active · HolobiomicsLab
Perspicacité — agentic AI literature assistant
An agentic AI assistant for scientists, researchers, and students who want to find and understand academic literature without getting lost in hundreds of papers — local-first, open-source, with multi-agent retrieval and an MCP server.
Perspicacité (French for "insight") is an agentic AI assistant designed for scientists, researchers, and students who want to find and understand academic literature without getting lost in hundreds of papers. It is local-first — only LLM inference calls leave your environment — and open-source (Apache 2.0).
It combines multi-agent, retrieval-augmented search over many sources (Semantic Scholar, OpenAlex, PubMed, arXiv, HAL, DBLP), a unified content pipeline, hybrid vector + BM25 retrieval, six RAG modes, an MCP server, and a REST API — supporting AI-assisted experimental planning, literature mining, and bibliography analysis across the lab's projects. Presented at the Metabolomics Symposium (2025).