Escuela Colombiana de Ingeniería Julio Garavito: catalog, repository, metasearch and reference assistant on a single infrastructure

Institutional context

The Escuela Colombiana de Ingeniería Julio Garavito is a private university based in Bogotá, specializing in engineering and related disciplines. Its library serves an academic community that works with technical literature, standards and its own research output, and that demands both an up-to-date catalog and fast access to the institution's own production.

The challenge

The library did not have one isolated problem to solve, but four fronts that conditioned one another. The catalog and the repository lived as separate systems, with different interfaces for users. Technical services depended on manual keying, both to catalog new material and to describe the documents entering the repository. Subscribed electronic resources, the catalog and the institution's own output were searched through different paths, so a student had to know in advance where to look. And the reference team spent a large part of its working day answering repetitive operational questions.

Solving only one of those fronts would have left the other three standing, so the project was conceived as an integrated platform rather than as four independent purchases.

What we implemented

• Koha as the integrated library system, with its online public catalog.
• BiblioScan AI for AI-assisted cataloging from photos of the material.
• DSpace as the open-access institutional repository for the university's academic output.
• MetaExtract API integrated into the repository's deposit workflow, to auto-fill the descriptive metadata of each document.
• BiblioMeta as the metasearch engine, with an administration panel and real-time usage statistics.
• Reference ChatBot (ChatBot Referencista), one of our AI chatbots, for the library's reference service.

How we did it

The project was carried out in phases rather than in a single simultaneous deployment, precisely because each component depends on the previous one being stable.

1. Infrastructure. A dedicated university server, administered by our team, was provisioned to host the catalog and the repository, with HTTPS, automatic backups and monitoring.
2. Information model. The MARC21 cataloging profile, the repository's structure of communities and collections, and the metadata schema with its mandatory fields were agreed.
3. Migration and loading. Bibliographic and patron data were migrated into the catalog, and the university's academic output was loaded into the repository.
4. Artificial intelligence layer. BiblioScan AI was integrated into the technical-services workflow and MetaExtract API into the deposit form, with mandatory human review before each record is saved.
5. Discovery. The library's sources were connected to BiblioMeta, which was published as the search entry point, together with its metrics panel.
6. Reference service. The institutional corpus of the conversational assistant was prepared, the rules for escalating to staff were defined and an internal pilot was run before opening the service to the public.

Result

The library now operates on a coherent platform: the catalog, the repository and the metasearch engine are publicly accessible, the repository exposes its content to academic search engines and aggregators, and descriptive work has moved from transcription to review. The relationship has been sustained over time with successive expansions of scope, and the reference assistant service is the most recent component to be added.

Technologies

Koha, DSpace, BiblioScan AI, MetaExtract API, BiblioMeta, Reference ChatBot (ChatBot Referencista), MARC21, Dublin Core, OAI-PMH and a dedicated Linux server administered by BiblioLabTech.