GroundTruth: Turning Everyday Observations Into Real-World Action 🌿
This is a submission for the Hacktoberfest Open-Source AI Challenge
Week 1: Touch
Grass.
What I Built
We walk past small problems every day.
A ramp is blocked. A public bin is overflowing. A tap is leaking. A
campus facility needs repair. We notice, think "someone should report
this," and keep moving.
It isn't always because people don't care. Sometimes they don't know who
to contact. Sometimes explaining the issue feels like a task of its own.
Sometimes they simply don't know where to begin.
So the observation stays an observation.
That is the gap I wanted to work on with GroundTruth.
GroundTruth is an AI-assisted public-space reporting tool. It helps
people turn a rough description of a real-world issue into a clear,
structured report they can review, edit, copy, and share with the
appropriate person or organization.
The workflow is simple:
Describe what you noticed and where it happened.
Let AI help organize the description into a structured report.
Review and edit the result so it reflects what you actually
observed.
Copy or save the report and decide what to do next.
GroundTruth is designed for students, residents, and anyone who notices
a problem in a campus or public space. It covers issues such as
accessibility, infrastructure, waste management, water problems, and
public facilities.
It won't repair a broken ramp by itself, and it doesn't automatically
submit reports to authorities. What it can do is make the first step
less intimidating.
The idea is simple: less friction between caring about a problem and
doing something about it.
Demo
🌐 Live application: https://client-steel-seven-78.vercel.app/
Try describing a real issue, generate a draft, review it, and copy or
save the report.
The generated text is a draft, so please check its accuracy before
sharing it.
Code
💻 GitHub repository: https://github.com/princequadai/GroundTruth
The project uses a lightweight full-stack setup:
Frontend: React and Vite
Backend: Node.js and Express
AI: Hosted inference through a model API
Storage: Browser localStorage for saved reports
I kept the architecture focused on the core workflow instead of adding
accounts, a database, or integrations that aren't necessary for the
first version.
How I Built It
The frontend collects the issue category, location, and the user's
description. The backend validates the request and sends it to the
configured hosted AI model. The model helps turn the description into a
structured report, including a title, a clearer description, a suggested
next step, and details the user may want to verify.
One important choice was to keep the result editable. AI can help
someone express an observation more clearly, but it should not invent
facts or pretend an issue has been verified. The person who saw the
problem remains responsible for checking the final report.
Saved reports are stored in the browser, so the initial experience
doesn't require account creation or a separate database.
The model runs remotely, which means generating a new AI report requires
an internet connection. The application does not run the model locally
or claim to be fully offline.
Model note: The exact model and provider should be named here only
after confirming what is configured in the deployed backend. The
implementation should not claim a particular open-weight model unless
the live version actually uses it.
Why Does Open Innovation Matter?
Public-space problems are local, but the challenge of reporting them is
familiar in many places. A student reporting a blocked campus ramp and a
resident reporting a damaged public facility may need the same thing: a
clear way to explain what they saw.
Open innovation makes it easier to experiment with how this workflow can
serve different communities. With an eligible open-weight model,
developers can compare models, adapt prompts, and explore local-language
support or community-specific report formats, subject to the model's
license and the provider's terms.
For GroundTruth, that flexibility could make it easier to explore
formats for campus maintenance, local public services, or languages
people are more comfortable using. It also gives other developers a
codebase they can inspect, adapt, and improve.
Open AI isn't valuable just because it is open. It becomes valuable when
that flexibility helps solve a specific problem and makes the result
more useful to real people.
Prize Categories
I will only list a partner prize category after confirming that the
model, provider, and implementation meet its current eligibility rules.
If the deployed version uses an eligible Gemma model, for example, I can
consider the relevant Gemma category; otherwise, I won't claim it.
A Small Step Is Still a Step
GroundTruth started with a simple observation: people can care about a
problem and still find it difficult to act on it.
I wanted to build something small that could help close that gap. The
next step is to test it with real users, learn whether the reports are
genuinely useful, and improve the workflow based on what they need.
Because the real measure isn't how many reports an AI can generate.
It's whether one of those reports helps someone take a useful step in
the real world. 🌱
Top comments (0)