VibOps

Manage GPU clusters and LLM deployments in plain language.

"Deploy llama3 on the prod cluster with 3 GPU replicas"
"Scale the gpu-ng node group to 5 nodes"
"Why is the inference service slow? Fix it."

VibOps is an AI-native GPU infrastructure operations platform. A single conversation replaces kubectl, Helm, and CI/CD pipelines for managing GPU clusters.

---

Install VibOps

Three paths — pick yours:

Option	Best for
A — Your PC	Local testing, demo
B — CSP internal server	Production — intranet, sovereign, air-gapped
C — Cloud server	Production — internet-accessible, remote teams

Full step-by-step instructions with screenshots: QUICKSTART.md

---

Option A — Your PC

What you need: Docker Desktop · An LLM API key (or Ollama for local AI)

git clone https://github.com/VibOpsai/vibops-install.git
cd vibops-install
make quickstart

Open http://localhost:8003 — your VibOps console is ready.

Full guide: QUICKSTART.md → Option A

---

Option B — CSP internal server

What you need: Linux server on internal network (4 vCPU / 8 GB RAM / Ubuntu 22.04) · Internal LLM endpoint or API key

# On your server
git clone https://github.com/VibOpsai/vibops-install.git
cd vibops-install
make quickstart

Access the console at http://INTERNAL_SERVER_IP:8003 — accessible to anyone on the internal network. No internet exposure required.

GPU clusters connect via outbound polling — no inbound ports needed on the cluster side.

Full guide: QUICKSTART.md → Option B

---

Option C — Cloud server

What you need: Linux cloud VM with public IP (4 vCPU / 8 GB RAM / Ubuntu 22.04) · LLM API key

# On your server
git clone https://github.com/VibOpsai/vibops-install.git
cd vibops-install
make quickstart

Share http://YOUR_PUBLIC_IP:8003 with your team — accessible from anywhere.

Full guide: QUICKSTART.md → Option C

---

Connect a GPU cluster

Once VibOps is running, connect your GPU clusters via the VibOps Connect gateway:

1. Console → Settings → Gateways → New Gateway
2. Copy both values it returns: the gateway id and the token (the token is shown once only). Connect authenticates as that gateway and exits without its id — it does not register itself.
3. On your GPU cluster:

helm upgrade --install vibops-connect charts/vibops-connect \
  --namespace vibops-connect --create-namespace \
  --set gateway.id="$GATEWAY_ID" \
  --set vibops.coreUrl="http://YOUR_VIBOPS_IP:8000" \
  --set vibops.token="PASTE_YOUR_TOKEN_HERE"

Gateway shows Online within 30 seconds.

---

Useful commands

make quickstart   # First-time setup — creates .env, generates secrets, starts stack
make check        # Health check — validates all components
make up           # Start the stack
make down         # Stop the stack
make hash PASSWORD=yourpassword  # Generate bcrypt hash for password setup
make pilot-create-client ORG="Acme" EMAIL=admin@acme.com PASSWORD=secret  # Provision a client org

---

Documentation

Document	Description
QUICKSTART.md	Step-by-step installation (PC / internal server / cloud)
Installation Guide	Full technical setup, server requirements, Helm production deployment
User Guide	Console walkthrough, agent commands, team management
API Reference	REST API endpoints with curl examples
Runbooks	Incident response, upgrade procedure, pilot go-live checklist
Roadmap	What's shipped and what's coming

---

Production deployment (Helm)

For production Kubernetes deployments:

helm install vibops ./helm/vibops \
  -n vibops --create-namespace \
  -f my-values.yaml \
  --set postgresql.auth.password=<strong-password> \
  --set core.secret.jwtSecretKey=<secret> \
  --set agent.secret.llmApiKey=<api-key>

See docs/installation.md for the full Helm configuration reference.

---

Support

• Documentation: docs/installation.md
• Health check: make check
• Support: david@vibops.ai

---

© VibOps 2026 — Proprietary licence. This repository contains installation files only. Source code is not included.