GRDN.AI is an AI-enhanced companion planting application that brings together traditional gardening wisdom and state-of-the-art technology. Now live in beta on Hugging Face, GRDN.ai leverages advanced algorithms and a rich tech stack to optimize plant pairings for healthier, more synergistic gardens.
Overview
GRDN.ai is all about companion planting—using artificial intelligence to analyze plant compatibility, environmental factors, and user preferences.
The application uses a custom, graph-based, AI-enhanced constrained genetic algorithm that shapes initial garden configurations using its deep contextual knowledge of a comprehensive plant compatibility matrix.
Recent Updates &
Key Innovations
Custom Graph-Based Genetic Algorithm:
Developed specifically for companion planting, this constrained genetic algorithm uses AI to generate optimal starting configurations rather than relying on arbitrary assignments.Enhanced AI Workflow:
Incorporates cutting-edge models including the Llama2 7b chat model from Hugging Face and DeciLM 7b (from Deci AI, now part of NVIDIA) to boost performance and contextual understanding.Open Source Transition:
The shift to open source was powered by leveraging tools such as LlamaIndex and LlamaCPP, enabling broader collaboration and transparency.Streamlit Unbounded:
A new network graph visualization component (agraph-component) has been integrated to display plant relationships dynamically.Image Generation for Visualization:
A preprocessing script was developed using OpenAI’s DALL-E 3 to generate nearly 100 plant images, enhancing the network graph visualization with rich, custom imagery.Simplified User Interface:
The UI has evolved from a chat-centric design to a streamlined interaction model where the LLM primarily augments the genetic algorithm’s performance.
Demo
Check out the demo video below to see GRDN.AI in action. The video highlights key updates, including a custom graph-based genetic algorithm and integration of AI models like Llama2 7b and DeciLM 7b, all wrapped in a simplified interface. It’s a snapshot of my ongoing work to blend traditional gardening insights with modern AI techniques and novel mathematical optimization algorithms.