Integrating a Database-Grounded Language Model Assistant Into a WebGIS for Cultural Heritage Information Access in Majene Regency, Indonesia

Muh. Rafli Rasyid (Universitas Sulawesi Barat)
Muzaki (Universitas Sulawesi Barat)
Astinawaty (Universitas Sulawesi Barat)

Journal of Vocational, Informatics and Computer Education
Vol 4, No 3 (2026): September 2026

Purpose – This study developed and evaluated a web-based geographic information system (WebGIS) that presents selected cultural and tourism information in Majene Regency, Indonesia, and connects map objects with a Gemini-based assistant. Methods – The prototype used Laravel 10.48.8, Leaflet.js 1.9.4, MySQL 8.0.36, Esri World Imagery, and the Gemini-2.5-Flash API. No fine-tuning, retraining, parameter adaptation, or retrieval-augmented generation was performed. Attributes from verified cultural records were inserted into a structured prompt before each API request. Evaluation included API functional and negative tests, two local load tests with 10 virtual users, browser-device testing, a seven-item user acceptance test with 30 purposively selected participants, a single-blind expert evaluation of 20 object-specific narrative-generation cases by three evaluators, and five stress-test scenarios. Findings – The system integrated 64 eligible records from an inventory of 170 locations. Local load testing completed all recorded requests without errors, although response times indicated limited capacity headroom in the local environment. In a single-blind expert evaluation of 20 object-specific narrative-generation cases, the database-grounded prompt configuration received substantially higher scores than Raw Gemini for factual correctness, source faithfulness, and cultural sensitivity, and produced far less unsupported content; paired Wilcoxon signed-rank tests confirmed that these differences were statistically significant with large effect sizes. Research Implications – Findings are limited to the reported records, cases, evaluators, and local environment. Wider implementation requires expert review, source provenance, content correction procedures, restricted-knowledge safeguards, cybersecurity controls, backup procedures, and external API monitoring. Beyond cultural-information access, the approach also offers an applied example of database-grounded generative-AI system development relevant to digital literacy and vocational information-technology education. Originality – This study evaluates a database-grounded WebGIS–Gemini integration using selected verified cultural records as a prompt context.