Weekend Creative Agent Challenge: Daily Music Drop
The idea
I wanted to build something simple for this challenge, but I also wanted it to actually fit the idea of an "always-on" creative agent.
I ended up making Daily Music Drop, a small application that recommends one song and one album every day. The interesting part is that you don't have to open the application and ask for a recommendation. The system does it by itself at a scheduled time, and the new recommendation is already waiting on the website.
The recommendation includes the song, artist, album, genre, mood, and a short explanation of why the song might be worth listening to.
I chose music because it is something that can be discovered continuously. There is always another album or artist to find, and I liked the idea of having a small "music drop" waiting for me each day.
How it works
The application has a pretty small architecture.
I created an S3 bucket called daily-music-drop to host the static website. The page is just HTML, CSS and JavaScript, so there is no frontend server that needs to be running.
The main logic is inside an AWS Lambda function called song-function . When Lambda runs, it sends a prompt to Amazon Bedrock asking the model to act as a music discovery agent and generate a recommendation.
The model returns the song, artist, album, genre, mood, and a short reason for the recommendation. It generates the mood and explanation in both English and Spanish. The website then detects the visitor's browser language and displays the appropriate version automatically.
I also added a small history.json file to keep track of previous recommendations. This gives the agent some basic memory and allows it to take previous recommendations into account instead of constantly suggesting the same songs.
Finally, I connected the Lambda function to Amazon EventBridge Scheduler. The scheduler runs once a day at the time I configured, which means the whole process happens without me having to do anything.
The flow is basically:
EventBridge Scheduler
↓
AWS Lambda
↓
Amazon Bedrock
↓
Music recommendation
↓
Amazon S3
↓
Updated website
AWS services
The project uses four AWS services:
- Amazon S3 for the static website and recommendation history.
- AWS Lambda for the application logic.
- Amazon Bedrock for generating the music recommendation and its descriptions.
- Amazon EventBridge Scheduler for running the agent automatically every day.
I also configured the Lambda execution role with the permissions required to access S3 and invoke the Bedrock model.
I deliberately kept the infrastructure small. Since the main goal was to build an always-on agent, I didn't want to add services that weren't really necessary for the idea.
What I learned
The biggest thing I learned from this project is that an agent doesn't necessarily need to be complicated.
The important difference here is that the AI isn't waiting for me to send a prompt. EventBridge decides when the process starts, Lambda handles the workflow, Bedrock makes the creative decision, and S3 publishes the result.
I also got more familiar with connecting AWS services through IAM roles and with using Bedrock from a Lambda function. Getting all the permissions and service interactions working correctly was probably the most annoying part of the implementation, but once everything was connected, the final workflow was surprisingly small.
Another thing I liked about this approach is that the application can keep working even when I'm not using it. If I don't open the website for a few days, the agent still keeps generating new recommendations according to its schedule.
The browser language feature was also a nice addition because it required no extra AWS service. The same generated page can adapt between English and Spanish depending on the visitor's browser settings.
Try it
The current version of Daily Music Drop is available here:
Every day, the agent generates a new recommendation and publishes it to the website automatically.
Loading article