WildProof: Go Outside With a Question, Come Back With Evidence

WildProof is an innovative, offline-first field observation application designed to transform outdoor walks into structured scientific investigations. Built as part of the Hacktoberfest Open-Source AI Challenge, the project emphasizes privacy and transparency by separating direct observations from AI-inferred deductions. The application utilizes React and TypeScript on the frontend, with a backend powered by FastAPI, Temporal, and the Gemma 2B model via Ollama. By prioritizing an offline-first architecture, WildProof allows users to collect notes and photo evidence in remote locations without connectivity. The system is designed to be honest about AI limitations, explicitly labeling uncertain claims and providing fallback states when models are unavailable. By combining durable workflow orchestration with local open-weight AI, WildProof offers a privacy-conscious tool for citizen scientists and hikers, ensuring that data remains under user control while providing a clear, evidence-based report of their outdoor findings.