DEV Community: Viraj Hudlikar

TouchGrass Buddy: An Offline AI-Powered Outdoor Adventure Generator 🌿

Viraj Hudlikar — Sat, 10 Oct 2026 18:38:57 +0000

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

What I Built

Most AI applications today are designed to keep users engaged with screens for as long as possible.

For the Touch Grass challenge, I wanted to explore a different idea:

What if AI could encourage people to spend less time on screens and more time outdoors?

That idea became TouchGrass Buddy, an offline AI-powered outdoor adventure generator built using Streamlit, Ollama, and Llama 3.2.

The application generates personalized outdoor missions based on:

Available Time
Location
Adventure Type

Users can choose adventure styles such as:

🌳 Nature Exploration
📷 Photography
🏃 Fitness
👨‍👩‍👧‍👦 Family Fun
🧘 Mindfulness

The AI then creates a personalized outdoor mission designed to get users exploring the real world instead of spending more time online.

Example mission:

🌿 Neighborhood Mindfulness Walk

1. Observe three different sounds around you
2. Find something with an interesting texture
3. Spend two minutes focusing on your surroundings
4. Notice a plant you've never paid attention to before
5. Walk a route you've never taken

The app also includes:

Mission generation
Progress tracking
Mission completion tracking
Start New Adventure workflow

The goal is simple:

Generate Mission
       ↓
Go Outside
       ↓
Complete Challenges
       ↓
Track Progress
       ↓
Start New Adventure

The screen should be the shortest part of the experience.

Demo

Home Screen

Generated Mission

Mission Complete

Code

GitHub Repository:

👉 https://github.com/vhudlikar/touchgrass-buddy

How I Built It

TouchGrass Buddy is built entirely using open-source tools:

Python
Streamlit
Ollama
Llama 3.2

The workflow is straightforward:

The user selects:
- Available Time
- Location
- Adventure Type
The application sends these inputs to a locally running Llama 3.2 model through Ollama.
The model generates a customized outdoor mission.
The user tracks progress using the built-in mission tracker.
Once completed, the user can start a brand-new adventure.

The entire application runs locally on a laptop without requiring internet connectivity after setup.

Why Does Open Innovation Matter?

Open-source AI was not just a feature of this project. It was the foundation of the entire solution.

By using Ollama and Llama 3.2, TouchGrass Buddy can run completely offline while still providing personalized experiences.

This open-source approach provides several benefits:

✅ No internet connection required

✅ No API keys

✅ No recurring costs

✅ No vendor lock-in

✅ User data never leaves the device

✅ Models can be swapped or customized

A closed AI service would have required cloud access, API management, and ongoing usage costs.

For this project, the ability to run AI entirely on a local machine aligned perfectly with the goal of creating a private, portable, and low-cost outdoor adventure generator.

Open innovation made it possible to experiment quickly, build freely, and keep the experience accessible to anyone with a laptop.

My Agent Session

I did not record a DevRelay session for this project.

Instead, I documented the development process through Git commits, project screenshots, and the public GitHub repository.

Prize Categories

🏆 Best Open-Source AI Project

TouchGrass Buddy uses open-weight AI as a core part of the application through:

Ollama
Llama 3.2
Local inference

The project demonstrates how open-source AI can power useful real-world experiences while remaining completely offline and privacy-friendly.

🏆 Best Use of GitHub Copilot

GitHub Copilot assisted with:

Streamlit development
Python code generation
Debugging
UI refinement
Development acceleration

Copilot helped speed up implementation while allowing me to focus on the overall user experience and challenge goals.

What I Learned

This project reinforced an interesting idea:

AI doesn't always need to increase screen time.

Open-source AI can be used to support real-world experiences rather than replace them.

Building TouchGrass Buddy also helped me deepen my understanding of:

Local AI workflows
Ollama
Llama 3.2
Streamlit application development
Prompt engineering
Privacy-focused AI design

Future Enhancements

Potential future improvements include:

GPS-aware adventures
Mobile-friendly experience
Nature scavenger hunts
Achievement badges
Group adventures
Local park recommendations
Wellness and fitness integrations

Conclusion

TouchGrass Buddy started with a simple question:

Can AI help people spend more time outdoors instead of keeping them online?

Using Streamlit, Ollama, and Llama 3.2, I built an offline AI-powered outdoor adventure generator that encourages exploration, movement, mindfulness, and real-world experiences.

