SIETK AI: A Fault-Tolerant Agentic RAG Architecture for High-Fidelity Institutional Retrieval

In the digital era, educational institutions require intelligent, zero-hallucination systems to provide instant, highly accurate information to diverse user bases. This paper presents the design and implementation of SIETK AI, a fault-tolerant Agentic virtual assistant for the Siddharth Institute of Engineering and Technology. The system utilizes a strictly controlled, two-stage Retrieval-Augmented Generation (RAG) pipeline to ensure high-fidelity responses. By leveraging an immutable, in-memory knowledge representation coupled with a dual-agent routing system powered by Groq and Gemini 1.5 Flash, the architecture achieves a 90% retrieval and generation accuracy rate with an end-to-end latency of just 2.0 to 3.0 seconds.