DEV Community: Isoken Ibizugbe

Threadwise: A Local-First AI Closet That Makes Getting Dressed Easier

Isoken Ibizugbe — Fri, 23 Oct 2026 10:23:59 +0000

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 3: Slow and Steady

Threadwise is a virtual closet that helps me keep track of clothes I already own, put outfits together, and plan what to pack for a trip. The idea is to make the small, repetitive decisions around getting dressed take less effort.

The problem

I wanted one simple place to remember what is in my wardrobe and make it easier to reuse those pieces. Instead of scrolling through photos, trying combinations from memory, or writing a separate packing checklist, I can add garments once and work with them in the app.

What I built

Add a clothing photo from a camera or photo library, give it a name, and choose its category, colour, and season.
Remove the photo background in the browser and save the garment in a local wardrobe.
Filter the wardrobe by clothing category.
Select pieces to create a look manually, or ask an optional local AI stylist for suggestions.
Create a trip packing list, enter a destination, and mark selected pieces as packed.

Demo

threadwise-d654.onrender.com

Code

Junie06 / threadwise

Threadwise

Threadwise is a mobile-friendly, local-first wardrobe app: photograph a garment, remove its background, organize your closet, build outfits, and make a trip packing list.

Threadwise uses an open-weight model through local Ollama inference; your photos stay in the browser.

What works in this MVP

Add clothing photos from a camera or photo library. Background removal runs in the browser, before the transparent image is saved to IndexedDB.
Sort and filter a personal wardrobe by tops, trousers, skirts, dresses, outerwear, shoes, and accessories.
Ask a locally running Ollama model for outfit ideas based on item names, colors, categories, and seasons.
Build a look yourself, or create a trip packing list and mark pieces as packed.
Use the app from a phone on the same Wi-Fi network as the computer running it.

Run it locally

Requirements: Node.js 20.19 or newer. Ollama is optional; without it, photo capture, your wardrobe, manual outfit…

View on GitHub

How the AI works

Threadwise uses AI for two distinct jobs:

Photo preparation: @imgly/background-removal runs in the browser to create a garment cutout before it is saved.
Outfit ideas: an optional Ollama service runs the qwen3:4b text model locally. It receives garment metadata—names, categories, colours, and seasons—not the photos. The model suggests combinations from the pieces already in the closet.

Manual outfit building, wardrobe organization, and packing do not require Ollama. If the local model is unavailable, those parts of the app remain usable.

Why open innovation matters here

Clothing photos can be personal. Using an open-source, browser-side image-processing library means the photos do not need to be uploaded to a third-party image API for background removal. Running the outfit model with Ollama also gives people a way to try AI styling without sending wardrobe metadata to a hosted model provider. The open tools make it possible to inspect, run, and adapt the parts that matter to this experience.

The photo-processing model assets are fetched by the browser when needed, but garment photos and wardrobe records are stored in that browser's IndexedDB. They are not synced between devices or backed up by the server.

How it is built

Frontend: React, TypeScript, and Vite
Local wardrobe storage: IndexedDB
Image processing: @imgly/background-removal
Optional text AI: Ollama with Qwen3
API and production serving: Express

graph TD
  Person[You] --> Browser[React app in browser]
  Browser --> Cutout[Browser-side background removal]
  Browser --> Wardrobe[(IndexedDB: photos and wardrobe)]
  Browser -->|Garment metadata only| API[Express API]
  API --> Ollama[Optional local Ollama + Qwen3]
  API --> Static[Production app files]

The project is licensed under AGPL-3.0. The background-removal library is also AGPL-3.0.

The Lazy Payoff

The payoff is less repeated searching and remembering: the wardrobe becomes a reusable list for outfit planning and packing. I have not measured a time-saved number yet, so I won't invent one; a useful next test is to compare preparing a real trip list with and without Threadwise.

What I learned and what is next

The most important product decision was to keep the closet useful without the AI. A local model can add convenience, but the basic task should not fail just because model setup is unavailable. Keeping photos in the browser also makes the privacy boundary easier to explain.

Prize Categories

Best Use of Render

Staffside: New staff onboarding has never been easier

Isoken Ibizugbe — Sat, 17 Oct 2026 16:03:50 +0000

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

What I Built

I built Staffside, a local-first people-ops workspace for small startup teams. It brings together a simple staff directory, per-person onboarding tasks, handbook Q&A, and payment-status tracking.

It’s for founders and small teams who end up doing HR work alongside their regular roles. Staffside helps keep onboarding and admin tasks in one place, while an AI assistant can answer questions from the team’s own handbook.

The payment tracker records amounts, periods, due dates, and paid/pending status. It does not process payroll, move money, calculate taxes, or store bank details.

Demo

Code

Junie06 / staffside

Staffside

Staffside is a small, local-first people-ops workspace for startup teams. It keeps staff profiles, onboarding tasks, payment-status records, and handbook documents on the machine running the app. An open-weight model served by Ollama drafts onboarding checklists and answers questions using retrieved handbook excerpts.

Run locally

Requirements: Node.js 20 or later. The staff directory, tasks, documents, and payment tracker work without an AI model.

npm start

Open http://127.0.0.1:3000. Workspace data is saved unencrypted in data/staff-data.json. The app binds to loopback only; it is a single-user local prototype, not a multi-user hosted service.

