This is a submission for the Hacktoberfest Open-Source AI Challenge Week 1: Touch Grass
What I Built
OutsideQuest AI is a Python-based application designed to turn screen time into real-world outdoor exploration.
The idea is simple: instead of spending more time scrolling through an app, users can generate a short outdoor mission based on their available time, preferred environment, and interests.
Missions can encourage users to observe plants, notice birds, explore colors and textures, listen to their surroundings, or take a mindful walk.
The goal is to make technology the starting point of an experience rather than the experience itself.
I designed the project around three principles:
- Less screen time: Generate a mission, then put the phone away.
- Open-source AI: Use an open-weight model as part of the actual mission-generation workflow.
- Privacy awareness: Minimize personal data collection and be transparent about whether inference runs locally or through a hosted provider.
OutsideQuest AI is intended for students, developers, and anyone who wants a simple reason to step away from their screens.
Demo
Live application: [INSERT YOUR DEPLOYED APPLICATION URL]
The demo shows the user selecting a mission duration and environment, generating an outdoor quest, and viewing the instructions before heading outside.
If you have not deployed the application yet, replace this section with a working video demonstration. Do not publish a fabricated demo link.
Code
GitHub repository: [https://github.com/abhra92/outsidequest-ai]
The repository contains the Python application, AI integration, mission validation, setup instructions, and tests.
The source code and installation instructions will be available for others to inspect, run, and extend.
How I Built It
I built OutsideQuest AI using Python and Streamlit, with a modular design that separates the user interface, mission-generation logic, validation, and local storage.
The planned technology stack includes:
- Python: Application logic and AI integration.
- Streamlit: User interface and mission presentation.
- Pydantic or equivalent validation: Validate the structure of generated missions.
- SQLite: Store completed missions locally.
- [INSERT THE ACTUAL OPEN-WEIGHT MODEL AND INFERENCE PROVIDER]: Generate personalized outdoor missions.
I used OpenCode as my AI-assisted development environment to help implement the application, organize the code, and develop tests.
The application's AI component is separate from the coding agent. The model used to write code is not automatically the model used to generate missions for users.
The application is designed to handle invalid model responses, API failures, and unsafe suggestions through validation and fallback missions.
Implementation status: [DESCRIBE WHAT ACTUALLY WORKS, WHAT YOU TESTED, AND WHAT REMAINS INCOMPLETE.]
Why Does Open Innovation Matter?
Open innovation matters because developers should have meaningful choices about how AI applications are built, deployed, and maintained.
For OutsideQuest AI, using an open-weight model creates the possibility of replacing the inference provider, adapting the mission-generation behavior, and inspecting the model's documented capabilities and licensing terms.
It also creates an opportunity to explore privacy-conscious alternatives to closed, proprietary AI services.
However, open weights alone do not guarantee privacy, zero cost, or complete openness. A hosted inference service may still process user prompts on an external server, and the model's license and availability must be checked individually.
My goal is to keep the AI component replaceable, minimize the information sent for inference, and make the application's limitations explicit.
The broader idea is that open AI should enable people to build tools that serve real-world needs, not simply produce more reasons to remain online.
My Agent Session
[OPTIONAL: INSERT YOUR DEVRELAY AGENT SESSION LINK OR EMBED IT USING THE REQUIRED DEV.TO AGENT_SESSION TAG.]
I used OpenCode during development to assist with implementation, debugging, and testing.
Prize Categories
[LIST ONLY THE PARTNER CATEGORIES THAT YOUR PROJECT ACTUALLY QUALIFIES FOR, BASED ON THE OFFICIAL RULES.]
Potential categories should be selected only after verifying the exact model, framework, and partner requirements. Using OpenCode as a development tool does not automatically qualify the finished product for an agent-framework prize category.
Top comments (0)