OnDemands New AI Changes AI Development Forever! (Ondemand AI Tutorial)

So welcome back to the AI Grid and in today's video we'll be taking a look at On Demand, which is turbocharging AI innovation and is the fast track to product development. On Demand is an AI driven platform enabling the rapid deployment of end to end products and this actually allows users to build, deploy and manage AI agents and AI plugins for various different applications. So let's actually go ahead and take a look here at what's going to build coming with this platform and how you can get started on using it today. So when you come into the On Demand menu, you'll see an overview right here. What you'll be able to see is an overview of the different things that you're going to need if you want to run your program successfully. What we can see here is the plan usage. You can see the storage, the retrieval, augmented generation calls. You can also see the amount of tokens and you can also see the vectors, transcription hours and of course your storage. What you can also see here is your plugins, media storage and your models. But this is just an overview. The fun begins when we start building. So now moving from the overview, let's click the left hand side to go on over to data sets. Now that we're here, this is where the fun begins. In the Datasets section of the On Demand platform. Here you can access a wealth of data that is crucial for training your machine learning models or enhancing the functionality of your plugin. You can use these datasets for model training, leveraging these datasets to train your own models directly within the platform. This data can serve as a foundation for building accurate and robust machine learning models tailored to your specific needs. What's also fascinating is that these datasets can also be used in conjunction with the plugins that you build. For example, if you're developing a plugin for text analysis, you can use a large corpus of text data from the data sets to enhance the accuracy of your plugin. Here you can simply download and use these extensive libraries of data sets at will. By utilizing the Datasets feature, you can significantly streamline the development and deployment of your AI model plugins, making your workflows more efficient and effective. Welcome to bring your own inference section of the On Demand platform. This feature allows you to host a language model and integrate it seamlessly into the on demand environment. And here's how you can get started with bring your own inference. You can utilize a language model hosted on an external server. This integration allows you to leverage your existing models within the On Demand platform, which without needing to migrate them. To get started, click on the Create Endpoint button. Here's where you'll need to configure several parameters in the endpoint. You'll need to give your endpoint a descriptive name. You'll have to look at the endpoint type, the endpoint URL, the bearer token, the model ID and the max token length. Once all of this is done, you can then create your own endpoint. Here's where things get more interesting. Taking a look at the bring your own model management section. So this feature actually allows you to integrate models from external sources such as Hugging Face directly into On Demand and and here's how you can get started. To do this, all you need to do is click the Create Model button. This is where you'll be prompted to search a model from Hugging Face and you simply enter the name of the model you want to use. Now all we need to do is proceed to configure this model. So now we can simply select Solar3.8 billion parameters instruct and you can see right here the model is supported. This is what happens when your model is working. So now we need to do is type in our model. Once we just click our model here you can see that the model on Hugging face is supported. Now here we'll have to enter the name. We'll just call this Llama three. Eight filling parameters instruct. And then once that's done, you can see the number of parameters will be auto detected and we'll just go ahead and click Create Model. Now that the model has been created successfully here, you can see that we can now use this model. This now means that our model is now ready for use within On Demand. Now once you've created the model, you're then going to want to go ahead and click Create Endpoint. Then of course we're just going to want to bring in the model that we literally just created. We can configure the endpoint name, the quantization mode, the engine arguments, the engine type, and of course for this one we'll be using the recommended one and and we can also automatically scale to zero. And once all of those things are figured out, we can simply click deploy. Now let's take a look at the Plugin Marketplace section of the On Demand platform. The Plugin Marketplace is a powerful feature that allows you to extend the capabilities of your AI applications with a wide range of pre existing plugins. Let's explore how you can actually make the most of this feature. The Plugin Marketplace actually offers over 100 pre existing plugins that you can browse and integrate into your generative AI application. This these plugins cover a variety of different categories and use cases, making it easy to find the tools that you need. You can see here that you can browse the plugins by top plugins. You can also browse the plugins by education plugins. You can also see that you can browse them by programming plugins and just any topic that you really wanted to discover. There are also different plugin types you can choose between chat plugins, file plugins, and knowledge based plugins. These filters make it easier to narrow down the list to the most relevant plugins for your needs. If you wish for a custom solution to you can create your own plugin. Click Create Plugin in the top right hand corner and then fill out the details. For example, to create a REST API plugin, give your plugin a descriptive name, add an icon for easy identification, provide a description of what the plugin does, and choose a category of what is appropriate for your plugin. You can configure the plugin by specifying the key name and key type and for the OpenAI schema, writing the code for the plugin following the OpenAI schema. Once this is all configured, you can then click Create Plugin. After you publish your plugin, you can then publish it to the marketplace. This is where it will be reviewed by On Demand's team of developers to ensure it meets the criteria for Marketplace inclusion and once it's approved, your plugin will be available for others