AWS Strands Agents · TechToday

👋 Hello World Agent ⓘ

Minimal single-turn Amazon Nova Lite agent invocation via the AWS Strands Agents SDK.

Send a prompt to a zero-tool Strands Agent(model=BedrockModel("amazon.nova-lite-v1:0")) and inspect its response.

LangGraph create_react_agent with ChatBedrockConverse (Nova Lite) and a @tool greet(name) function.

Compare Strands with LangGraph's explicit StateGraph cycle as the agent calls greet(name) for each person and synthesizes a reply.

@tool decorator reflection turning check_server_status(server_url: str) into a Bedrock tool schema.

Ask the agent to verify whether an HTTP endpoint is live; it executes a real requests.get() health check with a 5s timeout.

Multi-parameter @tool calculate_tip(bill_amount: float, tip_percentage: float = 15.0, num_people: int = 1).

Extracts typed numeric arguments and defaults from natural language to compute tip, total bill, and per-person split.

Sequential 3-tool chain: get_sales_data(quarter) → analyze_sales(revenue, deals, quarter) → send_email(to, subject, body).

Fetches quarterly sales metrics ($125,000 across 42 deals in Q4), computes average deal size, and emails a summary report.

Domain-specific @tool check_inventory(product_id: str) querying an in-memory warehouse catalog.

Check stock status and unit counts for tracked catalog SKUs: PROD-123 (15 in stock), PROD-456 (out of stock), and PROD-789 (8 low stock).

Off-the-shelf calculator tool from strands_tools paired with a math assistant system prompt.

Delegates exact exponentiation and arithmetic expressions to the prebuilt SymPy-backed calculator instead of token prediction.

Combining prebuilt http_request and calculator tools from strands_tools for live data analysis.

Fetches live JSON data from an external URL via http_request and pipes the extracted numbers into calculator.

Prebuilt use_aws tool from strands_tools translating natural language into live boto3 AWS SDK calls.

Inspect AWS resources (such as S3 buckets, EC2 instances, or DynamoDB tables) in us-west-2 using IAM credentials.

InventoryTools class encapsulating shared self.inventory state across @tool check_stock and @tool update_stock methods.

Mutate and verify live stock counts for PROD-123 (15), PROD-456 (0), or PROD-789 (8) within a stateful class instance.

Non-blocking async def check_warehouse_inventory(product_id, warehouse) via agent.invoke_async().

Queries east (50 units), west (30 units), and central (75 units) warehouses concurrently (~2s parallel vs ~6s sequential).

4-tool coordinator combining get_weather_forecast, suggest_packing_list, estimate_trip_cost (INR), and prebuilt calculator.

Plan a trip to Goa, Bangalore, Jaipur, or Manali with weather forecasts, packing checklists, INR cost breakdowns, and budget validation.