Meerkat is an observability platform where AI Agents directly analyze infrastructure logs and metrics to detect anomalies. Going beyond the limitations of traditional rule-based alerts, the AI independently understands context from natural language queries or events delivered via external webhooks, and queries tools like Prometheus and Loki to infer causes. Logs are collected with OpenTelemetry and stored in a vector database to support semantic-based search, and analysis results are delivered to external channels via webhooks.

The system is composed of two independent services, Analyzer and Vectors, and is deployed in a Kubernetes environment using Helm Charts. The Analyzer is designed to operate reliably even with large-scale log processing, featuring an asynchronous worker pool, duplicate analysis prevention, and context overflow recovery mechanisms.