AWS Builder Center
Weekend Creative Agent Challenge: BackBench AI Autonomous Daily Gazette

Weekend Creative Agent Challenge: BackBench AI Autonomous Daily Gazette

An always-on autonomous creative agent built with Amazon Bedrock (Nova Lite), Amazon SNS, and AWS Amplify that automatically generates daily 90s/2000s school gazette editions and emails daily digests to subscribers' inboxes.

AWS CB | DevOps Engineer

Vision & What the Agent Does

The premise of an autonomous agent is to transition software from a reactive, prompt-driven tool into a proactive background system that delivers creative value automatically. BackBench AI was initially designed as an interactive storytelling platform to relive late 1990s and early 2000s Indian school memories. For Level 200, the application has been transformed into an Always-On Autonomous Creative Agent: the BackBench AI Daily Gazette Agent.
Instead of waiting for a user to open the browser and manually type a memory, the agent executes autonomously on a schedule. Every day, it evaluates a nostalgia matrix driven by weather conditions, seasonal events, and historical school calendars (such as monsoon rain, surprise mathematics tests, annual sports day, and cafeteria samosa raids). Using these triggers, the agent autonomously synthesizes a complete 4-part Daily Classroom Gazette:
  1. School Diary: A handwritten diary entry written from the perspective of an Indian school student in 2002.
  1. Teacher Remarks: A red-ink exercise book correction box containing realistic teacher feedback, a performance grade, and a signature.
  1. School Magazine Column: An annual school souvenir article written in newsprint column format.
  1. Bollywood Screenplay: A dramatic script set in a 2000s classroom featuring dialogue in Hinglish.
When a user visits the live platform, today's edition is already generated, rendered, and ready to read. To fulfill the objective of a tool you never have to open, the agent integrates with Amazon Simple Notification Service (Amazon SNS). Every morning upon generating the daily gazette, the agent dispatches a structured text digest directly to subscribers' email inboxes.

Figure 1: The BackBench AI landing page displaying the live agent status header, execution cycle counter, and pre-loaded daily gazette edition.



How I Built It

The implementation focuses on decoupled, event-driven architecture combining client-side Web Audio synthesis with serverless AI execution and push notification delivery.

1. Autonomous Agent Engine & State Management

The core agent logic is contained in a TypeScript engine ( lib/agentEngine.ts ). It manages state persistence across execution cycles, tracks historical edition archives, and rotates through the nostalgia topic matrix. The agent can be executed automatically via background schedule or manually triggered on-demand via an API endpoint ( POST /api/agent/daily ).

2. Generative AI Processing via Amazon Bedrock

Each agent execution invokes Amazon Bedrock using the Amazon Nova Lite model ( amazon.nova-lite-v1:0 ). The agent passes structured prompt instructions to return strict JSON payloads containing all four nostalgia artifacts. This allows the system to generate formatted, contextual text across four different writing styles in a single inference call.

Figure 2: The email subscription modal allowing users to register their email address for daily SNS digests.

3. Email Digest Delivery via Amazon SNS

To implement background push notifications, the agent integrates with  @aws-sdk/client-sns . When a visitor submits their email address via the user interface, the system invokes  SubscribeCommand  against an Amazon SNS Topic ( BackBenchAIDailyDigestTopic ). When a new daily gazette is generated, the agent invokes  PublishCommand  to broadcast the formatted digest directly to all confirmed email subscribers.

Figure 3: Live email delivered to subscriber inbox via Amazon SNS containing the daily classroom diary, teacher remark, magazine column, and screenplay.

4. Technical Challenges & Resolution

During deployment to AWS Amplify, custom environment variables starting with  AWS_  were restricted due to reserved system prefixes. To resolve this, the backend configuration was updated to consume custom prefixed variables ( BEDROCK_AWS_ACCESS_KEY_ID ,  BEDROCK_AWS_SECRET_ACCESS_KEY , and  SNS_TOPIC_ARN ) and explicitly expose them in  next.config.mjs  for Next.js API routes.



AWS Services Used & Architecture Overview


Figure 4: Architecture.

Figure 5: AWS Amplify Console displaying successful deployment of the agents branch.

AWS Services Summary:

  1. AWS Amplify: Hosts the Next.js 15 SSR application, managing serverless API routes, edge caching, and automated deployment from GitHub.
  1. Amazon Bedrock: Executes generative inference using the Amazon Nova Lite model ( amazon.nova-lite-v1:0 ) to produce daily nostalgia text content.
  1. Amazon SNS (Simple Notification Service): Manages topic subscriptions and delivers daily email digests directly to user inboxes.
  1. AWS EventBridge & Lambda: Schedules automated daily execution ( cron(0 6 * * ? *) ) to trigger the agent engine without human intervention.



What I Learned

  1. Event-Driven Agent Design: Designing software that operates autonomously requires building resilient fallback mechanisms, clear state tracking, and automated delivery channels rather than relying on interactive user input.
  1. Amazon SNS Integration: Implementing Amazon SNS via the AWS SDK demonstrated how lightweight pub/sub systems can extend web applications into passive inbox utilities.
  1. Next.js 15 Environment Configuration on Amplify: Gained practical insight into environment variable scoping and Webpack bundling rules for SSR API routes hosted on AWS Amplify.



Links

Any opinions in this article are those of the individual author and may not reflect the opinions of AWS.
Enjoyed reading this content? Let the author know!

Your likes, comments, shares, and saves help creators reach more builders.

Loading recommendations

Loading article