Introducing Dify's New AI Workflow Update
Remember Dify? It's one of the best open-source large language model app development platforms where you can create various ranges of AI agents that can automate workflows. It's an ideal application that lets you quickly go from prototyping to production.
Now just last week, Dify released a large new update, which is the AI workflow update. This is something that is available both on cloud as well as on open-source versions. It's an update that has revamped by enhancing 100K lines of code where you can firstly develop large language model apps with the workflow, and this is with a simple drag and drop UI. You have the ability to visually debug nodes, you have flexible DSL export and import features where you can share the workflows, you have code nodes where there's Python and JS nodes which makes it easier to integrate programming languages into the application that you create. And then there is the introduction of the toolbox which is allowing you to have a wide range of tools and plugins to enhance the quality of your apps that you create within Dify.
Demo: Generating a React Component
Now just take a look at this example where the Dify workflow is basically creating this React component generator. And this is something where it creates certain components of the UI and this is just a video demonstration as to what you can create with the workflows on Dify.
Here I am going to select React burger menu and the text I'm going to include is "nature is beautiful." The colors I'd like to include should be a vibrant color scheme to match. Now from here I can click the start chat button and since I've already inputted all my variables, I can easily just type in the words "generate me my React component" and see what it comes out with.
Okay so here's the code it's generated for me. I will now test it for you all in React.
Testing the Generated React Component
All we have to do is make sure we're cd'd into the correct directory and as it says at my bottom burger menu, we are all good to go. So I can run npm start and see what it looks like. And here's the finished result. As you can see here, you have home, about, contact. It fits the pink and green gradient color style. It has the leaf icon that corresponds to nature. Really, it's got the animations as well. It's very, very well executed.
This is just a start, and it's something that we're going to be exploring throughout today's video. So with that thought guys, stay tuned and let's get straight into the video.
Sorry for being repetitive, but this month we had insane partnerships with big companies giving out subscriptions to AI tools completely for free. These are tools that will streamline your business's growth and improve your efficiency. Just being a Patreon this past month, you were given access to six paid subscriptions completely for free. Not only do you access these subscriptions, but you gain the ability for consulting, networking, collaborating with the community as well as with myself. You get access to daily AI news, resources, giveaways, and so much more. If you're interested, check out the Patreon link in the description below to gain access to these benefits.
Understanding the AI Workflow Components
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 Dify's new upgrade, which is the AI workflow. This is a tool that's going to help you create step-by-step plans for tasks that involve artificial intelligence, where you can use this easy drag and drop system where you're going to be able to connect different parts together. It's like building a whole Lego set, and this is where you're going to be building with these Lego blocks. And this is where you can work with all these different entities and features.
You have LLM nodes. This is the special block that contains the large language model and it basically uses its abilities for the generation as well as for the backbone of that tool or that application that you're creating. You also have tools where these are going to be like extra gadgets and they basically add more functions to your overall application. You have a question classifier. It automatically sorts out different types of questions or even inputs. You have knowledge retrieval, and this is bringing in extra information from other sources for helping with the decision-making process. You have different variations of code where you can add custom code in Python and Node.js. And you also have if and else blocks, and this is where it basically makes the decision in your plans, such as like if this happens, do that. And it's basically adding different criterias and filters.
Now overall, this is definitely a great tool that's going to make it easier to make complicated plans involving AI, and it's going to break them down into simple steps by arranging them like puzzle pieces, and this is by doing it within this simple drag and drop UI where you can create various ranges of software applications and so much more.
Getting Started: Cloud vs. Local Installation
So how do you get started? Well it's super easy. What you want to do is click on the get started button. This is if you're going to be following along with me on the cloud version. You can also install this locally which is possibly the ideal method for many of us. You can get started by going to the quick start section. You can start this up with Docker and it's super easy to do so. Make sure you just follow through with the command lines that they have basically updated and you're going to be able to install this.
But one cool, one important thing that you want to keep in mind is that you need to have a CPU that is greater than two cores and you should have more than 4 GB of RAM. This should be sufficient to have this operational on your local computer. But say if you do not have the requirements to fulfill that, then you can definitely have it hosted on the cloud, which is super simple. What you want to do is just simply click on the get started button. Once you have done that, create an account with GitHub or with Google. Once you have done that, you will be then sent to the next dashboard page.
Now once you have signed in, you'll be then sent over to this dashboard. You can see that this is the studio dashboard. You also have a knowledge one and a tool set dashboard where you can integrate various other third-party apps and tools where you can basically Implement all of them into the application that you're creating. You can also create your own custom tool and this is where you can give it a name, you can give it a schema, and you can even Implement available tools. You also have an authorization method as well as a privacy policy, and this is obviously for commercial use cases. So in this case if you're going to be building a custom tool you would just need to specify these certain criterias and then you can proceed forward. But in the case that you're going to be utilizing these other Integrations, you can easily get started by just simply clicking on something. For example, we can click on GitHub. Once you have done that, you can then click on search repositories, integrate something within Dify's dashboard and that's how you can quickly get started with all these other Integrations. And that's one of the great things about Dify as this is giving you more flexibility and it's giving you more access to other tools which are like these extra gadgets which is going to give you more functionality.
Now in the case that you're going to be building an application that will refer to the context that you wanted to basically generate upon, you can upload and create your own knowledge. You can do this by connecting an API or you can just create your own knowledge. You can simply click on this, import various file types, you can sync it with Notion as well as have it so that you can sync from the website, which is a feature that will be coming soon. You simply just upload the files, click next. You can easily do this with Notion where you can just connect your Notion account, but in the case you just simply upload a file, which I'll showcase right now.
