Advancing Scientific Workflows: A Human-LLM Note-Taking System with Case-Based Reasoning

Human-centered artificial intelligence (AI) systems are most effective when they foster collaboration with users, enabling iterative problem solving and creativity. However, many existing AI solutions operate as opaque, autonomous systems, limiting opportunities for human engagement, transparency, and refinement. This lack of integration between AI and human expertise often results in barriers to trust and hinders the development of innovative solutions, particularly in complex, data-intensive domains like scientific research. Here we introduce a novel human-LLM system built on the Obsidian note-taking application, designed to integrate large language models (LLMs) into a transparent, interactive, and collaborative framework. Key features of this system include case-based reasoning (CBR) and language-aware tools, with cases and tools represented as first-class notes. A Python executive program coordinates user interactions, while advanced search capabilities powered by embeddings and a graph database enhance the retrieval of relevant cases and tools. In addition, the system supports techniques and tools for incrementally building and refining solutions to novel problems, thereby facilitating both structured workflows and scientific discovery. The system tackles the challenges of integrating AI into human workflows by promoting transparency, adaptability, and meaningful collaboration. Initially tailored for the biological sciences domain, it enhances productivity and insight by enabling a seamless, interactive partnership between humans and AI, advancing the state-of-the-art in scientific workflows through a unified and accessible platform.