
Weekend Creative Agent Challenge: The Daily Grind Almanac
An autonomous, serverless AI agent that generates a daily Victorian coffee roast special and chalkboard decree using Amazon Bedrock Nova Micro, EventBridge, and DynamoDB. Output is pre-computed in the background and served instantly to an AWS Amplify digital display with zero user latency.
Vision & What It Does
Every morning before unlocking the front doors, artisanal café managers and head baristas face a small yet recurring creative challenge: concocting the day's special roast, curating distinct tasting notes, and penning an entertaining sidewalk chalkboard decree to catch the eye of morning commuters.
The Daily Grind Almanac automates this routine by functioning as an always-on, autonomous AI agent. Built directly around the core philosophy of the challenge - "The best tool is the one you never have to open" - this application requires zero manual prompting, button clicks, or interactive sessions during morning business operations.
Instead of waiting for human instruction, the agent wakes up autonomously on a scheduled background cadence, synthesizes a Victorian-themed coffee special using Amazon Bedrock (Nova Micro), and persists the generated state directly into Amazon DynamoDB.
When café baristas or customers open the digital chalkboard interface hosted on AWS Amplify, the pre-computed almanac entry is already rendered and waiting on the screen - ready to be displayed on counter screens or transcribed directly onto a sidewalk slate board with zero generative cold-start latency.
How I Built It
The primary objective was building a serverless, decoupled background pipeline where content generation is completely separated from consumer presentation.
Key Decisions & Engineering Process
- Foundation Model Selection: I selected Amazon Bedrock Nova Micro (
amazon.nova-micro-v1:0) for creative text generation. Nova Micro offers ultra-low latency, cost efficiency, and strict adherence to structured JSON schemas.
- Persona & Schema Conditioning: To ensure engaging output, I designed a system prompt conditioning the model as an eccentric Victorian London head barista. The prompt enforces a strict JSON schema:
{ "roast_name": "string (eccentric coffee blend name)", "tasting_notes": ["string", "string", "string"], "brew_guidance": "string (witty brewing instructions)", "morning_decree": "string (inspirational or satirical morning quote)" }
- Decoupled Architecture with Serverless State: Rather than invoking the foundation model on every page load (which would introduce latency and unnecessary token costs), the backend runs on a schedule and writes state to Amazon DynamoDB under the partition key
id: latest. The frontend fetches this cached item via a public Lambda Function URL in sub-second response times.
- Single-Bundle Responsive UI: The digital chalkboard was built with modern glassmorphism, ambient glow effects, responsive badges, and classical typography (
Cinzel,Playfair Display), bundled into a clean single-file structure to prevent single-page application asset routing issues.
Challenges Encountered & Solutions
- Amplify Single Page Application Asset Routing: During initial deployments with separated asset files (
index.html,style.css,script.js), default SPA rewrite rules routed stylesheet requests back toindex.html. To ensure smooth, bulletproof hosting with zero routing collisions, the markup, styles, and fetch logic were consolidated into a streamlined single-file artifact.
- Cold Table Initialization (Lazy Fallback): If the frontend was accessed prior to the initial EventBridge trigger executing, an empty table would return blank fields. I added a fallback routine within the Lambda handler: if
id: latestis not found, the function executesgenerate_daily_almanac()on the spot, stores the result, and serves the response immediately.
AWS Services Used / Architecture Overview
The solution runs entirely on AWS serverless primitives:

AWS Stack Details
- Amazon EventBridge (Schedules): Operates as the clockwork heartbeat, triggering the Lambda function automatically on a recurring schedule without manual initiation.
- AWS Lambda (Python 3.14): Orchestrates the pipeline, handles Bedrock invocations via
boto3, validates JSON schemas, and manages DynamoDB reads and writes.
- Amazon Bedrock (
amazon.nova-micro-v1:0): Fast generative AI foundation model used for structured creative text synthesis.
- Amazon DynamoDB: Fully managed NoSQL key-value store holding the current daily dispatch (
id: latest) and generation metadata (generated_at).
- AWS Amplify Hosting: Delivers global edge hosting, automated builds, and SSL distribution for the digital chalkboard UI.
📸 Verification & Proof of Autonomous Execution
Amazon EventBridge Schedule Configuration

Amazon DynamoDB State Persistence

Live AWS Amplify Digital Chalkboard

What I Learned
Building The Daily Grind Almanac for the Level 200 challenge provided valuable takeaways in designing autonomous cloud architectures:
- The Value of Decoupled Asynchronous Systems: Generative AI is often most practical when decoupled from the user request loop. Pre-computing output on an automated schedule removes user-facing model latency entirely.
- Operational Cost Predictability: Combining Amazon EventBridge, AWS Lambda, Amazon Bedrock (Nova Micro), and DynamoDB creates a resilient, always-on agent capable of running year-round well within AWS Free Tier limits.
- Structured Output Reliability: Providing strict JSON schemas and system personas in Nova Micro ensures dependable data formats suitable for direct programmatic database ingestion.
Link to App or Repo
- Live Application: https://main.d17sq6scu8kyy7.amplifyapp.com/
- GitHub Repository: https://github.com/ChakShubh/the-daily-grind-almanac
Enjoyed reading this content? Let the author know!
Your likes, comments, shares, and saves help creators reach more builders.
Loading recommendations
Loading article