Simple guide for RAG using Strands Agents and Amazon S3 Vectors

aws configure
# Enter your AWS Access Key ID, Secret Access Key, and region

{    
    "metadataAttributes": {        
    "tag": "Loan",        
    "year": 2022,        
    "borrower": "John Doe",        
    "customer_id": "CUSTOMER_NO_1"
    }
}

pip install strands-agents strands-agents-tools boto3 --quiet

from strands import Agent, tool
import osimport boto3os.environ["AWS_PROFILE"] = "YOUR_AWS_CLI_PROFILE"
os.environ["KNOWLEDGE_BASE_ID"] = "KB_ID"

@tool
def search_vector_db(query: str, customer_id: str) -> str:    
    """    
    Handle document-based, narrative, and conceptual queries using the unstructured knowledge base.    
    Args:        
        query: A question about business strategies, policies, company information,or requiring document comprehension and qualitative analysis        
        customer_id: Customer identifier    
    Returns:        
    Formatted string response from the knowledge base    
    """
    
    region = os.environ.get("AWS_DEFAULT_REGION", "us-west-2")    
    kb_id = os.environ.get("KNOWLEDGE_BASE_ID")    
    bedrock_agent_runtime = boto3.client("bedrock-agent-runtime", region_name=region)    
    try:        
        retrieve_response = bedrock_agent_runtime.retrieve(            knowledgeBaseId=kb_id,            
            retrievalQuery={"text": query},            
            retrievalConfiguration={                
                "vectorSearchConfiguration": {                    "numberOfResults": 5,                    
                "filter": {"equals": {"key": "customer_id", "value": customer_id}}}
                }
        )
        
        # Format the response for better readability        
        results = []        
        for result in retrieve_response.get('retrievalResults', []):    
            content = result.get('content', {}).get('text', '')  
            
        if content:                
            results.append(content) 
        
        return "\n\n".join(results) if results else "No relevant information found."    
    except Exception as e:        
        return f"Error in unstructured data assistant: {str(e)}"

agent = Agent(
    tools=[search_vector_db], 
    system_prompt="You are a rag agent. When a user asks you a question you will first check it in your knowledge base (if you can't answer it from the current conversation memory)."
    )

I'll help you retrieve the lender details for CUSTOMER_NO_1. I'll use the search_vector_db function to find the relevant information.
Tool #1: search_vector_db
Based on the search results, I can provide you with the following lender details for CUSTOMER_NO_1:
1. Borrower: SCHOOL SUPPLY (a Pennsylvania limited liability company)
2. Loan Amount: $1,300,000.00
3. Loan Documentation:   
    - Promissory Note dated as of the current date   
    - Mortgage and Security Agreement
4. Financial Covenants for the Guarantor:   
    - Maintain a Net Worth of at least $40,000,000   
    - Maintain Liquid Assets of at least $4,000,000
5. Reporting Requirements:   
    - Deliver Financial Covenants Certification within 30 days of each calendar year-end   
    - Provide annual financial statements within 90 days of each calendar year-en
    - Financial statements must be audited by an independent certified public accounting firm
6. Guaranteed Principal Amount:   
    - $1,050,000.00 when the unpaid principal balance is greater than $1,050,000.00  
    - Entire unpaid principal balance when it's equal to or less than $1,050,000.00
7. Title Insurance: Issued by Gotham Abstract & Settlement LLC

The lender has specific requirements and protections in place, including financial covenants and reporting obligations for the borrower and guarantor.

Is there anything specific about the lender details you would like me to elaborate on?