local-deep-research

Why we included this project

For people who need answers that draw on several sources at once, Local Deep Research pulls from arXiv, PubMed, general search engines, and your own documents, then hands the gathered material to whichever model you choose, whether that is a local llama.cpp or Ollama setup or a cloud API. Because everything stays on your machine and the database is encrypted, it suits researchers handling confidential material or anyone who does not want their queries leaving home. It runs comfortably on a single consumer GPU like a 3090 and posts around 95 percent on SimpleQA, which makes it realistic for individuals and small teams without a big hardware budget. The result is private, reproducible research with citations, a solid alternative to hosted services.