hey peeps is there a good resource for

The knowledge graph building in Llama-Index is pretty interesting.

Currently, when you make a knowledge graph index, it asks the LLM to extract triplets for each text chunk. The triplets have the format (subject, relationship, object). There is also the option to customize the prompt that extracts the triplets.

Then, at query time, the library extracts keywords from the query and uses those to find triplets where the keyword is the same as the subject. If you set include_text=True/False, it will or will not also include the text where the triplet was found to help answer the query

Alternatively, you can augment this process to use embeddings as well, by setting include_embeddings=True when constructing the index

The example notebook covers this all pretty well: https://github.com/jerryjliu/gpt_index/blob/main/examples/knowledge_graph/KnowledgeGraphDemo.ipynb

Thank you Logan, I appreciate your response! Looking forward to diving into the notebook.