Week 2 hacktoberfest challenge post

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 2: Side Hustle

What I Built

ReplyPal writes replies to customer reviews. You paste a review, say what your business is, pick a tone (warm, professional, or playful), and get a short reply back. It thanks the customer, answers their specific points, and if the review is negative, it apologizes and offers a fix.

It's for small business owners: Etsy sellers, cafe owners, freelancers. They know replying to reviews helps, but every reply takes time, and angry reviews are hard to answer calmly. A few dollars a month is an easy decision for someone who gets dozens of reviews.

Demo

Try it live: hacktoberfest-week-3.streamlit.app

It's on a free tier, so it may take a few seconds to wake up, and it may hit rate limits if a lot of people use it at once.

Code

Krak3n909 / HacktoberFest-Week-3

The whole app is one file, app.py, about 35 lines.

How I Built It

• Model: openai/gpt-oss-120b, an open-weight model, served through Groq's API.
• Interface: Streamlit, with a text input for the business, a tone dropdown, and a text area for the review.
• Logic: One prompt that includes the business description, the tone, a length limit, and the review text. It makes one API call and shows the result.
• Deployment: Streamlit Community Cloud, with the API key stored in the app's secrets rather than in the repo.

I kept it deliberately small. The whole product is one prompt and one call, so the model choice is the only real dependency. That made it easy to build in an afternoon and easy to reason about.

Why Does Open Innovation Matter?

Because the model is open-weight, I'm not tied to one vendor's pricing or terms. The model name is a single line of code. If Groq changes its prices or retires a model, I change that line, or I run the same weights somewhere else. For a tool that would sell for a few dollars a month, that matters: each reply costs a fraction of a cent, and no one else controls that cost.

Open weights also leave the door open for things a closed API wouldn't allow, like running it locally so a shop's customer data never leaves their machine, or fine-tuning it on one owner's past replies so it sounds like them.

Prize Categories

I'm not entering any partner categories.