AWS Bedrock: Native RAG Features

AWS Strengthens Enterprise RAG Offering

Amazon Web Services announces major improvements to Bedrock's RAG capabilities. Knowledge Bases become more powerful, agents more sophisticated, and integration with the AWS ecosystem more seamless.

"Enterprises want production-ready RAG without the complexity," explains Swami Sivasubramanian, VP AI/ML at AWS. "Bedrock Knowledge Bases v2 meets this demand."

New Features

Knowledge Bases v2

Knowledge Bases evolve significantly:

Simplified configuration:

DEVELOPERpython
import boto3

bedrock = boto3.client('bedrock-agent')

# Create a Knowledge Base
response = bedrock.create_knowledge_base(
    name="company-docs",
    description="Internal documentation",
    roleArn="arn:aws:iam::123456789:role/BedrockKBRole",
    knowledgeBaseConfiguration={
        "type": "VECTOR",
        "vectorKnowledgeBaseConfiguration": {
            "embeddingModelArn": "arn:aws:bedrock:us-east-1::foundation-model/cohere.embed-multilingual-v3"
        }
    },
    storageConfiguration={
        "type": "OPENSEARCH_SERVERLESS",
        "opensearchServerlessConfiguration": {
            "collectionArn": "arn:aws:aoss:us-east-1:123456789:collection/kb-collection"
        }
    }
)

# Add a data source
bedrock.create_data_source(
    knowledgeBaseId=response['knowledgeBase']['knowledgeBaseId'],
    name="confluence-docs",
    dataSourceConfiguration={
        "type": "CONFLUENCE",
        "confluenceConfiguration": {
            "sourceConfiguration": {
                "hostUrl": "https://company.atlassian.net",
                "hostType": "CLOUD"
            }
        }
    }
)

Chunking strategies are now directly configurable.

Improved RAG Agents

Bedrock agents now support complex workflows:

1. Multi-Knowledge Base

DEVELOPERpython
agent = bedrock.create_agent(
    agentName="support-agent",
    foundationModel="anthropic.claude-3-opus-20240229-v1:0",
    instruction="You are a support agent using multiple knowledge bases.",
    agentResourceRoleArn="...",
    knowledgeBases=[
        {"knowledgeBaseId": "kb-products", "description": "Product catalog"},
        {"knowledgeBaseId": "kb-support", "description": "FAQ and procedures"},
        {"knowledgeBaseId": "kb-internal", "description": "Internal documentation"}
    ]
)

2. Custom Actions

Agents can call Lambda functions:

DEVELOPERpython
bedrock.create_agent_action_group(
    agentId=agent_id,
    agentVersion="DRAFT",
    actionGroupName="order-management",
    actionGroupExecutor={
        "lambda": "arn:aws:lambda:us-east-1:123456789:function:OrderManagement"
    },
    apiSchema={
        "s3": {
            "s3BucketName": "api-schemas",
            "s3ObjectKey": "order-api.json"
        }
    }
)

3. Persistent Memory

Conversations persist automatically:

DEVELOPERpython
response = bedrock.invoke_agent(
    agentId="agent-123",
    agentAliasId="alias-456",
    sessionId="session-789",  # Persisted conversation
    inputText="What is my order status?"
)

These features align with our guide on agentic RAG.

Improved S3 Integration

S3 synchronization becomes real-time:

DEVELOPERpython
# Enable real-time sync
bedrock.update_data_source(
    knowledgeBaseId="kb-123",
    dataSourceId="ds-456",
    dataSourceConfiguration={
        "type": "S3",
        "s3Configuration": {
            "bucketArn": "arn:aws:s3:::my-bucket",
            "inclusionPrefixes": ["documents/"],
            "syncMode": "REAL_TIME"  # New
        }
    }
)

EventBridge automatically triggers reindexing on changes.

RAG Guardrails

New RAG-specific guardrails:

DEVELOPERpython
bedrock.create_guardrail(
    name="rag-guardrails",
    description="Guardrails for RAG applications",
    contentPolicyConfig={
        "filtersConfig": [
            {"type": "SEXUAL", "inputStrength": "HIGH", "outputStrength": "HIGH"},
            {"type": "HATE", "inputStrength": "HIGH", "outputStrength": "HIGH"}
        ]
    },
    contextualGroundingPolicyConfig={  # New
        "filtersConfig": [
            {"type": "GROUNDING", "threshold": 0.8},
            {"type": "RELEVANCE", "threshold": 0.7}
        ]
    }
)

Check our guide on RAG guardrails.

Architecture and Performance

Recommended Architecture

S3 / Confluence / SharePoint
           ↓
    [Data Sources]
           ↓
    [Knowledge Base]
           ↓
    OpenSearch Serverless
           ↓
    [Bedrock Agent]
           ↓
    Claude / Titan / Llama
           ↓
    [Application]

Benchmarks

AWS publishes benchmarks on standard RAG workloads:

Limits

Pricing

New Pricing Model

Comparison

*For 100K requests/month, 10GB of data

Check our guide on RAG cost optimization.

Use Cases

When to Use Bedrock KB

Ideal for:

• Enterprises already on AWS
• Need for native integration (S3, Lambda, etc.)
• Large volumes
• AWS compliance required

Less suitable for:

• Multi-cloud
• Startups/SMBs (cost)
• Need for open-source models

Complete Example

DEVELOPERpython
import boto3

# 1. Create the KB
bedrock = boto3.client('bedrock-agent')
kb = bedrock.create_knowledge_base(
    name="support-kb",
    knowledgeBaseConfiguration={...},
    storageConfiguration={...}
)

# 2. Add documents
bedrock.create_data_source(
    knowledgeBaseId=kb['knowledgeBase']['knowledgeBaseId'],
    name="docs",
    dataSourceConfiguration={
        "type": "S3",
        "s3Configuration": {
            "bucketArn": "arn:aws:s3:::my-docs"
        }
    }
)

# 3. Synchronize
bedrock.start_ingestion_job(
    knowledgeBaseId=kb['knowledgeBase']['knowledgeBaseId'],
    dataSourceId=ds['dataSource']['dataSourceId']
)

# 4. Query
runtime = boto3.client('bedrock-agent-runtime')
response = runtime.retrieve_and_generate(
    input={"text": "How do I configure product X?"},
    retrieveAndGenerateConfiguration={
        "type": "KNOWLEDGE_BASE",
        "knowledgeBaseConfiguration": {
            "knowledgeBaseId": kb['knowledgeBase']['knowledgeBaseId'],
            "modelArn": "arn:aws:bedrock:us-east-1::foundation-model/anthropic.claude-3-opus-20240229-v1:0"
        }
    }
)

Our Take

Bedrock KB v2 represents an important evolution:

Strengths:

• Native AWS integration
• Advanced guardrails
• Improved performance
• Multiple data sources

Points of attention:

• AWS lock-in
• High cost
• Configuration complexity
• Limited regions

For AWS-first enterprises, Bedrock KB becomes a serious option. For others, more agnostic alternatives exist.

Platforms like Ailog offer a cloud provider-independent alternative, with French hosting and simplified setup.

Check our guide to best RAG platforms to compare.