🌿 TerraQuest AI — Touch Grass, Not Screens

#devchallenge #hf26challenge #opensource #ai

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

What I Built

I built TerraQuest AI — Your Offline AI Companion for the Real World, an open-source, privacy-first outdoor exploration platform designed to help people spend less time staring at screens and more time connecting with nature.

Most digital applications are designed to maximize engagement and keep users online. TerraQuest AI takes the opposite approach: it uses AI to help people explore the real world, learn about biodiversity, and put their phones away.

The application combines offline-first functionality, local AI models, nature exploration tools, and gamified outdoor challenges to make stepping outside a more engaging and educational experience.

🌱 Key Features

Pocket Mode: Choose an outdoor quest, review your checklist, and switch to a minimal screen that encourages you to lock your phone and explore.
AI-Powered Nature Identification: Explore plants, birds, trees, and insects using guided morphological questions, image analysis, and optional local vision models.
Smart Adventure Planner: Plan outdoor activities lasting 15–60 minutes, with nature-focused checkpoints and accessibility considerations.
Touch Grass Challenges: Complete screen-free activities, observe wildlife, discover leaf structures, and earn nature-themed XP and badges.
AI Garden Planner: Get seasonal gardening guidance based on sunlight, soil type, and growing conditions.
Private Nature Journal: Save observations, photographs, notes, and optional location information directly in your browser.
Birdsong Analysis: Explore audio waveforms through the browser's Web Audio API.
Offline Field Guide: Continue using local taxonomy information and deterministic identification heuristics when internet access or a local AI model is unavailable.
Data Portability: Export and import journal data in JSON and CSV formats.

TerraQuest AI is designed around a simple philosophy: technology should help us reconnect with the world beyond our screens.

Demo

🌐 Live Application: https://terraquest-ai-murex.vercel.app

Explore the application and try Pocket Mode, the nature identification workspace, outdoor challenges, and the private field journal.

For a meaningful demonstration, start by selecting a quest, activate Pocket Mode, and experience how the application encourages you to step outside instead of continuing to browse.

Code

💻 GitHub Repository: https://github.com/GenX0Gravity/TerraQuestAI

The project is built with an open-source-first architecture, and contributions are welcome. Developers can explore the code, experiment with local AI models, improve nature identification, or add new outdoor activities.

How I Built It

I built TerraQuest AI using a modern web stack and a local-first AI architecture.

Technology stack:

Next.js 15 and React 19 — Application framework and user interface.
TypeScript and Tailwind CSS — Type-safe development and responsive styling.
Ollama — Optional local inference for open-weight language and vision models.
Qwen 2.5 and Gemma 2 — Open-weight language-model options for naturalist reasoning and assistance.
Moondream and LLaVA — Optional local vision models for image-based exploration.
Dexie.js and IndexedDB — Browser-based storage for observations, journal entries, and progress.
OpenStreetMap and Leaflet — Outdoor maps and nature-focused exploration.
Web Audio API — Audio waveform analysis for exploring birdsong.
Vitest — Unit and integration testing.
PyTorch, Torchvision, and ONNX — Supporting tools for the optional species-classifier training pipeline.

One of the central design decisions was to avoid making a cloud-based AI service mandatory. When Ollama is unavailable, TerraQuest AI can fall back to a deterministic local field-guide engine for supported features.

This makes the application more resilient in places where connectivity is unreliable and helps keep personal observations on the user's device.

The project also includes a machine-learning training pipeline for experimenting with datasets such as Oxford Flowers 102, Birds 525, and PlantVillage.

Why Does Open Innovation Matter?

Open innovation matters because AI-powered outdoor tools should not require users to surrender their personal data, depend on a single commercial provider, or maintain a constant internet connection.

TerraQuest AI demonstrates several advantages of an open-source-first approach:

1. Privacy by design

Nature observations, personal notes, photographs, and progress are stored locally in the browser using IndexedDB. The application is designed without mandatory accounts or cloud-based journal storage.

2. Freedom to choose AI models

By supporting local models through Ollama, the architecture gives users more flexibility to experiment with different open-weight language and vision models instead of relying exclusively on one proprietary API.

3. Accessibility beyond reliable internet

The local field-guide engine and offline-capable features help users continue exploring even when connectivity is limited. Some map content and local model capabilities have separate requirements, so not every feature operates identically in every offline environment.

4. Lower barriers to experimentation

Developers can inspect the code, modify the application, experiment with datasets, train classifiers, and contribute improvements without building the entire system around a paid AI service.

5. AI that encourages real-world experiences

Perhaps the most important part is the project's purpose. Rather than optimizing for endless scrolling, TerraQuest AI encourages users to observe birds, learn about plants, explore outdoor spaces, and disconnect from their devices.

Open innovation makes this approach easier to inspect, adapt, and improve collaboratively.

My Agent Session

I used AI-assisted development to help shape the project, refine its architecture, and organize its features into a cohesive application.

The development process focused on turning the challenge's central idea — Touch Grass — into practical functionality: Pocket Mode, offline-first nature exploration, local AI integration, outdoor challenges, and a private field journal.

I also focused on keeping the project maintainable through a clear README, documented setup instructions, testing guidance, and contribution guidelines.

The goal was not simply to add AI to another application. It was to explore how open AI models and thoughtful product design could encourage people to spend more time in the physical world.

Prize Categories

TerraQuest AI is especially relevant to the following themes:

Open-Source AI: The architecture supports local inference with open-weight models and a deterministic offline fallback.
Offline and Privacy-First Innovation: Core nature exploration and journal functionality are designed to work locally in the browser.
Touch Grass: Pocket Mode and screen-free outdoor challenges directly encourage people to step away from their devices.
Community Collaboration: The project welcomes contributions to its code, tests, nature database, and machine-learning pipeline.