DEV Community: Arkendu Kundu

TouchGrass AI: An Open-Source AI Companion to Help You Touch Grass

Arkendu Kundu — Sat, 10 Oct 2026 16:46:09 +0000

🌱 TouchGrass AI — a little less scrolling, a little more living

live demo: https://touchgrass-ai.vercel.app/

github repository: https://github.com/arkendukundu-dev/touchgrass-ai

#hf26challenge #hacktoberfest #opensource #ai #react #python

🌍 the problem

we spend so much time looking at screens that sometimes we forget to experience the world outside them.

many of us want to spend more time outdoors, explore new places, build healthier habits, or simply take a break from endless scrolling. but knowing we should take a break isn't always enough.

i wanted to turn that simple idea into something fun and actionable.

🌿 introducing touchgrass ai

touchgrass ai is an open-source, ai-powered wellness app that encourages people to step away from their screens and reconnect with the real world through small, achievable activities.

instead of just telling you to spend less time online, it gives you little outdoor quests to complete in real life.

✨ what can you do with it?

🌱 daily quests: discover activities that encourage you to get outside and explore.
🤖 ai-powered suggestions: generate personalized activity ideas using open-source ai.
⭐ xp and levels: earn experience points as you complete quests.
🔥 streak tracking: build consistency through small daily actions.
📝 field notes: record thoughts and experiences from your offline adventures.
🎯 custom quests: create your own challenges.
📊 progress tracking: see your journey toward healthier digital habits.
📴 offline-friendly fallback: keep using built-in quests when ai is unavailable.

🛠️ how i built it

i used the following technologies:

frontend: react, vite, and css
backend: python and fastapi
open-source ai: qwen 2.5 3b through ollama
storage: browser local storage
deployment: vercel for the frontend and render for the backend

🧠 why open-source ai matters here

for a project focused on digital well-being, i didn't want the solution to depend entirely on a closed-source ai service.

using an open-weight model such as qwen 2.5 3b through ollama gives developers the freedom to experiment with local inference, customize model behavior, and explore different models without depending on a paid proprietary ai api.

local inference can also help keep prompts on a user's own machine when configured that way. the model is replaceable, which makes the project easier to experiment with and extend.

i also wanted the basic experience to remain useful without ai. the built-in quests provide a fallback, so the app's core idea doesn't disappear when a model is unavailable.

🚀 try it yourself

🌐 live demo: https://touchgrass-ai.vercel.app/

💻 open-source code: https://github.com/arkendukundu-dev/touchgrass-ai

the project is built with an extensible architecture, and there's plenty of room to improve quest personalization, add achievements, and introduce more ways to encourage offline exploration.

🔮 what's next?

some ideas i'd love to explore:

more personalized quests based on interests and available time
achievements and community challenges
more outdoor activity categories
improved accessibility and mobile experiences

🌱 final thoughts

ai doesn't always need to keep us glued to a screen. sometimes, it can help us find a reason to put the screen down.

that's the idea behind touchgrass ai: use technology to make more room for real life.

i'd love to hear your feedback and ideas. what feature would make you more likely to step away from your screen and explore the world?

hf26challenge #hacktoberfest #opensource #ai

this is my submission for the hacktoberfest 2026 open-source ai challenge — week 1: touch grass.

hi

Arkendu Kundu — Sat, 10 Oct 2026 16:02:36 +0000

FriendStudy AI — Your Open-Source AI Study Companion

Arkendu Kundu — Wed, 07 Oct 2026 15:52:01 +0000

friendstudy ai — an open-source ai study companion 📚🤖

what if studying had a personal ai companion that could help you plan, learn, practice, and stay consistent?

that's the idea behind friendstudy ai.

i built friendstudy ai as an open-source ai-powered study companion for students who struggle with planning their studies, understanding difficult topics, and staying consistent.

✨ what it can do

💬 ask ai questions and get explanations
📅 generate personalized study plans
🧠 create practice quizzes
🎯 manage daily study goals
📊 track study progress
🔥 maintain study streaks
📚 make studying more structured and consistent

🤖 powered by open-source ai

instead of relying entirely on proprietary ai APIs, friendstudy ai uses qwen 2.5 3b through ollama.

the goal was to experiment with open-weight ai and understand what it takes to build a complete ai-powered application around an open model.

