Articles in this section

Using AWS Bedrock and RAG with Document Upload

Avatar
Molly
  • Updated
Follow

This article explains the AI Provider settings in Brilliant Assessments. It covers how to switch your AI interactions from Azure OpenAI to Anthropic Claude on Amazon Bedrock, how to upload reference documents that ground the platform's AI features, and how to associate those documents with individual assessments.

By default, Brilliant Assessments uses Azure OpenAI for all AI interactions. Enabling AWS Bedrock + RAG routes those interactions to Anthropic Claude on Amazon Bedrock and unlocks document-grounded responses. RAG stands for Retrieval Augmented Generation, a technique in which the AI retrieves relevant passages from your uploaded documents and uses them as source material when generating output.

Tip: Use AWS Bedrock + RAG when you want AI features to answer from your own source material (for example a methodology, a framework, or a book you have written) rather than from general knowledge. It also provides improved handling of large AI queries.

On this page

1. What AWS Bedrock and RAG provide

Enabling AWS Bedrock + RAG changes two things about how your site uses AI.

  1. Provider. AI interactions are routed to Anthropic Claude on Amazon Bedrock instead of Azure OpenAI.
  2. Document grounding. You may upload reference documents that the platform's AI features draw on as source material. When a document is associated with an assessment, the AI answers primarily from that document.

This setting is configured per site and applies across the platform once enabled.

2. Which AI features use uploaded documents

Once AWS Bedrock + RAG is enabled and documents are associated with an assessment, the following AI features may use those documents as grounding material:

When writing the prompt for any of these features, you may direct the AI to a specific part of a document (for example a particular chapter or section) to focus the output. For guidance on prompt wording, see Writing Effective AI Interpret Prompts.

3. Enable AWS Bedrock + RAG

  1. Navigate to Site Settings > Integrations > AI Provider.
  2. Select the Use AWS Bedrock + RAG checkbox.
  3. Save your settings.

Once enabled, your AI interactions are routed to Anthropic Claude on Amazon Bedrock, and the Upload panel becomes available on the same tab.

Site Settings, Integrations, AI Provider tab showing the Use AWS Bedrock + RAG checkbox, the Upload panel, and the documents table

Note: Switching the AI provider affects how AI features generate output across your site. If you have existing AI Interpret content, AI Discuss content, or AI-built assessment material, review the output after enabling to confirm it still meets your expectations.

4. Upload reference documents

Use the Upload panel on the AI Provider tab to add the documents you want AI features to use as grounding material.

  1. In the Upload panel, click Choose File.
  2. Select a document from your computer.
  3. Click Upload.

The document is sent to Amazon Bedrock for ingestion and appears in the document list below the panel.

Supported file types: PDF, DOCX, TXT, MD, and HTML.

Note: Excel files are not supported at this time.

You may upload many documents. There is no practical limit on the number of reference documents you can add.

5. Confirm a document is indexed

After upload, Amazon Bedrock ingests the document. A document is only available for AI retrieval once its Status shows Indexed.

  1. Locate the document in the list below the Upload panel.
  2. Review the Status column.
  3. If the status is not yet Indexed, click Refresh status to update it.

The document list displays the file name, type, size, status, upload date, and the user who uploaded it.

Note: Indexing is not instantaneous. If a newly uploaded document is not yet indexed, wait a short time and click Refresh status again.

6. Associate documents with an assessment

Uploading a document makes it available to your site, but each assessment must specify which documents it should use. You associate documents from the assessment itself.

  1. Open the assessment and navigate to Assessment Settings.
  2. Select the Options tab. Read more about the Options Tab.
  3. Locate the AI Project Documents section near the bottom of the tab.
  4. Select the checkbox beside each document this assessment should use.
  5. Click Save associations.
Assessment Settings, Options tab showing the AI Project Documents section with the document checkbox list and Save associations button

Note:

  • The AI Project Documents section only appears when AWS Bedrock + RAG (Anthropic) is enabled for your site.
  • Only documents in the Indexed state can be associated with an assessment. If a document you uploaded does not appear in the list, confirm its status is Indexed on the AI Provider tab (see Confirm a document is indexed).

Once associated, the assessment's AI features (AI Build, AI Interpret, and AI Discuss) draw primarily on the selected documents. For example, if you have written a book and uploaded it, you may associate it with one assessment or with several, and the AI will answer the assessment's questions primarily from that source material.

7. Manage and remove documents

You may remove a document at any time from the document list on the AI Provider tab.

  1. Locate the document in the list.
  2. Click Delete in the Actions column.

Removing a document makes it unavailable for future AI retrieval. Review any assessments that referenced the document to confirm their AI features still produce the output you expect.

FAQs & Troubleshooting

Use the questions below to troubleshoot common setup issues.

+ What is the difference between Azure OpenAI and AWS Bedrock + RAG?

Azure OpenAI is the default AI provider. AWS Bedrock + RAG routes AI interactions to Anthropic Claude on Amazon Bedrock and adds the ability to ground AI output in documents you upload. AWS Bedrock + RAG also provides improved handling of large AI queries.

+ Do I have to upload documents to use AWS Bedrock + RAG?

No. You may enable AWS Bedrock + RAG to route AI interactions to Anthropic Claude without uploading any documents. Document upload is an optional capability for grounding AI output in your own source material.

+ Why is my document status not Indexed?

The document is still being ingested by Amazon Bedrock. Indexing is not instantaneous. Click Refresh status to update the status. A document cannot be used by AI features until its status is Indexed.

+ Why can I not see the AI Project Documents section in Assessment Settings?

The AI Project Documents section in the Options tab only appears when AWS Bedrock + RAG is enabled for your site. Confirm the setting is enabled under Site Settings > Integrations > AI Provider.

+ Why does an uploaded document not appear in the AI Project Documents list?

Only documents in the Indexed state can be associated with an assessment. If a document does not appear in the list, return to Site Settings > Integrations > AI Provider, confirm the document's status is Indexed, and click Refresh status if it is not.

+ Can I upload an Excel file?

No. Supported file types are PDF, DOCX, TXT, MD, and HTML. Excel files are not supported at this time.

+ Can I focus the AI on a specific part of a document?

Yes. When writing the prompt for AI Build, AI Interpret, or AI Discuss, you may direct the AI to a particular part of a document, such as a specific chapter or section, to focus the output.

+ Can one document be used by more than one assessment?

Yes. A document is uploaded once at the site level and may be associated with any number of assessments from each assessment's Options tab.

Related articles

Comments

0 comments

Please sign in to leave a comment.