Prophet 🚀

The AIOps-Powered Self-Healing Multi-Cloud Karpenter Platform

Prophet is an open-source GitOps monorepo that makes Kubernetes clusters across AWS, GCP, and Azure predictive and self-healing.

At its core are custom Golang operators that automatically detect issues, reason about them with AI, and remediate common failures without human intervention.

Core Self-Healing & AIOps Capabilities

• AnomalyRemediator Operator
  Detects anomalies via OpenTelemetry + Grafana ML → automatically restarts pods, scales deployments, or drains nodes.

• PredictiveScaler Operator
  Uses Grafana ML forecasts to pre-provision Karpenter nodes before demand spikes.

• SLOEnforcer Operator
  Monitors error budgets and takes action when SLOs are at risk.

• AutonomousAgent Operator
  In-cluster LLM inference + Model Context Protocol (MCP) server for agentic autonomy: diagnose → decide → act.

• K8sGPT Integration
  AI-powered plain-English diagnostics on every alert.

• eBPF Observability
  Cilium + Hubble for kernel-level network visibility with zero overhead.

All managed declaratively through ArgoCD and Kustomize overlays.

Quick Demo: Watch Self-Healing in Action (<5 minutes)

git clone https://github.com/prophet-aiops/prophet.git
cd prophet/demo/remediation-chaos
./demo.sh

See a crashing app get automatically detected and healed by our operators.

Key Stack

• Autoscaling: Karpenter (multi-cloud)
• GitOps: ArgoCD + Kustomize
• Observability: OpenTelemetry → Prometheus → Grafana (with ML forecasting)
• Diagnostics: K8sGPT
• Networking: Cilium + Hubble (eBPF)
• AI Reasoning: Local LLM inference + MCP protocol
• CI: GitHub Actions (manifest validation, security scanning, operator builds)

Getting Started

1. Clone and explore
2. Try the quick demo above
3. Deploy operators using Helm charts or GitOps overlays

Testing K8sGPT Integration

See aiops/diagnostics/K8SGPT-TESTING.md.

UI: Headlamp (Prophet AIOps Console)

Deploy Headlamp to browse Prophet CRDs and “trust-but-verify” self-healing actions:

• clusters/common/ui/headlamp/README.md

Great as a starter kit, learning resource, or foundation for production AIOps.

---

This Repository

This is an active development fork focused on:

• Custom Go operators for self-healing automation (operators/)
• Multi-cloud Kustomize overlays for AWS, GCP, Azure (clusters/)
• Chaos engineering experiments with AI validation (resilience/, demo/)
• Rancher & Headlamp UI extensions for K8sGPT diagnostics (rancher-k8sgpt-extension/, headlamp-k8sgpt/)

See the operator-specific READMEs for current status and documentation.

---

For Contributors

Quick Links:

• Contributing Guide - Dev setup, PR guidelines
• Operator Reference - How to build and run operators
• Cluster Overlays - GitOps structure and deployment

Tech Stack:

• Go 1.22+ (operators, controller-runtime)
• Kubebuilder v3.x (CRD scaffolding)
• Kustomize + ArgoCD (GitOps)
• Tilt (local operator development)

Local Development:

# Run all operators locally (requires kind cluster)
make dev-up

# Run specific operator
cd operators/label-enforcer && make run

# Run demo
cd demo/remediation-chaos && ./demo.sh

Helm Deployment:

# Add Helm repository (if using hosted charts)
# helm repo add prophet https://charts.prophet-aiops.dev
# helm repo update

# Install operators via Helm
helm install prophet-label-enforcer operators/label-enforcer/helm/label-enforcer
helm install prophet-health-check operators/health-check/helm/health-check

# Or use GitOps with ArgoCD/Flux pointing to chart directories

---

Design Documents

Document	Description
PRD-ALIGNMENT.md	Requirements mapping
AIOPS-UPGRADE.md	V5 upgrade path
V6-AGENTIC-AUTONOMY.md	Agentic architecture

---

GitHub: https://github.com/prophet-aiops/prophet

Interested in self-healing operators or agentic Kubernetes? Let's connect!