IntentContinuum is a novel framework for intent-driven resource management in the compute continuum, enabling intelligent deployment and real-time adaptation of distributed IoT applications across edge and cloud environments. This system leverages Large Language Models (LLMs)—specifically OpenAI GPT-4o—to monitor performance, analyze root causes of violations, and recommend reconfiguration actions, all while maintaining adherence to user-defined Service Level Objectives (SLOs) such as response time for image processing.

Key Features

Intent-Aware Monitoring: Tracks application-level intents like latency or throughput and checks for violations in real time. LLM-Powered Root Cause Analysis: Uses GPT-4o to pinpoint issues (e.g., CPU bottlenecks, memory limits, network congestion). Automated Reconfiguration: Applies corrective actions suggested by the LLM to restore compliance with user-defined intents. Closed-Loop Feedback System: Continuously adapts to workload, topology, and environmental changes without human intervention. Edge–Cloud Optimization: Coordinates between edge and cloud resources to ensure optimal performance, low latency, and system reliability.

We provide an open-source prototype implementation of IntentContinuum, developed using widely adopted tools like Kubernetes and ONOS. The system is modular, extensible, and can be easily integrated into real-world IoT environments. Our extensive experimental evaluations demonstrate that IntentContinuum:

Outperforms traditional methods in maintaining SLOs under dynamic conditions Reduces manual intervention through automated diagnosis and action Enhances scalability and robustness of distributed IoT deployments

If you use this project in your research, please cite the associated paper (to be added once published).