Create Endpoints

Overview

Endpoints expose your services and models to external clients, providing HTTP/REST API access for inferencing and application integration.

Endpoint Types

Jupyter Endpoint

Expose Jupyter notebooks for remote access:

• Use Case - Remote development, collaboration
• Access Method - Web browser
• Authentication - Tenant credentials

LLM Endpoint

Expose Large Language Model services:

• Use Case - API-based LLM access
• Access Method - REST API
• Authentication - API keys

Model Inference Endpoint

Expose deployed models for inference:

• Use Case - Model serving, predictions
• Access Method - REST API
• Authentication - API credentials

TLS configuration

When you create an endpoint, you choose the TLS configuration method

Option	Description
CA SIGNED	Certificate signed by a trusted Certificate Authority (CA). Use when you have a CA-issued certificate.
SELF SIGNED	Certificate generated and signed by the system. Use for internal or test environments; browsers may show a security warning until the certificate is trusted.
UPLOADED	Your own certificate and key uploaded to Bridge. Use when you have a custom certificate (e.g., from an internal CA or another provider).

Select the option that matches your environment and security requirements.

Prerequisites

• Tenant Admin access — You must log in as a Tenant Admin to create endpoints.
• Cluster (for JupyterHub endpoint) — A JupyterHub cluster must already exist so you can attach it to the endpoint.
• Cluster (for model endpoint) — A Ray-enabled Kubernetes cluster used for model serving must exist so you can attach it to the endpoint.

Endpoint Types

Type	Use case	Access method	Authentication
JupyterHub Endpoint	Remote access to JupyterHub notebooks; collaboration	Web browser	Tenant credentials
Model Endpoint	Required to deploy a model; API-based access once deployed	REST API	API keys or credentials

Choose the endpoint type: JupyterHub endpoint for a JupyterHub cluster, or model endpoint when you need to deploy and access a model.

---

Create JupyterHub Endpoint

Use this flow to create an endpoint that exposes your JupyterHub server for remote browser access.

Step 1: Navigate to Endpoints

1. Log in to Armada Bridge as a Tenant Admin.
2. In the left sidebar, click Endpoints.

Step 2: Create New Endpoint

1. Click Create New Endpoint.

Step 3: Configure JupyterHub Endpoint

1. Enter the Endpoint name (e.g., jupyter-endpoint-01) and Description.
2. Select the JupyterHub cluster you want to expose (the cluster must already exist).
3. Select TLS configuration — CA SIGNED, SELF SIGNED, or UPLOADED (see TLS configuration above).
4. Click Create.

The endpoint is created and appears in the endpoints list.

Step 4: View Endpoint Details

Click the created endpoint name to view its details, including endpoint name, hostname, the cluster attached to the endpoint, and created time.

---

Create Model Endpoint

Use this flow to create a model endpoint. You need this endpoint to deploy a model; create it before deploying a model. After deployment, the model is accessed through this endpoint via the API.

Step 1: Navigate to Endpoints and Create New Endpoint

1. Log in to Armada Bridge as a Tenant Admin.
2. Open Endpoints in the left sidebar.
3. Click Create New Endpoint.

Step 2: Configure Model Endpoint

1. Enter the Endpoint name (e.g., llmendpoint or model-endpoint) and Description.
2. Enter the Hostname (e.g., your chosen hostname for the model API).
3. Select the cluster used for model serving (the cluster that will serve your deployed models).
4. Select TLS configuration — CA SIGNED, SELF SIGNED, or UPLOADED (see TLS configuration above).
5. Click Create.

The endpoint is created and appears in the list.

Step 3: View Endpoint Details

Click the created endpoint name to view its details, including endpoint name, hostname, the cluster attached to the endpoint, and created time.

Hosts file entry (when hostname is not resolvable)

To access models deployed through this endpoint from your local machine, you may need to add an entry to your /etc/hosts file.

This is required if the endpoint hostname is not resolvable via public or internal DNS.

Add an entry that maps the endpoint hostname to the public IP address of the GPU VM or Bare Metal node associated with the selected cluster. For example:

<GPU_VM_Public_IP> <endpoint_hostname>

Replace <GPU_VM_Public_IP> with the actual public IP and <endpoint_hostname> with the hostname you configured for the endpoint.

Using Endpoints

Access Jupyter Endpoint

1. Open endpoint URL in browser
2. Login with credentials
3. Start using Jupyter

https://jupyter-endpoint.your-domain.com

Call LLM Endpoint

Make API requests to your LLM:

curl -X POST https://llm-endpoint.your-domain.com/v1/completions \
  -H "Authorization: Bearer <your-api-key>" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "model-name",
    "prompt": "Your prompt here",
    "max_tokens": 100
  }'

Call Model Endpoint

Get predictions from deployed models:

curl -X POST https://model-endpoint.your-domain.com/predict \
  -H "Authorization: Bearer <your-api-key>" \
  -H "Content-Type: application/json" \
  -d '{
    "data": [1, 2, 3, 4, 5]
  }'

Manage Endpoints

View Endpoints

List all active endpoints:

• Endpoint status
• Access counts
• Resource usage
• Configuration

Monitor Usage

Track endpoint usage:

• Request counts
• Response times
• Error rates
• Bandwidth

Update Configuration

Modify endpoint settings:

1. Select endpoint
2. Click Edit
3. Update settings
4. Save changes

Security Best Practices

API Key Management

• Rotate keys regularly
• Use different keys for different services
• Store keys securely
• Don't commit keys to version control

Access Control

• Restrict endpoint access by IP
• Use authentication tokens
• Enable HTTPS/TLS
• Monitor access logs

Rate Limiting

• Set appropriate rate limits
• Monitor for abuse
• Adjust limits based on usage
• Alert on suspicious activity

Next Steps

• Deploy Models
• Access Model Playground
• Configure Workloads