Update data

Target

One of the biggest advantages of RAG compared to Fine-tuning (retraining) a model is the ability to update data quickly. When a business has new regulations, you simply need to ingest them into the Knowledge Base, and the AI will “learn” them immediately.

In this section, we will simulate the following scenario:

1. Ask the AI for a piece of non-existent information (The AI will answer that it doesn’t know).
2. Provide that information to the system by uploading a new file.
3. Ask the same question again to witness the AI answer correctly immediately.

Implementation Steps

Step 1: Verify the initial “lack of knowledge”

We need to confirm that the current AI knows nothing about the confidential information we are about to create.

1. Return to the Streamlit Chatbot interface (created in Part 5) or use the Test Knowledge Base window on the Console.
2. Ask a question about hypothetical fake information.
   • Example: “What is the activation code for Project Omega?”
3. Observe the result: The AI will answer that it cannot find the information in the provided documents or will attempt a generic answer (if not restricted).

Step 2: Create new data

We will create a text file containing this “secret” to ingest into the system.

1. On your computer, open Notepad (Windows) or TextEdit (Mac).
2. Copy and paste the following content into the file:

   CONFIDENTIAL NOTICE:
   The secret Project Omega officially launches on 01/12/2025.
   The activation code is: "AWS-ROCKS-2025-SINGAPORE".
   The Project Manager is Mr. Robot.
   Please keep this information strictly confidential.
   

3. Save the file as: secret-project.txt.

You can download the file here: TXT format file

Step 3: Upload and Sync

Now, we will feed this new knowledge into the AI’s “brain”.

1. Access the S3 Console, navigate to your old bucket (rag-workshop-demo).

2. Click Upload -> Add files -> Select the secret-project.txt file -> Upload.

3. Switch to the Amazon Bedrock Console -> Select Knowledge bases from the left menu.
4. Click on your Knowledge Base name.
5. Scroll down to the Data source section, select the data source (s3-datasource).
6. Click the Sync button (Orange).
7. Wait: Wait for about 30 seconds to 1 minute until the Status column changes from Syncing to Available.

Step 4: Verify again (The “Wow” Moment)

The system now possesses the new knowledge. Let’s challenge the AI once again.

1. Return to the Streamlit Chatbot interface (No need to reload the page or restart the server).
2. Ask the exact same question as before: “What is the activation code for Project Omega?”
3. Expected Result:
   • The AI answers correctly: “The activation code is AWS-ROCKS-2025-SINGAPORE”.
   • The AI cites the source as the secret-project.txt file.

Conclusion

You have just witnessed the true power of RAG!

• No code editing required.
• No model retraining required.
• Simply Sync the data.

Your Chatbot has become smarter and updated with the latest information in just a few simple steps. This is exactly why businesses choose this solution to build internal virtual assistants.