AWS Builder Center

Weekend Creative Agent Challenge

Here is a small, ready-to-use description for your challenge: *MuseBot — Always-On Daily Creative Agent* MuseBot is an autonomous, serverless AI agent that creates a new poem + visual-art prompt every day at 8 AM UTC — without needing a prompt. It wakes up via EventBridge, checks real-world context via Open-Meteo (weather, date, season), recalls its last 7 creations from DynamoDB to avoid repetition, generates with Gemini Flash

Weekend Creative Agent Challenge: MuseBot — Always-On Daily Creative Agent Vision & What It Does
For the Weekend Creative Agent Challenge, I built MuseBot — Always-On Daily Creative Agent, an autonomous AI-powered creative system that generates a new piece of creative content every day without requiring the user to manually open an application or provide a prompt.
The idea behind MuseBot is simple: creativity does not always need to start with a blank page. Instead of waiting for a user to decide what to write, MuseBot automatically creates a daily creative experience based on real-world context.
Every day at 8:00 AM UTC, the agent starts its workflow. It retrieves current weather information, identifies the date and day, considers seasonal context, and reviews its recent creative history. It then uses Google Gemini Flash to generate a themed poem together with a visual-art prompt. The generated content is stored in the cloud and made available through a gallery-style website.
One of the most interesting parts of the project is its memory. MuseBot reviews its previous seven creations so that the new poem does not simply repeat the same ideas or style. This gives the agent a basic form of creative continuity and allows its output to evolve over time.
The project solves a common problem with generative AI applications: users often have to repeatedly provide prompts and manually interact with an AI system to create content. MuseBot changes this into an autonomous workflow where the agent takes responsibility for generating and publishing content on a schedule.
The final creative output consists of:
A daily themed poem A visual-art generation prompt Weather and date context A persistent creative history A web-based gallery for viewing generated creations
The project demonstrates how an AI agent can combine external information, memory, scheduled execution, generative AI, cloud storage, and a user-facing interface into one automated creative system.
How You Built It
I designed MuseBot as a serverless application so that the project could remain lightweight, automated, and inexpensive to operate.
The first major decision was to use AWS Lambda as the main execution environment. Instead of maintaining a continuously running server, the creative agent runs only when it needs to perform its daily workflow.
I then used Amazon EventBridge Scheduler to trigger the Lambda function automatically every morning. This removes the need for manual execution and gives MuseBot its “always-on” behavior.
Inside the Lambda workflow, the agent collects weather information through the Open-Meteo API. It also obtains configuration and secrets through AWS Systems Manager Parameter Store. The creative generation step sends the relevant context to Gemini Flash, which produces the poem and visual-art prompt.
Another important design decision was adding memory. I used Amazon DynamoDB to store previous creative outputs. MuseBot can retrieve its recent history before generating new content, helping it avoid unnecessary repetition.
The generated content is then stored in Amazon S3, which also hosts the static gallery website.
One challenge was designing the workflow so that each service had a clear responsibility. Instead of putting everything into one large application, I separated the system into scheduled execution, AI generation, persistent memory, storage, and presentation layers.
Another challenge was keeping API credentials secure. Rather than placing the Gemini API key directly inside the source code, the project uses AWS Systems Manager Parameter Store with a secure parameter. This makes the application safer and also makes configuration easier to manage.
I also created deployment scripts and local testing support so that the agent can be tested without depending entirely on the scheduled production workflow. The repository includes separate directories for the frontend, infrastructure, Lambda code, and scripts, making the project easier to understand and maintain.
AWS Services Used / Architecture Overview
MuseBot uses several AWS services to create a completely automated serverless architecture.
AWS Services Amazon EventBridge Scheduler — Starts the creative agent automatically on a daily schedule. AWS Lambda — Runs the Python-based agent logic without requiring a dedicated server. Amazon DynamoDB — Stores creative history and provides memory for the agent. Amazon S3 — Stores generated JSON content and hosts the static gallery website. AWS Systems Manager Parameter Store — Securely stores configuration values and the Gemini API key. External Services Google Gemini Flash — Generates the poem and visual-art prompt. Open-Meteo API — Provides weather information used as creative context.
The overall architecture can be represented as:
Daily Schedule
│
▼
Amazon EventBridge Scheduler
│
▼
AWS Lambda (Python)
/ | \
/ | \
▼ ▼ ▼
Open-Meteo DynamoDB SSM Parameter
Weather Memory Store
\ | /
\ | /
▼ ▼ ▼
Gemini Flash
│
▼
Poem + Art Prompt
│
▼
Amazon S3
/ \
▼ ▼
Generated Content Web Gallery
The architecture is designed around serverless AWS components, which keeps the infrastructure simple and suitable for a small creative application. The repository documentation also estimates that the architecture can operate within AWS free-tier usage for the intended workload.
What You Learned
This challenge gave me practical experience in building an AI agent rather than simply building a conventional application that calls an AI model.
One of my biggest learnings was understanding how agentic workflows combine multiple services. The intelligence of MuseBot does not come only from the language model. The agent becomes useful because it can collect information, remember previous results, make a generation decision, create content, and publish the result automatically.
I also learned more about AWS serverless architecture. Working with Lambda, EventBridge, DynamoDB, S3, and Systems Manager helped me understand how different cloud services can work together without requiring a traditional backend server.
Another important lesson was the value of memory in creative AI applications. Giving the agent access to its recent creations makes the system more consistent and helps reduce repetitive outputs.
Security was another area where I gained practical knowledge. Keeping API credentials in Parameter Store instead of hard-coding them into the application showed me how cloud applications can separate configuration and secrets from source code.
Most importantly, this challenge changed how I think about AI applications. A creative AI system does not have to be a simple chatbot waiting for instructions. It can operate autonomously, respond to external context, maintain memory, and continuously produce new content.
Link to App or Repo
The complete source code for MuseBot is publicly available on GitHub:
The repository contains the frontend, AWS infrastructure, Lambda agent implementation, deployment scripts, and testing utilities.
Conclusion
MuseBot started as an experiment in automated poem generation, but it became a practical demonstration of how AI agents, cloud services, memory, scheduling, and creative generation can work together.
The goal was not simply to make an AI write a poem. The goal was to create a system that can independently decide when to create, understand its surrounding context, remember what it created previously, generate something new, and publish the result for people to enjoy.
This Weekend Creative Agent Challenge gave me an opportunity to explore that complete lifecycle and strengthened my understanding of building autonomous AI applications with AWS and generative AI. #agents #Productivity
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