DEV Community: Mridul Bagla

Good Ground: A Local AI Garden Planner That Gets You Outside

Mridul Bagla — Sun, 11 Oct 2026 12:03:14 +0000

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

What I Built

Good Ground is a garden planner that turns your frost dates, sunlight, and chosen crops into a short list of gardening tasks for the week.

It’s for anyone who wants a little help getting started in the garden—and a reason to put their phone down and get outside. No gardening account or location sharing required.

Demo

Good Ground runs locally on your computer.

Demo video: [Add a link to your demo video]

To try it, install Ollama, download the model with ollama pull qwen2.5:1.5b, then run python server.py from the project folder and open http://127.0.0.1:8000.

Code

Good Ground on GitHub

How I Built It

Good Ground uses the open-weight Qwen2.5 1.5B model with Ollama for local inference. A small Python server sends the garden details to Ollama running on the same computer, then returns a structured list of suggested tasks to the browser.

After the one-time model download, the planner can work offline. It doesn’t need a Hugging Face token, a hosted inference API, or a per-request payment. The model and prompt can also be changed to suit a different garden or setup.

Why Does Open Innovation Matter?

Open weights and a local inference runner let this project make garden plans without sending someone’s frost dates or notes to a hosted AI provider. There are no per-request inference fees, and the prompt and model are replaceable.

The trade-off is that users need to download the model and have enough local resources to run it. In return, they get more control over their data and can keep using the planner without an internet connection after setup—something a hosted API alone wouldn’t provide.

My Agent Session

Not included.

Prize Categories

Built for friend: A Local-First Medication Tracker with Open-Weight AI

Mridul Bagla — Sun, 04 Oct 2026 17:42:44 +0000

This is a submission for the Hacktoberfest Weekend Challenge: Build for a Friend.

What I Built

I built Naman’s Local Med Tracker, a small command-line medication checklist for my friend Naman.

It helps him keep a list of medicines and label- or clinician-provided instructions, see the day’s reminder-time checklist, and record doses as taken or skipped. The information is saved locally, so it’s still there after the app closes.

It’s a simple routine aid—not a medical device. Naman enters reminder times based on instructions from a healthcare professional. The app doesn’t calculate doses, send background notifications, or replace medical advice.

Code

MRIDUL-MB / hacktoberfest-2026-med-track

Created a medicine tracker usign a open source model.

Naman's Local Med Tracker

A small medication-routine helper built for my friend Naman. It keeps a local medicine list, shows a daily dose checklist, and records doses as taken or skipped so it's easier to keep track of a routine.

What it does

Saves medicine names, label/clinician instructions, and reminder times to a JSON file on this computer.
Shows today's reminder checklist and keeps a taken/skipped history across restarts.
Optionally reformats instructions with the open-weight HuggingFaceH4/zephyr-7b-beta model.
Keeps the tracker usable without an API token or an internet connection.

Reminder times must come from Naman's label or healthcare professional. This app does not calculate doses, send background notifications, or provide medical advice. It is a checklist, not a substitute for a prescription or care.

Run it

Install dependencies:

python -m pip install -r requirements.txt

Start the tracker:

python med_tracker.py

Use Check today's doses to see the checklist and…

View on GitHub

How I Built It

I built the project in Python using Cursor Pro during Global Hack Week.

The core tracker works offline and doesn’t require an AI service. I added an optional feature that uses the open-weight Zephyr-7b-beta model through Hugging Face’s inference router to reformat text pasted from a label or clinician’s instructions.

The AI feature is opt-in. It sends only the text entered for reformatting to Hugging Face and its inference provider; it does not automatically send the locally saved medication list or dose history. The result is only a readability aid and should always be checked against the original instructions.

Why Does Open Innovation Matter?

Medication routines are personal, so I wanted the useful part of this project—the checklist and dose history—to work locally without sending that information to an AI service.

Open-weight models also make the optional AI feature easier to inspect and potentially swap or run differently in the future. For this version, inference is hosted, so using that feature requires internet access and a Hugging Face token. The project’s core tracking features don’t depend on it.

Prize Categories

Best Use of Open Source Models
Most Creative Solution for a Friend

How to change terminal colors and styling in Vs COde??

Mridul Bagla — Sun, 31 Jan 2021 08:10:23 +0000

I want my terminal like this but I don't know how to do it. Can anyone help me out with this??