🛠️ tech stack

frontend: react + javascript

backend: python + fastapi

ai: qwen 2.5 3b + ollama

deployment: vercel + render

🔗 try it

🌐 live demo:

https://friendstudy-ai-hf26.vercel.app/

💻 github:

https://github.com/arkendukundu-dev/friendstudy-ai-hf26

🧠 what i learned

building this project taught me how different pieces of an ai application come together — from frontend and backend development to model integration, api communication, debugging, and deployment.

the biggest takeaway for me was:

building with ai isn't just about using a model. it's about building a useful product around it.

i built friendstudy ai for the launch weekend dev challenge and i'm excited to keep improving it with better personalization, smarter study plans, stronger quizzes, and more open-source models.

if you're a student, give it a try and let me know what you think! 🚀

I built friendstudy ai — an open-source ai study companion for students

Arkendu Kundu — Sun, 04 Oct 2026 17:02:49 +0000

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

What I Built

I built FriendStudy AI, an AI-powered study companion designed to help students study more effectively.

It provides students with an interactive study workspace where they can:

ask AI questions about any topic
generate study plans
practice with AI-generated quizzes
track study progress
maintain study streaks
organize their daily study goals

The idea came from a simple problem: students often have to switch between multiple tools for planning, learning, practicing, and tracking their progress.

FriendStudy AI brings these features together in one place.

Who I Built It For

I built FriendStudy AI for students who want a simple study companion that can help them understand topics, practice questions, and stay consistent with their studies.

The goal was to create something that feels less like a generic chatbot and more like a study companion that students can use as part of their daily routine.

Demo

Live Demo:

https://friendstudy-ai-hf26.vercel.app/

The AI features are powered by Qwen 2.5 3B, running through Ollama.

How It Works

The application uses:

React + JavaScript for the frontend
FastAPI + Python for the backend
Qwen 2.5 3B as the AI model
Ollama to run the open-weight model
Render for backend deployment
Vercel for frontend deployment
ngrok to connect the deployed backend with the locally running AI model

The production architecture looks like this:

React → Vercel → FastAPI → Render → ngrok → Ollama → Qwen 2.5 3B

The frontend communicates with the FastAPI backend, while the backend sends AI requests to the locally hosted Qwen model through Ollama.

Features

AI Study Companion

Students can ask questions and get AI-generated explanations using Qwen 2.5 3B.

Study Planning

FriendStudy AI can help students organize their study sessions and create structured study plans.

AI Quiz

Students can practice topics using AI-generated quiz questions.

Progress Tracking

The dashboard provides an overview of study progress, completed topics, daily goals, and study streaks.

Daily Goals

Students can set study goals and keep track of their daily progress.

What I Learned

This project helped me learn a lot about taking an AI application from local development to a deployed application.

Some of the biggest things I learned were:

connecting a React frontend with a FastAPI backend
deploying a FastAPI application on Render
deploying a Vite frontend on Vercel
working with Ollama and open-weight LLMs
connecting a deployed backend to a locally running AI model
exposing a local AI service for development using ngrok
debugging CORS and API connection issues
working with environment variables
troubleshooting deployment and build issues
connecting different services together into one working application

Challenges

The biggest challenge was getting the local Qwen model to work with the deployed application.

Initially, the frontend and backend worked separately, but the deployed backend could not directly access Ollama running on my computer.

I eventually built this architecture:

React → Vercel → FastAPI → Render → ngrok → Ollama → Qwen 2.5 3B

Getting every part of this pipeline working together required quite a bit of debugging.

I also had to troubleshoot frontend build issues, CORS configuration, environment variables, API connectivity, and the connection between the deployed backend and the local AI model.

Seeing the deployed application successfully send a request through the entire pipeline and receive a response from Qwen made the debugging worth it.

Code

The complete project is available on GitHub:

https://github.com/arkendukundu-dev/friendstudy-ai-hf26

Feel free to explore the code and try the application.

What's Next

I would like to continue improving FriendStudy AI by adding:

better personalized study plans
improved quiz generation
persistent student profiles
better progress analytics
faster AI responses
more AI-powered study tools
improved personalization based on student progress

Final Thoughts

Building FriendStudy AI was a great experience because I wasn't just building another AI chatbot.

