VynOps
v0.3.0  ·  Open Source  ·  MIT License

VynOps — AI-Powered Kubernetes Operations Platform

Real-time visibility, intelligent incident management, and autonomous remediation — in a single open-source platform your team owns.

Built on Next.js 15 · Groq Llama-4 · Prometheus · Loki · Jaeger · Alertmanager

Get Started Free View SourcePricing
Real-time visibility
AI root cause analysis
Smart alert management
Auto remediation
SLA & on-call mgmt
Cloud & cost analytics
Multi-cluster
Secure by design
Team collaboration
Live pod exec

Features

Everything, documented

Real-Time Cluster Visibility

Live pod, node, and namespace monitoring across every connected cluster. Drill from cluster health score down to individual container metrics in two clicks. Supports multi-cluster views with per-cluster and aggregate dashboards.

  • Health scoring (0–100) based on pod restarts, resource pressure, and alert severity
  • Real-time pod status: Running, Pending, CrashLoopBackOff, OOMKilled, Evicted
  • Node resource utilisation: CPU, memory, storage per node
  • Namespace-level breakdown: workload counts, resource quotas, cost estimates
  • Live kubectl exec and log streaming via WebSocket

AI Copilot & Root Cause Analysis

Groq Llama-4-Scout-17b is the default for fastest available inference — OpenAI, Anthropic, Gemini, Azure OpenAI, Ollama and any OpenAI-compatible endpoint are also supported. The AI Copilot correlates Prometheus metrics, Loki logs, Kubernetes events, and alert history to surface the most likely root cause. Bring your own API key.

  • Natural language queries: "What's broken in production right now?"
  • Automatic event correlation across metrics, logs, and alerts
  • Blast radius estimation before any remediation is applied
  • Confidence scoring on all AI suggestions (0–100%)
  • Persistent incident memory — learns from past incidents over time
  • Sub-second inference via Groq ultra-fast LLM runtime (meta-llama/llama-4-scout-17b-16e-instruct)

Intelligent Alert Management

Ingests alerts from Prometheus Alertmanager via webhook. Automatically deduplicates, groups, and suppresses noise. Shows you the 3 alerts that actually need attention, not the 300 that are firing.

  • Alertmanager webhook receiver — no agent installation needed
  • Automatic deduplication and grouping by namespace, severity, and workload
  • Suppression rules: silence by label, time window, or alert name
  • Alert history and firing frequency analytics
  • Escalation to email, Slack, Microsoft Teams or a custom webhook with configurable routing rules

Autonomous Remediation

The AI generates step-by-step remediation plans with kubectl and Helm commands. Each plan includes a dry-run preview, blast radius estimate, and risk rating. Nothing runs without explicit confirmation.

  • Full kubectl command generation: patch, scale, delete, rollout restart
  • Helm chart upgrade suggestions with diff preview
  • Risk rating per remediation step: LOW / MEDIUM / HIGH
  • Dry-run mode: see what will change before applying
  • Full audit trail: who approved what, when, and what the outcome was
  • Rollback instructions included in every remediation plan

SLA & On-Call Management

Built-in SLA tracking with configurable thresholds and automatic escalation policies. Manage on-call rotations, schedule overrides, and escalation tiers directly in VynOps.

  • SLA definition by severity: Critical (<30 min), High (<2h), Medium (<8h), Low (<48h)
  • Automatic escalation when SLA breach is imminent
  • On-call schedule editor with rotation and override support
  • Slack integration: alert notifications, @-mentions, and escalation messages
  • SLA compliance reporting: breach rate, MTTR, MTTA by team and service

Cloud & Cost Analytics

Unified cost and infrastructure visibility across Kubernetes clusters and cloud nodes. Track compute spend, storage costs, and efficiency scores at namespace and workload level — with AI-generated optimisation recommendations.

  • Node inventory: CPU cores, memory, pod count, uptime, instance type, OS, and arch
  • Custom cost rate editor: USD/vCPU/hour, USD/GiB-hour, USD/GiB-month
  • Namespace and workload cost breakdown with efficiency scores
  • Over-provisioning detection: CPU/memory requests vs actual usage delta
  • PVC storage inventory: capacity, usage %, storage class, and per-PVC cost
  • Optimisation recommendations with kubectl right-size commands
  • Monthly cost projection and wasted spend identification

Automation Runbooks

A library of pre-built runbooks for common Kubernetes operations. Each runbook is a step-by-step automated workflow — run from the UI with a single click, with full output visibility and run history.

  • Diagnose CrashLoopBackOff: scan events, pull previous container logs, restart analysis
  • OOMKilled patch: detect OOM pods, bump memory limit +256Mi, trigger rollout restart
  • Rollback failed deployment: view revision history, execute rollback, post-health check
  • Force delete stuck Terminating pods with gracePeriodSeconds=0
  • Scale deployment replicas up or down with rollout watch
  • Cleanup evicted/failed pods to free namespace quota
  • Cordon & drain node: cordon, list evictable pods, generate drain command
  • Audit TLS certificates: decode secrets, report expiry, renewal recommendations
  • Debug ImagePullBackOff: surface pull errors, output fix commands
  • Full run history with per-step output, duration, and pass/fail status

Installation

Quick Start

From zero to running cluster in under 10 minutes.

