DEV Community: Arya More

What If Your Rental Photos Could Help Protect Your Deposit?

Arya More — Sun, 18 Oct 2026 13:52:59 +0000

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

What I Built

Moving into a rental home is exciting, but moving out can be stressful. Tenants may struggle to prove whether a wall mark, damaged fixture, or other issue was already present when they moved in.

That's why I built DepositProof, an AI-powered rental documentation application that helps tenants record property conditions, organize photos, track maintenance issues, and prepare evidence for future reference.

The idea is simple: Document it before. Prove it later.

With DepositProof, users can:

Create property records and organize rooms.
Upload photos during move-in, maintenance, and move-out inspections.
Record existing damage and add notes to condition checklist items.
Maintain a log of maintenance issues.
Generate PDF evidence reports.
Use AI to describe property photos and suggest possible damage categories.

Live Demo

Try DepositProof here: https://rental-evidence-app-4t0s.bolt.host

Explore the dashboard, create a property record, document a room, and see how rental evidence can be organized in one place.

Why I Built It

A rental deposit can represent a significant amount of money. However, documenting a property's condition is easy to forget until the day a tenant moves out.

I wanted to build something practical that encourages tenants to record existing conditions early, keep their evidence organized, and make it easier to review their records later.

DepositProof is my attempt to turn a stressful, disorganized process into a simple documentation workflow.

How Open-Source AI Powers DepositProof

DepositProof connects to Ollama and uses LLaVA 7B (llava:7b), an open-source vision-language model, to analyze property photos.

The workflow is:

A user uploads a photo of a room or property feature.
The application stores the photo using its configured storage integration.
When the user requests AI analysis, the image is sent to their configured Ollama instance.
The model suggests a description and possible damage categories based on visible details.
The user reviews the suggestions and documents the property's condition.

Running inference locally gives developers greater control over where AI processing happens and reduces dependence on paid external AI APIs.

AI observations are suggestions, not definitive findings. The model cannot reliably establish when damage occurred, who caused it, or who is legally responsible.

Tech Stack

React and TypeScript — frontend application
Vite — development and build tooling
Tailwind CSS — styling
Supabase — configured storage and data integration
Ollama — local AI model runtime
LLaVA 7B — vision-language model for property photo analysis

What Makes DepositProof Useful?

DepositProof brings several parts of rental documentation together:

Move-in evidence: Capture property conditions at the beginning of a tenancy.
Maintenance tracking: Keep a record of issues that need attention.
Move-out documentation: Organize photos and compare recorded conditions.
PDF reports: Prepare a printable summary of inspection evidence.
AI-assisted observations: Get a starting point for describing visible property conditions.

The goal is not to replace careful inspection or professional judgment. It is to make documentation easier to maintain and review.

What I Learned

Building DepositProof gave me practical experience connecting a web application to a locally running vision-language model.

It also taught me that building an AI application involves more than adding a model. The surrounding experience matters: file uploads, storage, inspection workflows, error handling, and making the limitations of AI clear to users.

I learned to treat model outputs as suggestions that people should review, rather than automatically treating generated descriptions as facts.

What's Next?

I want to continue improving DepositProof by:

Refining photo uploads and AI analysis.
Making move-in and move-out comparisons more useful.
Improving PDF reports and evidence organization.
Strengthening deployment and AI backend configuration.
Making the project easier for other developers to run and contribute to.

Why Open-Source AI Matters

Open-source AI gives developers the opportunity to experiment, learn, and build practical tools without depending entirely on paid AI services.

With DepositProof, I can run a vision-language model locally, experiment with its capabilities, and build a useful application around it.

My goal is to make rental documentation more organized, accessible, and practical with the help of open-source AI.

I'd love to hear your feedback and suggestions for making DepositProof more useful!

Built for Hacktoberfest 2026 using open-source AI.

🌿 EcoSnap: From Waste to Action

Arya More — Sat, 10 Oct 2026 18:44:39 +0000

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

🌿 EcoSnap: From Waste to Action

What I Built

EcoSnap is an AI-powered waste identification tool designed to turn a simple photo into an opportunity to take action for the environment.

We see litter on roadsides, in parks, and around our neighborhoods every day. We know waste is a problem, but sometimes we don't know what kind of waste we're looking at or how to dispose of it responsibly.

I built EcoSnap to make that first step easier.

With EcoSnap, users can upload or capture a photo of waste and use AI to identify its likely category and get guidance on responsible disposal.

The idea is simple: instead of just scrolling through environmental awareness posts, people can step outside, observe their surroundings, and learn how to make a small difference.