to use. If you wish to create a different type of plugin. When you're creating your plugins, you can choose a simple knowledge based plugin. This is a plugin that will simply be allowing you to access a lot more data. You can see right here you can upload different files and do the same things that we just did previously. Now if you want to manage your plugins, just come over to this right here. Then click the My Plugin section and you can now see where you what type of plugins you actually have. This section allows you to see all the plugins you're subscribed to and those you have created. You can manage your subscriptions and publish or update your plugins in this section. Welcome to the Servers Application section of the On Demand platform. This feature allows you to deploy and manage applications without the need to maintain your own server infrastructure. Here's how you can get started so serverless applications allow you to run your applications in the cloud without managing server infrastructure. This ensures your applications are always available and scalable and free from the risk of downtime with associated local servers. Let's look at creating a serverless repository. To start, click on Create Repository. Then you're going to need to fill out the following repository name, Repository Platform, Repository Visibility, and Repository URL. Once the repository is created, you can now configure your serverless application. Click on the Configurations tab in the left hand side. Then all we need to do now is click on Create Configuration and fill out the form here in the Serverless Configuration tab. This is where you'll need to fill out your application name, select the repository, the branch name, and the Docker script path. Now configuring the endpoint is actually relatively easy too. After configuring the application, you'll just need to set up the endpoints once again. Click Create Endpoint. Then of course, all you'll need to do is fill out these easy details. All you'll need to do is just fill them out and then click Deploy welcome to the Cloud Section Services of the On Demand Platform Cloud services provide essential functionalities that enhance the capabilities of your applications. Let's explore the available cloud services and how you can leverage them. Cloud services are integral components that offer advanced capabilities such as tech, Text to Speech and speech detection. These services are designed to be easily integrated into your applications, providing robust and scalable solutions without the need for complex infrastructure. As you can see, we currently have Text to Speech. This Text to Speech service converts written text into natural sounding speech. This is useful for applications that require audio output such as virtual assistants, accessibility tools and interactive voice response systems. You can also enable this and for the Speech to Text service. This is ideal for applications involving voice commands, transcription services and voice activated controls. Now let's go ahead and take a look at the Agent Builder. This is something that isn't yet available, but it currently will be in the coming week. This feature is designed to help you create customized AI agents by combining various plugins and inference models. Let's explore what you can expect from this upcoming feature. The Agent Builder will essentially allow you to create tailored AI experts by integrating multiple plugins and different inference models. You can see here on the right hand side we've got the Aviation so Search plugin, we've got the Aviation News plugin, and we've got the Aviation Live Tracker plugin. When we combine those with an inference model, we can get a super AI agent that can do many different tasks. You can tailor your AI agent's behavior and responses to suit specific requirements and this allows you to automate complex workflows, saving time and resources. The Agent Builder will continue to evolve with new features and improvements being added based on user feedback and other advancements. And by leveraging the upcoming Agent Builder feature, you can deploy powerful AI agents tailored to your specific needs. Now let's take a look at automations, which are also coming soon. This feature is designed to streamline and automate various tasks, making your workflows more efficient and productive. Let's dive into what you can expect from this upcoming feature. Automations essentially enable you to set up tasks that can be performed automatically based on predefined triggers and actions. This feature helps in automating repetitive tasks, reducing manual intervention and improving overall efficiency. You're going to be able to create automation workflows by defining specific triggers and corresponding actions. Triggers conclude events such as receiving new data, reaching a specific time, or changes in certain conditions. Actions are the tasks that the system will perform in response to the triggers, such as sending notifications, updating databases, or executing scripts. Right here on the right hand side we've got an example Automation Workflow. Imagine you want to stay up to date with the latest cryptocurrency news. You you can set up an automation to fetch the latest crypto news every hour. Once the news is fetched, it can be analyzed by a financial analyst agent based on the analysis. The automation can then place the trades on your finance trading platform, keeping you informed and proactive in your investments. Now let's take a look at the Playground section of the On Demand platform. The Playground is where you can build, test and experiment with your generative AI applications. Let's dive into the features and capabilities of this powerful tool. The Playground is designed to be an interactive environment where you can develop and test your AI applications. It provides a space to experiment with different plugins, models and configurations to create sophisticated AI solutions. Start by importing your plugins that you have subscribed to from the Plugin Marketplace. So when you've subscribed to plugins in the Plugin Marketplace, all you need to do now is click Add right here. Then you'll just need to add these plugins to your chat conversation. Just click Add and you can see I've now got the Flight Detail plugin added which searches for one way flights using skyscanner, offering details like departure location and destination information. Once that's there, I can now use this to test my application in real time. So let's walk through an example scenario to demonstrate the Playground's capability. So for Example, let's say we wanted to create a travel assistant application