404: Nature Not Found? WildQuest Fixes That.

Iqra Khan — Fri, 09 Oct 2026 19:22:40 +0000

🌿 WildQuest: Touch Grass, For Real — An Open-Source AI Adventure Generator

Less scrolling. More living.

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

What if AI didn't try to keep you on your screen longer, but helped you put your phone away?

That's the idea behind WildQuest — a browser-based outdoor adventure generator that uses open-weight AI to turn your mood, available time, interests, and surroundings into a small, practical outdoor mission.

Instead of endlessly scrolling for something to do, you can generate an activity, step outside, explore your surroundings, and reconnect with the world beyond your screen.

What I Built

WildQuest — Touch Grass, For Real is a lightweight, interactive web application designed to help people take meaningful breaks from screens and discover the outdoors in everyday life.

The idea is simple: going outside shouldn't require an elaborate plan, expensive equipment, or an entire free afternoon. Sometimes, all you need is a little inspiration.

WildQuest lets users personalize an outdoor quest using four choices:

Setting: A nearby park, a garden or balcony, a neighbourhood walk, or a nature trail.
Available time: 5, 15, 30, or 60 minutes.
Energy level: Choose the mood that fits how you feel that day.
Interests: Explore themes such as birds and wildlife, plants and small details, movement, mindfulness, or spending time with a friend.

Based on these preferences, the application uses an open-weight AI model to generate a short outdoor adventure with a title, introduction, mission, and actionable steps.

For example, someone with only 15 minutes to spare might receive a sensory exploration challenge, a tiny wildlife observation mission, or a short neighbourhood discovery activity.

The goal isn't to make people interact with another app for hours. It's to make the decision to go outside easier — and then encourage them to put the phone away.

Key features

Personalized outdoor quests: Activities are generated from the user's selected setting, available time, energy, and interests.
Open-weight AI in the browser: The model performs inference locally in the browser instead of relying on a private AI inference API.
No paid inference API key: The demo does not require a dedicated AI backend or paid inference service.
Fallback activities: If the AI library or model fails to load, users can still receive a ready-made outdoor quest.
Responsive interface: The application is designed for desktop and mobile screens.
Safety-conscious prompts: Generated activities are guided to avoid disturbing wildlife, entering private property, eating unknown plants, or suggesting risky activities.

WildQuest is built around one small but meaningful idea: technology can be useful without demanding more of our attention.

Demo

🌐 Live website: https://iqrakhan2909.github.io/WildQuest-Touch-Grass-Project/

💻 GitHub repository: https://github.com/iqrakhan2909/WildQuest-Touch-Grass-Project

Try choosing different settings, time limits, energy levels, and interests to see how the outdoor quest changes.

The first AI model load requires an internet connection and may take some time. If local model generation is unavailable, the application falls back to a ready-made quest. Fully offline operation on the first visit is not guaranteed.

Code

The complete project is available on GitHub:

👉 Explore the WildQuest source code

The project uses a simple three-file frontend structure:

index.html — The page structure, outdoor quest planner, input controls, and results interface.
style.css — The visual design, responsive layouts, animations, and nature-inspired styling.
script.js — Preference handling, prompt construction, local AI model loading, output processing, fallback quests, and user interactions.

The repository also includes a README explaining setup, features, the AI implementation, safety considerations, and future improvements.

The project is released under the MIT License, making it available for others to explore, adapt, and build upon under the license terms.

How I Built It

I built WildQuest using HTML, CSS, vanilla JavaScript, Transformers.js, and the open-weight FLAN-T5 Small model.

The goal was to create a practical AI-powered experience without depending on a paid hosted language-model API or building an unnecessarily complicated backend.

1. Frontend and user experience

I used HTML to structure the application and collect the user's preferences, CSS to create a responsive, nature-inspired interface, and vanilla JavaScript to manage form submissions, model interactions, and generated quests.

I wanted the experience to be straightforward: choose a few preferences, generate a quest, and head outdoors.

2. Open-weight AI model

The application uses:

Model: Xenova/flan-t5-small
Inference library: Transformers.js
Task: Text-to-text generation
Runtime: The user's browser

Transformers.js loads the model and enables it to generate text in the browser. There is no separate private AI inference server handling the user's activity preferences.

3. Turning preferences into an outdoor mission

When a user submits the form, JavaScript collects their choices and builds a prompt containing the selected setting, duration, mood, and interests.

