this is amazing. Now we're going to see about GraRag and how you can visualize in Neo4J. At the back end, we'll be using Gro to power the large language model. As you probably know, GraRag is released by Microsoft. This will increase the quality of the response when you chat with the AI compared with the basic rack. So, by the end of this video, you will learn what is behind graph rack, that is the list of prompts, and how the GraRag is generated. Then, how you can integrate that with Gro, how you can set up Neo4J locally. So, Neo4J is to visualize the graphs. As you can see here, this is the Neo4J to visualize the graph, and also we can explore how these entities are related to each other. That's exactly what we're going to see today. Let's get [Music] started. Hi everyone, I'm really excited to show you about GraRag, Neo4J, and Gro. So, when you take a GraRag compared to

the basic rag the gra rag extra entities and other information from the data we provide, and hence the quality of the response is going to be higher. For a basic rack, we are just searching the database with semantic search. So how this entity extraction happens? So the first important prompt which is used is entity extraction. So here's the prompt: given a text document that is potentially relevant to the activity, identify all entities with entity name, entity type, entity description, extract The Entity relationship. Also, it will provide the relationship strength. So it's just the normal prompt sent to the L language model to different entities, their relationships, and their strength. Next important prompt is a community report prompt. So when you see in this image, the first step is to extract entities. The next key extraction is about Community extraction. That means the overall picture of the whole data set which we provide. Without a community, it's more specific. With community, that gives an overall picture. So when we see a community prompt, the goal is to write a comprehensive report of a community given a list of entities that belong to

the community as well as their relationships an optional Associated claims so this is just to extract the overall theme from the data provided so here we're getting the title summary how this community is impacting the overall data and the rating explanation and detailed findings fight to 10 key insights about the community so technically the data is grouped together and each group is given a topic and how this topic or the category impacts other topic or other category is used as the key when the retrieval process happens next I'm going to take you through step by step on how you integrate that with Gro set up near 4J locally and finally visualize but before that I regularly create videos in regards to Artificial Intelligence on my YouTube channel so do subscribe and click the Bell icon to stay tuned make sure you click the like button so this video can be helpful for many others like here I've already covered basics in regards to gra Frack which I will link that in the description below as a quick summary first pip install gra Rag and then click

enter after this export your GRA RAG API key. So this is your GRA API key. You can generate this from GRA website. So if you don't know about GRA, GRA provides multiple models with high inference speed. So here we are going to use MixL because it contains 32,000 tokens. So after exporting the API key, just type python hym GRA RAG.index in it root and the dot dot means it's the current folder and then click enter. This will automatically create the settings.yaml file where you can set up the model name. That's what we saw just now. Then we need to set up GRA API base, then the tokens per minute, requests per minute, max retries, and max retry weight. We might need to change this to lower numbers based on the ti you are in. So ideally, it's better to keep it low because this will automatically generate multiple tokens. So the next thing is that when you see embeddings, GRA doesn't provide any embeddings model. So in that case, we are going to use OpenAI embedding model.

here I'm going to change this to OpenAI API key. Then going to modify a few other things, such as Community reports. Let's say 1,000 and maximum input is 4,000, just to keep this tutorial quick. And that's it. Now in your terminal, export your OpenAI API key like this and then click enter. This is to use the embedding model. You can even use embedding model from other providers, even using LM Studio that you can run that locally and for free. Next, I'm going to create a folder called input. And in this folder, I'm going to place a file called book.txt. Here I got a book titled A Christmas Carol by Charles Dickens. And we are going to run GraRAG over this book. After this, in our terminal, GraRAG index and the root folder is the current folder. And then click enter. Now this will automatically create the required index, extract all the key entities, and save it in a folder. Yeah, now it's all done. Now I can ask questions like this with query method local. And who is this?

person and what are his main relationships. And I'm going to get the response. And this is local search, which means we are not using any community-generated report. Now, step number three is to set up Neo4J locally. To do that, you have multiple OS versions supported, such as Linux, macOS, or Windows. I'm going to use macOS. First, downloading the Neo4J, the latest version Community version, using this command. I'll put all the information on the code in the description below. After this, click enter. Next, I'm going to extract the zip folder like this and then click enter. Now, it got extracted. Now, navigating into that folder. Inside that, I'm going to type bin Neo4J start and then click enter. This will automatically start the Neo4J application and it will automatically take you to this page and you can see that it is running in this URL. Now, it might ask you to connect. So, to connect, you might need to enter the username and password. So, both username and password is Neo4J and you are able

to set your own password. After that, next we need to import data into this Neo4J. So that is the final step to visualize. To do that, if you see the output folder, there you go, the artifacts, and these are the files. So these files need to be converted to CSV. So if I open an example file, you can see this is how it's going to look like. So I've created a script which goes into the artifacts folder and converts that to CSV file inside Neo4J folder. Now I'm going to run this in our terminal python convert to CSV and that will automatically convert those files. And if I go back to the folder structure in the import folder, you will see all the files there. And if I open, you can see those content. Now we need to import that data into Neo4J. So at the top here is a terminal, I'm going to paste a command. So this will point to the import folder and the file. And after that, it's going to create a document. Similarly, I'm importing all the files as you can see.

here, then I'm going to click the play button on the right-hand side. Side. Then you can see it's automatically importing, and it seems like the index already exists. That's because I've already run this. So I'm going to modify the code by going here, coming down, just deleting these index, and then running it again. Now you can see everything got imported, and it's finished. Now, we can query the data. To query the data, I got a few commands here. So these are Neo4J specific, so you might need to understand this syntax to run this. But even if you want to, you can use ChatGBT or Clot to generate these commands. So after entering, clicking the play button, and you can see run, you can see, for few completed after 70 milliseconds. It's better to run these commands separately. For example, I'm using this command and then clicking enter, and you can see the graph structure here. Similarly, by specifying the command, you can clearly see how these are linked with each other. You can also see the node labels here. For example, if I choose entity, making it

bigger than you can see. How the entities are linked with each other. This is just the basics. I haven't really covered on Advanced queries in regards to newor 4J. Which I will leave that in your hands. So get started with this Basics and explore further. I'm really excited about this. I'm going to create more videos similar to this. So stay tuned. I hope you like this video. Do like, share, and subscribe. And thanks for watching.

Easy GraphRAG with Neo4j Visualisation Locally

Easy GraphRAG with Neo4j Visualisation Locally

스크립트