Day 11: Invoking Amazon Titan from AWS Lambda (Serverless GenAI).

Hello World, AI 🧠
Welcome to Day 11. Today, we successfully connected our serverless backend to a Foundation Model.

I chose Amazon Bedrock because it allows me to access models like Titan, Claude, or Llama via a simple API call, without managing GPU servers.

The Code (Python Boto3)

Here is the working snippet to invoke Amazon Titan Text Express. Note: The JSON structure for Titan is different from Claude!

Python
import json
import boto3

1. Setup the Client (us-east-1 usually has all models)

bedrock = boto3.client(service_name='bedrock-runtime', region_name='us-east-1')

def lambda_handler(event, context):
user_prompt = "I spent $90 on Cloud Servers. Is this a good investment?"

# 2. Titan-Specific Payload
# Titan expects 'inputText', not 'messages' like Claude.
body = json.dumps({
    "inputText": user_prompt,
    "textGenerationConfig": {
        "maxTokenCount": 512,
        "temperature": 0.7,  # Creativity level
        "topP": 0.9
    }
})

try:
    # 3. Invoke the Model
    response = bedrock.invoke_model(
        body=body,
        modelId="amazon.titan-text-express-v1",
        accept="application/json",
        contentType="application/json"
    )

    # 4. Parse Response (Titan returns 'results' -> 'outputText')
    response_body = json.loads(response.get("body").read())
    ai_reply = response_body.get('results')[0].get('outputText')

    return {
        'statusCode': 200,
        'body': json.dumps(ai_reply)
    }

except Exception as e:
    return {'statusCode': 500, 'body': str(e)}

Key Learnings

Region Matters: I kept my Lambda in Stockholm but pointed the Bedrock client to N. Virginia (us-east-1) to ensure model availability.

Guardrails: Amazon Titan is very "safe" by default. It refused to give personalized financial advice, which is actually a good feature for a fintech app compliance.

See you on Day 12!