lab_goals.txt

2) Lab Goals (What “Good” Looks Like)

You don’t need line-by-line keys; you need architecture + behavior.

Lab 1 – Event-Driven Image Processor
    Architecture:
        S3 ObjectCreated:* → Lambda (metadata + EventBridge)
        EventBridge routes a “IMAGE_PROCESSED” event to a logging Lambda
        Logging Lambda writes to Aurora (image key, size, user, timestamp)
        SNS sends user/admin success notification
    Lab correctness signals:
        Lambda idempotent (can process same S3 event twice without duplicate DB entries)
        Aurora writes use parameterized queries or Data API
        SNS topic subscription works (test via email)

Lab 2 – Serverless Order System
    Architecture:
        API Gateway REST/HTTP → Lambda (validation, enrichment) → SQS
        Worker Lambda pulls from SQS, writes to Aurora orders table
    Lab correctness signals:
        Orders with invalid fields are rejected at API layer
        SQS DLQ configured and demonstrably catches poison messages
        Aurora schema supports indexing by order_id / customer_id

Lab 3 – Real-Time Inventory
    S3 CSV → Lambda → Aurora update → SNS notify
    Correct CSV parsing, graceful handling of bad lines, transactional upserts.

Lab 4 – Event Stream Router
    3+ EventBridge rules with different patterns (e.g. source/type/region)
    Each pattern routes to different Lambda/SQS/SNS targets
    DLQ configured at bus or target level

Lab 5 – Aurora AI Query Bot
    API → Lambda → Bedrock (or GPT) → safe SQL → Aurora
    SQL constrained to a limited set of templates (no arbitrary SQL)
    Queries logged (for auditing)

Labs 6–10 follow same style: you just check that:
    Required services are present
    Error paths are handled
    Scaling/throughput or failover behavior is demonstrated (for 9 & 10)