01

Clone the repository

git clone https://github.com/vynops/VynOps cd vynops/apps/web
02

Configure your environment

cp .env.local.example .env.local # Required AUTH_SECRET=<generate: openssl rand -base64 32> NEXTAUTH_URL=http://localhost:3030 GROQ_API_KEY=<from console.groq.com — free tier available> K8S_API_URL=https://your-cluster-api-server # Optional but recommended PROMETHEUS_URL=http://prometheus:9090 ALERTMANAGER_URL=http://alertmanager:9093 LOKI_URL=http://loki:3100 SLACK_WEBHOOK_URL=https://hooks.slack.com/...
03

Install & run

npm install npm run dev # App running at http://localhost:3030 # Default login: set up via AUTH_SECRET
04

Connect your first cluster

# Add cluster via the UI dashboard, or via API: curl -X POST http://localhost:3030/api/clusters \ -H 'Content-Type: application/json' \ -d '{ "name": "production", "apiUrl": "https://your-k8s-api-server", "token": "your-service-account-token" }' # Response: {"id":"clus_1","name":"production","status":"healthy"}

Requirements

Node.js18.17.0 or later
Kubernetes clusterAny v1.24+
kubectlConfigured with cluster access
Groq API KeyFree tier — console.groq.com
PrometheusOptional but strongly recommended
LokiOptional — enables AI log correlation
JaegerOptional — distributed trace integration
PM2Optional — recommended for production

Environment Variables

Full reference for .env.local. Copy from .env.local.example to start.

VariableDescriptionRequired
AUTH_SECRETRandom 32-byte secret for NextAuth session signingrequired
NEXTAUTH_URLFull URL of your VynOps deployment (incl. port)required
GROQ_API_KEYGroq API key — free tier at console.groq.comrequired
K8S_API_URLKubernetes API server URL for default clusterrequired
K8S_TIMEOUT_MSRequest timeout for k8s API calls (default: 10000)optional
PROMETHEUS_URLPrometheus server URL for metrics queriesoptional
ALERTMANAGER_URLAlertmanager URL for alert webhook receiveroptional
LOKI_URLLoki URL for log queries (enhances AI analysis)optional
JAEGER_QUERY_URLJaeger Query URL for distributed tracingoptional
GRAFANA_URLGrafana URL for dashboard deep-link integrationoptional
SLACK_WEBHOOK_URLSlack incoming webhook for alert notificationsoptional
ALLOWED_DEV_ORIGINSComma-separated IPs for remote dev accessoptional
PORTHTTP port for the server.mjs process (default: 3030)optional

Architecture

Frontend
  • Next.js 15 App Router
  • React Server Components
  • Zustand + React Query
  • Framer Motion UI
API Layer
  • Next.js API Routes
  • NextAuth v5 session
  • Per-cluster header routing
  • Node.js WebSocket server
Integrations
  • Kubernetes API server
  • Prometheus HTTP API
  • Loki query API
  • Groq LLM inference
Multi-cluster routing

Each API call from the frontend passes X-Cluster-ID and X-Cluster-URL headers. The API layer uses these to route requests to the correct Kubernetes API server. No session-level cluster affinity — fully stateless.

WebSocket pod exec

Real-time pod log streaming and kubectl exec run through a custom Node.js server (server.mjs) that proxies WebSocket connections to the Kubernetes API. Must run via node server.mjs or PM2 — not npm start.

Tech Stack

Frontend

  • Next.js 15App Router + RSC
  • TypeScript 5.5Strict mode
  • Tailwind CSSDark-first
  • Framer MotionAnimations
  • Zustand 5State management
  • React QueryServer state

Auth & API

  • NextAuth v5AUTH_SECRET
  • Node.js server.mjsWebSocket host
  • PM2Process manager

AI & Inference

  • Groq AILLM inference
  • Llama-4-Scout-17bModel

Observability

  • PrometheusMetrics
  • LokiLogs
  • JaegerTraces
  • GrafanaDashboards
  • AlertmanagerAlert routing

Troubleshooting

ProblemSolution
Login redirects to 404 after credentialsSet NEXTAUTH_URL to the actual host:port you're accessing. Must match exactly.
All API routes return 404 after Linux deployDelete .next/ and rebuild: rm -rf .next && npm run build
AI Copilot returns "API key invalid"Use GROQ_API_KEY not OPENAI_API_KEY. Get a free key at console.groq.com
Cluster shows as "unreachable"Verify K8S_API_URL is reachable from the VynOps server. Test: curl -k $K8S_API_URL/healthz
No metrics appearing in dashboardsSet PROMETHEUS_URL to your Prometheus instance. Check /api/metrics/test endpoint.
WebSocket pod exec not workingEnsure you're running via server.mjs (node server.mjs), not npm start. npm start does not support WebSockets.
AUTH_SECRET error on startupVynOps uses NextAuth v5 — use AUTH_SECRET, not NEXTAUTH_SECRET. Generate: openssl rand -base64 32

Production Deployment

With PM2 (recommended)
npm run build NODE_ENV=production pm2 start server.mjs \ --name vynops \ --env production pm2 save && pm2 startup
With Docker
docker build -t vynops . docker run -d \ --env-file .env.local \ -p 3030:3030 \ --name vynops \ vynops
Free · Open Source · Self-hosted

Deploy in 10 minutes

No account. No cloud signup. Clone, configure, connect your cluster, and go.

Get Started View PricingTalk to us