AWS Builder Center
Weekend Creative Agent Challenge: DawnDraft

Weekend Creative Agent Challenge: DawnDraft

DawnDraft is an autonomous, always-on AWS creative agent that wakes up before you do. Powered by Amazon EventBridge, AWS Lambda, and Amazon Bedrock (Nova Lite & Titan Image Generator), it transforms live local weather conditions into a bespoke morning vignette, minimalist color palette, and digital postcard—delivered fresh to your inbox every dawn.

Aws sbg geu technical core member
Vision & What It Does:
We spend most of our mornings opening multiple apps just to brace ourselves for the day: weather forecasts, calendars, news feeds, and to-do lists. Most of these interfaces are functional, but few are inspiring.
DawnDraft:
flips that routine by acting as an autonomous morning muse. It is an always-on creative agent that runs in the background while you sleep. Every day at 6:00 AM local time, DawnDraft gathers local atmospheric data (weather condition, cloud coverage, temperature, and sunrise timing), synthesizes the mood, and uses generative AI to produce a unique, reflective three-line vignette paired with a four-color harmonious palette and a generative minimalist illustration. When you wake up, you don't open an app to generate anything—your personalized morning canvas is already waiting in your inbox or synced to your digital picture frame.
How I Built It
Building an agent that creates autonomously requires removing user prompts from the equation and replacing them with environmental context.
The Workflow:
  1. Trigger & Data Ingestion: Amazon EventBridge triggers an AWS Lambda function on a daily cron schedule. The Lambda function pulls live environmental metrics from a free weather API.
  1. Contextual Reasoning & Prompt Crafting: Instead of static prompting, the agent evaluates the meteorological data (e.g., dense morning fog, 14°C, 82% humidity) and maps it to nuanced emotional archetypes (e.g., contemplative, grounding, quiet anticipation).
  1. Multimodal Generation: The agent invokes Amazon Bedrock using Amazon Nova Lite to craft the micro-poetry and a structured JSON specification for a color palette. Next, it calls Amazon Titan Image Generator via Bedrock to render a minimalist, atmospheric visual based on the synthesized mood.
  1. Storage & Delivery: The output assets are stored in Amazon S3, and Amazon Simple Email Service (SES) formats and dispatches the HTML postcard directly to the subscriber.
Key Challenges & Decisions:
  • Avoiding Cliché Output: Early test runs yielded repetitive poetry whenever it rained. I refined the Bedrock system prompt to adopt concise, imagist poetic forms (reminiscent of William Carlos Williams or Bashō) and explicitly banned generic morning idioms.
  • Cost & Performance Optimization: Using lightweight serverless functions and Amazon Nova models kept execution well within the AWS Free Tier, running under 15 seconds per execution cycle.
Architecture & AWS Services Used
  • Amazon EventBridge: Serves as the autonomous heartbeat, scheduling daily triggers across time zones.
  • AWS Lambda (Python 3.12): Orchestrates data collection, payload transformation, and model calls without persistent compute costs.
  • Amazon Bedrock (Amazon Nova Lite & Amazon Titan Image Generator G1): Powers both the contextual text reasoning/poetry and the generative art creation.
  • Amazon S3: Archives daily generated visual postcards and metadata.
  • Amazon SES: Delivers the curated morning email digest.
[EventBridge Cron] │ ▼ [AWS Lambda] ──(Pulls Local Weather)──> [Open-Meteo API] │ ├──(Generates Poetry & Palette)──> [Bedrock: Nova Lite] ├──(Renders Artwork)───────────> [Bedrock: Titan Image Gen] │ ├──(Stores Artifacts)─────────> [Amazon S3] └──(Dispatches Digest)─────────> [Amazon SES] ──> [User Inbox]
What I Learned
Building an agent designed to run unattended shifted my mindset from interactive user experience (UI/UX) to ambient autonomous design. The biggest takeaway was understanding how deterministic contextual inputs (like barometric pressure and dawn light) can be turned into rich, dynamic prompts that give generative models genuine creative personality.
Project Links & Code
  • GitHub Repository:  https://github.com/pratyush34/dawndraft-agent 
  • Live Demo / Showcase Postcards:
  •  http://dawndraft-assets-yourname-101.s3-website.ap-south-1.amazonaws.com 
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