Publoom Press for Libraries: a collection that answers, shown in screenshots

Purpose and the people it serves

A university repository holds decades of theses, articles and reports, and almost nobody reads it as a whole. A search box finds a title. What a librarian, a student or a researcher wants is to ask the collection what it knows and be shown the evidence. Publoom Press for Libraries is that front door. The engine indexes every record and every full text the repository can serve, classifies each paper against a canonical spine of domains, fields and subfields, and answers a question by retrieving the passages that bear on it and writing a cited answer from them.

The custodians are the first users: the answer, the map and the trend are how a library sees its own collection. Readers are the second: the same question box, the same citations, and a reader that opens the paper. Nothing leaves the institution's server that was not already public in its repository, and the model that writes the answer is chosen by the venue.

Sign-in and the workspace

The workspace leads with the question box. Beneath it are the model in use and the account's allowance, then the quieter tools: search the corpus, funding fit, a collection brief, the account's own documents, creation from a prompt, and the working documents. A question is typed and sent with Enter, and the answer lands as blocks in the document on the right, where it can be named, extended and exported.

The answer, and the search by meaning

The answer is written from the retrieved passages and every claim carries a bracketed number pointing to a record in the list beneath, each with its handle in the repository. Search is the other route: a phrase is matched against every record by meaning rather than by keyword, and the results carry a match percentage, so a weak match is labelled as one rather than shown as if it were the answer.

The collection map

The map is the whole collection classified: four domains, twenty-six fields and 252 subfields in a treemap sized by paper count. Selecting a field opens the rail with its papers, newest first, and its authors, and a research guide or a digest can be started from that field. The links view draws the fields as a graph, edges where papers share subfields, and the trend view counts the holdings by year and domain.

A paper and its reader

Any record opens as a paper page: title, creators, date, subjects, abstract and the link back to the repository. Where the repository serves the full text, the reader shows the document itself beside the record, so a citation in an answer is one click from the page it came from.

Where it runs, and what it is not

The venue is one configuration of the Publoom Press codebase, the same engine and interface that serve the researcher venue and the Giants' Shoulders literature component. A library deployment points the engine at the repository by its address, and the harvest and classification run in the background and resume where they stopped. Several repositories can sit side by side in one venue, each in its own corpus.

It is not a discovery layer over the open web, and it does not invent a source. Every citation resolves to a record the repository holds, and a question the holdings cannot answer says so.

State on 15 September 2026

Every screen was captured by the venue's own harness on the live demonstration, with each task asserted before its picture was taken: the sign-in, the answer landing as blocks, the search results and their match percentages, the map summary, a treemap selection, the field's papers and authors, the links graph, the trend bars, the paper detail and the reader rendering the document. The research guide template and the usage page were not captured on this run and are not shown.