Building an AI App Fast: Monitoring, Orchestration, and More Under One Roof with Eden AI

--- type: transcript date: 2025-02-19 youtube_id: G0ylKloEIPk --- # Transcript: Building an AI App Fast: Monitoring, Orchestration, and More Under One Roof with Eden AI in this video I'm going to be showing you Eden AI which is a platform that simplifies AI integration by handling monitoring observability orchestration as well as provider management all within a single interface what in AI allows you to do is effectively unify the core components that are involved with creating a production grade AI application the first thing that you can do is once you made an account on the platform you will be able to get some free credits to try all of this out this is what it looks like the first neat thing at the top is there is a co-pilot say if you're trying to figure out and isolate a potential idea that you might have you can try out that co-pilot and it will give you suggestions on some of the different services within Eden that you could potentially build your application with these are some of the different templates say if you have something that needs to purse documents or if you want to have some generative AI use case whether it's a custom chatbot or code generation or what have you those are all within here now there are also multimodal capabilities if you want to pass in things like images and remove the background or if you want to classify different images you can do that and additionally within the platform there's both speech to text models as well as textto speech models pretty much regardless of the use case of what you're trying to build they have a huge array of different offerings that are built into the platform so of course there are a ton of text examples on how you can get started with different applications and then finally there are video use cases if I just click into one of these I just want to show you what this looks like now one of the nice things with the platform that I want to highlight is the way that the billing works it's going to be passed through to these providers but you're going to be build at the same rate at what the provider would charge you so just to give you an idea these are the rates for what Amazon charges you to use this model the economics of this makes sense is we'll actually get a discounted rate from the different providers to be able to provide these at cost on par with what you would have otherwise had to do in setting up with whether it's GC P or AWS or what have you and the other nice thing with this platform is honestly not having to worry about all of the devops that go into managing a platform whether it's gcp or AWS where you have to be very Mindful and often times almost like a subject matter expert in being able to just navigate and utilize those platforms effectively another thing that I want to mention within the platform is they do have caching available within the apis for instance where this can be helpful is if you have a common request for your application the caching mechanism will be stored for 5 minutes and you won't have to infer the cost of making that request to the llm and you will have that cash response which is going to be faster as well as cheaper another great feature that's built into the platform as well are fallback providers for instance we have seen open ey go out before we've seen anthropic go down basically all of these different services at one time or another can go down so if your application is dependent or highly dependent on one of these services this will allow you to be mindful about when there are outages to have a fallback provider and potentially not even have your users experience or notice that there might have been an outage next what I'm going to do is show you the workflows tab here within the workflows tab there are a number of different examples as well what I'm going to do here is use the llm chat template I'm going to fire that up here if I go over to the Builder we'll see a very simple workflow effectively how this one works is we have the input we can choose from the different providers if I just show you this quickly here we have the models from Gro deep seek perplexity anthropic open AI meta cohere Amazon Google we also have mistol in here basically all of the state-of-the-art models and the top AI labs are all built into the platform let's say for instance we want to use deep seek R1 we can go and configure that now to test this it's super simple we can go ahead and make a request with the piece of JavaScript or python with in our application or we can test it within the interface here I can say hello world just have a simple example here and if I go and query that we'll see it will go through and run through this workflow now the other nice thing with this is you can easily create reg chatbots we can go ahead and create if I just call this demo and I go within create project within here all that we have to do to create our chatbot we can go and upload our files and we will be able to start to have a conversation with the documents if you want to chat amongst proprietary data or if you want to even upload a book or a series of Articles or papers or what have you that can be a really useful use case for the rag chatbot next I want to show you a couple features that are currently still in development but they will be available in the future be able to set up both guard rails as well as fine-tuning let's say we want to fine-tune GPD 40 we can go and upload the CSV of these structured prompts and we can have a fine-tune model in just a number of clicks next are guard rails so this is something that especially in production use cases is something you hear about much more with guard rails what you can do is set up your application in a way where only certain things are going to be able to be discussed in this example here we can see that the topics are cats and dogs if allowed you can basically describe which different topics are or are not allowed and then you'll be able to use that within your application now next is monitoring as you start to use this appc this will all become a lot more Rich within here you can see all of the different metrics you can see the cost over time the cost per sub feature you can export this and this is just nice to have the visibility with not having to worry about setting up any of this effectively you will be able to see all of this as a part of using the platform now the other nice thing with this and I always have to call this out when it is the case setting up the API key for Eden is super straightforward this isn't always the case with some providers especially especially with hyperscalers you do have the option where you can bring your own API Keys as well another thing to note is with in account management you can go ahead and include all of your different team members within here and like I said billing is going to be straightforward as soon as you verify that you are a human like it says here you'll be able to get $10 worth