DEV Community: RAKSHIT KHANNA

TerraQuest AI: Your AI Outdoor Adventure Companion 🌿

RAKSHIT KHANNA — Sat, 10 Oct 2026 04:35:38 +0000

Hacktoberfest Open-Source AI Challenge — Week 1: Touch Grass

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

🌱 What I Built

TerraQuest AI — Turn Your Free Time Into an Outdoor Adventure

TerraQuest AI is an AI-powered outdoor companion designed to help people disconnect from their screens and reconnect with nature.

Instead of spending free time scrolling through social media, users can generate personalized outdoor adventures based on their available time, preferred activities, interests, and exploration style.

Whether someone wants a peaceful nature walk, an energetic hike, a wildlife discovery mission, or a photography adventure, TerraQuest AI helps them find an activity and gives them a structured plan to get started.

The application combines AI-generated outdoor missions with semantic search to recommend relevant recreational routes from California State Parks data.

✨ Key Features

Personalized Outdoor Missions: Generate activities for walking, hiking, running, cycling, nature exploration, and relaxation.
Interest-Based Adventures: Choose from nature, fitness, wildlife, photography, and adventure.
Flexible Time Planning: Get an outdoor mission that fits your available free time.
AI-Generated Activity Plans: Receive a warm-up, main activity, and cooldown tailored to your preferences.
Semantic Trail Discovery: Find relevant route records using text embeddings and vector similarity instead of relying only on exact keyword matches.
Location-Aware Search: Optionally use manually entered coordinates to support location-based route discovery.
Digital Detox Philosophy: Get your mission, put your phone away, and enjoy the real world.

TerraQuest AI is built for anyone who wants to make their free time more meaningful by exploring the world beyond their screen.

🎥 Demo

The project includes screenshots of the working Streamlit application.

1. TerraQuest AI Interface

View the application interface

2. Personalized Outdoor Missions

View a personalized mission

3. Trail Recommendations

View trail recommendations

The application currently runs locally and has not yet been publicly deployed.

To install and run it, follow the setup instructions in the project README.

💻 Source Code

🌿 GitHub Repository: TerraQuest AI — Source Code

📸 Project Screenshots: View the screenshots

The repository contains the Streamlit application, AI mission-generation logic, semantic retrieval components, database integration, and installation instructions.

Note: The repository and screenshot links currently point to the original project. Update the repository name, screenshots, and documentation if you create a separate TerraQuest AI version.

🛠️ How I Built It

TerraQuest AI combines open-weight language models, semantic search, vector databases, and an interactive Python interface.

Technology Stack

Python: Application logic and time allocation.
Streamlit: Interactive web interface.
Ollama + Gemma 3 4B: Local AI-powered mission generation.
Sentence Transformers: Text embedding generation.
BAAI/bge-base-en-v1.5: Semantic search embeddings.
Tiger Cloud PostgreSQL: Storage for route information and vector embeddings.
pgvector: Similarity-based retrieval of relevant routes.
psycopg and python-dotenv: Database connectivity and environment configuration.

⚙️ How It Works

User Preferences: The user chooses an activity, interests, mission style, and available time.
Search Query Creation: The application builds a semantic query from the selected preferences.
Embedding Generation: The BGE model converts the query into a numerical vector representation.
Route Retrieval: PostgreSQL and pgvector search for relevant route records using vector similarity.
AI Mission Generation: Gemma 3 4B generates activity instructions organized into a warm-up, main activity, and cooldown.
Time Allocation: Python manages the activity's time distribution and assembles the final mission.
Mission Display: Streamlit presents the generated plan and relevant route recommendations.

I also included fallback instructions so that the application can still provide a useful outdoor activity plan when AI generation fails or its response cannot be parsed correctly.

The route dataset comes from California State Parks recreational routes. These records represent route segments rather than necessarily complete trails, so individual segment lengths should not be interpreted as the total distance of an entire trail.

🌍 Why Open Innovation Matters

Open innovation makes it possible to experiment with AI systems, understand their underlying components, and build applications without depending entirely on proprietary hosted AI services.

With Gemma 3 4B running through Ollama, TerraQuest AI can generate outdoor missions locally. This makes experimentation with prompts and mission-generation logic more accessible.

The BGE embedding model and pgvector provide another important capability: finding relevant recreational routes based on semantic similarity rather than relying exclusively on exact keyword matches.

Together, these technologies demonstrate how open models, accessible infrastructure, and open-source tools can be combined to build a practical AI-powered application.

TerraQuest AI is not completely offline because route retrieval depends on its configured PostgreSQL database service. Its recommendations are also limited by the available California route dataset. Users should verify trail accessibility, weather, current conditions, and safety information before heading outdoors.

🚀 What Makes TerraQuest AI Different?