The project demonstrates how open-source AI can be used not only to create intelligent applications, but also to encourage users to step away from the screen and engage with the world around them.

GitHub Repository:

👉 https://github.com/vhudlikar/touchgrass-buddy

Built for a Friend: ServiceNow Interview Buddy Using Ollama and Llama 3.2

Viraj Hudlikar — Fri, 02 Oct 2026 19:06:50 +0000

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

ServiceNow Interview Buddy

What I Built

I built ServiceNow Interview Buddy, an AI-powered interview preparation tool designed for ServiceNow professionals.

The idea came from helping a friend prepare for ServiceNow interviews. While reviewing potential interview topics together, I realized that finding role-specific questions that matched both the target role and experience level often required searching multiple resources and manually organizing relevant content.

As someone who regularly participates in ServiceNow technical evaluations and interview discussions, I saw an opportunity to create a lightweight AI-powered assistant that could instantly generate realistic interview questions tailored to a candidate's role and experience level.

The application allows users to select:

ServiceNow Administrator
ServiceNow Developer
ServiceNow Consultant
ServiceNow Architect

and specify their years of experience.

The tool then generates customized ServiceNow interview questions using a locally running AI model.

Demo

Application Screenshots

Landing Page

Role Selection

Generated Questions

GitHub Repository

View the GitHub Repository

Code

The complete source code is available in the GitHub repository.

Key project files include:

app.py
requirements.txt
README.md

How I Built It

Technology Stack

Python
Streamlit
Ollama
Llama 3.2 (Open-weight AI Model)

The user interface was built using Streamlit.

Ollama was used to run the open-weight Llama 3.2 model directly on my laptop.

When a user selects a ServiceNow role and years of experience, the application dynamically generates a prompt and sends it to the local Llama 3.2 model, which then produces tailored interview questions.

The workflow is:

User selects a ServiceNow role.
User enters years of experience.
The application builds a dynamic prompt.
Llama 3.2 generates interview questions.
The questions are displayed through the Streamlit interface.

Everything runs locally without requiring any cloud-hosted AI service.

Why Does Open Innovation Matter?

Open innovation was a key factor in making this project possible.

Using an open-weight model provided several important benefits:

No API costs
No subscription fees
Full control over the model
Local execution
Better privacy
No vendor lock-in

Because the model runs through Ollama, the application can continue functioning even without internet access once the initial setup is complete.

A closed AI service would have introduced recurring costs and external dependencies. Open-source AI allowed me to build a practical and accessible solution that anyone can run on their own hardware.

Feedback From My Friend

The feedback was positive because the generated questions adapted to both the selected role and years of experience.

Instead of relying on static interview question lists available online, the application generated questions that felt more relevant to the selected ServiceNow role and experience level.

This helped make interview preparation more focused and highlighted areas that deserved additional attention before an interview.

Challenges

One of the challenges was learning how to connect a Streamlit application with a locally running AI model through Ollama.

Another challenge was ensuring that the generated questions remained relevant across different ServiceNow roles while keeping the application simple and easy to use.

My Agent Session

I used Microsoft 365 Copilot together with local development tools to plan, build, troubleshoot, document, and prepare this project for submission.

During development I:

Installed and configured Ollama
Downloaded and tested the Llama 3.2 model
Built the Streamlit user interface
Connected the application to a locally running AI model
Created project documentation
Prepared the GitHub repository and screenshots
Drafted and refined the submission article

This demonstrates how AI-assisted development can accelerate the creation of practical applications while still requiring design, implementation, testing, and iteration by the developer.

Future Enhancements

Planned improvements include:

Interview answer evaluation
Candidate scoring
ServiceNow certification preparation
Learning recommendations
Mock interview simulation
Expected answer generation
Personalized preparation plans

Conclusion

Building ServiceNow Interview Buddy was an excellent opportunity to combine open-source AI with a real-world professional challenge.

What started as an effort to help a friend prepare for ServiceNow interviews evolved into a practical tool that generates role-specific interview questions for administrators, developers, consultants, and architects based on their experience level.

Using Ollama and the open-weight Llama 3.2 model made it possible to build the solution without API costs while ensuring that everything runs locally and remains under the user's control.

This project demonstrates how open innovation can be used to create practical learning and career-development tools. It also provides a strong foundation for future enhancements such as interview answer evaluation, scoring, certification preparation, and mock interview simulations.