Open-source tool alerts when your agent starts looping,drifting,burning tokens (github.com)

A new open-source tool called DriftShield has been launched to help developers monitor AI agents in real-time, alerting them when issues like looping, goal drift, or resource spikes occur. By wrapping around existing agents, DriftShield provides alerts on platforms like Slack or Discord without requiring any complex setup or cloud infrastructure. It tracks three critical aspects: loop detection to identify excessive repetitive API calls, goal drift to ensure the agent's output remains aligned with its original task, and resource spike detection to monitor unexpected resource consumption. This tool is significant for the AI/ML community as it addresses a common problem in AI agent deployment—unexpected behaviors that can lead to wasted resources and operational inefficiencies. DriftShield operates entirely on the user's machine, using local embeddings and storing traces in a SQLite database, ensuring data security while minimizing overhead. As a lightweight solution, it aims to provide quick and actionable insights, making it particularly useful for solo developers and small teams who lack access to heavy observability platforms. The roadmap includes future enhancements such as auto-correction hooks and enhanced multi-agent support, indicating potential for greater functionality in managing AI deployments.