Name: Marvin
Uploaded: 2024-12-20T14:02:02.982Z
Duration: 2 min 33 s
Description: An open-source AI engineering framework for building reliable, scalable, and trustworthy natural language interfaces.

marvin

An open-source AI engineering framework for building reliable, scalable, and trustworthy natural language interfaces.

Marvin

Hey guys. We got another video today. We're going to be checking out askmarvin.ai. So this is a batteries included library for building AI powered software. Marvin's Drive is to integrate AI directly into your code Base by making it look like any other function. 

So they have these functions here um where you can just use natural language to create the definition. And there's a little example right here generate a list of fruits. So we have a decorator here and then our functioned and then right there with the text generated list of n fruits. That's the actual text we're going to send to our model for our function. 

So instead of writing out code for our function, we're just going to use natural language. That's a great or that's one of the main features of Marvin. We're going to create a simple python script to take a YouTube url summarize the transcript. Here I have my code editor with my main file and we're going to start importing. 

We're going to start with OS. We'll need this to set our environment variable. This is where we'll set our API key and then from dot m will import low.v. This is also to share AI Agent Store with colleagues.

Handle our API key and from here we'll import YouTube transcript API. So this is an API that will allow us to download the transcripts from our urls. This is what we'll use to get the text and send it to our AI model. From here we'll import Marvin. So we'll say from Marvin import AI underscore FN for AI functions. This is what we'll use to create our language functions. 

And then we'll also need to create an environment, a python environment, and install all these dependencies. So we'll say python3 Dash M fan n. We'll activate the environment. Then from here we can start installing. I have these in my history because I was testing it out a little bit earlier. 

So here we can see Python and our pip install python dash dot ends and then also YouTube Dash transcript Dash API. At the end I'll just type Marvin as well so we can install everything in one command. We'll just let this finish here while we keep on coding. This might take a while. I think Marvin takes the longest out of our all these. 

So we'll start with typing load.mf to start the process for getting our environment variable and then right.

Here we're going to set our open AI API key. This will equal OS Dot and then right here we can see copilot's kind of finishing it for us. So os.getme and then in there we'll put in all uppercase open AI underscore API key. We can see Marvin still downloading so let's keep coding for now. 

We do need to create a DOT and file but for now um we'll just wait until all the dependencies are done downloading and then we'll go into our terminal and create that. So for now we'll just comment this so we know what's going on. 

Okay so down here we're going to start creating our AI function. So it takes a decorator at and then AI underscore FN just for AI function. Below this we can start creating a normal python function. So we'll say def we'll put the name of our function. We're just going to call it summarize video and then we'll input or the uh we'll give this function text in a string. 

So this will be the transcript that we're downloading. We're going to pass that into this function and then right here this is where we can type our prompt. So we can just use the language to define this function. We'll say.

Summarize summarize the video based off the provided text or let's just write summarize the provided text. So instead of creating a function definition, we're just going to use natural language to tell the function what we want it to do. So that's pretty much all we have to do for that function.

Now I'm down here. We'll start getting our video URL. We'll make this an input so the user is prompted to enter a URL at the terminal. And then down here we're going to get the ID from that URL. So we'll use a split to just get that last little part of the URL. This is what we'll give into. This is what will feed into the YouTube transcript API to get that transcript. 

So we'll say split. We'll type watch question mark V equals. This is what the top URLs at the top of your browser usually look like. We can see up at the top here has watch question mark V equals. So we'll just make sure we copy that same format in our python script and autofill. 

Put this as a positive one, but we'll make this a negative one here so we're splitting correctly. And then down here we're going to just print. We're going to print some lines.

So we know what the actual program is doing at any given time. So we're going to start fetching the transcript here. This is where the program might take a little bit time to load. So we just want to inform the user on what's going on. 

So we're fetching the transcript down here. We'll get the transcript from YouTube transcript API with the function get transcript. And then in here we'll just pass in that video ID. Keep in mind we just have to pass in the ID, not the actual URL. That's like the last part of the URL, just that string of characters and numbers. 

We'll set the language to English for now, and you can set this to whatever language you need. And then this transcript is going to be separated, so we do need to join this. So we'll create a new variable transcript underscore text. We're going to join everything from the spaces. 

And then down here we'll just print another line to let the user know what's going on. We're summarizing, so first we're fetching and then we're summarizing. Down here is when we can start using our AI function. So we'll say summary equals, we'll use our AI function summarize video.

And then we'll pass in the transcript text so this function will take that text and just summarize it. 

And then at the bottom here we'll just print summary and after this we will print the actual summary of the text and that should be all of our file here. 

We have everything set up so now we can try and test this out. 

We do need to set up uh we do still need to set up our API key and our M file so make sure we do that right now. 

Let's go ahead and get so here we will create our DOT end file. 

So we'll say touch Dot m.emv and then in here we can create our variable for our key. 

So we're going to type in all uppercase open AI underscore API underscore key and then we can paste in the key that we got from the open8 open AI website right in here. 

Quick note here if you check the docs they actually check for Marvin underscore open AI underscore API key first. 

I actually had run Marvin before with a different API key so when I set my open API key it wasn't correctly getting the right key. 

And that's because it's always checking for the Marvin key first and then if it can't find that it'll check open API.

Key so make sure you're using the correct variable name. Use Marvin underscore open AI underscore API underscore key. So I'm going to update all these right now. I already updated my end file. So now I'm just going to update all the variables here as well. 

Okay so we should be able to run our program now. One thing to keep in mind is we did right here in our video URL split. We are splitting a specific URL format. So make sure you're copying this from the top of your browser. If you click share, you can see it doesn't have the same watch question mark V. So just make sure you either update your code to work with both URLs or you use the correct URL. 

And this one, the code I wrote only works with the top URL because of our split, our splitting line. Anyway, make sure you have your API key set and we can start running this again. We have the URL copied. We're going to run python main.pi here. We can see it's asking us to enter the URL. We'll enter that and then also make sure you have that correct format just like we have typed up here in the code. 

And then here we can see it's.

Fetching transcript and now it's summarizing. This is probably the longest part of the program here while it just summarizes. Once it's done, it'll spit out a summary of the video. 

Keep in mind this is actually it's not watching the video. It's just taking that transcript, all the text from that video, and summarizing what the video is about. So it does a pretty good job about it. You can update this function to do more things than just summarize, but that's the scope of this video. 

Have fun with it. Check it out and see what else you can do with Marvin.

Note, that there may be slight differences in text and video. Watch video for the original content.

Marvin Video