Agentic RAG for Administration
Overview
This service provides an agentic Retrieval-Augmented Generation (RAG) system integrated with a user-friendly web interface for company/institute administration. It enables users to query complex internal datasets---including project details, employee records, travel planning, leave policies, and institutional regulations---using natural language. The system bridges the gap between static internal documentation and actionable intelligence by providing context-aware summaries and structured answers via a browser-based dashboard. The system utilizes a multi-stage AI pipeline: a responsive web interface built with FastAPI and a modern frontend framework (e.g., Streamlit or React) for UI/UX; semantic search powered by vector embeddings (e.g., Gemma-3) and indexed via PGVector for retrieval; and integration with LLMs specialized for the Slovenian language (GaMS, Gemma-3, or GPT-OSS-20B) for generation.
Outcome
A fully operational RAG prototype capable of retrieving relevant administrative context and generating coherent, factually accurate responses in Slovenian.
Service Card
| Category | AI products |
| Domain | Administrative Automation, Internal Knowledge Management |
| Vertical | Digital Society |
| Target users | Company/Institute Administrators, Management Staff |
| Service type | Model & System Integration |
| Outputs | Specialized vector index and fine-tuned retrieval prompts for setup; natural language answers with cited sources/references from internal documents for inference. |
| Requirements | Hardware: NVIDIA GPU node (for local LLM inference and embedding generation). Data: Access to internal administrative documents, policies, and project databases in structured or semi-structured formats. |
| Limitations | Access is restricted to internal institute data; requires strict adherence to data privacy and GDPR protocols regarding personnel information. |
| Availability | Q4/2026 |
| Licensing | MIT (Software) / Internal data remains proprietary |
| Visual content | assets/service-visuals/agentic-rag-for-administration.png |
| Visual caption |