Name: BaseAI.dev
Uploaded: 2024-12-20T05:58:59.975Z
Duration: 2 min 33 s
Description: An open-source web AI framework for building serverless autonomous AI agents with memory and tools.

baseai-dev

An open-source web AI framework for building serverless autonomous AI agents with memory and tools.

BaseAI.dev

Hey everybody. I am Ahmed.  I'm the founder of Langbase and we are a surveill AI agent platform. So if you have ever thought about building agents completely serverless, a rag store, file store, memory, you name it, we got it all right. 

But what today I'm going to do in this demo is I'm going to show you a couple of things. How you can build an agent, connect it to your company's data, your own data, and do agentic calling with a rag store. Right? So uh if you see it right now, this is what our langage studio looks like. Uh it should feel very familiar. Uh we take a good deal of inspiration from GitHub in fact.

The founder of GitHub Tom Preston Warner is our backer. This is a rag agent that knows a little bit about nothing. Right as you can see I'm asking what are the features of language, who created it, how tall is the founder, right stuff like this. 

And if I run it, it basically knows nothing. Right and if I were to give it memory, let's say I add memory about Lang Bas and memory about me, these are two separate rag stores. It has separate data set, separate files. All of this is, you can imagine, just two rag stores. 

And I'm using GPD for o. Right I could go on and use any of the top 100 llms, open source or not.

And all I have to do is now run this again and hope the demog gods work. There you go. Now what's happening is right now it figured out all the features of Lang base. It figured out that I am the founder of Lang base and then there's nothing about my height in the company documentation. Believe me, right? It automatically went ahead, used another R store which is named after me and figured out that I'm 185 cm tall, right? 

And you can actually prove all of this by going down to the Run, opening it up and looking at exactly what was augmented inside of this particular run, right? You can see all of this, right?

I'm not going to show you. So a couple of weeks ago, what we have done is we have op sourced and created a base AI framework which is built for React developers, TypeScript developers. How many people are here? WR JavaScript or ever have written? There you go, right? 

So we have purpose built for web. You can use Lang Bas for pretty much any language. For Bas AI is for web, right? And it is completely local. You don't have to download any Dockers or anything. You can use AMA to create and use with any LLM in just one command based AI deployer. 

All of your agents and memory become serverless.

But here's what I will show you a little bit uh more about it. It is not only is this framework local but it has agentic memory and agentic Tool calling. 

Here's a very simple uh nextjs project. When you start Bas AI, it basically becomes a folder inside of nextjs. It works inside of remix, wheat, whatever you know you name it right. 

And it basically downloads all of your pipes, tools, and memory in your code base right. In fact, you can chat with your GitHub repository. All you have to say is, well here are the documents, only look at the markdown files or typescript files and we will index it for you.

Automatically right and now this is here's a fun thing that happens right now. What I'm doing here in this demo is I'm asking to it to convert 185 cm to feet and it is correctly converting it into 6.07 which is my exact height. 

But behind the scene there's something fun happening right. What's happening is this is GPD 40 mini not the very best model out there. This is exactly what I asked and if you go down here it discovered this is a tool called the calculator. 

But it hallucinated in the tool call it's trying to convert 187 185 cm into inches instead of feet. Right behind the scene we got this.

Particular result of 72 in and without you having to do anything completely agentically it figured out that oh the user was asking for cm to feet and we just got back 72 in which is a hallucination it's not not the correct result right. 

So without you having to do anything it went ahead and did another call 72 in into feet this is happening automatically right and then it's getting back a big answer I think something like 6.69 or whatever right. 

And it's you know because it's an llm it's converting that into 6.07 ft and I can run it again and you can confirm that this is actually happening live.

Right there you go. Now what I can do is I can also show you memory. So who found Lang Bas when and why? And something like you know how all is the founder. There you go. There you go. Right, it knows. Right, so right now it is using the memory and the tool calling in one call. There you go, that's Lang base indaia.

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

BaseAI.dev Video