AquiLLM: Open-Source RAG-LLM Framework for Tacit Knowledge Capture in Research Groups

A recent publication on arXiv (2608.08883) presents AquiLLM, an open-source, modular framework that integrates retrieval-augmented generation (RAG) with large language models (LLMs) aimed at facilitating tacit knowledge capture within research teams. This initiative addresses issues related to transparency, reproducibility, and privacy associated with proprietary AI systems. AquiLLM employs open-weight models to uphold scientific integrity. The framework boasts enhancements such as local embedding and reranking, multimodal features, interfaces compatible with OpenAI, improved user interfaces, as well as semantic and episodic memory capabilities. Insights from experts, including astrophysicists and environmental scientists, guided the development of these features, advancing AI systems for better scientific collaboration and knowledge management.