TouchGrass AI: An Open-Weight AI That Gets You Off the Screen

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

TouchGrass AI: An Open-Weight AI That Gets You Off the Screen

What I Built

It is an offline-first outdoor quest companion designed to solve a simple problem:
What if AI didn't keep us on the screen, but actually gave us a reason to leave it?

Instead of using AI to generate another piece of content to consume, TouchGrass AI uses an open-weight AI model to generate short, personalized outdoor quests.

Users chooses:

• Activity type
• Difficulty
• Available time

The AI then generates a structured outdoor quest containing several real-world tasks.

Key Features

• AI-generated outdoor quests
• Open-weight local AI inference
• Offline-first quest execution
• Optional photo evidence for tasks
• XP and streak-based gamification
• Safety-aware quest generation
• Local persistence using IndexedDB
• No cloud database or account required for the MVP

Demo

Live Deploy Link - https://touch-grass-gilt.vercel.app/

The recommended demo flow is:

1. Open TouchGrass AI
2. Select the quest preferences
3. Generate an outdoor quest
4. Save the quest
5. Go outside
6. Complete the tasks
7. Add photo evidence where required
8. Return and complete the quest
9. Earn XP and continue the streak

Code

GitHub Repo Link - https://github.com/Pravesh165/TouchGrass-Challenge
This repository contains the complete source code, setup instructions, architecture details and documentation.

How I Built It

1. TouchGrass AI is built around local open-weight AI inference using Ollama.

2. The current quest-generation model is:
   qwen2.5-coder:7b

3. The application sends the user's quest preferences to a local Ollama instance and asks the model to return a structured JSON quest.

4. The generated quest contains:

5. Quest title

6. Description

7. Duration

8. Difficulty

9. Outdoor tasks

10. Task instructions

11. Photo evidence requirements

12. AI is used when creating the experience, but once the quest is saved, the user can complete it without continuously relying on an internet connection.

Why Does Open Innovation Matter?

Open innovation matters because an application like TouchGrass AI should not require sending every outdoor experience request to a closed AI service.

Using an open-weight model with local inference gives the project several important advantages:

1. Privacy Quest generation can happen locally instead of sending every request to a third-party AI provider.
2. Offline Potential The AI model can run locally, which fits naturally with the project's goal of reducing dependence on constant connectivity.
3. Experimentation Developers can change the model, prompts, safety rules, and generation pipeline without being locked into a single proprietary API.
4. Ownership The application can be built around an AI model that developers can inspect, run, modify, and integrate into their own workflows.

This is especially meaningful for TouchGrass AI because the entire idea is about reducing digital dependency.
It would be ironic to build an application encouraging people to disconnect while making the application itself completely dependent on a remote closed API.

My Agent Session

I used AI-assisted development while building the project, particularly Claude.

Prize Categories

This project is submitted for the Hacktoberfest Open-Source AI Challenge Week 1: Touch Grass.

I am solo submitting this project.