GeoSim Lab: A Virtual Geography Laboratory with Local AI

A developer released GeoSim Lab, an open-source virtual geography laboratory that pairs a deterministic rainfall-runoff simulation and a reproducible synthetic terrain explorer with GeoTutor, a geography assistant that routes experiment context through an Express backend to a configurable Ollama model. The tool runs inference against a local Ollama endpoint by default, letting learners use open-weight models without a paid hosted AI API, while the numerical simulation stays deterministic and independent of the AI service. The developer notes the runoff model is an educational simplification, not a flood forecast or engineering tool, and that terrain data is synthetic.

GeoSim Lab https://github.com/Habeebullahi0/GeoSim-Lab is an open-source virtual laboratory for Geography students, teachers, and independent learners. It combines a rainfall-and-runoff experiment, a reproducible synthetic terrain explorer, and GeoTutor, a geography assistant that sends experiment context through a backend to a configurable Ollama model. The goal is to make hands-on geographical learning more accessible without requiring physical lab equipment or a paid AI API. The rainfall laboratory uses a deterministic, documented water-balance model. Learners adjust rainfall, vegetation, soil infiltration, surface storage, and slope, then compare the calculated results with a baseline. The terrain explorer generates a reproducible synthetic surface; it is not real geographic data. GeoTutor receives the actual experiment inputs, calculated outputs, units, and model limitations through the Express backend. The application does not ask the model to calculate authoritative results. php flowchart LR Learner -- UI React and TypeScript interface UI -- Sim Shared deterministic simulation Sim -- UI UI -- API Express API with request validation API -- Ollama Configurable Ollama model Ollama -- API API -- UI Instead of sourcing equipment or manually repeating a simple classroom water-balance exercise, a learner can change one variable at a time, run the same scenario again, compare saved scenarios, and ask GeoTutor to help interpret the result. It is intended to support self-paced exploration; it does not replace fieldwork or a complete hydrological model. GeoTutor uses Ollama, with a local endpoint as the default, so a learner can use an open-weight model without sending experiment context to a paid hosted AI API. The model is configurable, while the numerical simulation remains deterministic and transparent. Ollama and each model have separate licensing terms; the repository's MIT licence does not cover model weights. Watch the GeoSim Lab demo on YouTube https://youtu.be/YnXT3A46HPE The repository includes genuine application screenshots and local setup instructions. GeoTutor requires Ollama and a compatible configured model for inference. React, TypeScript, Vite, Express, Zod, Recharts, SVG, Vitest, and Ollama. The runoff calculation is an educational simplification, not a flood forecast or engineering tool. Terrain is synthetic, not a measured digital elevation model. GeoTutor is optional; lab calculations work without an AI service. Model compatibility and response quality depend on the selected Ollama model and available hardware. No Week 3 partner prize category is claimed: the current project does not meaningfully use the listed partner technologies. AI disclosure: This post was prepared by an AI assistant using project information and the demo link supplied by the author.