Many AI applications encourage users to spend more time interacting with technology. TerraQuest AI takes a different approach.

It uses AI to help people plan an activity, discover a place to explore, and then step away from their devices.

The technology is meant to support the outdoor experience—not replace it.

The long-term vision is to make discovering outdoor activities easier, encourage people to explore their surroundings, and turn small amounts of free time into memorable real-world experiences.

🤖 My Agent Session

I have not recorded a DevRelay agent session for this project yet.

🏆 Prize Categories

Best Use of Gemma: Gemma 3 4B generates personalized outdoor mission instructions locally through Ollama.
Best Use of Tiger Data: Tiger Cloud PostgreSQL and pgvector power semantic retrieval across route records and their embeddings.

🌿 The Final Goal

Technology should not always demand more of our attention. Sometimes, it should help us put our phones away.

TerraQuest AI is my attempt to use open-source AI to encourage more outdoor exploration, healthier digital habits, and meaningful experiences beyond the screen.

Find your adventure. Step outside. Explore more. Touch grass. 🌱

I Built SafePlate for a Friend, Making Everyday Meal Safety Simpler

RAKSHIT KHANNA — Sun, 04 Oct 2026 23:12:30 +0000

This is a submission for the Hacktoberfest Weekend Challenge: Build for a Friend

What I Built

SafePlate is a food-safety and meal-planning app that helps households manage meals when different people have different food allergies and safety requirements.

I built it for a friend who wanted a simpler way to plan meals without constantly checking ingredients and worrying about whether a meal was safe.

SafePlate lets a household:

Create multiple household profiles
Store individual allergens and their severity
Check ingredients for potential safety issues
Generate personalized weekly meal plans
Understand why a meal is considered safe
Swap or lock meals
Generate a grocery list from the weekly meal plan
Keep each household member's meal plan separate
Create an account and securely store household data

I also focused heavily on making the interface easy to use. Instead of asking users to fill out complicated forms everywhere, SafePlate uses cards, selectable options, chips, and clear safety indicators.

The goal was to build something that could actually be used by a real household rather than stopping at a hackathon prototype.

Demo

Live app: https://safe-plate-iota.vercel.app/
(Sign-In required)

Code

GitHub: https://github.com/RAKSHIT-0511/SafePlate

The project includes the React/TypeScript application, Supabase integration, safety engine, household profile management, meal planning, grocery-list functionality, and automated tests.

How I Built It

SafePlate uses React, TypeScript, and Vite for the web application, with Supabase providing authentication, PostgreSQL storage, and Row Level Security. Vercel is used for deployment.

I built the application through an AI-assisted development workflow using Antigravity. I used the agent to help implement features, restructure the application, debug issues, improve the user experience, and test the application as it evolved from a prototype into a real multi-user application.

One of the most important parts of the build was moving away from demo-only data and making the application work with real authenticated users.

The final architecture allows one account to have multiple household profiles, with each profile having its own allergens and meal plan:

Account
├── Rakshit
│ └── Allergens + Meal Plan
├── Mom
│ └── Allergens + Meal Plan
└── Dad
└── Allergens + Meal Plan

Each profile is stored independently in Supabase while remaining associated with the authenticated account.

The safety logic itself is deterministic. The application checks ingredients against the active profile's allergen information rather than relying on an AI model to make a safety decision.

I also added automated tests for the safety engine and multi-profile functionality. The final test suite contains 30 passing tests, with TypeScript compilation and the production build also passing successfully.

Why Does Open Innovation Matter?

Open innovation made it possible for me to combine different technologies and development tools to turn SafePlate from an idea into a working application.

Instead of depending on one closed platform, I was able to build different parts of the product using technologies that fit the problem: React and TypeScript for the interface, Supabase for authentication and persistent data, automated testing for reliability, and Vercel for deployment.

AI-assisted development also made it possible to iterate much faster. I could describe a problem, inspect the implementation, test it in the real application, and continue improving it.

One example was the multi-profile system. During development, an old database constraint from the original single-profile architecture prevented new household members from being created. The issue was identified while testing the real application and fixed without rebuilding the entire project.

For me, open innovation is about being able to combine tools, experiment quickly, understand how the pieces work together, and keep improving the product until it becomes useful to real people.

My Agent Session

I used Antigravity throughout the development of SafePlate as an AI-assisted coding agent.

The development process included implementing the application, improving the UX, integrating Supabase, building the multi-profile architecture, debugging runtime issues, writing tests, and preparing the application for production deployment.

I don't have a saved DevRelay session link yet, so I'm leaving the session link out rather than adding a placeholder.

Prize Categories

-Best Use of Render
-Best Use of DigitalOcean
-Best Use of Gemma
-Best Use of MongoDB Atlas
-Best Use of Temporal