Name: TaskingAI
Uploaded: 2024-12-18T12:42:10.268Z
Duration: 2 min 33 s
Description: A developer-friendly cloud platform for building and running LLM agents for AI-native applications.

taskingai

A developer-friendly cloud platform for building and running LLM agents for AI-native applications.

TaskingAI

A couple of weeks back we had covered TaskingAI, which brings firebases simplicity to AI native app development. It's a new platform that enables you to have the ability to create AI applications that use a wide range of large language models from various providers. For example, you can use hugging face models as well as open AI models. You have the flexibility to choose amongst a lot of different providers, which is the flexibility that TaskingAI provides. 

This is a platform that unifies modular functions like inference, retrieval, assistance, and tools into one seamless ecosystem, which is going to streamline the creation and the development of applications for developers. Well, now Tasking AI has just introduced a serverless cloud platform that lets you access and test it out completely for free. This is where you're going to have the ability to create AI native applications with their cloud platform. 

Now Tasking AI is providing this user-friendly UI in a framework with the ability to work on the front end as well as the back end of applications, which is ideal for any user. This is for efficient and flexible large language model app development. You can access hundreds of AI models as well as unified APIs, and it will expand your ability to create high quality AI native apps. 

Now with the release of tasking AI serverless cloud platform, we're going to be exploring its capabilities. We're going to be comparing it to L chain and so much much more. So with that thought, guys, definitely stay tuned throughout today's video. So let's get straight to it. 

Hey, what is up guys? Welcome back to another YouTube video at the world of AI. In today's video, we're going to be taking a look at TaskingAI's new cloud platform. Now before we basically go into the playground, I want to emphasize and briefly touch up on the new cloud features. Basically, if you are to go over and sign up to their waitlist, you're

Going to be able to soon access this Cloud platform. Now there's five key concepts behind TaskingAI and their Cloud platform. You have different models, which is where TaskingAI incorporating various chat completion models. You can create these new models where they have different providers such as Mistral AI, open AI, anthropic, and many others. You can see that this is like serving as a core, like such as the brain of the AI assistant. 

These models offer distinct, like basic capabilities, attributes, and support different strengths in terms of input token lens. So it can be used for reasoning, logic capabilities, and various other categories. You can select and switch between models based on your needs as well as your task complexities. Next up, we have assistant. 

So Tasking AI's assistant feature is going to refer to the customizable AI entities that are capable of performing various tasks from customer service for internal training. You have different features such as providing different assistance to your chatbot. Now their functionality is basically based on the model as well as the tools that are provided. So you can select a particular model, give it a description, you can add different system prompts like we saw at the start of the video in that example, as well as adding memory. 

This is where you're also able to integrate different tools and retrieval, which makes the bot a little bit more solidified in terms of what it's able to do. Now, like I said, it's based on the model and the tools that are provided and has the ability to access a broader range of information and capabilities through the retrieval system. Next up, we have tools in tasking AI. This is where it enables AI assistants to interact with external resources and performing specific actions such as fetching live information or communicating with external systems. It's defined in open API schema format, so these tools basically.

Enhance the assistance capability now. Something I want to clarify is that only actions require the provision of open API schema. Whereas plugins on the other hand is something that TaskingAI has preconfigured. Coming in in number fourth we have the retrieval core concept. This is where you're going to be enabling AI systems to access external knowledge bases and this will obviously enhance their ability to provide more accurate and context relevant answers. It's a feature that they've integrated so that additional information can be poured into responses for crucial tasks that require specific and up-to-date information. 

And lastly you have projects. This is where they're organizing units within TaskingAI and it helps for group related activities and resources being merged together. They are providing the structured way to manage different like initiatives as well as brands which will ensure that there's a clear segregation and management of information. 

So you can see that this is a rag chatbot that I'm actually developing and this is a project that I've worked on. So I can simply click on this and I'll be able to work on all these other core components that are part of the project. 

So now let's actually take a look and create a practical AI assistant. I'm going to be demonstrating the practical application of this platform by creating an AI assistant which will showcase different features that we've mentioned. So I'll be actually going through and making an archive assistant which will be able to find archive papers on AI as well as answering questions about recent AI papers or concepts. 

So firstly I'm going to start off by selecting a model. So I'll click on new. I'm going to basically use open AAS model and you can see that there's various options to choose from. But in this particular case I think a simple question and answer bot that is made for answering papers off of archive will be.

Sufficient to provide an open AI API key for one of their models so I can then proceed. Next, I can choose from these different models. I think GPT 3.5 turbo should be more than enough. Going to click confirm. Going to give it a model name Q&A bot or no, I'm just going to leave it open AI gbt 3.5 turbo. 

And then I'm going to input my open AI API key. So once I've done that, I'll be right back. Next up, we want to go over to tools and then we want to create a new plugin. This is where we're going to create an archive search plugin where we're going to then click on this and then we're going to click next. 

So as I talked about, we're going to be creating an archive assistant that will help us find research papers on AI as well as answering questions on those recent AI papers. So we can then see that it will basically demonstrate what it can do. Search for academic papers on archive. It will give you a query which is required and this is the search query. 

And now what we can do is just simply click confirm and we have our tool ready. Now for this particular case, we're not going to be using any sort of retrieval system as this is just going to be something that's going to help us answer questions. But if you want a retrieval system set up so that the AI assistants can access external knowledge bases, you can definitely set this up quite easily. 