EcoSnap is designed for students, communities, and anyone who wants to become more conscious of the waste around them.

I wanted to build something that connects technology with the physical world. AI doesn't have to keep us staring at screens. It can help us understand what's happening around us and encourage us to do something about it.

The goal isn't just to recognize waste. It's to turn recognition into action.

🌱 EcoSnap in Action

Here's EcoSnap analyzing a waste image using the LLaVA 7B vision-language model. The goal is to help people identify waste and make more informed disposal decisions.

Demo

🌎 Try EcoSnap: https://ecosnap-ai-environme-hks8.bolt.host

The demo lets you explore the EcoSnap interface and try the waste-scanning experience.

Code

💻 GitHub Repository: https://github.com/aryamore05/EcoSnap

I'd love for others to explore the project, suggest improvements, and contribute ideas for making environmental action more accessible.

How I Built It

I built EcoSnap with a focus on open-source AI and a straightforward architecture.

The project uses:

React, TypeScript, and Vite for the frontend.
Python and FastAPI for the backend API.
Ollama to run an AI model locally.
LLaVA 7B, an open-weight vision-language model, to analyze images.
SQLite for lightweight data storage.

The core workflow is:

A user uploads or captures a photo of waste.
The frontend sends the image to the backend.
The backend passes the image to the vision AI model.
The model analyzes the image and helps identify the likely waste category.
EcoSnap presents the result and relevant disposal guidance.

I used Bolt to help build the application and bring the frontend and backend together.

One of my goals was to make the project practical rather than unnecessarily complicated. I wanted a working application that demonstrates how an open-weight AI model can be integrated into a real-world environmental use case.

Why Does Open Innovation Matter?

Open innovation matters because useful AI should not be limited to people who have access to expensive APIs or proprietary platforms.

For EcoSnap, using an open-weight vision model gave me an opportunity to experiment with image understanding while running inference locally through Ollama. I could explore the AI workflow, connect it to my own backend, and build around a model I could run and test myself.

This also gave me more control over the architecture and helped me understand how the pieces fit together: image input, model inference, backend logic, and the user experience.

Beyond the technical benefits, open innovation makes experimentation more accessible. Students and independent developers can build projects around real problems without having to start with a large budget or a closed AI ecosystem.

EcoSnap is a small project, but it represents something I believe in: open-source AI can help people build practical tools for problems in their own communities.

And when the problem is environmental waste, the best outcome isn't another hour spent online. It's someone learning something useful and taking that knowledge into the real world.

My friend kept saying ‘I know I wrote this somewhere’ — so I built Somewhere

Arya More — Sat, 03 Oct 2026 17:44:45 +0000

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

1.What I Built

I built Somewhere, a local-first AI search tool for notes you already wrote down.

I built it for a friend who writes everything down.

Project ideas. Things to buy. Weekend plans. Random thoughts. Reminders. Little ideas that seem important at the time.

The problem is that writing something down doesn't mean you'll be able to find it later.

A few weeks after writing something, the memory usually sounds like:

"I know I wrote this somewhere..."

Maybe you remember that it was about an AI project.

Maybe you remember that it worked offline.

But you don't remember the exact words, the filename, or which note it was in.

Opening Notes and scrolling through everything isn't really search.

So I built Somewhere.

Instead of asking:

"What exact word did I write?"

you can ask:

"I remember something about an AI app that works offline."

Somewhere searches the notes by meaning, finds the relevant information, and uses a local AI model to explain what it found.

It also shows the original source note and snippet.

The idea is simple:

You already wrote it down. I'll help you find it.

2.Demo

The main workflow is:

Write something down → forget where you put it → describe what you remember → find it.

For example:

"I remember something about an AI app that works offline."

Somewhere searches the existing notes and retrieves the relevant information.

Another example:

"What did I want to buy?"

And even:

"tomatoes"

can find the relevant shopping and weekend notes.

The important part is that Somewhere isn't relying only on exact keyword matching.

It is trying to understand what you mean.

I also added a + New Note flow.

That means the complete workflow can happen inside Somewhere:

Create a note → save it → index it → forget about it → find it later.

3. Screenshots

Here are a few screenshots of Somewhere in action.

Search interface

Search results with sources

My notes section

4.Code

The complete source code:

GitHub Repository — Somewhere

5.How I Built It

The core of Somewhere is a local RAG pipeline:


text
Notes
  ↓
Chunking
  ↓
Local Embeddings
  ↓
Qdrant
  ↓
Semantic Search
  ↓
Relevant Note Chunks
  ↓
Qwen3
  ↓
Answer + Sources