that finds flights, checks visa requirements and suggests activities at the destination in the playground. We would need to select the relevant plugins such as the Flight Details plugins and of course we would need to configure the model. We would need to get whichever model we wanted to use, like GPT4 Turbo. Now we would just need to test the application by inputting a query such as with your model you can see that you can choose any of the models that are currently available. If you do want a model that isn't currently here, you can always bring this in with the byom. Bring your own model section where you can add any model from GitHub and then use it in this application and then use it in the playground. So I'm just going to stick with GPT4O. Now for testing and debugging, you can enable debug mode to see the multi step processes in action and troubleshoot any issues. This helps you ensure that your application functions correctly and efficiently before any deployment. If you want to test and save certain presets, don't forget to save your presets for easy access and reuse. For example, right here what I can do is I can click Save preset, but of course I need to put in a name as you can see. All you need to do now is put in a name and then of course click Save presets. You can see that the preset was created successfully so easily. What I can do is I can go ahead select a preset which is preset flights and then I can easily test anything I want to within the playground without having to load everything once again. So let's walk through an example scenario to demonstrate the playground's capabilities. Let's say we wanted to create a travel assistant that finds flights, checks, visa requirements and suggest so let's say we wanted to create a travel assistant application that finds flights, checks visa requirements and suggests activities at the destination in the playground area. Select the relevant plugins such as the flights detail and of course ensure you have the right language model like GPT4 Turbo. What we can do now is test the application by inputting a query. So I've said find flights from New York to Tokyo, check if a US passport needs a visa, and suggest activities from Narita Airport. So we're going to click Run now and you can see it's now interacting, retrieving the relevant plugins and it's going to give me exactly what I need. Currently you can see it's finding the flights from New York to Tokyo and you can also see the direct flights. It also checks the visa requirements for US Passport holders and it actually gives me all the activities near Narita Airport. Once your application is built and tested, you can now integrate this into your system. All you need to do is to click this button right here. Click Get Code to generate the necessary code for your programming language. For example, if we needed JavaScript, all we would type in is Java and then we would click Get Code. You can now see here this application is working to get us the code and the code has been generated successfully. All we would now need to do is to implement the code in our system, configure the API key and our Generative AI application is ready to use the benefits of the Playground are absolutely outstanding. This allows you to experiment with different configurations and setups to find the best solution for your needs. This also allows you to gain immediate feedback on how your application performs, allowing for quick adjustments and improvements. By leveraging the On Demand Playground, you can create, test and refine powerful gen AI applications tailored to your specific requirements, ensuring optimal performance and functionality welcome to the Storage Management section of the On Demand Platform. This feature allows you to efficiently manage your documents, videos and images, ensuring that your resources are organized and accessible. Let's explore how you can utilize this feature. The Storage Management helps you keep track of all of your uploaded files, providing a comprehensive overview of your storage usage. This feature ensures that you can easily locate, organize and manage your resources. You can sort them by date, sort them by descending, and of course you can view all your files here. This is where we have the Usage tab. This feature allows you to monitor and analyze your usage across various resources, helping you to optimize your operations and manage costs effectively. This provides detailed insight into your consumption of different resources on the on demand platform. This helps you understand your usage patterns, manage costs, and make some formed decisions about the resource allocation. This dashboard displays key metrics such as the total number of API calls, storage used, and model inference time. If we switch over to the Cost tab, you can see this allows you to monitor how often each large language model is used. This cost breakdown includes expenses related to API calls, storage model inferences and other services. Now this is the API Keys Management section of the On Demand Platform. This feature allows you to create, manage and secure API keys, which are essential for integrating on demand capabilities into your applications. Now, API keys are crucial for authenticating and authorizing requests for made to the On Demand platform for your applications. Proper management of these keys ensures secure and efficient access to the platform's resources and services. To create a new API key, navigate to the API Keys Management section and click Create New Key. Provide a new name for your API key to easily identify its purpose or the application it might be used for. Once your name is there, just click Create Secret Key. You can now see it says Success. Your API key is created. Now remember, you're going to need to copy and paste this down because you're not going to be able to view it again for security reasons. Now, in this dashboard, you can view and manage all your existing API keys from this API Key Management Dashboard. Each key will be listed with details such as the name and the small details of the key. Here's where you can revoke keys that are no longer needed and have been compromised to maintain the security of your platform. Access now hopefully this video allows you to navigate and utilize the On Demand platform. From everything to managing datasets, leveraging cloud services, building and deploying serverless applications, to creating custom plugins and automation. By following this guide, you should have a clear understanding of how to make the most of On Demand's powerful features. Hopefully you enjoyed the video and I'll see you in the next one.