Warc Gpt

WARC GPT is an open source retrieval augmented generation tool from the Library Innovation Lab at Harvard Law School Library. It lets people build a custom chatbot that uses a collection of web archive files, known as WARC files, as its knowledge base, so a collection can be explored by asking questions in natural language instead of relying on keyword searches and metadata filters. The tool performs multi document full text search across the archive and produces a summarized answer. It lists the sources behind each response and shows the relevant excerpts from the archived records, so a user can check the answer and find points of interest within the collection. It is customizable, and the lab frames it as a way for web archiving practitioners to understand how retrieval augmented generation applies to specialized digital archives. It is intended for librarians, archivists, researchers and tinkerers who work with web archives and want to experiment with AI on them. The lab describes the release as experimental software and a prototype, published as part of its wider series on how artificial intelligence changes the relationship to knowledge, and the repository carries a disclaimer to that effect. The software is free to run yourself. Installation instructions are on GitHub, and no paid plans or hosted service are offered.