Building No-code Agents with Tasking AI: A Quick Guide

In today’s rapidly evolving tech landscape, the emergence of no-code platforms has democratised access to advanced technologies, including artificial intelligence (AI). Tasking AI is at the forefront of this revolution, offering a robust no-code framework that enables individuals and businesses to build AI agents effortlessly. This comprehensive guide will walk you through the essentials of creating your first AI agent with Tasking AI, leveraging its powerful features and user-friendly interface.

Introduction to Tasking AI

Tasking AI is an innovative no-code platform designed to facilitate the creation of AI agents. These agents can perform a variety of tasks by coordinating with each other, ultimately streamlining processes and enhancing productivity. As an open-source tool, Tasking AI is accessible to everyone, allowing users to either use the platform directly online or host it locally for more control and customization.

Getting Started with Tasking AI

Signing Up and Initial Setup

To begin using Tasking AI, you first need to sign up on their platform. This can be quickly done through a streamlined signup process, including options like signing up with Google for convenience. Once registered, users are directed to the home screen, which bears a resemblance to the interface of Autogen Studio, but with enhanced functionality and user experience.

Creating Your First Project

The journey starts with creating a new project. You can easily name your project (e.g., Newsletter Project), add a brief description, and proceed to explore hosting options, which promise effortless scalability. This initial setup is crucial as it lays the foundation for your specific AI agent-based tasks.

Building Your AI Agent

Understanding Agents and Tools

AI agents in Tasking AI act like virtual managers, each designed to handle specific tasks ranging from content writing to complex data analysis. These tasks are accomplished through "tools" within the platform, which include actions (API calls) and plugins (predefined functionalities).

Adding Plugins and Actions

Adding a plugin is straightforward; Tasking AI provides a range of pre-built functionalities, from web scrapers to content summarizers. For instance, if you need to summarize a URL, simply select the appropriate plugin, configure it with your specifications, and it’s ready to be used by your agent.

Configuring Agents

Once your tools are defined, the next step is to bind them with an agent. This process involves specifying what tasks the agent is responsible for and configuring it to use the selected tools effectively. During this stage, you can set parameters like memory (how much past interaction the agent should recall) and retrieval settings if your application requires real-time data fetching.

Experimenting in the Playground

Testing and Iterations

Tasking AI features a "Playground" where users can test their agents live. This environment allows for real-time interaction with the agent, enabling you to fine-tune its settings and performance based on immediate feedback. For example, you might input a URL to see how well your agent summarizes content or generates social media posts from it.

Enhancements and Advanced Features

The Playground also proves invaluable for iterating on your agent’s capabilities. Whether you’re adjusting the memory settings, switching models (like moving from GPT-3 to GPT-4 for enhanced performance), or adding new tasks, this interactive platform ensures that your agent evolves in line with your operational needs.

Deploying and Scaling Your Agents

With your AI agents configured and tested, the final step involves deployment and scaling. Tasking AI supports various deployment options, including hosting on your local server or using Tasking AI’s cloud service for hassle-free scalability. This flexibility ensures that as your business grows, your AI agents can scale accordingly without skipping a beat.

Conclusion: Harnessing the Power of No-code AI

Tasking AI stands out as a potent tool for anyone looking to integrate AI into their operations without delving into the complexities of coding. By providing an intuitive, scalable, and robust platform, Tasking AI empowers businesses and individuals to automate tasks, enhance productivity, and drive innovation. Whether you are a seasoned tech enthusiast or a business professional seeking to leverage AI, Tasking AI equips you with the tools you need to succeed in the digital age. Try building an AI agent today and experience firsthand the transformative potential of Tasking AI.

[h3]Watch this video for the full details:[/h3]


In this video, we’ll look at TaskingAI’s revolutionary approach to AI-native app development. Get our 2024 AI Course: https://www.udemy.com/course/2024-generative-ai-masterclass-beginners-to-expert/?referralCode=65445B91BF0DB591B523

💪Always ahead in AI
Newsletter: https://explainx.substack.com/
Discord: https://discord.gg/TkBBwXHYHU
Twitter: https://twitter.com/goyashy
Subscribe: https://bit.ly/3NayPyC
Instagram: https://www.instagram.com/goyashy.ai/

TaskingAI is setting new standards in AI-native app development by offering a platform that combines the simplicity of Firebase with the power of large language models (LLMs). With TaskingAI, developers can build responsive assistants supported by stateful APIs, manage memory, and integrate tools for augmented generation systems. The platform’s user-friendly UI and framework cater to all developers, ensuring efficient and flexible LLM application development.

