This Way: Build Community in Your Own Apartment Building

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 1: Touch Grass

What I Built

This Way helps apartment neighbors find common ground and turn an introduction into an outdoor plan.
The name comes from the Friends theme song. Growing up watching that show, I loved the idea of having friends close enough to share everyday life with. This project asks: what if those relationships could start in your own apartment building?
In the demo, you introduce yourself, choose your availability, and say what kind of friendship you’re looking for. This Way finds neighbors with matching schedules and prioritizes shared friendship goals. Then Gemma suggests a low-cost outdoor activity based on your interests.
The goal is to make the screen a starting point: meet someone nearby, find something to do together, and get outside.

Demo

Try This Way : https://this-way-cj3b.onrender.com
No account is needed:

1. Click Try the demo.
2. Enter fictional details or choose Use example answers.
3. Click Find My Matches.
4. Ask Gemma for an outdoor activity with a match. The building contains six fictional neighbors. The email reveal demonstrates a future contact feature; those addresses do not reach real people.

Code

The source code is on GitHub:
https://github.com/powtog/this-way

How I Built It

I built This Way with HTML, CSS, browser JavaScript, and a small Node.js server.
The homepage shows an apartment building with six illustrated neighbors behind lit windows. I wanted the interface to feel welcoming and connected to the place where these friendships could begin.
Matching uses straightforward rules: shared availability comes first, followed by shared friendship goals. AI handles the next step—helping two people decide what to do together.
For that, the backend calls Gemma 4, Google’s open-weight model, through Google’s hosted API. It sends the two introductions and preferences with instructions to suggest one low-cost outdoor activity in two short sentences, without inventing shared interests or specific local venues.
Render hosts the frontend and backend together. The API key stays on the server. I also added request limits and caching so repeated requests can reuse recent suggestions. A persistent disk preserves the shared daily request count across restarts.
Profile preferences stay in the visitor’s browser. When someone requests an activity, the two profiles are sent to Google, so the demo asks visitors to use fictional details.

Why Does Open Innovation Matter?

Open-weight models give a small project like this room to grow.
I’m using hosted inference for this demo, but Gemma’s available weights create a path toward running the model on infrastructure I control or adapting it for this specific task. That matters for a project built around personal introductions and community.
A closed API could also generate activity suggestions. The difference is the ability to take the model beyond that provider’s endpoint. This version does not run locally or offline, but an open-weight foundation makes those future directions possible.
It also gives other builders something they can study and adapt: a small app where predictable matching rules work alongside AI, with a clear purpose for each.

My Agent Session

I built the project with help from Codex, working through the matching flow, Gemma integration, homepage design, and deployment.

Prize Categories

• Best Use of Gemma — Gemma generates personalized outdoor activity suggestions.
• Best Use of Render — Render hosts the public demo and its Node.js backend.