Quick start

From a fresh checkout to a working chat at http://localhost:3080 in about ten minutes.

This guide stands up the full NUFI platform — chat, console, gateway, observability, monitoring — on your laptop. For the chat-only deployment, see nufi-chat.

1. Authenticate to the container registry

The NUFI images ship as private container images. You only have to log in once per machine.

# Create a Personal Access Token with `read:packages` scope at
# https://github.com/settings/tokens/new
echo ghp_xxxxxxxxxxxxxxxxxxxx | docker login ghcr.io -u <your-gh-username> --password-stdin

2. Clone the platform repository

git clone https://github.com/dudaji-vn/npuops-platform.git
cd npuops-platform

3. Run the bootstrap script

./scripts/bootstrap.sh

It will ask:

Which AI backend? Pick ollama if you have Ollama running locally, remote for a model server on your network, cloud for an OpenAI / Anthropic / Together / Groq API key, or skip to bring the stack up without a model.
Which model? (Ollama only.) Pick from installed models or let the script pull one for you.

Non-interactive alternatives:

./scripts/bootstrap.sh --backend ollama --model qwen2.5:3b   # local
./scripts/bootstrap.sh --backend skip                         # stack only

The script:

Generates every secret in .env.
Pulls every container image (~2 GB on first run).
Brings the stack up.
Waits for the gateway to report healthy.
Runs the smoke test (skipped if --backend skip).
Prints the URLs and the initial admin credentials.

4. Open the URLs

URL	What	Default credentials
http://localhost:3080	Chat	Register on first visit
http://localhost:3001	Developer console	SSO from the chat
http://localhost:3000	Trace viewer	Email + password printed by bootstrap
http://localhost:3030	Dashboards	Admin credentials in `.env`
http://localhost:4000/ui	Gateway admin	Master key in `.env`

5. Send your first message

Open http://localhost:3080.
Register an account. The first registered user gets the ADMIN role automatically.
Pick a model from the dropdown (fetched live from the gateway).
Type a message.
Open http://localhost:3000 → Traces to see the request you just sent.

What is next

The smoke test (./scripts/smoke-test.sh) is a good idempotent regression check after any change. See Verify install.
To register another model later, use ./scripts/add-model.sh.
To call NUFI from your own code, see End user → API keys.
For production deployment, see Deployment.