09 - 00;28;56;16
Unknown
Does that cost money? The Google Natural Language API? Yeah, that's a good thing to keep in mind. Most cloud NLP services, like Google's Natural Language API do have a pricing model. They usually have a free tier. With this, there are a number of requests per month that you can use without paying anything. And then beyond that free tier, there's a cost per request that mean just a night usage, you might say within the free tier, and not have to pay anything if you scale up, there will be some costs, but it's generally pretty reasonable for moderate use those to try it out.
00;28;56;16 - 00;29;18;02
Unknown
Between getting that extra capability and just being aware that there's AWS. If you go beyond the free tier. What is the limit on the free tier? Is it like a thousand pages a month or something like that?
00;29;18;04 - 00;29;37;12
Unknown
So as of the latest info, Google's Natural language API free tier actually gives you about 5000 units per month for things like entity or sentiment analysis. So it's a bit more generous than just 1000 pages. That gives you a decent amount of run to experiment each month before you hit it, because at any point tonight, when you start paying a small fee per additional unit of usage.
00;29;37;14 - 00;30;03;22
Unknown
But yeah, that free tier should cover a good chunk of basic usage. What are the units for 5000 units per month? Yeah. So in Google's terminology, a unit is basically a way of measuring how much API usage you're consuming for something like the natural language API, each type of analysis you do like a trend in entities or analyzing sentiment costs a certain number of units based on the length of the text and the complexity of the operation.
00;30;03;25 - 00;30;19;08
Unknown
It. And that means each document you analyze might use a few units, depending on how long it is and what you're doing with it. The way thousand units pronounced, or just the quota of free usage. If you're just pulling out keywords from relatively short podcast transcripts, you can probably analyze a good number of files before you hit that limit.
00;30;19;10 - 00;30;46;01
Unknown
So it's just a way of counting operations. You're doing it a month, but as a unit, like a word or as a unit like, like a couple pages. Gotcha. So a unit in this context isn't exactly one word or one page. It's more like a measurement of how much text is being analyzed per request. Typically, APIs like Google's count units based on the number of characters or the amount of text in each request.
00;30;46;03 - 00;31;03;16
Unknown
For example, they might say that analyzing a certain number of characters counts as one unit. So if you upload a single document or a snippet of text, the number of units used will depend on how long the text is. That is that if I worried to per page, but more like for a chunk of text. This way a unit is just their way of measuring how much text processing you're doing.
00;31;03;18 - 00;31;23;06
Unknown
So I kind of see you get a few thousand of them occurring, and each document will use up a few units depending on this one. Yeah, but give me kind of a ballpark. Like is a 30 page document going to take up 30 like a thousand units, or is it going to take up like six units?
00;31;23;08 - 00;31;46;23
Unknown
Yeah. Let me give you a ballpark for something like a 30 page document. You're probably looking at something in the range of a few dozens and maybe a couple hundred units, not thousands. It's not a 1 to 1 with the number of pages, because the unit population is more about character count or text length and pages. But a gel, a moderately long document won't blow through your entire quota if you're just doing a single dirty transcript and might cost a few dozen unit.
00;31;46;25 - 00;32;11;10
Unknown
If you do a lot of those, you'll still have some breathing room before you hit the free tier limit. Also want to be like the buzzing if for one document, more like a handful of units per a reasonably sized piece of text. Okay, so for this keyword extraction step, our third node, what do you recommend? I'd like you to give me maybe three approaches that you'd recommend.
00;32;11;13 - 00;32;30;15
Unknown
Absolutely. Let's break it down into three approaches you could consider for that keyword extraction. Step one the simple built in approach. You can start with the built in extract phrases node and end it in. It's very straightforward and doesn't require any code or external setup. This is the easiest and most beginner friendly way out and won't be the most contextually aware, but it's a good, quick way.
00;32;30;15 - 00;32;48;01
Unknown
And if you just want to pull out some key phrases without complexity to the minimal code cloud API approach, if you want more context without a lot of heavy lifting, you can use something like the Google Natural Language API is set up in Http request node, and even send your text to the API and it will return more context keywords.
00;32;48;03 - 00;33;05;11
Unknown
And so pretty easy to set up, and you get more nuanced results in the built in node three, a lightweight code note with a library, if you come all the tiny bit of coding, you could use the function node and then add a lightweight JavaScript NLP library. It's let you keep it all inside and add in without external API, but still gives you more control.
