The architecture of RAG for knowledge-graph-based semantic search
As established, we will build a graph-based RAG program in this chapter. The graph will enable us to visually map out the relationships between the documents of a RAG dataset. It can be created automatically with LlamaIndex, as we will do in the Pipeline 3: Knowledge graph index-based RAG section of this chapter. The program in this chapter will be designed for any Wikipedia topic, as illustrated in the following figure:
Figure 7.1: From a Wikipedia topic to interacting with a graph-based vector store index
We will first implement a marketing agency for which a knowledge graph can visually map out the complex relationships between different marketing concepts. Then, you can go back and explore any topic you wish once you understand the process. In simpler words, we will implement the three pipelines seamlessly to:
- Select a Wikipedia topic related to marketing. Then, you can run the process with...
