A High-Availability Multi-Agent RAG Architecture for Automated Expert Assistance and Multimodal Technical Onboarding
Abstract
This paper presents the design, deployment and evaluation of a high-availability, multi-agent Retrieval-Augmented Generation (RAG) architecture acting as an expert assistant for technical training. In sectors such as cyber-physical systems and automated instrumentation, onboarding depends on synchronous supervision by senior engineers, which creates bottlenecks and increases technical downtime. Retrieval is restricted to a closed, curated corpus of manuals, datasheets and approved guidelines, keeping answers traceable and reducing large language model hallucinations. The interaction layer runs on WhatsApp through the Evolution API and accepts text, documents, circuit images and voice notes; the backend combines n8n orchestration, a dedicated virtual private server and a Supabase PostgreSQL instance with the pgvector extension, with inference served by paid-tier models through Google Vertex AI with provider fallback. Retrieval scope is bound to the caller’s resolved identity rather than chosen by the model, enforcing need-to-know separation, and the interaction policy is configurable by user profile: a Socratic mode that withholds finished solutions, and a direct mode for experienced practitioners. An adoption survey administered to final-semester engineering students at the end of an intensive training cycle (n = 24) obtained a global Cronbach’s alpha of 0.90. Perceived competence (M = 4.38, d = 1.93) and autonomy (M = 4.14, d = 1.39) were significantly above the scale midpoint, whereas relatedness was not (M = 3.24, p = 0.125): the assistant substitutes for the informational function of mentorship but not for its social function. Industrial deployment remains a projected application.
Keywords: Retrieval-Augmented Generation, Multi-Agent Systems, Engineering Education, Technical Onboarding, Self-Determination Theory, Human-Machine Co-Agency
DOI: 10.54941/ahfe1008394
Cite this paper
More from this volume
- AI-Enhanced Live-Streamed English Language Learning: Designing and Evaluating an Intelligent Platform for Chinese Learners
- "They Will Move": Confirmation Bias, Expectation Frames, and Premature Monitoring Drop Off in COLREG Encounters
- Automation transparency communication does not improve supervisory control performance, but shapes operator trust
- The influence of neuroticism on startle and surprise response: a flight simulator study
- EEG-Based Classification of Reward, Penalty, and Movement Events Across Four Video Games of Varying Complexity
- The Effects of Task Similarity and Time Pressure on Task Performance Following Interruptions in Complex Human-Computer Interaction Systems
- Trusting the Chain, Not the Black Box: A Provenance-Based Architecture for Human-Machine Teaming in Military Autonomy
- Design of a Simplified Humanoid Robot Face: Cultural Differences and Emotional Synchronization
- From Visual Observation to Auditory Interaction: A Human-Centered Tangible Music System Based on Physical Objects
- The Impact of Scarcity-Based Dark Patterns on User Performance, Perceived Darkness, and Workload in Subscription Services
- A Behavioral Observation Tool for Diagnosing Human–AI Interaction Profiles in Engineering Design Education
- The Mediated Moon: Semantic Mediation Between Gesture and Generative Image


AHFE Open Access