Skip to main content

​
Introduction

Providers are external repositories and services that store the AI models you want to deploy to nodes. ZEDEDA Edge Inference Services connects to these providers to browse available models and import them into your model library. Providers were previously called Sources. Supported provider types include:
  • Hugging Face: public and private model repositories on the Hugging Face Hub.
  • NVIDIA NGC: NVIDIA’s catalog of GPU-optimized models and containers.
  • Azure ML: models hosted in an Azure Machine Learning workspace.
  • Azure Blob Storage: model files stored in an Azure Blob Storage container.
  • MLflow: models tracked in an MLflow model registry.
  • Amazon S3: model files stored in an Amazon S3 bucket.
  • Amazon SageMaker: models registered in Amazon SageMaker.
  • Qualcomm AI Hub: models from the Qualcomm AI Hub catalog.
  • Local (file upload): model files uploaded directly from your local machine.

​
Prerequisites

  • You have the Organization Admin or Super Admin role in your organization. See Organizations, Users, and Roles.
  • You have the credentials required by the provider type you want to add (API token, access keys, or service principal details).

​
Add a Provider

  1. In the bottom navigation bar, select Inference.
  2. In the sidebar, select Inference > Deploy > Providers.
  3. Select the contextual + create action to open the Add Provider wizard.

​
Provider Type

Select the provider type that matches your model source. The credential and configuration fields in the following steps adjust based on the type you select.

​
Basic Information

Enter a Name and optional Description for the provider. The name identifies the provider in lists and log entries. For provider types that require a base URL (such as MLflow), enter the URL of your provider endpoint.

​
Credentials

Enter the authentication credentials for your provider type:

​
Configuration

Enter additional connection details for your provider type:

​
Review

Review the provider details and select Create Provider to save the configuration. The new provider appears in the Providers list.

​
View the Providers List

  1. In the bottom navigation bar, select Inference.
  2. In the sidebar, select Inference > Deploy > Providers.
The list shows all providers in your organization with the following columns: Name, Type, URL, Status, Description, and Updated. The Status column shows:
  • Active: the provider is available for browsing and imports. New providers start in this state, and a successful connection test returns a provider to it.
  • Inactive: the provider has been set to inactive. Imports from the provider are blocked.
  • Error: the most recent connection test failed. Imports from the provider are blocked until a test passes.
To filter the list, use the Type or Status filter controls at the top of the page.

​
View Provider Details

  1. Open the Providers list.
  2. Select View Details from the row actions menu, or select the provider name.
The provider detail page has the following sections:
  • Identity: name, description, and provider type.
  • Credentials: the credential values on file (sensitive fields are masked).
  • Configuration: additional connection settings such as region, bucket, or workspace.
  • Endpoint: the base URL for the provider connection.
  • Model Browser: available for all provider types. Browse and import models from the provider directly in the console.

​
Test a Provider Connection

Use the Test Connection action to verify that ZEDEDA Edge Inference Services can reach the provider using the credentials and configuration on file. Only Super Admin can run it.
  1. Open the provider detail page.
  2. Select the Test Connection button (plug icon).
A confirmation message appears indicating whether the connection succeeded or failed. A failed test sets the provider to an error state, and imports from it are blocked until a test passes. If the test fails, verify that the credentials and configuration values are correct and that your network allows outbound access to the provider endpoint.

​
Browse Provider Models

The Model Browser is available for all provider types.
  1. Open the provider detail page.
  2. In the Model Browser section, use the search field to find models by name, or navigate the directory structure using the breadcrumb trail. Search is available only for Hugging Face, MLflow, NVIDIA NGC, and Qualcomm AI Hub providers.
  3. Select a model row to expand it and view file details. Version lists are available only for NVIDIA NGC and Qualcomm AI Hub models.

​
Import a Model from a Provider

After locating a model in the Model Browser, you can import it into your ZEDEDA Edge Inference Services model library.
  1. In the Model Browser section, expand the model row you want to import.
  2. Select the Import button (the label reads “Import Model”, “Import File”, or “Import Directory” depending on the provider and item type).
  3. Complete the import wizard to specify model details and the target organization. For NVIDIA NGC models, the model identifier must include a version, in the form org/[team]/model:version.
When the import starts, ZEDEDA Edge Inference Services redirects you to the Jobs page in Infrastructure Monitor, where you can track the progress of the import job.

​
Edit a Provider

  1. Open the provider detail page.
  2. Select Edit Provider.
  3. Update the provider name, description, credentials, or configuration as needed.
  4. Save the changes.

​
Delete a Provider

  1. Open the Providers list.
  2. Select Delete from the row actions menu, or open the provider detail page and select the Delete Provider button (trash icon).
  3. Confirm the deletion in the dialog.
Deleting a provider does not remove any models that were already imported from that provider.

​
In This Section