Creating a Knowledge Base
Now once you have uploaded your file, you can just simply click next and you will be then sent over to the text pre-processing and cleaning. You can just click on the chunk setting which is automatic or you can chunk it with the custom chunk rules. You also have index mode where you can select between these two options. You have high quality as well as economical. You can go to the retrieval settings and configure it to the best possible method as to how you would want the retrieval settings for your data. Once you have done that, once you have uploaded it, you can then simply just go to the document and you're going to be able to reference the document that you have uploaded. And there you go, you have your document or your knowledge base uploaded.
Now if you are to click on the export tab, you're going to be able to see all the other applications that have been developed on the Dify platform on the cloud. You can see that there's various other apps that have been made by Dify themselves or by other community members. You can see you have a code converter, you have meeting minutes, you have a fully SEO optimized article including FAQs, you have website generators and so many other options.
For example, we can just take a look at this, add it to our own workspace and just simply click on create. You can see that there has already been an instruction given to it and we can already go to the overview and we can even deploy this in other areas where you can preview this onto your own website. You can also have it so that it's integrated into your own workflow by embedding it and you can even customize it so that it could be better suited to your own need.
Now in this case we're going to just simply just start chatting with it, where you have created a new conversation. We're going to ask it, "Hello, create me a basic HTML sidebar," and then we'll click enter. And you can see right away that it's going to start generating a code and it'll be able to do this within a couple seconds. Now this is just a small little application that was just developed within Dify and you can see that it is easy as that to utilize any of the explore applications that have already been created.
Creating a Workflow from a Template
Now the great thing with Dify is that you can create chat Bots, different agents, as well as now this new feature which is the workflow. So what we're going to be doing is we're going to be starting from creating from blank, but you can also create from templates. You can see that you can utilize these other templates in the workflow section where you can utilize the workflow planning assistant, the text generation or summarization workflow, the SEO blog generator and so many other options.
Now one thing that looks really enticing to me is the sentiment analysis, which is going to be helping us analyze text followed by JSON output and sentiment classification along with the scores. So this is something that I could actually use. I'm going to be duplicating this template by simply clicking create and you can see over here that within the dashboard you see this UI where you're going to be able to drag and drop different nodes to create your application. This is something that is new with Dify and this is where we mentioned that you can utilize this drag and drop UI to create your apps.
Building a RAG Chatbot from Scratch
Now for example, if you want to test out the application that you have created, you can just simply click on publish and just click on run app and it'll take you to this page where you're going to be able to test it out. You can see that you can run it once or run it with a batch by uploading your files, and you're going to be able to get the sentiment analysis based off the context that you give it.
Now what we're going to be doing is clicking on the workflow tab back in the studio dashboard. We're going to create one from blank. Now what you can do is that you can specify what type of app that you want to create. You can create a simple chatbot, text generator, an agent, or workflow. In this case we're going to be working with a workflow where we can see that you can build an application that generates high quality text based on workflow or orchestration with a high degree of customization. This is suitable for experienced users. So this is the new feature that we have, and you can see that we're going to be able to create this application with this drag and drop system and it's going to be able to connect all these different parts together so that it's going to be able to build this intricate application that is potentially going to be a little bit better than previously as to how we're able to create applications.
So what we're going to be doing is creating this simple app, which is going to be having this knowledge retrieval system that will be able to answer questions on PDFs. So what we're going to do is create this workflow by creating it a RAG chatbot and the description is going to be, we're going to give it knowledge retrievable chatbot. And once that is done, I can just simply click create and we can see over here that we can start off with this starting node. This is where we can add different orchestrations. You can add a question understanding or question classifier, which is one of the tools that has been integrated with workflows, where it automatically categorizes user inputs and this is to root conversations and processes that are powered by the large language model. You can connect it to a large language model, a knowledge base, a knowledge retrieval system. You can end the node, which is something that you would want to have at the end. These two are foundational for any application, and in this case if you want to add other things, you can just simply click on the addition button and you can add other components.
So to save some time I basically went ahead and added a couple nodes. We have a knowledge retrieval node which is going to allow you to query text content related to user questions from the knowledge, which is then being sent over to the large language model. And this is where the LLM, it's going to answer the questions or process any natural language queries, and then it's going to then provide the answer as the end node. So we can actually test this out. Click on publish, you can click on run application and we're going to be able to start this chatbot and we can start it off by answering questions or having it so that it will retrieve information based off the content that we have provided. And this was something that I developed within a couple of minutes. It doesn't take that long, you're able to do it within a couple minutes.
Conclusion and Final Thoughts
You have API access so that you can have it so that it's embedded into other areas or people can access your API. You have it so that you can track the logs and annotations, and you have a basic overview as to what is happening as well as what sort of interactions are being made with your application.
And that's about it for today's video on Dify. That's just a brief overview of this new update. I truly recommend that you take a look at it further in detail because it is something that is quite amazing and it will basically allow you to create quite intricate applications within minutes. So I'll leave all the links as to what I use in today's video in the description below. Make sure you check out the repo. Make sure you check out the video or the links that I have basically pasted in the description below.
And with that thought guys, definitely recommend that you take a look at the Patreon page because this is a great way for you to stay up to date with the latest AI news. And make sure you follow us as well on Twitter. And lastly subscribe, turn on 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, thank you guys so much for watching, have an amazing day, spread positivity and I'll see you guys fairly shortly. Peace out.
