shikharmishra1/gridmind

GridMind is a platform for analyzing and visualizing electrical power networks, aimed at power system engineers, operators, and researchers. It pairs an interactive graph based visualization tool in the browser with a Python backend that runs a range of power system calculations, and adds an AI agent on top that can answer plain language questions about the network being studied. The visualization side lets you explore networks of any size, from a handful of buses up to networks with tens of thousands of them, with different layout styles, zooming and filtering, and overlays that highlight things like structural weak points, rule violations, the shortest path between two points, and voltage levels compared to their limits. You can export a snapshot of the graph as an image. On the analysis side, GridMind can run power flow calculations, contingency analysis (checking what happens if one or two pieces of equipment fail), optimal power flow for cost related decisions, load shedding analysis, and topology checks such as finding weak points or redundant paths in the network. It can also take in live SCADA and PMU sensor data and use a graph neural network to predict network states quickly. The AI agent is built with LangChain and has access to more than sixty specialized tools, so you can ask it things like what happens if a particular bus goes down, which lines are overloaded, or what the cheapest way is to relieve congestion, and it responds with an explanation plus highlighted parts of the network graph. The project includes its own evaluation suite, using the RAGAS framework, to measure how accurately the agent picks the right tool and how well its answers are backed by real data. You can also upload an image of a single line diagram and GridMind will use computer vision to detect the components and rebuild the network model automatically. Saved networks are stored in MongoDB. To run it locally you need Python 3.13 or newer, MongoDB, and optionally Azure OpenAI credentials for the AI agent, then you install the Python and frontend dependencies and start both servers.