An open-source digital library for research and academia
ELVIRA puts a library’s publications in one catalog, opens them in a reader in the browser, lends protected e-books, and lets readers search inside every book. It is developed at the Slovak University of Technology in Bratislava, and any library can run it on its own servers.
Run it yourselfHow it worksAdopting ELVIRA
Libraries using it include the Slovak University of Technology and the Catholic University in Ružomberok.
- Inside
- catalog, web reader, e-book lending, search, an assistant
- Works with
- Thorium and other OPDS apps, Calibre
- Code
- open source on GitHub
- Licence
- MIT, BSD-3-Clause
- Latest
- catalog 0.15.0, 12 Nov 2026
What it does
- Read in the browser
- Every PDF opens in the built-in reader. Search inside the text, page through thumbnails, highlight and take notes. Nothing to install.
- Borrow protected titles
- E-books are lent with Readium LCP, the open DRM public libraries use. Readers borrow, renew, return and join a waiting list when all copies are out.
- Use any e-reader app
- Every catalog is also an OPDS 1.2 and 2.0 feed. Add it to Thorium or another OPDS app and browse, borrow and read there.
- Cite and share
- Copy a citation as BibTeX, BibLaTeX or RIS. Share a page range instead of the whole file.
- Many libraries, one install
- One server hosts separate catalogs, each with its own publications and permissions. Sign in with a local account or your institution’s LDAP.
- An API for everything
- Whatever the portal can do, a script can do over the documented REST API. AI agents connect over the Model Context Protocol.

Search and AI
The catalog reads its own books: it searches inside them, explains them and cites the page. See catalog 0.15.0 for what changed and how to set it up.
- Multilingual search
- Results are ranked by what a query means, so a question in Slovak can find a paragraph in English. Word search handles Slovak and Czech word forms and suggests corrections.
- Search inside publications
- A search returns the paragraph that matches, not only the book, with a link that opens the reader on that page.
- Structure from every PDF
- Workers add a text layer to scanned PDFs, split each publication into paragraphs, crop its figures and tables, and turn its formulas into LaTeX.
- An assistant that cites
- Ask about the collection, or about a passage or figure in the reader, and get an answer with a link to each page it used. Titles lent with Readium LCP are never read by AI.
Getting books in
Besides uploading in the portal or through the API, two tools fill a catalog: one from a Calibre library, one from a book scanner.

Calibre plugin
Keeps a Calibre library and a catalog in sync in both directions. It shows what it will do to every book first, asks when both sides changed, and never deletes anything. Released as version 2.1.

Books digitalizer
Cleans and OCRs scans of printed books, renders a searchable PDF, a PDF/A master and an EPUB, and publishes them to the catalog. A person reviews only the pages the machine was unsure of.
Latest releases
| Project | Version | Released | Links |
|---|---|---|---|
| Catalog server | 0.15.0 | 12 Nov 2026 | AnnouncementNotes |
| PDF reader | v0.7.9 | 12 Nov 2026 | Notes |
| Calibre plugin | 2.1.0 | 8 Nov 2026 | AnnouncementNotes |
| PostgreSQL image | r202611161245 | 16 Nov 2026 | Notes |
Where to start
See it in use
The Slovak University of Technology runs ELVIRA at elvira.stuba.sk, and the Catholic University in Ružomberok runs its own instance.
Plan a deployment
What you get, what you need to bring, and where the limits are: Adopting ELVIRA. To talk it through, write to us.
Run your own
The catalog starts with Docker Compose on localhost:8000. Its code is in the EvilFlowersCatalog organization on GitHub; EvilFlowers is the name of the catalog server at the centre of ELVIRA.
git clone https://github.com/EvilFlowersCatalog/EvilFlowersCatalog.git
cd EvilFlowersCatalog
docker compose up -d --build django
docker compose exec django python3 manage.py setup
docker compose exec django python3 manage.py createsuperuser