The full-stack TaskingAI is now open-sourced, incorporating Docker for straightforward deployment. This move towards open accessibility allows for unlimited collaboration and adaptability within the open-source community. TaskingAI’s roadmap includes strategic plans for open-source expansion and feature-rich enhancements, ensuring continuous integration of new features and innovations shaped by community feedback.

TaskingAI excels with its API-centric architecture, offering a powerful assistant API, vector-based retrieval system, and comprehensive LLMs integration solutions. The platform enables autonomous decision-making and augmented generation, providing updated and reliable outputs. With TaskingAI, developers can drive their AI projects with innovation, not complexity, leveraging the platform’s edge over others in scalability, collaboration, and enhanced privacy.

#generativeai #aidevelopment #opensourceai

🐱‍🏍About Me
Experience the Power of Generative AI, Startups, and No-Code Tools. Our goal is to equip you with the tools and knowledge you need to thrive in your industry and become a confident, competent product manager. With our cutting-edge platform, you’ll have access to the latest in generative AI technology, insights from top startups, and the best no-code tools. Whether you’re just starting out or looking to take your skills to the next level, we have everything you need to succeed. Subscribe now!

[h3]Transcript[/h3]
so in this video I’m going to be working you through tasking AI which is a agent framework in order to build AI agents but it’s no good now I’m assuming you already know what AI agents are so effectively to give you a summary they can take in a task do a bunch of things by coordinating with each other and then give you the output so tasking a is a new open source tool so you can find their entire code here you can see this is the entire platform so you can also run it locally and host it and run it yourself without going on the website but effectively this framework allows you to create agent s to autogen studio so the UI looks similar to autogen Studio but it does work as opposed to autogen Studio which I tried a few weeks back and it was just a nightmare right so let’s start building our first agent which will take in a URL and then summarize that URL for us right so in order to use the platform just go ahead and sign up click on get started and then you’ll go to the O signup screen going to be signing it up with Google so once you sign up you should you know land on something called home and this UI should look similar to the autogen now here we can create new project let’s say newsletter project and then we can add a sample description then you have the option for hosting it which is experience effortless scale with tasking a surve hosting I don’t know what this is so I’m just going to go ahead and proceed let’s click on newsletter project and then this is where you will start building agent so now this looks familiar to the autogen studio because all of these things you’ll also find there to give you a recap of what it is so agents are like your managers that do tasks right so think of an agent like marketing manager who can do SEO research content writing and so on and all of these things that it does are its tasks which we will add in tools so let’s say tools will have two things actions and plugins actions is where you can call the API plugins is where you can use predefined set of plugins they have on the platform I guess let me quickly add a plugin and you can see there are a bunch of things that it can already do crew AI is a great framework to do this uh because there while you can see all of this in the UI cre has automatically created what you call tools for this right where you can programmatically mention and I have a complete master class on creating a completely free agent using your own local llm using cre if you can if you want you can check the master class out for now let’s focus on the tasking AI plug-in feature and let’s build a new summarizer or a URL summarizer for us you can also choose models by the way where you can see you can you have a bunch of options here you have anthropic open Google Gemini and so on going to be using open gp4 so let’s click here this should hopefully open it mod is okay I need to click next and then I’m going to be using GPT 4 Turbo then let’s click on confirm and then here you have to put the model details so G4 turbo and then we’ll add our open API key so in order to get your keys go to platform. open.com and then create a new key I’m going to be calling it tasking AI test so let’s create this key copy this go back to console add it here and then click on confirm so the model is now created we can go ahead and create the tasks before we create our agent we need to create a task that we can assign to the agent or a tool that we can assign to the agent so in our scenario because we want to create a very simple agent that can take a URL and summarize it for us click on new plugin and then you can see there are a bunch of different ones here there is news API there is STP advisor and so on but we just need the web reader that will we read the video from web we’ll also try creating you know a transcript of our YouTube videos but for now let’s just create a web reader and then you can see we’ll need to get a URL from the user and then which will be a string and then that will be processed so we will create the web reader here and now a task or an action or a tool is created now we need to create an assistant who will then use this tool in order to do that task so let’s say new assistant and then say web summarizer and you can see things like you are a website summarizer go to the following get the URL from the user and go to the website and summarize the website for us so very basic prompt for model we’ll use the model or model that we created gp4 turbo click on confirm and then there are it seems that you can also put in a system prompt going to be keeping this empty and then this is the memory meaning how much of the previous shed do you want the model to remember now in our scenario I don’t because I just want to put the news URL and then get the summary so you can have message window which means it can remember up to 20 messages you can have zero which means it will remember nothing and you can have naive which means it will still remember some things now the