DriftSentry

DriftSentry is a web app (with optional lightweight desktop agent) for anomaly detection focused on “slow-burn” issues: gradual latency creep, memory bloat, cost spikes, and config drift that standard alert thresholds miss. It ingests time-series metrics and logs from common stacks (Prometheus, CloudWatch, Datadog exports) and builds per-service baselines that adapt to seasonality and deploy cycles. Instead of flooding teams with alerts, it groups anomalies into incident narratives (what changed, when, and likely blast radius) and links them to deploys, feature flags, and infra changes. The product is an AI + traditional app: classical statistical detectors for reliability, with an LLM layer for summarization, root-cause hypotheses, and “next checks” runbooks. It’s realistic for small teams: start with 5–10 key signals, then expand. Pricing is per service/host with a free tier for one environment.