I wanted to build something that could actually become a useful study companion for students.

This project also gave me hands-on experience with AI, React, FastAPI, deployment, open-weight models, Ollama, and connecting local AI infrastructure with a deployed web application.

Thanks for checking out FriendStudy AI!

If you try it, I'd love to hear your feedback.

I built friendstudy ai — an open-source ai study companion for students

Arkendu Kundu — Sun, 04 Oct 2026 11:11:21 +0000

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

What I Built

I built FriendStudy AI, an AI-powered study companion designed to help students study more effectively.

It provides students with an interactive study workspace where they can:

ask AI questions about any topic
generate study plans
practice with AI-generated quizzes
track study progress
maintain study streaks
organize their daily study goals

The idea came from a simple problem: students often have to switch between multiple tools for planning, learning, practicing, and tracking their progress.

FriendStudy AI brings these features together in one place.

Who I Built It For

I built FriendStudy AI for students who want a simple study companion that can help them understand topics, practice questions, and stay consistent with their studies.

The goal was to create something that feels less like a generic chatbot and more like a study companion that students can use as part of their daily routine.

Demo

Live Demo:

https://friendstudy-ai-hf26.vercel.app/

The AI features are powered by Qwen 2.5 3B, running through Ollama.

How It Works

The application uses:

React + JavaScript for the frontend
FastAPI + Python for the backend
Qwen 2.5 3B as the AI model
Ollama to run the open-weight model
Render for backend deployment
Vercel for frontend deployment
ngrok to connect the deployed backend with the locally running AI model

The production architecture looks like this:

React → Vercel → FastAPI → Render → ngrok → Ollama → Qwen 2.5 3B

The frontend communicates with the FastAPI backend, while the backend sends AI requests to the locally hosted Qwen model through Ollama.

Features

AI Study Companion

Students can ask questions and get AI-generated explanations using Qwen 2.5 3B.

Study Planning

FriendStudy AI can help students organize their study sessions and create structured study plans.

AI Quiz

Students can practice topics using AI-generated quiz questions.

Progress Tracking

The dashboard provides an overview of study progress, completed topics, daily goals, and study streaks.

Daily Goals

Students can set study goals and keep track of their daily progress.

What I Learned

This project helped me learn a lot about taking an AI application from local development to a deployed application.

Some of the biggest things I learned were:

connecting a React frontend with a FastAPI backend
deploying a FastAPI application on Render
deploying a Vite frontend on Vercel
working with Ollama and open-weight LLMs
connecting a deployed backend to a locally running AI model
exposing a local AI service for development using ngrok
debugging CORS and API connection issues
working with environment variables
troubleshooting deployment and build issues
connecting different services together into one working application

Challenges

The biggest challenge was getting the local Qwen model to work with the deployed application.

Initially, the frontend and backend worked separately, but the deployed backend could not directly access Ollama running on my computer.

I eventually built this architecture:

React → Vercel → FastAPI → Render → ngrok → Ollama → Qwen 2.5 3B

Getting every part of this pipeline working together required quite a bit of debugging.

I also had to troubleshoot frontend build issues, CORS configuration, environment variables, API connectivity, and the connection between the deployed backend and the local AI model.

Seeing the deployed application successfully send a request through the entire pipeline and receive a response from Qwen made the debugging worth it.

Code

The complete project is available on GitHub:

https://github.com/arkendukundu-dev/friendstudy-ai-hf26

Feel free to explore the code and try the application.

What's Next

I would like to continue improving FriendStudy AI by adding:

better personalized study plans
improved quiz generation
persistent student profiles
better progress analytics
faster AI responses
more AI-powered study tools
improved personalization based on student progress

Final Thoughts

Building FriendStudy AI was a great experience because I wasn't just building another AI chatbot.

I wanted to build something that could actually become a useful study companion for students.

This project also gave me hands-on experience with AI, React, FastAPI, deployment, open-weight models, Ollama, and connecting local AI infrastructure with a deployed web application.

Thanks for checking out FriendStudy AI!

If you try it, I'd love to hear your feedback.