retrieval is for rag use cases and by the way talking about rag I published a complete master class on rag last week so if you’re interested go ahead and check that out basically you can add the links to the website I guess collections is something that we I think have to create somewhere but you can basically create the collection and which will contain a bunch of websites and we’ll also try creating one rag based chatboard but for now let’s focus on the tool ones and let’s assign a tool to this agent which is the web reader tool so let’s click on this and then read the web page task then click on confirm the UI is kind of unintuitive because I have to scroll down anyway this is just click on confirm and this should create the agent for us now you can to use this agent you have to go to playground so you can either go from here or you can click here and then goes to the playground here and where you can start a new chat again similar to the autogen UI and now when we um I don’t know if you’re using the right assistant oh yeah so this is the assistant that we created and it’s using that let’s open up a website let’s go to explan x. and put this URL in there so I’m going to be adding this URL and hopefully it should take this URL and summarize it for us you can see it’s summarizing the website without any further detail about the website now you can customize the agent to summarize the website or create say tweets for those website or highlight important points of those website and so on so what we can do is you can go back to assistants and then let’s say you are a social media content generator accept URL of from the user and write or tweet based that right so now if we put this we go to playground and then start a new chat with the same assistant then I put the URL now this should hopefully automatically generate tweets for us or did be a LinkedIn post is it still summarizing the chat or let’s see if we Chang this it seems we changed it but might not have taken effect let’s go from here and then let’s start a new chat then put the link again it’s still summarizing the we site I guess it’s because of the name going to go back click here and let’s say the Tweet right so now hopefully it should be able to take the context I don’t know if name really has that impact let’s start a name chat and then put this in again for all of this that you’re running right now you are consuming your opening API key so ensure that you have enough credits or don’t spend if you don’t want to spend money running all of this again it seems like it summarized it but it’s not really you know doing what we told it to which is right tweet so let’s start a new chat okay it seems that we have to click here let’s see if that changes anything we put the link now we were chatting or starting the chat from here this time I clicked here to start the chat and to summarizing the content for us anyways I think it’s cast it somewhere and now it’s sending the same query so again it seems like a early development of the product it’s still not developed but you get the point right so it’s an agent that automatically take certain details and do something for you based on the action that you define in the background now let’s start generating something else so let’s try doing something else where we you know you can also generate the images or you can summarize now let’s create blob posts right or blog images when we put the link to the blog post we want the to generate the image for us so we use Del 3 and then confirm this then we have to put open AI API key again copy this and this is because Del it’s not going to use the direct model and this is a tool what we set up is a model for text not for images so and if you want to do image generation you’ll have to set up the API key separately anyway so now that the image is set up let’s set up a new assistant and blog image generator your job is to accept a URL and based on the text generate a Blog IM Mage for that blog in landscape format cool so now we use the model which is gbd4 Turbo and then we don’t want anything to be retained you can also put this in the system prompt which will clarify this further so you can add this as the system prompt here and then inside of the plugin you can add d three and confirm this and then confirm this now that this is generated let’s try it out in the playround so we click on new chat and then we type we look for a blog post so when we now go and put the blog link here and we click on send and generate this should hopefully generate an image for us I me these are our basic use cases of you know agents in a real world scenario where we can also use actions post the blog or the images generated we should be able to also push that to say some other platform let’s say in a a newsletter or something on those lines right and you can see the imagees generat it’s rather sharing a link with us going to this and then you can see the image is ready and it was able to take this blog post about Jamba which is the new model apparently and then based on this it was able to generate this image for us so really interesting stuff let’s try something relatively more advanced now so let’s go go back to uh the retrieval here or sorry let’s go back to tools here and then see what we can create in action not PL so in action is similar to you what what you would see on the assistants or gbd custom gbd assistant cbii where you would typically mention something like for example get data from some other website and based on what you receive from that website respond to the user so let’s say if you are a weather platform and you want to give out customized notes to your end users based on how the weather is in that area now the weather changes everything and I mean the weather changes every day and the model will not be up to date with the new weather data so what you can do is use action where every time before generating the response to the model or response by the model the ABI or action will be first run to get the realtime weather data for that region and then post that the blog post or note will be generated so again don’t think I want to dive deeper in this because this is very simple and I’ve used this before rather let’s try something else let’s try summarizing the YouTube video using this and then confirm it or we need the API key so it’s not going to get the python library that can currently get videos without the API key