20.500.12556/DKUM-97649 Integracija pogovornega agenta v aplikacijo Integration of a conversational agent into an application Diplomsko delo se posveča razvoju inteligentne rešitve za podporo procesom mednarodne izmenjave Erasmus+. Glavni namen naloge je premostitev težav pri iskanju informacij v obsežni dokumentaciji s pomočjo klepetalnega robota, ki temelji na sodobni arhitekturi RAG (angl. Retrieval-Augmented Generation). V okviru praktičnega dela je bila v okolju Oracle Cloud Infrastructure vzpostavljena avtonomna baza podatkov, v kateri je s pomočjo modelov ONNX in tehnologije AI Vector Search omogočeno generiranje ter shranjevanje vektorskih vložkov neposredno v bazi. To zagotavlja visoko stopnjo zasebnosti, saj se podatki za iskanje ne pošiljajo zunanjim ponudnikom, temveč se iskanje po podobnosti izvaja lokalno. Za varno komunikacijo z zunanjim vmesnikom OpenAI API, ki skrbi za končno generiranje naravnega jezika, so bili konfigurirani varnostni protokoli ACL in ustrezne poverilnice. Celoten proces, od uvoza datotek iz shrambe Object Storage do avtomatske pretvorbe besedila v vektorsko znanje, je avtomatiziran s pomočjo SQL procedur in namenskih sprožilcev. Rezultat dela je delujoča aplikacija v okolju Oracle APEX, ki uporabnikom nudi natančne in vsebinsko podprte odgovore na podlagi naloženih dokumentov. S projektom je dokazano, da uporaba nizko-kodnih platform v kombinaciji z umetno inteligenco bistveno optimizira administrativno delo in izboljša dostopnost informacij v akademskem okolju. The diploma thesis focuses on the development of an intelligent solution to support the Erasmus+ international exchange processes. The primary objective of the work is to overcome the difficulties of retrieving information from extensive documentation using a chatbot based on the modern RAG (Retrieval-Augmented Generation) architecture. Within the practical framework, an autonomous database was established in the Oracle Cloud Infrastructure environment, where the generation and storage of vector embeddings directly within the database were enabled using ONNX models and AI Vector Search technology. This ensures a high level of privacy, as the search data is not transmitted to external providers; instead, the similarity search is performed locally. To ensure secure communication with the external OpenAI API, which handles the final natural language generation, ACL security protocols and appropriate credentials were configured. The entire process, from importing files from Object Storage to the automated transformation of text into vector knowledge, was automated using SQL procedures and dedicated triggers. The result of the work is a functional application in the Oracle APEX environment, providing users with accurate and context-supported answers based on the uploaded documents. The project demonstrates that the use of low-code platforms in combination with artificial intelligence significantly optimizes administrative tasks and improves the accessibility of information within the academic environment. Oracle APEX umetna inteligenca RAG arhitektura AI Vector Search Oracle Cloud Infrastructure. Oracle APEX artificial intelligence RAG architecture AI Vector Search Oracle Cloud Infrastructure. true true false Slovenski jezik Angleški jezik Diplomsko delo/naloga 2026-03-29 12:33:48 2026-07-29 21:14:37 2026-07-30 03:09:40 0000-00-00 00:00:00 2026 0 Kranj 0 0000-00-00 NiDoloceno NiDoloceno NiDoloceno 0000-00-00 0000-00-00 0000-00-00 286386435 VS_Oblak_Aljosa_2026.pdf VS_Oblak_Aljosa_2026.pdf 1 D39368666CA8B9C66AF6F2A54EAB641C 18241d9dbce2101c7e58be88c4906ab8e3f8bb6d7c6018e44a0f1078389d2f59 01658539-2c1f-11f1-b04b-00155d0001a1 https://dk.um.si/Dokument.php?lang=slv&id=204895 Fakulteta za organizacijske vede 0 0 0