I Built TouchGrass AI: A Little Less Scrolling, A Little More Outside

Description and use

A lightweight web app that suggests outdoor activities using a local Qwen 2.5 1.5B model via Ollama, encouraging screen breaks without paid APIs.

Problem

People spend too much time on screens and need simple ways to encourage outdoor activity.

Outcome

The app provides quick, personalized outdoor activity ideas without relying on paid APIs, demonstrating how local open-weight models can power web experiences.

Workflow

• 1. User visits the web app
• 2. User inputs mood, available time, and preferred setting
• 3. JavaScript sends input to local Qwen 2.5 1.5B model via Ollama
• 4. Model generates activity suggestions
• 5. Suggestions are displayed to the user

Evidence

It is a simple web app that suggests outdoor activities based on a user's mood, available time, and preferred outdoor setting.

Source

The app uses HTML, CSS, and JavaScript for its interface and connects to a locally running Qwen 2.5 1.5B open-weight model through Ollama to generate suggestions.

Source

Running an open-weight model locally helped me learn how AI can be integrated into a web app without relying on a paid hosted AI API.

Source