The prompt instructs the model to produce four sections:

A short, engaging title.
A friendly introduction.
A clear outdoor mission.
Three simple, practical steps.

The instructions also encourage safe, accessible activities that require no special equipment and help the user spend less time looking at a screen.

The generated response is parsed and displayed in the interface.

4. Making the experience resilient

A model can fail to load, take too long to initialize, or return output that doesn't match the expected format. I didn't want those situations to make the entire application unusable.

WildQuest therefore includes a collection of ready-made quests. If local generation fails or the model output cannot be parsed, the application can display a fallback activity instead.

This means the core experience remains available even when the AI component isn't working.

5. Deployment

The frontend is deployed through GitHub Pages, making the application accessible through a public website without requiring a dedicated application server.

The initial page, Transformers.js library, and model download require internet access. Although browser caching may allow some resources to be reused later, WildQuest does not claim guaranteed offline operation.

Why Does Open Innovation Matter?

For WildQuest, open innovation isn't just about using a different model. It changes how the application can be built, inspected, adapted, and used.

1. More control over the technology

Using an open-weight model gives developers the opportunity to experiment with prompts, adjust application behavior, and explore compatible models rather than depending entirely on a closed provider's API and product decisions.

WildQuest can be extended with different types of outdoor activities, community-specific suggestions, or alternative models as the project evolves.

2. A more privacy-conscious architecture

People should be able to choose a simple outdoor activity without needing to send their preferences to a private AI inference service.

WildQuest processes the selected preferences and runs model inference in the browser. This reduces the need to transmit those preferences to a separate AI inference provider.

This does not mean that every part of the website is offline or that the initial model download requires no network connection. It means that the AI inference itself is designed to happen on the user's device.

3. Lower barriers to experimentation

A small project shouldn't necessarily need a paid API key, a backend server, and ongoing inference costs just to demonstrate an idea.

By using browser-based inference and open-weight AI, WildQuest offers a way to experiment with AI-powered personalization without a dedicated paid inference service.

The model download and the user's device capabilities still matter, but the architecture keeps the project relatively accessible to learners and independent developers.

4. The community can build on it

Open-source projects create opportunities for others to inspect the implementation, report problems, improve prompts, suggest safer activities, and contribute new features.

Someone could extend WildQuest with local weather information, a nature journal, bird and plant observation tools, or more reliable installable offline support.

These are potential future improvements, not features currently claimed by the demo.

5. AI should serve people, not just screen time

Perhaps the most important reason open innovation matters here is the purpose behind the technology.

AI is often experienced through products designed to maximize engagement. WildQuest explores a different direction: use AI to reduce the effort needed to make a decision, then encourage the user to leave the interface and experience the real world.

The best outcome isn't a longer session on WildQuest. It's someone closing the tab and discovering something outside.

My Agent Session

I haven't included a DevRelay agent session in this submission.

The project source code is available in the GitHub repository linked above. If I record and publish a DevRelay session, I'll add its link here.

Prize Categories

Hacktoberfest Open-Source AI Challenge Week 1 — Touch Grass: Overall challenge submission.

I'm not claiming eligibility for any specific partner prize category because the current project is built with Transformers.js and FLAN-T5 Small rather than the listed partner technologies.

What's Next?

WildQuest is a starting point, and there are several directions I'd like to explore:

Weather-aware quests based on local conditions.
A nature journal to record discoveries after returning indoors.
Bird and plant observation activities.
Better model-loading feedback and more consistent AI output.
Optional installable PWA support with tested offline capabilities.
More opportunities for community contributions.

These improvements could make WildQuest more useful while keeping its central purpose intact: helping people spend less time on screens and more time noticing the world around them.

Final Thoughts

Building WildQuest made me think about a different way to approach useful AI.

Instead of asking how to make people spend more time with a product, I wanted to explore how technology could help them do something meaningful beyond it.

An outdoor adventure doesn't have to be a mountain hike or an entire weekend away. It could be noticing a bird outside your home, taking a different route through your neighbourhood, or spending five quiet minutes observing the plants around you.

Sometimes, a small nudge is enough to get started.

🌱 Less scrolling. More living. Touch grass, for real.

hacktoberfest #hf26challenge #devchallenge #opensource #ai #webdev

DEV Community: Iqra Khan