Skip to main content
Version: Next

Quick Start

This guide walks you through the full flow of logging into NuFi for the first time, deploying an AI model, and testing it, in 5 steps.

Estimated time: about 10 minutes (excluding model loading time)

Prerequisites​

Confirm the following before starting the Quick Start.

ConditionHow to check
NuFi installation completeAsk your administrator, or see the Installation Guide
Dashboard accessibleOpen the NuFi dashboard URL in a browser and verify access
Account issuedReceive a login account (email) from your administrator

Step 1: Log In​

Dashboard login

Click the ① Login button in the upper right. OAuth-based authentication is supported.

After login, the project selection dropdown in the top header becomes active.

Step 2: Create a Project​

A project is the unit of resource isolation. Labs, Servings, and Volumes are all managed per project.

Project creation

  1. Click the ② New Project button in the top header.
  2. Enter the project name.
    • Lowercase letters, digits, and hyphens (-) / at least 3 characters (e.g., my-first-project)
  3. Click the Create button.

The created project is automatically selected in the dropdown.

Step 3: Deploy a Model​

In the left sidebar, click Serving.

Serving list

  1. Click the ① Create button.
  2. Fill in Step 1 — Basic information.
    • Service Name: deployment name (e.g., llm-demo)
    • Service Template: the inference framework to use (e.g., vLLM)
  3. Click Next to move to Step 2 — Detailed settings.
    • Image: select the inference server container image
    • Accelerator Type / Count: select the GPU or NPU to use
    • When using vLLM, enter the Huggingface model name in the Model field
  4. Click Next and review the settings in Step 3 — Review & Deploy.
  5. Click the Deploy button.

When you return to the Serving list, the status changes from Pending → Running. Model loading may take several minutes.

Step 4: Test in the Playground​

Once the status is Running, click Playground in the left sidebar.

  1. From the dropdown at the top of the panel, select the Serving you just deployed.
  2. Enter a message in the bottom input box and click the Send button.
  3. Watch the model response stream in.

Using the + button you can add panels and place multiple Servings side by side to compare responses.

Step 5: Find the API Endpoint​

A deployed model provides an OpenAI-compatible API endpoint.

Serving list

Click the ② Connect button to find the endpoint URL.

The endpoint URL has the form https://<host>/<project>/<deployment-name>. Append the inference framework's API path (e.g., /v1/chat/completions) to call it.

curl -X POST 'https://<deployment-endpoint>/v1/chat/completions' \
  -H 'Content-Type: application/json' \
  -H 'Authorization: Bearer <token>' \
  -d '{
    "model": "<model-name>",
    "messages": [{"role": "user", "content": "Hello"}]
  }'