> ## Documentation Index
> Fetch the complete documentation index at: https://docs.blinkops.com/llms.txt
> Use this file to discover all available pages before exploring further.

# AI Settings

## AI Features

**Enable Builder Copilot test data access**-  When enabled, the Builder Copilot can access workflow testing execution data from the current session to generate more context-aware and precise configuration suggestions. When disabled, suggestions are based only on the workflow definition and user input.

<Frame>
  <img src="https://mintcdn.com/blinkops-2/agJo1XZ4GmiqPHjk/img/AccountAdmin/AIFeatures.png?fit=max&auto=format&n=agJo1XZ4GmiqPHjk&q=85&s=8abac1a4e0c8890c1bd623f5b80bcf9f" width="1727" height="1082" data-path="img/AccountAdmin/AIFeatures.png" />
</Frame>

***

## AI Models

Blink supports two options for AI infrastructure:

* **Blink** *(recommended)* – Uses Blink-managed AI infrastructure through Blink's Amazon Bedrock account. This is the default option and provides a fully managed experience, with all AI requests routed through Blink.

* **Amazon Bedrock** – Uses your own Amazon Bedrock account. When selected, all AI-related requests are executed through your AWS infrastructure, giving you full control over your AI resources and billing.

<Frame>
  <img src="https://mintcdn.com/blinkops-2/agJo1XZ4GmiqPHjk/img/AccountAdmin/AIModel.png?fit=max&auto=format&n=agJo1XZ4GmiqPHjk&q=85&s=1b884e088694704fe79fb19d3bd10c9e" width="1727" height="1082" data-path="img/AccountAdmin/AIModel.png" />
</Frame>

***

### Amazon Bedrock Prerequisites

1. An AWS account with permission to deploy CloudFormation stacks and create IAM roles.

### Limitations

**Important to Note**: [Knowledge sources](/docs/agent-builder/building-the-agent), in the Agent Builder, are not currently supported when using [Amazon Bedrock as the AI provider](/docs/blink-platform/account-management/ai-settings#configuration-amazon-bedrock-as-the-ai-provider). Knowledge functionality is available only when using [Blink as the AI provider](/docs/blink-platform/account-management/ai-settings#ai-models)

***

### Configuration Amazon Bedrock as the AI Provider

<Steps>
  <Step title="Open AI Settings">
    Navigate to **Account Settings → AI Settings → AI Models**.
  </Step>

  <Step title="Select &#x22;Bedrock&#x22; as the AI provider">
    In the **AI Provider** dropdown, select **Amazon Bedrock**.
  </Step>

  <Step title="Note your External ID">
    Blink automatically generates a unique **External ID** for your account and displays it on screen.

    <Note>
      **What is the External ID?**

      It is a secret token that Blink embeds in every `AssumeRole` request. Your IAM trust policy must require it, which prevents any other AWS account from assuming your role even if they know your Role ARN. Do not share it.
    </Note>
  </Step>

  <Step title="Enter your AWS Account ID and IAM Role Name">
    Fill in:

    * **AWS Account ID** — the 12-digit numeric ID of the AWS account where Blink should operate (e.g. `123456789012`).

    * **IAM Role Name** — the name you want to give the IAM role that Blink will assume. The default suggestion is `BlinkBedrockExecutionRole`; you can use any name that fits your naming conventions.
  </Step>

  <Step title="Deploy the IAM Role via CloudFormation">
    Click **Open in AWS Console**. This opens the AWS CloudFormation quick-create page with all parameters pre-filled:

    | Parameter   | Value                          |
    | ----------- | ------------------------------ |
    | Stack name  | `blink-bedrock-access`         |
    | External ID | your External ID from step 3   |
    | Role Name   | the name you entered in step 4 |

    In the AWS Console, scroll to the bottom, acknowledge the IAM capability checkbox, and click **Create stack**. Wait for the stack status to reach `CREATE_COMPLETE` (usually under 2 minutes).

    <Note>
      **What does the stack create?**

      A single IAM role in your account. The role's trust policy allows Blink's AWS account to assume it — but only when the correct External ID is presented.

      The role's permission policy grants the following actions, scoped to Anthropic foundation models and cross-region inference profiles only:

      * `bedrock:InvokeModel`
      * `bedrock:InvokeModelWithResponseStream`
      * `bedrock:Converse`
      * `bedrock:ConverseStream`
      * `bedrock:CountTokens`
    </Note>
  </Step>

  <Step title="Test the connection">
    Return to the Blink AI Settings page and click **Test connection**. Blink will attempt to assume the role and invoke a lightweight Bedrock call using the values currently in the form - **no need to save first**.

    A success confirmation means the role and permissions are correctly configured.
  </Step>

  <Step title="Save the configuration">
    Click **Save changes** to activate the integration.
  </Step>
</Steps>

<Note>
  **Reverting to Blink's Default Provider**:

  Switch the **AI Provider** dropdown back to **Blink** and save. The IAM role in your AWS account is not deleted automatically — you can clean it up by deleting the `blink-bedrock-access` CloudFormation stack.
</Note>

***

### Required Bedrock Model Access

The customer must enable access to the following cross-region inference profiles in their AWS account before the integration will work. All are from the **Anthropic** provider in Bedrock.

**How to enable:** AWS Console → Amazon Bedrock → Model access → Manage model access → select each model below → Request model access.

<Note>**Note**: All models in the table must be enabled </Note>

| Model             | Bedrock Model ID                                  |
| ----------------- | ------------------------------------------------- |
| Claude Sonnet 4.6 | `global.anthropic.claude-sonnet-4-6`              |
| Claude Haiku 4.5  | `global.anthropic.claude-haiku-4-5-20251001-v1:0` |

<Note>
  **How to Confirm a Model is Ready**:

  Go to **AWS Console → Amazon Bedrock → Playgrounds → Chat**, select the model, and send a test message. A successful response confirms the model is accessible.
</Note>

***