Where you can create a new collection, you can name it something, add a description to your collection, and you can input an embedding model. You can have it so that it's going to be able to refer for various use cases and it's going to enhance your ability to provide more accurate and context relevant answers with the app that you create. 

So what we can do now is go over to assistants and create our application. We're going to click on new assistant. We're going to give this a name, basically going to give it an archive Q&A bot. The description we can give it.

Is so that it can find uh relevant archive papers as well as answering questions on AI papers. The language model we're going to choose is the one that we basically connected our API with. We're going to confirm that it's the open AI GPT 3.5 turbo model and then click continue. 

Now you can give it a system prompt. So in this case, this is a good system prompt already, so I'm going to just simply provide it. You can add multiple ones up to 10, so you can have it so that it can be categorizing your answers a little bit more refined. You can add a memory system where the context memory is the assistance ability to retain and use information from previous interactions. 

You can also integrate retrieval, which is something that we're not going to be touching right now, but we have a tool which is the archive plugin. So we're going to select this one and then confirm it. Once we have done that, you can see that it has now basically now that we have selected the archive search, we can then simply just click confirm and then you can then click confirm again. 

And like that, you have your AI chatbot created, which is going to help you with archive question answering. Now that we have created our assistant, I'm going to actually test it out, and you can do this afterwards as well by clicking on the playground button over here, which will take you to this place over here. 

Now, as we have created an assistant, we're going to select this option. But if you are testing out a model, you can go over to chat completion and test it out over there. But since we have created and tested, want to test out this model, we can just select assistant. 

Select our assistant, which is archive Q&A bot, click confirm, create a new chat, and we can now simply just start asking it different things. So since this is an assistant that will help me find relevant AI related papers on archive, I'm gonna ask it to find different.

Trending papers on fine tuning. So here is the prompt that I thought about. Find me the best trending research papers on fine tuning for this year 2024. I'm going to click Send and generate. So let's see if it's going to be able to find different things. 

We can see that the assistant is using the chat completion tool as well as different tools which will help me find the trending papers. So it found me five different research papers. We first have cross lingual learning versus low resource fine tuning a case study with factchecking in Turkish. Now in this case, if we want it so that it only finds relevant information in English, we can actually do this as well. 

So we can ask it. So now that I have wrote this prompt, find me f tuning papers in English only please. I'm going to send it and generate it. So it's actually quite smart. It's going to be going back and finding me more papers that are in English and they are trending. 

You can see that this is a functional tool that will help me find these research papers. It tells me the authors, the summary of the paper. So if it aligns to what I'm looking for, I can then simply go to the research paper. This is by going over to this link. 

So I need to copy this and then click on go to this page and it will take me to the archive page which showcases this research paper which is showcasing me this fine tuning method. So I can view this PDF and I can obviously observe it. 

So this is just a simple Q&A chat bot that I recently just created within a couple of seconds. It didn't take me long. I was able to do it quite frankly within a couple seconds. This is something that you can do and create various things. 

This is a simple assistant that I created. There are various functionalities that you can create with it. So definitely let me know what you actually create. Post it on Twitter as well as on my comment section so I can see the creativity of what you.

Guys are able to do with it. Let's now take a look at how you can interact with the TaskingAI platform through the python SDK. Before heading to the actual code, we're going to need to simply create an API key first. Copy that and then paste it into the code block within Google collab that had prepared earlier. 

Now to start using the SD, the first thing is to install the TaskingAI packages using pip. After that has been initialized, the tasking AI client is going to then showcase the API key that we had just copied, and it should be able to connect to the tasking AI platform. 

Now first we're going to try listing the assistant, and we can see over here that it has listed our archive Q&A bot as well as the description that we had given previously within the TaskingAI platform. Now what we're going to do is try having a round of conversations with it. 

So we first created a new chat session over here, and we can see that you can send a user message to that chat message. So what I'm doing is that I'm going to ask it please find me papers on rag in 2024. Once that user message has been successfully sent by clicking on the initialize button, and within Google collab, it's going to be able to successfully find the archive papers utilizing the archive plug-in to search for the papers and generate that answer for me. 

So it takes a couple of seconds. Once that is done, it's going to then find me the best papers within 2024 on Rag, and we can see over here that it found a couple sources on that. Now for more advanced usage of the S, you should definitely take a look at the Documentation Center at do docs.

TaskingAI cuz you're going to find a more detailed explanation as to how you can use their product and it will basically also keep updating the integrations of models and plugins and list them there as time goes. 

And that's about it for today's video on TaskingAI. I hope you found this video to be very beneficial and you got some sort of idea as to what you can do with it. I'll leave all these links in the description below. 

Make sure you check out their repo, which I'll leave a link to in the description below. Make sure you follow us on Patreon if you haven't already. This is a great way for you to access different subscriptions completely for free. 

For example, we actually provided TaskingAI subscriptions completely for free for our Patreons. So if you're interested, take a look at the link in the description below. If you haven't already, make sure you follow us on Twitter. This is a great way for you to stay up to date with the latest AI news. 

And lastly, make sure you guys share the AI Agent Store with colleagues. Turn on the notification bell, like this video, and check out our previous videos so you can stay up to date with the latest AI news. 

But with that thought, guys, have an amazing day, spread positivity, and I'll see you guys fairly shortly. Peace out, fellas.

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

TaskingAI Video