of free credits to try all of this out and then you have the option on whether you want to autore refill credits or if you want to add credits to your balance you can do that as well what I want to do is I want to create an image gener ation application I'm going to use this image generation endpoint and then I'll show you how it incorporates into some of the features that I just demonstrated here I'm going to take this curl request and I'm going to hop over to cursor here if you're not familiar with cursor I'm just going to npx create next app we'll create it within the root of our directory here I'll just go through and accept all of the different prompts then once we have that what I'm going to do with in here is I'm just going to add in the page as the context use app router for the API and I'll just plug that in there now the other thing that I'm going to grab are also all of the providers my thought is I'm going to take all of these and put them within the front end and have a bit of a drop down I also want to have a drop down on the front end to show the different models and I'll paste that in as well finally I want to be able to choose from one two three or four generations and display them in a grid when the results are returned I'll go ahead and I'll send that through this is how I personally would build and start to build out something like this but you could obviously go through and plug in the different details I find for stuff like this a lot of these AI code generation tools do actually do a pretty good job at getting you from 0 to one like here it can very easily spin up a post request endpoint for us and it should be able to handle the updates to the page as well it's going ahead and setting up ourv as well with our environment variables now the other thing to note any environment variables that you did see on screen I am going to be spinning these down and rotating them so don't worry about any of them that you see I'm going to go ahead and accept that and let's go and run our application npm runev and test this out a sign that reads hello world we'll go and we'll generate this we can see that we do have the option for resolution number of images as well as the provider and we also have this loading State here as well we have our working endpoint and if I swap this out let's try a different provider so maybe let's try three photos and let's try out stability AI now it didn't plug in all of the different models but that's something you can always go in after the fact and update as well so here we go we have three different Holo worlds from stability Ai and it does all look to be working now that we have our basic image generator working what we can do is we can go over and look for another feature that we might want to build let's say I want to remove the background from some of these images what I can do here is I can select this endpoint and I can just specify I'm going to say I want to now have the ability to remove the background from each respective photo with the click of a button on on Hover here is the Endo once complete fade in the new image with the background removed I'll paste in that endpoint we'll have the composer go and do its thing you can access all of the different state-of-the-art llms within here and you can do it all with one API key you don't need to go and hop between a number of different services and you do have that option to build out those no code agentic flows within that workflow Builder that we saw earlier we'll go ahead and I'll say a photo of a panda I'll go and I'll send this in now ideally it has the request of what we asked for we're going to be leveraging this AI image generation Ino that's passing through Eden to Leverage The specified model and also remove the background so here we now have this remove background button and if I click this hopefully we have this removed here we see the loading state within the button that we just clicked there we go it removed the background for us so I do see there is a little slight issue there and what I could do here is instead of using this API for a go and potentially use another model here the provider was set to API for AI but just to show you how easy this is to swap out what I could very well do is choose a different provider within here and I could just go instead use this clip dropped if I go over to API I go to remove background and within here if we just take a look for for the provider and I swwa that out for clip drop now the other option within here you can use the AI apis but if you choose to use a workflow you can find one that you might particularly find interesting let's say you want to convert a video into a Blog what's great with this is you can have this whole process run on the Eden infrastructure within here if we take a look at the Builder basically what this one does is you have a video and it goes through the process of extracting the text from the video and finally it generates this transcript for you it generates an image and by the end of it you'll have a blog post one thing to note if you are going to be using the workflows they are slightly different than the AI apis and the way that they work since they can be long running tasks what you'll have to do for these is either set up a web hook integration so your app can alert your user when those results are finished or alternatively what you can do is you can pull to get the execution results here we see the example we're similar to the launch execution we also also had the get execution and this will be the final result of that finished workflow that you have set up just a couple other pieces that I do want to mention within Eden one of the great features of the platform and being able to have access to all of these different providers is you can set up fallback providers let's say for instance you run into an issue with a rate limit on a particular provider like open AI or anthropics pretty notorious for having pretty strict rate limits what you can set with your application is to have these fallback providers if the request fails to anthropic or whether it's an outage or you your rate limits you can have the fallback providers ready where if there is a failure it's going to automatically default to those providers that you've specified that can be a very helpful feature especially in the case of outages because in some cases opening eyes API has been out for several hours and having the fallback provider just makes your application more reliable for your users overall that's pretty much for this video I encourage you to check out Eden AI there's a ton of different use cases within here they also have a ton of content on their site that is very relevant to different use cases in a ton of different domains pretty much regardless of what you're looking to build there is likely going to be something within Eden AI that you're going to find helpful I encourage you to check out the platform if you found this video useful please comment share and subscribe otherwise until the next one