Ad-hoc scoring

Detect — score a traffic snapshot

POST a single network flow or a batch envelope to /api/detect and receive a per-flow anomaly score, severity (low → critical, or null for clean), and a list of triggered reasons. The TS scorer in src/lib/business/anomaly.ts is a single source of truth for live scoring; the Python pipeline at python/pipeline.py trains the offline model that feeds /api/anomaly-ml.

Submit traffic

Pick a preset scenario or paste raw JSON. Scores are computed against a sliding 30-minute window of submitted flows.

Choose a sample

One-click scenarios that mirror the four pattern silhouettes. Pick a preset, submit, and inspect the per-flow anomaly score + reasons.