An efficient graph-vector database powered by ArcadeDB for tree-of-thought and graph-of-trought knowledge representation and agentic reasoning.
Agentic LLM inference with a graphRAG. Using ArcadeDB as the DB for knowledge representation.
11/5 Meeting with Dr. Clark Overview Graph RAG as a Retrieval system
Build graph with embeddings as nodes
Representation of nodes (embedding) and edges(chucks? Topics? Other metadata or related files?)
During inference: get embedding + other metadata related. MORE context rich, have things related to the embedding as well. (LLM has the potential to crawl the data to achieve better results)
Other: Try out different graph rag systems, maybe. Deepseek OCR as an encoder for document level or page level vector embedding.