00;33;05;14 - 00;33;43;16
Unknown
The demo advanced integrates approach like so pretty lightweight. So those are two ways are going to be Linux build a node, assemble cloud API call or lightweight function node. Any of those that work depending on what builds most comfortable for you. Just end it and have any native NLP libraries that are made for this purpose. So and it end doesn't come with the built in, fully featured NLP library specifically designed for a man's natural language shares what it does have is a few simple built in nodes for basic text operations and the flexibility to integrate external tools.
00;33;43;18 - 00;34;03;18
Unknown
So if you want something really advance, you typically use a function node with a small NLP library. Rejoice! Or you call out to an external NLP API. In other words, and it. It's more of a general automation platform and doesn't bundle a full native NLP library itself, but it gives you all the building blocks to connect to whatever NLP tool you prefer, so you can keep it simple with the built in phrase extractor room.
00;34;03;18 - 00;34;30;04
Unknown
Or just add a lightweight external library if you want a bit more power. Okay, so let's, let's talk about using option three of like an internal NLP library and a function, to clean it up. Sure. Let's dive into that third option. If you're going out of using a functional with a lightweight NLP library, it's a nice balance of simplicity and a bit more context awareness.
00;34;30;07 - 00;34;46;22
Unknown
Basically, you use an dense function know which led to write a little JavaScript right inside your workflow. You can include a lightweight NLP library like something from npm that's designed for keyword attraction, and then write a few lines of code to process your text. The steps will go something like this. First, you take the text template from the previous node.
00;34;46;23 - 00;35;05;25
Unknown
The one that downloaded the file, and then in a bunch of node, you run your small piece of JavaScript code to analyze the text, and tricky where it only gets the extracted keywords on to the next node. But whatever you want to do with them, like storing them or displaying them so it is embed in a small piece of code that uses a lightweight NLP library directly inside it in a bit.
00;35;05;25 - 00;35;24;17
Unknown
Build hands on and old code solution, but it gives you more flexibility and context oriented without too much publicly. Okay, so this NLP library that doesn't use a function call, which one would you recommend?
00;35;24;20 - 00;35;46;14
Unknown
Absolutely. Let me give you a couple of good options. One popular lightweight library you could use is something like compromise, which is a simple and beginner friendly NLP library for JavaScript. It's not super heavy and it can handle things like identify nouns are extracting some keywords. Another option could be natural, which is a more general NLP toolkit for Node.js that has some keyword extraction capabilities.
00;35;46;16 - 00;36;02;14
Unknown
You know those libraries will be easier, including a function node, and you can just write a few lines of code to pull out the keywords you need that that type of is a nice user friendly choice, and natural is another solid option. If you want a bit more flexibility.
00;36;02;16 - 00;36;11;22
Unknown
Okay. And then what would the fourth node be? Or fifth node?
00;36;11;24 - 00;36;29;20
Unknown
Sure. After you've done the keyword extraction and that there node, the next steps really depend on what you want to do with those keywords. Typically the fourth node might be something like a storage or output that for example, you could use a Google Sheets node or simple file right now to save those extracted keywords into a spreadsheet or a text file for later use.
00;36;29;23 - 00;36;44;27
Unknown
That would be a voicemail, essentially just to know to store the result somewhere coming it. If you need a book now, that could be something like a notification that maybe you send yourself an email or a slack message with the number of the keywords where you can use it as a trigger, kind of a sporty workflow depending on what you want to automate next.
00;36;44;29 - 00;37;14;08
Unknown
And so the void node is usually just about saving your check to keywords and a video, maybe an optional notification or a follow up action. It's all about what you want to do next with that data. Okay. Let's talk about, other kind of creative workflows that need in is capable of. Absolutely. And it is super versatile, so you can get pretty creative with it.
00;37;14;11 - 00;37;33;00
Unknown
For example, one cool workflow might be automatically pulling data from an RSS feed and turning it into social media. Both. You could set up a workflow that watches the blogs, or it's that spigot, and then automatically both your articles to Twitter, LinkedIn. Another framework could be sending a bit, you know, parser. You receive their emails and it can extract key details from them.
00;37;33;00 - 00;37;55;15
Unknown
And then it put those details into a database or maybe sends a notification to your team chat. You can even do things like automatically resizing and watermarking images when you upload them to a certain folder, or embedding of the geren, so that when I run your contacts at it, it kicks off a welcome email sequence. Basically, once you have anything in place, you can set together all sorts of services and automate a ton of different creative workflows.