Enable the local AI

Install Ollama, then download and start the default model:

ollama pull qwen3:4b

Keep Ollama running, then start or refresh Staffside. The default model is qwen3:4b; set OLLAMA_MODEL to use another Ollama model. If Ollama runs at a non-default local URL, set OLLAMA_URL (for example, http://localhost:11434). Non-loopback model URLs are rejected…

View on GitHub

How I Built It

Staffside uses Ollama 0.40.5 with Qwen3 4B (Q4_K_M). It is a dependency-free Node.js web app with a small browser-based frontend and local JSON persistence. Its AI features are built around Ollama and the configurable open-weight model qwen3:4b.

For handbook Q&A, the app searches local documents for matching passages and sends only those excerpts, along with the question, to the local model. Staff profiles and payment records are not included in the AI prompt. For onboarding, the model drafts a short checklist grounded in retrieved handbook passages; the user reviews the suggestions before adding them.

Why Does Open Innovation Matter?

This prototype makes it possible to keep inference on the same machine as the app, rather than sending handbook excerpts to a hosted AI API. The model can be swapped by configuration, giving the team more control over where inference runs and which model it chooses.

That’s a useful privacy and flexibility trade-off for sensitive workplace information.

Unsit: Helps you get off your screen and go outside

Isoken Ibizugbe — Fri, 09 Oct 2026 16:12:22 +0000

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

What I Built

A lot of apps ask us to spend more time looking at a screen. UnSit is meant to do the opposite: help someone choose a small outdoor break, then put their phone away.
UnSit creates gentle movement-break options based on how much time someone has, their energy, where they’re headed, and what feels comfortable.
Choose a 5-, 10-, or 15-minute break and set a focus and outdoor setting. UnSit offers three options: an outdoor walk, a park or bench reset, and a porch or balcony reset. Pick one to start a timer; after the break, an optional check-in and a small activity history stay in the browser.

Demo

https://unsit.onrender.com

Code

Junie06 / unsit

UnSit

UnSit is a guest-first, local-first web app for short outdoor movement breaks. Pick a time, energy level, focus, and setting to get three options: an outdoor walk, a park/bench reset, or a porch/balcony reset.

The dashboard shows progress toward a five-break weekly goal and a seven-day streak ring. After a break, users can optionally record a private, subjective check-in (better, the same, or not so great); it is not a pain score or medical assessment.

The interface is a mobile-ready PWA. On a secure origin, users can opt into running Gemma 3 1B directly in a WebGPU-capable browser; inference stays on the device. The first use downloads about 600 MB of model data, so the app keeps its built-in plans available as a no-download fallback. Personal plans and check-ins stay in that browser's local storage. The Node/Ollama endpoint is only used on localhost when browser Gemma is unchecked. The…

View on GitHub

How I Built It

Vanilla JavaScript, HTML, and CSS
Node.js for local Gemma/Ollama integration
Gemma 3
A service worker, web app manifest, and browser local storage

Why Does Open Innovation Matter?

For an app involving personal comfort preferences, local inference is useful because the user can try model-generated suggestions without sending those preferences to a hosted AI provider. Using an open-weight model also makes it possible to run the generation path on a device the user controls.

Prize Categories

Best Use of Gemma
Best Use of Copilot
Best Use of Render

Kinsync: Sync your family activities

Isoken Ibizugbe — Sun, 04 Oct 2026 20:23:56 +0000

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

What I Built

It happens constantly in our house: Dad mentions an event or a trip the day before it’s supposed to happen. My sisters and I get so annoyed trying to reshuffle our schedules at the last minute!

During one of those classic frantic moments, someone said, "We seriously need a shared family planner so everyone is actually on the same page." That was the spark for KinSync.

With KinSync, anyone can just drop a casual message in the family chat ("Dad has a dentist appointment on Tuesday at 3 PM" or "Someone pick up groceries tomorrow"), and our AI assistant automatically pulls out the details and puts them straight onto our shared family schedule.

Demo

Try it out here:

kinsync-8u8b.onrender.com

Code

Junie06 / kinsync

This is a Next.js project bootstrapped with create-next-app.

Getting Started

First, run the development server:

npm run dev
# or
yarn dev
# or
pnpm dev
# or
bun dev

Open http://localhost:3000 with your browser to see the result.

You can start editing the page by modifying app/page.tsx. The page auto-updates as you edit the file.

This project uses next/font to automatically optimize and load Geist, a new font family for Vercel.

Learn More

To learn more about Next.js, take a look at the following resources:

Next.js Documentation - learn about Next.js features and API.
Learn Next.js - an interactive Next.js tutorial.

You can check out the Next.js GitHub repository - your feedback and contributions are welcome!

Deploy on Vercel

The easiest way to deploy your Next.js app is to use the Vercel Platform from the creators of Next.js.

Check out our Next.js deployment documentation for more…

View on GitHub

How I Built It

Framework: Next.js (App Router), TypeScript, Tailwind CSS
AI Engine: Gemma 2 (gemma2-9b-it via Groq API for lightning-fast cloud response; with full local support via Ollama)
Deployment: Render

Why Does Open Innovation Matter?

Relying on closed models means accepting sudden API price hikes, model deprecations, or sudden policy changes that can break an app overnight. Open innovation guarantees that our intelligence layer remains accessible forever, whether served via low-latency cloud providers (like Groq) or run directly on edge devices inside the home.

Prize Categories

Render
Gemma
Copilot