restack

An open-source framework empowering developers to build, launch, and scale autonomous AI products efficiently.

Restack

Hello everyone. Let's look at Restack, which is the open-source framework for autonomous intelligence helping developers build resilient autonomous AI applications. Wow, so they've got TypeScript and Python examples. So I thought we'd just look at these, see what the state of the union is currently with some of the AI frameworks. I've done enough Lang chain examples recently or in the past, so I thought I'd just look at this.

So this uses something called an engine, which is Docker. Then you go into the developer UI. We won't worry about the UI side of things for the time being, but let's just look at the intro. The Restack framework has polyglot long developers to build in both. Okay, so they designed it to address the challenges while scaling AI applications, creating truly autonomous applications. Blah blah blah. Okay, sounds good. Let's look at the architecture.

So you've got various clients including Flask, FastAPI, Django, and then you've got the services to Kubernetes. Let's have a look down here, see if we can find some Python code to get dirty with. So the Restack Python library provides functions for your AI application. So why no built-in API integrations? Direct function implementation enables full control over API interactions without framework-imposed integration constraints. I think that's probably a good point because as good as Lang chain is, the framework can be frustrating.

I would say so. Let's look at some examples anyway. So what we've got here is an audio transcript basic example using the Restack AI library to transcript an audio file to text and then translate it. Let's view the code. So P project Tomo looks like somebody's been using UV SRC EMV example. Let's look at EMV files. They, yeah, they're using open a Restack transcript. 

So Docker run Docker locally create MV file with Stripe secret key. I don't know why you need Stripe. Okay, let's look at source code, Tre client, and services. Let's look at client Cloud connect options.

Restack ai.

Restack okay services translate workflow so. That's doing a lot of the heavy lifting then you've got your functions and it's running async iio asynchronous right. Then let's look inside workflows transcribe translate.

Py here we get to the nitty-gritty. So we're providing a file path and a target language. Okay, passing in the file path and it's waiting. It's waiting workflow step and then it returns transcription in the target language. Okay, transcription translation. Let's go back a step, see what else we can do. Basic example using restack AI library to send emails with send grid API. Uh, stripe a basic example using stripe to create product and payment link. Okay, sounds good.

And autonomous projects this is more interesting. So monitor HN for interesting projects. Create does to reach out to their creators. Human in the loop trigger or receive calls. Autonomous PR creator. Autonomous let's have a look at. Okay, python flask fast API stream lit wv8 Lang chain. So flask and streamlit we've all had a go with those. Fast API kind of the linga franer for writing APIs in Python. 

Let's look at autonomous browser that sounds interesting. So the following tutorial can find get clone repo. This project tackles FOMO on Co Hacker News projects. It uses an autonomous AI to keep an eye and ping something interesting pops up. So next.js front ends for python back end poetry and all level zero automated open front end local host. Notice the difference between. Uh, that's quite a lot. Python's quite short but it's okay. 

Right, so it is using a we is asynchronous. So lots of Geary stuff going on here with agentic. The user has a chat where they can ask anything to the AI. In example, the user asks to search projects in healthcare and create tools to contact creators. Yeah, I don't know if anybody else has kind of got a YouTube account but um as soon as you your there's a little button that you can click to reveal business address business email address. I'm getting the idea that uh some of these things are um scraping the YouTube channels and emailing all the us or the channel owners because I'm getting stuff types script. So that's typescript that's for the front end when.

The llm decides to use a tool. The function sends an event to the assistant workflow. There typescript as well. Okay, so it's all going quite typescript. There more and more type script. Let's uh go back to restack cloud with restack framework. You're free to self host or deployed with restack Cloud deployer up. 

Okay, so there's an API that helps you deploy it to the cloud. That's nice, right? Let's look at libraries typescript npm types script blah blah blah. So it says polot but a lot of it's python. But it typically ends up back at typescript. So polyglot is maybe slight exaggeration. 

