TrailFlora & Garden AI — Breaking Screen Fatigue with Offline Open-Weight Botanical Intelligence

# devchallenge# hf26challenge# touchgrass# python
TrailFlora & Garden AI — Breaking Screen Fatigue with Offline Open-Weight Botanical IntelligenceDien

This is a submission for the 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

Modern life traps developers and creators behind glowing screens for 10–14 hours a day. The prompt for Week 1 — "Touch Grass" — challenges us to build an open-source AI companion that makes the screen the shortest part of the user journey, nudging people outside into sunlight, woods, and backyard gardens.

I built TrailFlora & Garden AI: an offline-first botanical scout and autumn planting sentinel designed for trail hikers, backyard gardeners, and outdoor walkers.

Key Capabilities:

  1. Trail Flora & Toxic Weed Sentinel: Identifies wild species from simple descriptive observations (leaf shape, berry clusters, stem geometry) and alerts hikers against contact dermatitis hazards (Poison Ivy, Poison Oak) or lethal ingestions (Deadly Nightshade/Belladonna).
  2. Backyard Garden & Frost Predictor: Calculates zone-specific autumn planting schedules and warns when ambient evening temperatures threaten frost-sensitive crops.
  3. Touch Grass & Sunlight Scout: Evaluates real-time ambient temperature and cloud coverage to compute an "Outdoor Vitality Score", suggesting optimal trail walk windows that reset digital eye fatigue and circadian rhythm.

Demo

You can also run it locally on zero pip dependencies:

git clone https://github.com/dieennn/hacktoberfest-2026.git
cd hacktoberfest-2026
python main.py
# Open http://localhost:8080/week-1/
Enter fullscreen mode Exit fullscreen mode

Code

The full project is open-source and part of my unified Hacktoberfest 2026 monorepo:

GitHub logo dieennn / hacktoberfest-2026

🎃 Hacktoberfest 2026 — DEV Challenges Monorepo

Repository containing project submissions for the Hacktoberfest 2026 DEV Challenges ("AI belongs to everyone").
Built by @dieennn.

📁 Repository Structure










































Round Challenge Theme Folder / Project Status

Launch Weekend (Oct 2 - Oct 5)		 Build for a Friend 00-weekend-allergy-guard ✅ Completed

Week 1 (Oct 5 - Oct 12)		 Announced Oct 5 01-week-1 ⏳ Upcoming

Week 2 (Oct 12 - Oct 19)		 Announced Oct 12 02-week-2 ⏳ Upcoming

Week 3 (Oct 19 - Oct 26)		 Announced Oct 19 03-week-3 ⏳ Upcoming

Week 4 (Oct 26 - Nov 1)		 Announced Oct 26 04-week-4 ⏳ Upcoming

🚀 Projects Overview

00-weekend-allergy-guard — Allergy & Diet Guard (SafeBite AI)

  • Theme: Build for a Friend
  • Problem: Protect friend Sarah (severe peanut allergy, celiac disease, and lactose intolerance) from hidden allergens (arachis oil, seitan, casein) and cross-contamination when eating out.
  • Tech: Local…
View on GitHub

Directory: 01-week-1/

How I Built It

1. Zero External Dependencies (Standard Library Architecture)

Out on a wilderness hiking trail, you have zero cellular data and cannot run pip install on bloated 500 MB cloud SDKs.
TrailFlora is built using pure Python standard library (http.server, urllib, json, re). It loads in milliseconds on any battery-constrained laptop or offline field device.

2. Google Gemma Open-Weight AI Bridge

TrailFlora interfaces with Google's Gemma open-weight models (gemma2:2b / gemma:2b) running via local Ollama inference (http://localhost:11434/api/generate). When hikers encounter ambiguous foliage descriptions, Gemma generates concise 2-sentence outdoor ranger field advice.

3. Resilient Offline Graceful Degradation

If the user is deep in the backcountry without an active local LLM instance, the engine automatically falls back to a deterministic botanical rule matrix and hardiness database (PLANT_TAXONOMY and GARDEN_SCHEDULES). The application never crashes or leaves the user stranded on the trail.

Why Does Open Innovation Matter?

  1. True Off-Grid Capability: Proprietary cloud AI endpoints (e.g., OpenAI or Claude APIs) are completely useless when you lose signal 4 miles into a forest trail. Open-weight models (like Google Gemma) belong to the user, run on local hardware, and never demand an active Wi-Fi connection.
  2. Zero Recurring Infrastructure Costs: By pairing an open-weight model with a lightweight Python standard library gateway, anyone can self-host this application for free on Render's free tier.
  3. Data Privacy in the Wild: Your personal trail coordinates, garden habits, and schedule notes stay strictly on your local device.

Prize Categories

  • Hacktoberfest Open-Source AI Challenge: Week 1 (Theme: Touch Grass)
  • Best Use of Render: Deployed as part of a unified multi-application monorepo web gateway running seamlessly on Render's cloud platform.
  • Best Use of Gemma: Integrated Google's open-weight Gemma model to deliver trail ranger field guidance and botanical safety insights.