meishiwhy/literature-mind

LitMind is a Python toolkit that turns a researcher's Zotero library of academic papers into a searchable AI knowledge base that can answer questions, find supporting or opposing evidence, and help draft parts of a research paper. It works as a pipeline of eight connected pieces. The first pulls paper metadata straight out of your local Zotero database. The next parses the actual PDF text, cleaning up page headers, footers, and page numbers, and splitting each paper into its standard sections like abstract, methods, and results. A third piece sends that text through an AI model to pull out structured facts: the research question, the methods used, the variables studied, and the paper's claims. Those extracted facts get stored in a local knowledge base built on SQLite and ChromaDB, which supports both keyword and meaning based search. On top of that knowledge base sit four more tools. One lets you ask natural language questions about your paper collection and get answers with citations back to the source papers. Another takes a scientific claim and automatically finds which papers in your library support it, oppose it, or are neutral, along with an overall strength rating. A third takes your own study's results and drafts a structured seven section discussion section, citing relevant literature for context. The last one takes a research topic and writes a full structured literature review, grouping papers into themes, spotting where the field agrees or disagrees, and pointing out gaps that have not been studied. To guard against AI generated content inventing fake papers, every citation the AI produces is checked against the real paper IDs already stored in the knowledge base, and anything that does not match is automatically dropped. LitMind installs with pip, supports Anthropic's Claude and OpenAI's models as the underlying AI provider, and can be called directly as slash commands inside Claude Code. It is released under the MIT license.