This is a submission for the Hacktoberfest Open-Source AI Challenge Week 1: Touch Grass
Most outdoor apps demand that you stare at a screen in the forest: following digital map lines, maintaining streaks, or logging stats. Groundfall is built on the inverse principle: make the screen the shortest part of the experience.
Groundfall is an offline field naturalist and sensory walk companion that actively encourages you to keep your phone in your pocket. Before you step outside, it crafts a sequence of tactile and acoustic micro-observations. Once you are out, it uses a unique "Eyes-Up" Screen-Time Auditor to measure how much time your screen was actually locked or hidden versus active. Every glance penalizes your Eyes-Up Naturalist Ratio, rewarding you for truly experiencing nature.
Check out Groundfall in action here:
You can view the complete source code and run it locally:
"Build something with open-source AI at its core that gets people off the screen and into the world."
Submission for Hacktoberfest Open-Source AI Challenge Week 1: Touch Grass (#hf26challenge, #devchallenge)
👉 https://aditya-prog-bit.github.io/groundfall/ Anyone can open and test Groundfall immediately on any device (phone or desktop) with zero installation, zero server required, and 100% on-device client privacy.
Most "outdoor" apps do the exact wrong thing: they ask you to look at a screen while you're standing in front of an ancient oak tree. They flash notifications, show turn-by-turn map lines, and track gamified streaks that keep your gaze glued to glass.
Groundfall is built on the inverse principle: Make the screen the shortest part of the experience.
Before you step outside, Gemma crafts a tailored sequence of tactile, acoustic, and botanical…
Groundfall relies entirely on local, on-device intelligence and native web APIs to function deep in the backcountry:
Google's Gemma (Open-Weights): Acts as the core sensory engine. Before departure, Gemma crafts custom tactile and acoustic micro-observations (e.g., noticing leaf chlorosis transitions or identifying high-frequency bird calls).
Backboard Memory: Used for the Field Notebook feature, seamlessly summarizing walk debriefs, awarding botanical specimen badges, and storing observations in a persistent .backboard/field_memories.json format.
Page Visibility API (document.visibilityState): Powers the "Eyes-Up" auditor to track and penalize active screen time.
Web Audio API & Real-Time FFT: Synthesizes zero-asset trail bells (587 Hz & 880 Hz) directly in memory to avoid network downloads. It also monitors environmental sound frequencies to compute an ambient Acoustic Green Index, filtering out mechanical hums.
FastAPI & Python: Bridges the local Ollama inference with the frontend PWA.
Open innovation is the only way a project like Groundfall can exist. Challenging the cloud-API-first orthodoxy, local inference proves that intelligent applications don't need subscriptions, credit cards, or continuous network connectivity.
For backcountry tools and privacy-first outdoor companions, zero-network open-weight models represent a genuine paradigm shift. By utilizing open-source tools, I was able to build an AI application that respects user privacy and actually works when you are miles away from the nearest cell tower, keeping walkers connected to nature rather than their screens. I am submitting Groundfall for the following prize categories:
Best Use of Gemma
Best Use of Backboard