Alerts are triggered by user-defined and unsupervised patterns that can incorporate a wide range of conditions: outputs from temporal graph queries (e.g., new high-risk connections, specific sequences of events), results from our graph analytics algorithms (e.g., a node's centrality score exceeding a threshold, changes in community structure), and even predictions from your traditional Machine Learning or statistical models, thanks to Pometry's ability to run embedded in Python environments. This allows you to fuse diverse analytical techniques into a single, powerful alerting mechanism.

Alerts can be configured to trigger various actions, including notifications and, most powerfully, the materialisation of the alert and its associated contextual subgraph directly into the Pometry User Interface with just a single line of code for immediate visual investigation.

By leveraging Pometry's underlying speed, flexible temporal query language, pre-built graph analytics, and easy UI integration, you can now rapidly define, test, and iterate on alert logic without extensive custom development or data engineering.

Absolutely. This is a core strength. The Alerts Engine is built on our native temporal graph, allowing you to define alerts based on evolving patterns, rates of change, sequences of events, and comparisons of graph states across different time windows.

When entities are flagged by alerts as part of a pattern, they can be tagged. Users can then explore these tags to see all entities involved in similar risks or trends, and investigate the broader subgraph context, effectively synthesising individual signals into a more complete picture.

Yes. Alert-driven insights, tagged entities, or identified subgraphs can be easily pushed to downstream systems, such as Machine Learning pipelines for further analysis (e.g., generating embeddings, training models) or other operational workflows.