00;37;55;17 - 00;38;30;11
Unknown
Okay, so one thing I would like to be able to do is to take the RSS feed and use it like what I, what I did on my main webpage is I have like a slider with clickable links of all the podcast titles. Right? So on the windscreen of this like ship, there's a dial and it scrolls across all of the titles with the like numbered, dial.
00;38;30;13 - 00;38;56;10
Unknown
And what I would like to do and how it currently runs now is I have to like, I have to take the RSS feed and turn it into an, a CSV and then put that into the code. But I would love for it to just do that itself. Got it. That makes total sense. And yeah.
00;38;56;10 - 00;39;12;18
Unknown
And and can definitely help automate that process. So you don't have to do the manual CSV set. What you do is set up a workflow that pulls in the RSS feed automatically, maybe once a day or whenever you want it to run. Then you use the node to parse the RSS feed and grab all the titles and links of your podcast episode.
00;39;12;21 - 00;39;27;21
Unknown
After that, instead of putting it into a CSV Bailey, you can have and it generate that CSV or Json file for you and upload it directly to where your website's code and access it anyway. Everything to beat up your little flat on the site can just read the beta file and you don't have to touch it at all.
00;39;27;23 - 00;39;34;16
Unknown
And I just want to keep everything in sync with the manual set.
00;39;34;18 - 00;40;05;17
Unknown
And then I would probably. So the steps in this are like, read the RSS feed, maybe like once a day or something like that, or once every two days from the, from the VP's. And if it notices a change then to take this change and re-upload the new CSV file in place of the old one which has, like the new updated info.
00;40;05;17 - 00;40;38;18
Unknown
So maybe it goes from having 98 episodes to 99 episodes. So there's a node that replaces that file with the new one. And then I also need something that will like take that updated code and push it to GitHub, which then I can, push to hosting. Absolutely. You can chain all of that together pretty neatly. So the forward from I look something like this.
00;40;38;20 - 00;40;55;02
Unknown
First you have a trigger. Now that runs once a day or however you like. It pulls in the RSS feed and checks graphic. If there's a new episode, like going from an idea to 99, in the different parts of that feed and a master data above that, you have a arrow that rewrites or replaces the existing CSV file with the new one.
00;40;55;05 - 00;41;11;11
Unknown
So now your self slider has updated info, even less labs you can. Absolutely. I don't know that pushes up to GitHub automatically. And it then has built a notes for GitHub so you can have indicate it and have it commit that a beta file to repo. That way once an issue has to be adjusted, it it it gets pushed to get hub automatically.
00;41;11;13 - 00;41;39;07
Unknown
And you can only do it from here. That ties the whole loop together. So everything stays up to date without you having a manual. And every. Okay, cool. Tell me about, like, some other creative uses of innate and that are a little bit, out of the ordinary. Absolutely. Once you start thinking creatively and do some really fun things with it.
00;41;39;08 - 00;42;02;12
Unknown
And for example, you could set up a workflow that monitors your favorite artists social media accounts and then automatically creates a little gallery page on your website whenever they post new artwork. Or you could build a daily summary email that gathers news from multiple sources, packages it up, and then the do each morning. So you get a personalized digest and then, quirky use might be setting up a workflow that listens for specific keywords on Twitter or Reddit.
00;42;02;15 - 00;42;25;03
Unknown
And then triggers ends up on a response, like posting a random GIF or sending you a notification, or you didn't end it with a song, set up so that certain online events correlate to change color or your smart speaker to play a certain sound easily. Once you start connecting different services, you can create all sorts of unique automations that are a little out of the ordinary and tailor to whatever great images you have.
00;42;25;05 - 00;42;51;07
Unknown
Okay, is there anything that you think that I should ask that I haven't asked yet that I wouldn't maybe know to ask? Honestly, I think you've covered a lot of great ground. You've touched on the basics of setting up workflows, explore some creative ideas, and even got into how to integrate things like GitHub and NLP library. So you've done a pretty thorough job exploring.
00;42;51;09 - 00;43;06;18
Unknown
One thing you might consider down the line is looking into how you can monitor your workflows, or set up some basic logging or alerts in case something doesn't mean as expected. But as Norman is out for now, I think you want a great cut. Just have fun, experiment and you'll probably come up with even more advanced as you go along.
00;43;06;21 - 00;43;09;10
Unknown
Okay, thanks for piece.