Bowerbird

Inspiration

I enjoy collecting inspiration of code, design, and tools on X using Bookmarks. However, it is challenging at times to go back through the bookmarks to find specific designs, tools, or code I'm looking for. So I created Bowerbird, a tool that helps me solve that issue.

What it does

Bowerbird allows you to bookmark any link and provide a reason for why you're collecting that link. Then Bowerbird will automatically collect metadata like the author, source link, tags, etc and display the media in a grid for you to view. Bowerbird also records your reason into memory and learns your design/tool preferences over time. Using Bowerbird, you can easily find the tools

How we built it

I built Bowerbird using Claude Code for my primary AI coding tool. For Bowerbird itself, I used AWS Bedrock + AWS Bedrock AgentCore for creating the orchestrator AI agent along with tools such as a self-hosted SigLip 2 model on an AWS Lightsail instance and Exa for internet search. Then for the backbone of the application storing all the data and embeddings, I used CockroachDB. I also used the CockroachDB skills tool to configure and manage my database.

Challenges we ran into

One challenge I ran into was image search performance using the AWS Titan embedder. I realized it was more effective to host a SigLip 2 on AWS Lightsail. Another challenge was figuring out how to make the UI interface clean and intuitive. For example, I iterated on the AI agent panel several times before I was satisfied with it.

Accomplishments that I'm proud of

I'm proud of the polished user interface I created. I tried to focus as much on the tiny details that matter for interactions. I'm also proud that I was able to create Bowerbird in such a short notice after discovering the hackathon.

What I learned

I learned that the Bower bird is a species of bird that collects mostly shiny blue things to create structures that attract mates. I thought that made sense for the application because designers collect inspiration to create better interfaces that attract users. I also learned that it is easier than I thought to setup an AI agent using AWS Bedrock AgentCore and CockroachDB.

What's next for Bowerbird

If the idea of Bowerbird resonates with people, I would work on improving the quality of the search results Bowerbird gets and make it better at collecting inspiration.

Built With

• aws-bedrock
• aws-bedrock-agentcore
• aws-lightsail
• cockroachdb
• docker
• exa
• nextjs
• siglip-2
• typescript
• vercel