Principle:Evidentlyai Evidently Snapshot Storage
| Knowledge Sources | |
|---|---|
| Domains | ML_Monitoring, Infrastructure |
| Last Updated | 2026-02-14 12:00 GMT |
Overview
A persistence mechanism that stores evaluation snapshots within monitoring projects for time-series dashboard visualization.
Description
Snapshot Storage adds computed Snapshot objects (from Report.run()) to an Evidently monitoring project. Each snapshot is timestamped and persisted in the workspace's file storage, enabling time-series visualization in the monitoring dashboard.
This is the bridge between the evaluation pipeline (Report → Snapshot) and the monitoring dashboard (Project → Dashboard panels). Dashboard panels query stored snapshots to render metric trends over time.
Usage
Use after running a Report and within a monitoring loop to persist snapshots for dashboard visualization. Typically called in a loop with different data batches and timestamps.
Theoretical Basis
Snapshot storage follows the event sourcing pattern where each evaluation result is stored as an immutable event:
# Pseudocode: Event sourcing for monitoring
for batch, timestamp in monitoring_schedule:
snapshot = report.run(batch, reference, timestamp=timestamp)
workspace.add_run(project.id, snapshot) # Append immutable event
# Dashboard queries the event stream for visualization