Okay, let's go back and just see if we can find some actual working examples. So use cases let's a look company's building with restack Oscar orange CGI Stone. Let's have a look at another one then. So let's have a look at video production video use cases script generation. 

So use AI to generate video scripts based on. So I'm getting the impression that a lot of this you could actually just do anyway just conversing with chat GPT. Um autonomous video publishing. So schedule unpublish videos across based on engagement. Is this new clothes? It's always the question. 

So what's it Us in import YouTube? Why restack inte Integrations? Okay, so maybe we're getting somewhere. Let's have a look for let uh show me YouTube example earn by building. 

Okay, we're offering exciting opportunities to talented developers to create groundbreaking AI applications using restacks autonomous framework. Wow, so you can get $500 for contributing pre-built examples. Contribute 500 for each accepted submission. What's the 10K for creating open source projects? 

So let's have a look here are some of the projects and preu examples we'd love to see. So build Next Generation AI software engineering platforms streamline engineering workflows. So this is for developers. Share AI Agent Store with colleagues. Transform marketing by continuously analyzing campaign performance optimizing ads spend and providing realtime.

Insights on customer behavior. Okay, build an autonomous trading product. Good luck with that. You would, uh, yeah, I don't even know enough to comment about that. But yeah, if you can do high-frequency trading, then you probably don't need to be tinkering with code. 

Um, product teams, data analysis, task creation, insights creation, enabling product teams to focus on right handles routine tasks. Behavior, okay, whatever that is. Video production pipelines, autonomous AI that automates video creation by generating, editing, and optimizing content. 

Okay, start with our quick start guides. Join Discord. Okay, awesome, autonomous. Let's have a look at that. That sounds good, and, uh, there's nothing there. Share your project. So nobody shared a project yet. You can probably get the idea that I'm not being paid to review this or promote this because, um, if I was, I wouldn't be doing a very good job. 

Restack versus Lang chain. Let's have a look at that. We're often asked about the difference between Restack and Lang chain. Learn how they differ and the problems they solve. So Lang chain, chat with my data, applic autonomous, durable air. 

Okay, well, this is going to be kind of a bit biased anyway, isn't it? But use Lang chain in a Restack workflow. Okay, so they're trying to draw in people who already like Lang chain, I would guess. So, Pantic for your model A, Open AI chat for your LLM, Lang chain for your framework, and then your functions from Restack. 

GPT 40, 40 mini. No, yeah, that's not really telling me very much. With the Lang chain function wrapped in the Restack decorator, it can be used as a step in a Restack workflow. So basically, it's a framework on top of a framework. Um, not my C, not my preferred way of going about things. 

But I think it's early days to commit yourself to one framework and putting all your eggs in one basket because these things are going to evolve very quickly. And there's, you get what's called the network effect where a bit like Facebook and...

Twitter became kind of the dominant and everything else fell by the wayside for a few years. Arguably that may change with sort of Nosta and uh all these other new versions of competitors to Twitter. But um there's a Docker file. So I don't know about you but I'm I am finding myself experimenting more with Docker files at the moment. Especially when I'm doing, if you watch my recent video on um using natural language to SQL, I wanted to use a realistic postgrad database and MySQL database. So I used the MySQL was 4 million records or I think it was 4 million, yeah. And um you don't want to be installing the database and creating all those records from scratch. So Docker file is great for that.

Docker compose I had an issue with Docker when a Docker compose up. And I think um sometimes there's a hyphen and sometimes not. I think you have a symbolic link if there's a hyphen you just create a symbolic link whereas there's no hyphen. But that's uh yeah that's once a niche. But that's um just going off on a tangent there. Um hackathons so that's already happened. Let's see what the €150 in cash for the top implementation of Restack Hacker KN. 

Let's see right this is my pet hate with videos when the font is too small. My other one is bad sound which I'm really guilty of myself. But um oh go away that's too much. Right okay oh that's not too bad. So I think there's still a long way to go with a lot of these AI things where it's still kind of one to one. It's a one user talking to one LLM. I think the real power is going to be with distributed and with obviously agents, lots of agents all talking to each other. 

