The Persistent Knowledge Base: A New Approach to AI-Assisted Troubleshooting
When an unexpected alert fires, engineers typically turn to an AI assistant for help. But without prior context, the assistant must start from scratch—asking about data sources, services, connections, metrics, and labels. This discovery process consumes precious minutes during an incident. Grafana Assistant eliminates that friction by building a persistent knowledge base of your infrastructure in the background, so it already knows your environment before you ask a single question.
How Assistant Builds Its Understanding
Assistant runs an automated infrastructure memory process with zero configuration. A swarm of AI agents works continuously to discover and document your entire observability stack. This involves four key steps:
1. Data Source Discovery
The system identifies every connected Prometheus, Loki, and Tempo data source in your Grafana Cloud stack. This creates a complete map of where your metrics, logs, and traces live.
2. Metrics Scans
Agents query your Prometheus data sources in parallel to find services, deployments, and infrastructure components. They capture which metrics matter and what labels are available.
3. Enrichments via Logs and Traces
Loki and Tempo data sources are correlated with their corresponding metrics. This adds context about log formats, trace structures, and service dependencies—linking all telemetry together.
4. Structured Knowledge Generation
For each discovered service group, agents produce documentation covering five areas: what the service is, its key metrics and labels, how it's deployed, what it depends on, and relationships to other services. This becomes the persistent knowledge base.
Benefits for Incident Response
With this pre-loaded context, conversations become faster and more accurate. When you ask about a service, the assistant already knows, for example, that your payment system talks to three downstream services, its latency metrics live in a specific Prometheus data source, and its logs are structured JSON in Loki. You skip straight to troubleshooting.
Speed matters during incidents. Preloaded context can shave valuable minutes off your response time, even if you're an experienced engineer. But this capability is especially powerful for teams where not everyone has full infrastructure knowledge. A developer investigating an issue in their service can ask about upstream dependencies and get accurate answers, even if they've never looked at those systems before.
Zero Configuration, Maximum Context
Assistant requires no manual setup. The background agents automatically discover, scan, enrich, and document your infrastructure. The result: by the time you ask your first question, the assistant already has a complete map of your world—services, connections, metrics, logs, traces, and dependencies—all ready to support rapid incident resolution.