Chat and RAG with Tabular Databases Using Knowledge Graph and LLM Agents

Рет қаралды 6,891

Күн бұрын

In this video, together we will go through all the steps to construct a #knowledgegraph from Tabular Datasets and design a ChatBot APP to interact with the Knowledge Graph using natural language. For this purpose, we will use Knowledge Graph LLM agents and the GPT model. We will design a Chatbot that can:
1. Chat with Graph DB using an improved LLM agent
2. Chat with Graph DB using a simple LLM agent
3. RAG with Graph DB
Moreover, in this video, I will show you the second RAG approach for interacting with Tabular data but this time, using the knowledge graph. The code is available on the Github repository.
🚀 GitHub Repositories:
Advanced Q&A and RAG series: github.com/Farzad-R/Advanced-...
LLM-Zero-To-Hundred Series: github.com/Farzad-R/LLM-Zero-...
00:00:00 Intro - (Presentation)
00:00:17 Table of Contents - (Presentation)
00:01:55 Why Knowledge Graph? - (Presentation)
09:35 Project schema walk-through - (Presentation)
00:06:14 LLM Model Matters - (Presentation)
00:07:26 Series schema (RAG vs Q&A) - (Presentation)
00:08:05 Knowledge Graph Fundamentals - (Presentation)
00:10:29 How to Construct Knowledge Graph - (Presentation)
00:14:12 ChatBot Schema walk-through - (Presentation)
00:16:02 Knowledge Graph Agent schema walk-through - (Presentation)
00:18:00 Second RAG approach for tabular data - (Presentation)
00:18:24 Knowledge Graph for Movie Dataset - (Presentation)
00:21:41 Knowledge Graph for Microsoft medical chatbot - (Presentation)
00:22:52 ChatBot demo
00:23:36 Graph database installation and configuration
00:32:27 Code structure walk-through
00:33:25 Verify your OpenAI and Neo4j connection
00:34:39 Download the Movide dataset and generate synthetic data
00:37:15 Construct the knowledge graph from the Movie dataset
00:45:50 Creating and populating the Vector Index in Graph Database
00:51:23 Q&A with GraphDB populated with Knowledge Graph of the Tabular Data (designing the simple and improved agent)
01:07:47 RAG with GraphDB
01:13:22 Testing the ChatBot
01:17:10 Microsoft Medical Chatbot walk-through
01:22:52 Ending notes
Frameworks: #langchain , #openai, gradio, #neo4j,
#chatbot #rag #llm #agent #python #gpt

Пікірлер: 34

@alexanderroodt5052 28 күн бұрын

A++ video. Very informative and detailed.

@airoundtable 28 күн бұрын

Thanks! I am glad you liked the video

@TooyAshy-100 27 күн бұрын

Thank you to Farzad-R, for providing such insightful content on the latest advancements in AI, including RAG using Knowledge Graphs and LLM Agents; it has been incredibly informative and inspiring!

@airoundtable 27 күн бұрын

Thanks for the kind words! I'm glad the content was helpful

@thedatascientist-lg4ls 24 күн бұрын

You are super!!!. No 15 mins BS others call projects. Thank you so much.

@airoundtable 24 күн бұрын

Appreciate it! I am glad you liked the video. Thanks for watching

@DefensorVieira 15 күн бұрын

you have no idea how useful this tutorial is...thank you very much.

@airoundtable 15 күн бұрын

Thanks, I am glad it helped!

@sumitpawar000 2 күн бұрын

Solid content as usual 🙂🚀

@airoundtable 2 күн бұрын

Thanks!

@ChathurangaBW 26 күн бұрын

Excellent video. thank you so much!

@airoundtable 26 күн бұрын

Thanks! Glad it was helpful!

@sachinrajora6753 24 күн бұрын

Great video! Totally worth watching 💯

@airoundtable 24 күн бұрын

Thanks! I am glad you liked it

@awakenwithoutcoffee 27 күн бұрын

appreciate it brother. Looking forward learning more together.

@airoundtable 27 күн бұрын

Thanks!

@wassfila 24 күн бұрын

really great content, useful, focused, original !

@airoundtable 24 күн бұрын

Thanks! I am glad you liked the video and the content

@fionalau2920 12 күн бұрын

This video is really helpful to those who are stuck with RAG and tabular data. Quick question, when would you use Graph agent, and when would you use SQL agent? And what if you do if you have a mix of text and tabular data?

@airoundtable 12 күн бұрын

Thanks! The main difference is that a SQL LLM agent is good for querying databases. However, if you are planning to extract specific information and details from a series of data and you are looking to connect some data points to create more meaningful data paths, knowledge graph (KG) is the way to go. KG is for more specialized use-cases imo.

@just-another-man 25 күн бұрын

Thank you for the thorough course. You saved me a lot of time and effort to start working with Neojs. Here's an idea for improvement: Some questions may require thinking step by step, and for each step you may need to make a query to retrieve information from the database. Although there may be one question, and to answer it you may need to make several queries to the database, including, in some cases, queries to a vector database and in some to a graph database. You might consider using function calling and further improving the prompt

@airoundtable 25 күн бұрын

Thanks for watching! I am glad the video was helpful. You are right and that is a very good point. As I mentioned in the video all these agents can be combined and used in bigger systems together to solve more complex problems as the one that you mentioned. Thanks for the insight!

@ChathurangaBW 26 күн бұрын

keep it up !

@mohsenghafari7652 8 күн бұрын

thanks farzad.

@airoundtable 7 күн бұрын

Thanks Mohsen!

@thedatascientist-lg4ls 9 күн бұрын

How about connecting to an existing database rather than creating one as you showed? And also, what if the existing database contains both numeric values and letters. For example, what is the status of the customer order? and How many orders did the customer requests? What framework is best to use?

@airoundtable 9 күн бұрын

It depends on the structure of the data. If it is a tabular data or sql data the best way to interact with it is using sql agents. I have a detailed video for those agents. If for getting the answer, the knowledge among multiple databases need to be used together, then graph agents can be the better choice. And finally, graph agents only work with graph database. In case you have other databases such as SQL, I recommend you to watch my other video that is focused on those databases. "Chat with tabular data using sql agents"

@ahmed_hefnawy 26 күн бұрын

As usual, a useful and powerful video.. This is exactly what we need - I have a suggestion to make a detailed video about: Chat With Document with Knowledge graph data base | Converting Document to KG and query to cypher. - personally I need it, because I'm working on my Master project in critical data that Doesn't accept any LLM hallucination, in addition to RAG retrieving limitations ... kindly keep in that in your mind, and I'm waiting for that. Thanks for your effort :)

@airoundtable 26 күн бұрын

Thanks! Happy to hear the content was useful. That is indeed the subject of the next video (RAG with knowledge Graph on PDF and text files). I am writing the code for it. Just to mention, it won't be focused on Q&A as I described how we can perform Q&A with documents using knowledge graph through the Microsoft project that I explained in this video. The next video will be focused on RAG so it will have some uncertainty at the end due to the interinsic characteristics of RAG. But if you want your project to be accurate, I recommend a similiar approach as the Microsoft project on text data.

@drm2005 10 күн бұрын

Can we use the Grok api or the performance will not be the same with open ai ?

@airoundtable 10 күн бұрын

i never tested Grok. I am not sure which one would perform better (I have a sense that GPT4 would perform better). But please feel free to test and check the results