In um on Noster there's something called um data vending machines where you can use um web LN lightning payments of exosis which are like a millionth of a penny or something like that. And um yeah you could automate the payment then. So um that's the problem because a lot of this AI you can build but if it needs payment or it.

Needs paid for API then you know you could before you know it you could be subscribing to you know 10 different APIs that are all $50 a month or something. Um so yeah that's uh it's interesting but I think that's there's reasons to look at it and be aware of it. But I think there's also reasons to be hesitant about getting too too deep into some of the proprietary stuff at the moment. 

I mean how many of us really want to receive calls with Twilio maybe? Um Bell AIS SDK M okay. Um that's that's it for today I think I'm just going to leave it there really. But it's just been a little bit of a whistle-stop tour of one of the competitors to LangChain. 

Um and if you are learning Python then stick with it because it is really the lingua franca of AI still. Um for more performance things I am using Rust for doing embeddings um with multithreading and async and all that good stuff. But for prototyping I think Python is there's a lot to be liked for doing things easily and quickly. 

CLE Python just to get proof of concepts up and running and so on. Um and yeah Docker. So if there's anything you can take away from today's video I would say uh don't shy away from Docker. If you can write a Docker file or edit a Docker file or even just clone a Docker file from GitHub just use it. 

Um much of the things you will do will need a database. If you need a data space then a Docker file is the easy way to get that up and running. If you're not using the Docker file what are you using? You using a big text file? A big flat text file? I don't know but [Music] um Docker Docker file YAML file it's all stuff that it's worth knowing. 

And I think um Poetry or UV there's a lot be said for those as well. I think requirements.txt feels a bit old-fashioned now already. I know is that ARG codes did a video about um sea engines. I created an engine so I signed up for the um let's just zoom in on that I'll show you it. Yeah ARG codes did a.

Video on UV and it is written by Astral who are backed by VC Venture Capital. So not 100% sort of the open source model that we all know and love. But I think if you can be familiar with using UV or Poetry or Ry, um I did some videos on Ry. One video, let's um let's just uh what day is it today? THS, let's just call it Thursday. 

Runs on Rest Cloud, one day retention for completed workflows, CR engine production Cloud. It's a great engine. So then we get an engine ID and an engine address and create an API key, which I won't do for obvious reasons. Um get $1,000 of credits. Okay, you have to schedule Jo an interview with one of the salespeople. I won't do that right now. 

Um so you create an engine and then you have stacks build and deploy your autonomous application stack with Restack Cloud SDK Python. Okay, it takes you back there. Poetry, poetry install Local Host hello world. I did do this a while back. So yeah, you can just do the hello world example. Um if you follow from, yeah, you clone that, get started from it. Sorry, clone the examples d Python and then go into the examples Python, get started there. 

That's a hello world thing. Docker back end, um created engine, deploy stack, get token, blah blah blah. So yeah, it's all kind of quite nicely presented and easy to use, I would say. But uh yeah, chat with us to enable. So create a cluster on your AWS, GCP or Azure to deploy your autonomous application stack. 

Right, so there we go. That's just a quick look at uh Restack. It's got some Python stuff on it, but um yeah, the key takeaway from this video I would say is just uh use Docker. Because more and more you're going to be needing a database for doing any AI stuff. And uh if you're going to be um even if you're going to be doing stuff with sort of vector databases, you may still need a traditional relational SQL database just to do your message history or history and all that kind of stuff. 

And plus user accounts if you want like uh different.

Name spaces and all that sort of stuff. I don't really know what I'm going on about now, but um yeah I'll leave that there. 

And yeah look at restack. It's like um a framework similar to Lang chain and there's some goodies on offer. So if you're feeling creative then you could go and build something. 

Thanks for watching. I'll be back soon.

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

Restack Video