AWS Builder Center

Builder Center

Level Up Your Document Intelligence: Building Your First Amazon Bedrock Knowledge Base

Level Up Your Document Intelligence: Building Your First Amazon Bedrock Knowledge Base

Building a RAG (Retrieval-Augmented Generation) System

Pranali Khose

Solutions Architect
Welcome to Day 3 of Challenge #2!  Today you'll learn how to build an AI-powered knowledge base using Amazon Bedrock that can understand, search, and answer questions about your documents with the precision.
In about 60 minutes, you'll go from scattered documents to a fully functional RAG (Retrieval-Augmented Generation) system. We're not just storing files - we're building a comprehensive AI assistant that can analyze content, extract insights, and provide contextual answers.

What You Will Accomplish Today

✅ Enable Amazon Bedrock model access for AI-powered document processing
✅ Set up S3 bucket for document storage and vector embeddings
✅ Create and configure your Bedrock Knowledge Base
✅ Upload and synchronize your documents
✅ Test your intelligent knowledge system with real queries

What You Will Need

💻 Your AWS account with appropriate permissions
📄 Document ready for upload. Download this file .
⏰ About 60 minutes
🔍 Basic understanding of AI/ML concepts
🧠 Curiosity about intelligent document processing
Note : Excited to learn about the visual effects design of League of Legends? Download this file  - we're going to use this comprehensive VFX style guide later in our blog to explore how Riot Games creates stunning spell effects that balance visual spectacle with competitive gameplay clarity.

Step 1: Enable Amazon Bedrock Model Access 🔑

Before we can harness the power of AI, we need to unlock access to the foundation models that will drive your knowledge base. Think of this as your Ascension moment - you're gaining access to abilities that will transform how you interact with documents. We'll use embedding models to convert documents into vectors and language models for generating responses.

REQUEST MODEL ACCESS

  1. Navigate to Bedrock Console - Go to the AWS Console and search for "Bedrock"
  2. Access Model Access - In the left sidebar, click "Model access" under "Configure and learn"
  3. Modify Model Access - Click the "Modify Model Access" button to modify your model permissions
  4. Select Required Models:
    • Titan Text Embeddings V2 - For converting documents to vector embeddings
    • Claude 3 Haiku - For generating responses (cost-effective option)
  1. Submit Request - Click "Save changes" and wait for approval (usually instant for these models)
Important: Model access may take a few minutes to activate. You'll see a green "Access granted" status when ready.

UNDERSTANDING THE MODELS

Titan Text Embeddings V2: This model converts your text into numerical vectors (embeddings) that capture semantic meaning. Documents with similar content will have similar vector representations, enabling intelligent search and retrieval.
Claude Models: These are the language models that will read retrieved document chunks and generate human-like responses to your questions. Think of them as the "brain" that understands context and formulates answers.

Step 2: Create Your Document Storage Infrastructure 📦

Time to establish your Nexus - the central hub where all your documents will live securely. We'll create an S3 bucket that serves as the foundation for your entire knowledge system.

Let's set up the S3 bucket to store the documents

CREATE THE DOCUMENT BUCKET

  1. Navigate to S3 - Search for "S3" in the AWS Console
  2. Create Bucket - Click "Create bucket"
  3. Bucket Name: Enter something like rift-rewind-ai-documents-[your-name]
    • Example: rift-rewind-ai-documents-alex
    • Bucket names must be globally unique across all AWS accounts
    • Use lowercase letters, numbers, and hyphens only
  1. Keep default security - Ensure "Block all public access" is checked ✅
  2. Create Bucket - Click "Create bucket"
You'll know you're done when you see your new bucket in the list with a green "Successfully created bucket" message

Step 3: Upload Your Documents to S3

  1. Access your bucket - Click on the bucket you created in Step 2
  2. Upload documents - Click "Upload" then "Add files"
  3. Select your files - Upload your downloaded PDF file (e.g., VFX file). (Refer Download this file . link if not downloaded earlier.)

Step 4: Create Your Bedrock Knowledge Base 🧠

Now for the exciting part - creating the knowledge base that will make your documents searchable and queryable.

This is where the real magic happens. We're building your Hextech Core - the intelligent system that will power all your document interactions. Every piece of information will be processed, indexed, and made instantly accessible through AI.

