FlyBoxLab
A real-time closed-loop sandbox around a simulated fruit-fly connectome: 166,700 neurons and ~25.6 million synapses, with live stimulation, silencing, deterministic lesions, state forking, checkpoints, structured experiments, and 3D activity inspection.
FLYBOX is a real-time, closed-loop sandbox built around the MaleCNS fruit-fly connectome graph used by FlyBrain. The graph contains 166,700 neurons and roughly 25.6 million synapses. The neuroscience data and core FlyBrain simulation are upstream work; this project builds the interactive infrastructure around them: a FastAPI + WebSocket backend, a React/Vite 3D frontend, controlled neural interventions, reproducible experiment tooling, Python and JavaScript SDKs, documentation, testing, and deployment. The interface explicitly distinguishes source connectome data and simulated neural dynamics from experimental sensory encoders, motor decoders, and game mechanics so experimental mappings are not presented as established biological mechanisms.
- Python
- FastAPI
- WebSockets
- React
- Vite
- FlyBrain