Documentationbreadcrumb arrow Grafana Cloudbreadcrumb arrow AI and machine learningbreadcrumb arrow Grafana Assistantbreadcrumb arrow Guidesbreadcrumb arrow Infrastructure memory
Grafana Cloud

Infrastructure memory

Grafana Assistant automatically builds a knowledge base of your infrastructure by analyzing your telemetry data. A swarm of AI agents discovers your data sources, scans your metrics, logs, and traces, and infers knowledge about your services, dependencies, and topology. This context helps the Assistant answer questions faster and with more accuracy.

Understand how it works

Infrastructure memory runs automatically for all active tenants and refreshes weekly. The system discovers your connected data sources, scans the telemetry they store, and generates structured memories that describe your services, clusters, namespaces, and their relationships.

Discovery is driven by Prometheus data sources. The system scans your Prometheus data sources to find services, then enriches each service with context from logs and traces. This means the Assistant can only discover services that have metrics available in Prometheus data sources it has access to.

The process follows these steps:

  1. Discover data sources: The system identifies all available Prometheus, Loki, and Tempo data sources in your Grafana Cloud stack.
  2. Scan metrics: AI agents query your Prometheus data sources in parallel to discover services, deployments, and infrastructure components.
  3. Enrich with logs and traces: The system correlates Loki and Tempo data sources with their corresponding metrics data sources and gathers additional context about each service group.
  4. Generate memories: For each discovered service group, agents combine metrics, logs, and traces to produce structured knowledge about identity, key metrics, deployment topology, dependencies, and observability patterns.

Generate memories manually

If your stack doesn’t have infrastructure memories yet, you can trigger the first scan manually.

  1. Open Grafana Assistant.
  2. Click the menu icon and select Settings.
  3. Navigate to the Infrastructure memory section.
  4. Click Generate.

The scan runs in the background. Depending on the number of data sources and services, it may take several minutes to complete.

View and review memories

You can view the generated memories and provide feedback on their accuracy.

  1. Navigate to Grafana Assistant > Settings > Infrastructure memory.
  2. Browse the list of discovered service groups.
  3. Expand a group to review its content.
  4. Use the feedback controls to flag inaccurate or missing information.

Your feedback helps improve future scans.

Review permissions

Infrastructure memories respect your organization’s RBAC policies. Each memory is linked to the data sources that were used to generate it. Users can only see memories generated from data sources they have access to.

Understand how the Assistant uses memories

Infrastructure memories give the Assistant context about your specific environment. When you ask a question or run an investigation, the Assistant searches relevant memories to understand which services exist, how they relate to each other, and where to find the right telemetry.

This is especially useful for:

  • Investigations: Agents use memories to quickly identify affected services and correlate signals across data sources.
  • Querying: The Assistant maps natural language like “How is the checkout service?” to the correct metrics and labels.
  • Troubleshooting: Memories provide service dependency context that helps the Assistant trace issues across your stack.

Next steps