Quick Start
Get Klarsicht running in your cluster in under 5 minutes.
Prerequisites
- Kubernetes cluster (1.26+)
- Helm 3
- Grafana with alerting enabled
- Prometheus (optional, for metrics correlation)
1. Install with Helm
helm install klarsicht oci://ghcr.io/outcept/klarsicht/helm/klarsicht \
--namespace klarsicht --create-namespace \
--set agent.llmApiKey=<your-api-key> \
--set agent.metricsEndpoint=http://prometheus.monitoring.svc:9090
This deploys:
- klarsicht-agent — FastAPI service that receives webhooks and runs investigations
- klarsicht-dashboard — React UI for viewing incidents and RCA results
- klarsicht-postgres — PostgreSQL for incident storage
2. Configure Grafana
Create a webhook contact point in Grafana:
- Go to Alerting → Contact points → Add contact point
- Name:
klarsicht - Type: Webhook
- URL:
http://klarsicht-agent.klarsicht.svc:8000/alert
Set it as the default receiver in Alerting → Notification policies.
Or use the one-click setup in the Klarsicht dashboard at /setup.
3. Test it
Send a test alert to verify the pipeline:
# Option A: Mock alert via API
curl -X POST http://klarsicht-agent.klarsicht.svc:8000/test
# Option B: Deploy a pod that actually crashes
kubectl apply -f https://raw.githubusercontent.com/outcept/Klarsicht/main/examples/test-crashloop.yaml
The investigation result will appear in the dashboard within 60 seconds.
4. Clean up test
kubectl delete -f https://raw.githubusercontent.com/outcept/Klarsicht/main/examples/test-crashloop.yaml
What’s next
- Architecture Overview — understand how the agent investigates
- Integrations — configure Prometheus, Grafana
- Helm Values Reference — all configuration options