How I Built "The Sentinel": My AI Supply Chain Agent

Last December, I decided to challenge myself. I joined a hackathon because I wanted to see if I could build a real AI tool in just 48 hours.

My goal was to solve a problem I see often in logistics. Companies have too much data and not enough time to understand it. They usually rely on slow spreadsheets to make decisions. I wanted to build something faster.

I built "The Sentinel." It is an intelligent agent that watches inventory levels and tells you exactly when to restock.

Here is how I built it.

The Tools I Used
I needed tools that were fast to set up. I chose these four:

• Google Gemini: I used this as the "brain" to analyze the data.

• Kestra.io: I used this to connect everything together. It handles the workflow.

• Next.js: I used this to build the website so users can see the results.

• Vercel: I used this to host the website online. It made deployment very fast.

How It Works

The system follows a simple path.

First, Kestra wakes up and pulls the inventory data from my database. It sends this data to Google Gemini.

I wrote a specific prompt for Gemini. I told it to act like a Senior Logistics Manager. It looks at the numbers and decides if the stock is too low.

If Gemini sees a problem, it sends a message back. Kestra takes that message and triggers an alert. Finally, my Next.js website (hosted on Vercel) updates to show the user the warning immediately.

The Hardest Part

The project was not easy. The biggest challenge was the connection between the AI and the workflow.

At first, Gemini would give me long paragraphs of text. Kestra could not understand that. I had to tweak my prompts many times. I finally got Gemini to reply with "JSON" format. Once I did that, Kestra could easily read the data and take the next step.

It was a frustrating few hours, but fixing it felt great.

What I Learned

This project taught me that you don't need a huge team to build powerful software. By combining a tool like Kestra with AI, I built a working system in a weekend.

I am still on my journey to becoming a DevOps expert. Projects like "The Sentinel" prove to me that I can build things that solve real problems.