Skip to main content
The following information applies only to dedicated instance and in-VPC deployments of Unstructured Business.
Dedicated instance and in-VPC deployments require you to supply your own AI provider credentials. Unstructured does not provision shared API keys — you bring access credentials from the AI provider of your choice. This lets you use your preferred provider and apply any volume discounts you have negotiated with that provider to your Unstructured workloads. Credentials are required to use any of the following features: You must have an account with each AI provider you configure. Your provider bills usage directly to your account — these charges are not included in your Unstructured bill. If you do not pay these provider bills, the features that depend on that provider may stop working without prior notice.
The following steps do not apply to Let’s Go, Pay-As-You-Go, or Business SaaS accounts. This is because the costs for using the models that Unstructured provides are already factored into the pricing structures for these account types.For Unstructured Business dedicated instance and in-VPC deployments, completing the following steps enables you to better manage and reduce your overall costs by negotiating lower overall billing rates with Unstructured for its core features, while also leveraging any separately negotiated volume discounts with the AI providers that you want to use.

Keeping model provider traffic private

If your network policy requires AI provider traffic to stay off the public internet, use a model provider accessible through your cloud provider’s private networking. AWS — Use Amazon Bedrock to access foundation models through VPC endpoints. Unstructured configures the Bedrock VPC endpoint on its side; you grant access via IAM policy. Traffic never leaves the AWS network. For dedicated instance setup, see Connecting to services with AWS PrivateLink. To grant Bedrock access by assuming a cross-account IAM role instead of supplying static credentials, see Enterprise Connect for Amazon Bedrock. Azure — Use Azure OpenAI Service to access models through a private endpoint on the Microsoft backbone. Unstructured creates a private endpoint in its VNet toward your Azure OpenAI resource. For dedicated instance setup, see Connecting to services with Azure Private Link.

Configure an AI provider

To have Unstructured use an AI model, you must first configure the model’s AI provider for each target Unstructured workspace within each corresponding Unstructured organizational account in your dedicated instance or in-VPC deployment, as follows. This procedure assumes that you have already created an account with the model’s AI provider, generated access credentials for your account with that AI provider, and obtained any other miscellaneous settings that the AI provider might also require. For more information, see the AI provider’s documentation.
1

Sign in

Sign in to your Unstructured account.
2

Select the workspace

In the top navigation bar, click the organizational account selector and select the account that contains the workspace. Then click the workspace selector and select the target workspace.
3

Open AI Providers

In the workspace navigation bar, click the AI Providers tab.
4

Open the provider configuration

In the list of available AI providers, click Configure next to the target provider.
5

Enter credentials and settings

In the Configure AI Provider dialog, enter your credentials and any other settings the AI provider requires.
6

Save the configuration

Click Save Configuration.
The configured AI provider is now ready to be used by all workflows in this workspace.

Update an AI provider’s configuration

For dedicated instance and in-VPC deployments, use Update configuration to replace an AI provider’s settings without deleting the provider. Re-enter every field, including every secret. Values already shown in the form are defaults, not your current settings. If you leave an optional field empty, Unstructured clears its value.
1

Sign in

Sign in to your Unstructured account.
2

Select the workspace

In the top navigation bar, click the organizational account selector and select the account that contains the workspace. Then click the workspace selector and select the target workspace.
3

Open AI Providers

In the workspace navigation bar, click the AI Providers tab.
4

Open the provider's actions menu

In the provider’s row, click the ellipsis (three dots) icon.
5

Choose Update configuration

Click Update configuration.
6

Enter the new configuration

Enter every setting required for the AI provider, including all secrets. Re-enter any optional settings you want to keep.
7

Confirm the update

In the provider configuration dialog, click Update configuration. In the confirmation dialog, click Replace configuration.
Workflows in this workspace that rely on this AI provider use the new settings on their next run.

Delete an AI provider’s configuration

You cannot delete an AI provider’s configuration while a workflow uses it. To change credentials or the access method, use Update configuration instead.
1

Sign in

Sign in to your Unstructured account.
2

Select the workspace

In the top navigation bar, click the organizational account selector and select the account that contains the workspace. Then click the workspace selector and select the target workspace.
3

Open AI Providers

In the workspace navigation bar, click the AI Providers tab.
4

Open the provider's actions menu

In the provider’s row, click the ellipsis (three dots) icon.
5

Choose Delete Configuration

Click Delete Configuration.
6

Confirm the deletion

In the confirmation dialog, click Delete Configuration.
You can configure the target AI provider again at any time.

No API support for AI provider management

You cannot use the Unstructured API to manage AI providers. You can manage AI providers only through Unstructured Pipelines.