# Hacktoberfest Submission: WildGuide Local _This is a submission for the Hacktoberfest Open-Source AI Challenge Week 1: Touch Grass_ ## What I Built **WildGuide Local** is a privacy-first, offline AI companion designed to turn your smartphone or laptop into a field guide for the great outdoors. It helps users identify plants and animals and provides regional ecological advice without requiring an internet connection. The goal is to lower the barrier to nature exploration by providing instant, local knowledge, while explicitly encouraging users to "put the screen away and observe" after getting their answers. It's built for hikers, gardeners, and nature enthusiasts who want the power of AI without the tether of a data plan or the privacy concerns of cloud-based APIs. ## How I Built It WildGuide Local is built entirely on an open-source, local-first stack: * **Local Inference** : Powered by **Ollama** , allowing the app to run heavy models on consumer hardware. * **Vision Model** : Uses **llava** for image recognition and species identification. * **Reasoning Model** : Uses **Llama 3** for generating ecological advice and interpreting local data. * **RAG (Retrieval-Augmented Generation)** : Implemented using **LangChain** and **ChromaDB**. This allows users to ingest their own regional ecological data (frost dates, local laws, native species lists) into a local vector store, making the AI an expert on their specific backyard. * **Frontend** : A clean, tabbed interface built with **Streamlit**. ## Why Does Open Innovation Matter? Open innovation is the heartbeat of WildGuide Local. By using open-weight models and local inference, we achieve three things that a closed API couldn't: 1. **True Offline Capability** : In the deep woods, there is no 5G. A closed API is useless in the very environment where this tool is needed most. 2. **Absolute Privacy** : Nature exploration is personal. By keeping images and queries on-device, users don't have to trade their privacy for knowledge. 3. **Community Customization** : Because the RAG pipeline is open, the community can share "Regional Knowledge Packs" (JSON/Text files) that others can drop into their local instance to immediately make their guide an expert on a specific National Park or biome. ## Prize Categories * Local AI / Edge Computing * Environmental Impact / Nature * Open Source Contribution