This is a submission for the Hacktoberfest Open-Source AI Challenge Week 1: Touch Grass
What I Built
I built SproutTime, an intelligent, data-driven gardening agent designed to maximize your time in nature and minimize digital screen interactions.
Instead of trapping users in infinite scroll feeds or complex setups, SproutTime dynamically tracks real-time meteorological conditions for any global city and pipes that context into a local open-weight model. It instantly generates a physical, printable three-tier markdown checklist tailored to the user's exact gardening experience level. It is built for absolute beginners, hobbyists, and experts who want actionable plant tasks so they can turn off their screens immediately and step into the dirt.
Demo
Interact with my live deployed application here:
👉 https://huggingface.co/spaces/laya-03/garden-planner
Code
You can view the source files, UI layout configurations, and dependency definitions directly on the Hugging Face Space repository:
👉 https://huggingface.co/spaces/laya-03/garden-planner/tree/main
How I Built It
SproutTime relies entirely on an open-source, local-first architecture:
-
Open Data Orchestration (RAG): The application uses
geopy(connected to OpenStreetMap) to securely convert city strings into global coordinates. It then calls the open-sourceOpen-Meteo APIto grab real-time climate telemetry (Temperature, Humidity, Rain) without encountering closed data firewalls. -
Local Inference Pipeline: Powered by the open-weight Qwen-2.5-0.5B-Instruct model running completely contained inside the container space via a Hugging Face
transformershardware pipeline (torch+accelerate). -
Deterministic Few-Shot Anchor Guardrails: Because lightweight 0.5B models are highly sensitive, I built a rigid conversational structural template block using specialized tokenizer system markers (
<|im_start|>and<|im_end|>). This forces the pipeline to deliver pure agricultural value and eliminates structural text clutter. - Lightweight UI: Built natively with the Gradio framework for smooth user interaction.
Why Does Open Innovation Matter?
My Agent Session
Open innovation was the absolute foundation of this project. A closed-source API ecosystem would introduce high request latency, cost walls, and unpredictable pricing matrices for casual outdoor hobbyists.
By leveraging open-weight structures like Qwen-2.5, I was able to download, optimize, and deploy an entire text generation pipeline directly inside a self-contained container memory environment. This architecture allows developers to build systems that work without relying on third-party cloud data brokers, perfectly aligning with a private, offline "Touch Grass" philosophy.
Prize Categories
- Main Challenge Category: Week 1 Challenge - Touch Grass
Top comments (0)