The goal was to create a version that can run without GPUs using only API services to search over PDF
The paper is available at: https://arxiv.org/pdf/2405.14831
The authors' implementation is available at: https://github.com/OSU-NLP-Group/HippoRAG
- OpenAI: GPT-3.5-turbo-1106 for (LLM)
- OpenAI: text-embedding-small 256 (embedding)
- Neo4j (auraDB): GraphDB / Vector DB
- Langchain