Sovereign Artificial Intelligence for Municipal Governance: A Case Study of Jazan Municipality's Local AI Ecosystem Implementation of Open WebUI, RAG, Oracle 26ai, and Multi-Agent Systems

Background: The enterprise intelligent assistant system has recently undergone a series of architectural and operational updates aimed at enhancing local inference efficiency, improving language model integration with internal work environments, and strengthening reliance on a sovereign infrastructure that operates entirely within the organization without data leakage to external services. Objective: To develop and evaluate a comprehensive AI platform that enables natural language interaction with municipal databases while maintaining complete data sovereignty and supporting Arabic language requirements. Methods: The system architecture integrates Open Web UI as the operating layer, Oracle 26ai for database intelligence, Retrieval-Augmented Generation (RAG) for document processing, and specialized AI agents for various municipal functions. Recent updates include the addition of Qwen3.5-35B-A3B-FP8 model using a specialized Docker image (hellohal2064/vllm-qwen3.5-gb10) with advanced features including Flash Infer, FP8 KV Cache, and Prefix Caching. Results: The pilot deployment with 12 internal users demonstrated significant improvements: 77 messages processed within 24 hours, 98% reduction in document search time (from 45 minutes to under 1 minute), and 65% improvement in task completion speed. The system achieved 100% data sovereignty with zero external data transmission. Technical challenges including 100% root partition utilization were successfully resolved through cache cleanup and storage optimization. Conclusion: This study demonstrates the feasibility and effectiveness of sovereign AI implementation in municipal governance. The proposed architecture provides a scalable model for government agencies seeking to leverage AI capabilities while maintaining data security and linguistic compatibility.