Fugee has been introduced—a specialized agentic AI assistant developed to support refugees and asylum seekers. The system is capable of conducting structured interviews in 10 languages, relying on international law to suggest realistic migration routes.

image

What Happened

Developers have introduced Fugee, which uses an agentic approach to assist people in complex humanitarian situations. The application conducts multilingual interviews and analyzes user data in accordance with the provisions of the 1951 Convention. Technically, the solution runs locally on small language models (SLM) with up to 32B parameters, such as lfm2.5:8b, via the Ollama platform.

Context

To ensure high accuracy and minimize hallucinations, the system uses a RAG (Retrieval-Augmented Generation) architecture based on official UNHCR guidelines. This allows the agent to work with a verified knowledge base without relying on irrelevant or unstable content from the open internet.

Why It Matters for the Industry

The project demonstrates the technical viability of Vertical AI Agents running on small open-weights models. This confirms the possibility of creating highly specialized tools within secure environments where data privacy and autonomy are critical in conditions of limited infrastructure.

Why It Matters for Users

Developers and researchers can study the "pure Python" architecture for building agentic systems capable of solving complex multilingual tasks in offline mode. For end users, this means the potential emergence of private and accessible tools that do not require constant access to the cloud APIs of tech giants.

Sources

Author

Look at AI, Editorial Team