DEV Community

Cover image for Building Repo Promoter Agent for the DigitalOcean Gradient™ AI Hackathon
Karlis
Karlis

Posted on

Building Repo Promoter Agent for the DigitalOcean Gradient™ AI Hackathon

I recently participated in the DigitalOcean Gradient™ AI Hackathon with a project I’m pretty excited about — Repo Promoter Agent. The idea came from a simple but recurring problem: promoting open-source projects on GitHub is often harder than building them.

💡 Inspiration

As someone who maintains several open source repositories, I noticed that my projects often struggle to reach the right audience. Writing promotional content — tweets, summaries, posts — takes time that could otherwise go into building features. I wanted a tool that could take a repository URL and instantly generate reusable marketing copy for it.

🚀 What It Does

Repo Promoter Agent does exactly that. Paste any public GitHub repo URL into the app, and it generates a complete promo pack:

  • A catchy headline
  • A short summary and key benefits
  • Tweet-sized snippets
  • A LinkedIn-style post
  • A call to action

All generated content is stored, indexed, and searchable — so you can quickly reuse good material across different projects.

🧩 How I Built It

The app combines several components:

  • Frontend: A simple web UI with “Generate” and “Search” screens
  • Backend: Written in Go, it fetches repo data and structures input for the AI agent
  • AI Layer: A Gradient AI agent processes the input and returns well‑structured JSON output
  • Search Engine: Apache Solr 10 indexes the generated content for fast retrieval

Once the agent returns the JSON promo pack, it’s stored and instantly searchable in Solr.

🧠 What I Learned

Using DigitalOcean’s Gradient AI platform was surprisingly smooth. The key insight was to think of prompt design as a contract — if the backend can trust the JSON format, the rest of the pipeline becomes predictable. It’s a great exercise in bridging LLM creativity with backend reliability.

⚙️ Challenges

Getting Solr 10 configured correctly took significant tuning, especially for schema and query logic. Another tricky aspect was finding the right balance between AI creativity and structure — ensuring the agent’s output stayed valid JSON without sounding robotic.

💰 Costs

Here’s what the first day of testing looked like:

  • App (29 hours): $1.07
  • Droplet (34 hours): $1.62
  • Database (58 hours): $1.69
  • Agents: $0.15

Total: $4.53 for a live day of experimenting. The cost will increase as I test more agents and models.

🛠️ Built With

  • GitHub
  • Go
  • Solr
  • Gradient AI

If you want to explore the project or see it in action, check out the Devpost page here.


Top comments (0)