so I’m not going to try this because this is going to take a very long time let’s see what else we can do let’s try the retrieval one and in retrieval one we’ll basically have to mention a bunch of webites before we respond so apparently you can also choose an embedding API so we can choose say open a embeddings embeddings are basically small pieces of your text so let’s say if your web page is 10 10,000 pages long you will have some sort of rule and break down your 10,000 pages into different smaller chunks and embeddings will tell the agent which text is available inside of which embedding it’s some probability based method let’s say we use these small because I haven’t really used embeddings of open let’s call this let’s add our open API key and then let’s call it ragp basically we want to put in a bunch of links that we can use to get data in real time so we selected this and then this like I said this is the number of chunks this can break down the text into and you can call this when AI because we using retrieval we will have to mention where to retrieve this data strong and you can see we have to add a record and the text content here okay so you can add a record or you can also add the chunk just going to copy this text so right now if you ask chat gbt about this model it will not be able to respond because this does not have the output for that model so I’m going to be copying this text and I’m going to be creating a new record or you can also directly add the website so let’s say we take xen x./ log I don’t think it’s going to uh scroll through each pages each page but let me just take the explan X AI here and then you can break it down in chunk and you can call it explain its source so now every time before it responds it will look to our website before it responds so we’ve already added the website here let’s close this and our retrieval is now ready right so our record is ready I don’t know if yeah it was able to break it down in three separate chunks the records show zero so okay we have now three records and three one record and three CH you can if you want more things right so let’s say we want that Jamba blog we will copy this go back to in here and add a new website inside this we’ll say Jamba explain X right so this will basically help you get data that does not exist in your database and in scenarios where there are more than one websites you can collect in real time you can you know generate a Sit map like this so let me show you so if you go to any website and do sitemap.xml you can see it’s going to generate the links for that website right and in in scenarios where you can let’s say you want the site app of this then you can copy this and this this should hopefully give you link of all blogs now you can see these are all the explan X blogs what we can do is copy all of this text puted inside gbd and extract all these links or write some code to extract all these links and then add all those links here in the tasking a console to keep it simple I’m just going to say this what this will do is before it responds to you it’s going to look for the data inside of this link so now let’s create a tool I think oh no I don’t think we need to create the tool so we need to create an assistant but this time we’ll use the retrieval and we’ve created open embedding collection so let’s confirm that and then let’s say explain xbot language model we use the gbd4 we earlier created and then context wise we need to have some some window we’re not going to be using any plugins and then you have an option here where you can do function calling which means you can run code you can use the memory or use user message so we want to use user message here and then click on confirm and you can see the chatboard is created now if you go here and start a new chat and ask things like what is chamba let’s first also go to GPT 3.5 and say CH and it’s it should not be able to give you the answer because it’s not updated I don’t know if gbt 4 will be able to give you the reason why I tried gbt 3.5 is when you use embeddings you would typically use it with gbd 3.5 because it’s cheaper gbd4 is more expensive blard already have has access to you know web browsing so it should be able to look it up but regardless let’s see if yeah it was able to give us the answer which basically means that it can access web so it can answer questions for you but gbd 3.5 can’t and in those scenarios having a retrieval sort of a chatboard can go a long way so you know what let’s create a new model but this time we’ll use gbt 3.5 so we’ll use gbd 3.5 turbo and then confirm this gbd 3.5 Turbo and then you put in the API key which we’ll get from here the reason why I’m using same API key because I can delete this all together and then I don’t have to worry about you know deleting multiple Keys later on the account won’t be able to access my data basically right and then now let’s create another assistant because we have to select the model again in gbd 3.5 model you notice that the earlier one we tried wasn’t able to answer us the question so now hopefully it should be able to answer questions like what is Jamba right let’s keep naive as memory and then confirm this and I’ve not named it but anyways let’s go ahead and use it what it is cha so hopefully it should be able to now answer the question by referring the blog post and you can see it is able to answer right so again it says it’s a model by xai with an impressive 3.14 billion parameters so the data is wrong the model is by2 z let me quickly take a look yeah this this data is incorrect this is correct I think this initial part is incorrect but anyways I think really interesting let me look up chamba Ai and this is by 21 Labs cool I think interesting rack based chatboard now you can also I think use this as an API where you can directly integrate this agent on your website and then let your users chat with your data in real time so really interesting the fact that it makes it easy for us to run these agents is very interesting now autogen was also interesting but it was very hard to run but this seems like an easy way to do this and if you folks have any questions let me know this is a very simple tool to use and it’s open source so if you if you know how to code you can actually go this locally and run it on your own computer without even logging somewhere right so cool I think that’s all for the video guys thank you so much