Recently, big model technologies have been changing rapidly, leading to a proliferation of R&D projects related to them. One of the most popular technologies is RAG (Retrieval Augmented Generation). Today, I would like to share with you an excellent project:GraphRag. Out of my keen interest in the technology, I decided to implement one myself using the .Net framework, and documented the entire process to share with you.
What is GraphRag?
GraphRagIt is based on the further development of RAG technology, combining the traditional RAG Q&A system with knowledge graph, which makes the Q&A system more efficient and accurate in dealing with complex problems. In this way, we are not only able to utilize powerful big models for information retrieval and generation, but also further enhance the accuracy and contextual understanding of information with the help of the relational network of knowledge graph.
Project Development History
After two days of work, I have implemented a preliminary version and hosted the code on Github. Interested parties can head over to view and learn from the code, and are welcome to exchange ideas and technical thoughts with me.
/xuzeyu91/
Here, let's take a look at the effect demo together:
/Wycqtur
How to run the project
Installing the .net8 SDK
To run the project, first pull the code from the Github repository:
git clone /xuzeyu91/
Then configurefile, as shown below:
"OpenAI": {
"Key": "",
"Endpoint": "",
"ChatModel": "",
"EmbeddingModel": ""
}
Make sure you have filled out the OpenAI information correctly. Next, start the project:
dotnet run
Project Features
In this project, we can use text import to build theGraphRagof the knowledge graph. Traditional RAG systems can only simply perform questions and answers, while theGraphRagBy transforming textual information into a knowledge graph, the relationships between individual pieces of data can be better understood to provide smarter responses. In this way, information retrieval does not only rely on text matching, but also allows for deeper analysis through knowledge nodes and relational networks.
Project plan and future outlook
The project is still in its early stages, and I will continue to work on feature and performance optimizations and updates. Here are the plans for the next few directions:
-
Enhance knowledge graph construction capabilities: optimize data processing and graph generation algorithms to improve the accuracy and relevance of the generated graphs.
-
Extending the intelligence of the Q&A system: enabling the Q&A system to handle more complex and diverse questions by introducing more semantic analysis techniques.
-
Optimize user interaction experience: Improve the friendliness of the interface and the smoothness of user interaction to build a more easy-to-use system.
Invitation to join our developer community
The progress of technology cannot be achieved without the joint efforts of the community. I am very much looking forward to discussing, researching and exploring GraphRag related technologies with you, and welcome you to join our exchange group to share your experience together. You can get the latest project progress and technology sharing at the first time by following my WeChat public number.
concluding remarks
Building a .Net-based from scratchGraphRagThe project is challenging but also very interesting. By practicing, you can not only deepen your understanding of RAG and Knowledge Graph technology, but also learn and grow in the project. If you are interested in the project, please feel free to visit my Github repository to check out the code, submit an issue or pull request, and let's work together to improve the project. Thank you for your attention and support!
Let's look forward to it.GraphRagContinuous iteration and progress of the project!