LAUNCH KNOWLEDGE BASE CREATION

  1. Navigate to Bedrock - Go back to the Bedrock console
  2. Access Knowledge Bases - Click "Knowledge Bases" under "Build"
  3. Create Knowledge Base - In the Create dropdown list, choose "Knowledge Base with vector store"

STEP 1: PROVIDE KNOWLEDGE BASE DETAILS

  1. Knowledge Base Details:
    • Name: Enter something like riot-docs-kb-[your-team]
    • Description: Enter something like "AI-powered knowledge base for document Q&A and retrieval" (optional)
    • IAM Permissions: Select "Create and use a new service role"
    • Data source type: Choose "Amazon S3"
  1. Choose Next to proceed to the data source configuration
Why a new service role? This creates the necessary permissions for Bedrock to access your S3 bucket, generate embeddings, and manage the vector store automatically.

STEP 2: CONFIGURE DATA SOURCE**:**

  • Data Source Name: Assign a descriptive name (e.g., riot-knowledge-base)
  • Data Source Location: Choose "This AWS account"
  • S3 URI: Click "Browse S3" and select your S3 bucket created in step 2.
  • Parsing strategy: Select "Amazon Bedrock default parser"
  • Chunking strategy: Choose "Fixed-size chunking"
  1. Choose Next to proceed to data storage and processing

STEP 3: CONFIGURE DATA STORAGE AND PROCESSING

  1. Embeddings Model:
    • Model: Titan Text Embeddings V2
    • Pricing: On-demand (pay per use)
  1. Vector Store Configuration:
    • Creation method: "Quick create a new vector store" ✅
    • Vector store type: "Amazon S3 Vectors - Preview"
  1. Click Next
Why S3 Vectors? It's optimized for cost-effective, durable storage of vector embeddings. Perfect for knowledge bases that don't need millisecond query response times but want to minimize ongoing costs.

STEP 4: REVIEW AND CREATE

  1. Review Configuration - Double-check all settings
  2. Create Knowledge Base - Click "Create knowledge base"
Creation Time: The knowledge base creation takes 5-10 minutes. AWS is setting up the vector storage, configuring permissions, and preparing the ingestion pipeline. Wait until the status changes to "Available."

Step 5: Synchronize Your Data Source 🔄

Time to Recall your documents into the system. This process transforms your static files into an intelligent, searchable knowledge network. The AI will read, understand, and index every piece of content, making it ready for instant retrieval.
  1. Access your knowledge base - Click "Knowledge Bases" under "Build"
  2. Select knowledge base- Select your knowledge base.
  3. Initiate sync - Click "Sync"
  4. Wait for completion - Wait for the "Sync completed for data source" message

Step 6: Test Your Intelligent Knowledge Base 🧪

Time to see your AI-powered knowledge system in action!

Your knowledge base is now ready for its First Blood - the moment you discover just how powerful intelligent document search can be. Let's put it through its paces with some targeted queries.

QUERY YOUR KNOWLEDGE BASE

  1. Navigate to your knowledge base - Select your knowledge base from the list
  2. Access test interface - Click "Test Knowledge base"
  3. Select Model - Choose "Anthropic" then "Claude 3 Haiku" and click "Apply"

TEST QUERIES

Try these sample questions:

  • "Tell me more about Areas of focus"
  • "What are the visual effects in LoL?"
  • "How to determine value range?"

UNDERSTANDING THE RESULTS

Each response will include:

  • Generated Answer - AI-crafted response based on your document
  • Source Citations - References to specific document chunks used
Pro Tip: If answers seem incomplete, try rephrasing your question or adding more context. The AI works best with specific, well-formed questions.

Mission Complete: Your Knowledge Base is Ready! ✅

✅ Verified "Active" Knowledge Base Status
✅ Successfully Tested queries via AWS Management Console
✅ Configured Document Processing Pipeline
✅ Testing & Optimization Complete
You've transformed from document storage user to advanced RAG practitioner! Your system now provides:
🎯 Precision Querying: You can extract specific insights from your documents with surgical precision, whether you're analyzing technical specifications, policy documents, or research papers.
⚡ Automated Analysis: Your knowledge base can process complex queries intelligently, enabling automated Q&A, content discovery, and insight extraction.
🔧 Advanced Configuration: You understand how to fine-tune retrieval settings and optimize query performance for different types of content analysis.

Next Level: Building Advanced Query Systems 🤖

Ready to take your document intelligence to the next level? In our next challenge, we'll build Lambda functions for programmatic access and create automated workflows that can conduct autonomous research using your Knowledge Base as the primary data source.
Your RAG knowledge base is the foundation - now we're building the intelligence layer that makes it truly powerful for enterprise document processing and analysis.
# rift-rewind# rift-rewind-challenge-2# hackathon# challenge
Any opinions in this article are those of the individual author and may not reflect the opinions of AWS.
Enjoyed reading this content? Let the author know!

Your likes, comments, shares, and saves help creators reach more builders.

Loading recommendations

Loading article