This is a submission for the Hacktoberfest Open-Source AI Challenge Week 1: Touch Grass
What I Built
NatureQuest AI is a local-first AI companion designed to turn screen time into outdoor discovery.
We often walk through parks, gardens, and neighbourhoods without noticing the fascinating details around us. I wanted to build something that uses AI not to keep people staring at their screens, but to give them a reason to put their phones away and explore the real world.
NatureQuest AI helps users:
🌱 Generate personalized outdoor quests based on available time, location setting, difficulty, and interests.
🍃 Discover the little things through guided observation activities focused on plants, birds, natural patterns, and everyday surroundings.
📓 Keep a field journal to record observations and preserve discoveries locally in the browser.
🐦 Ask a local AI naturalist questions about nature, observations, habitats, and seasonal changes.
📴 Explore with privacy using a locally running AI model, including after the initial setup and model download are complete.
The idea is simple: let AI help plan the adventure, then let the real world take over.
I designed it for curious people who want to spend more intentional time outdoors, whether they have 10 minutes in a nearby park or an afternoon to explore a trail.
Demo
My planned demo flow:
Generate a personalized 15-minute nature quest.
Head outdoors and complete the activity.
Return to record an observation in the field journal.
Ask the local AI naturalist a question.
Demonstrate that the AI can work without an internet connection after setup.
I want the demo to show more than a working interface. It should show the app accomplishing its real purpose: getting someone outside.
Code
GitHub repository: NatureQuest-AI
The project uses a React frontend, a FastAPI backend, and Ollama to run an open-weight language model locally.
I'll include the setup instructions, configuration, and source code so others can run the project on their own machines and adapt it for their own outdoor adventures.
How I Built It
I built NatureQuest AI around local inference rather than treating AI as an optional feature.
The stack:
Qwen2.5 3B: An open-weight language model for generating outdoor quests and answering nature questions.
Ollama: Runs the model locally and provides the interface between the application and the model.
FastAPI: Handles API requests, validates user input, constructs prompts, and returns structured AI responses.
React + Vite: Powers the interactive quest builder, field journal, and naturalist interface.
Browser localStorage: Stores field journal entries on the user's device, within that browser profile.
The workflow is straightforward:
The user chooses an outdoor setting, duration, difficulty, and interests.
The frontend sends those preferences to the FastAPI backend.
The backend constructs a prompt with instructions for safe, screen-free outdoor activities.
Ollama sends that prompt to the locally running Qwen model.
The model generates a structured quest that the frontend presents as actionable steps.
The user can complete the activity, record observations, and ask follow-up nature questions.
I also added safety-focused instructions to discourage touching, collecting, eating, feeding, or disturbing wildlife. AI-generated nature advice can be wrong, so the application treats the model as an educational companion rather than an authoritative species-identification tool.
Why Does Open Innovation Matter?
For NatureQuest AI, open innovation is what makes a private, local-first experience possible.
1. Your observations can stay yours
A field journal might contain personal reflections, photographs in future versions, or details about places someone visits. By running the language model locally, the app does not need to send AI prompts to a third-party hosted AI API.
Journal entries are stored in the browser, while AI inference runs through Ollama on the user's machine.
2. Offline exploration becomes possible
Once the application dependencies and model weights have been downloaded, the core AI experience can run without an internet connection.
That matters when someone is exploring a trail, visiting a garden, or spending time somewhere with poor connectivity. The initial setup requires internet access, but the local inference workflow does not require a cloud AI service.
3. No per-request cloud AI bill
There is no per-request charge from a hosted AI API in this architecture. Users still need suitable hardware and may incur electricity or other setup costs, but they can experiment without depending on a paid inference endpoint.
4. The model is replaceable
Using Ollama gives the project a practical path to experimenting with different compatible models. I can change the configured model, adjust prompts, and explore local vision-language models for future photo-based nature questions without redesigning the entire application.
5. AI becomes a tool for the real world
Many AI experiences are designed to keep users engaged with a screen. I wanted to explore the opposite direction: AI should help people begin an activity, then get out of the way.
Open models make that approach practical because the application can run on the user's own computer instead of depending on a permanently connected cloud service.
Final Thoughts
NatureQuest AI started with a question: What if AI helped us pay more attention to the world instead of demanding more of our attention?
The goal isn't to replace outdoor experiences with AI-generated content. It's to use AI to make getting outside easier, more curious, and more rewarding.
Generate a quest. Put the phone away. Notice something new.
🌿
Top comments (0)