MiroFish, an open-source swarm intelligence prediction engine incubated by China's Shanda Group, has surged on GitHub with over 500 stars in its first day of trending. The system takes a novel approach to simulation and prediction by combining multi-agent AI with what the developers call a digital world sandbox.
The workflow begins with seed extraction: users upload data analysis reports or narrative documents, and the system constructs a high-fidelity parallel digital world using GraphRAG (graph-based retrieval-augmented generation). Within this world, thousands of AI agents — each with independent personalities, long-term memory (powered by Zep Cloud), and behavioral logic — freely interact and undergo social evolution.
Users can inject variables from a God's-eye view to observe how changes propagate through the agent society, then generate prediction reports via a dedicated ReportAgent. The system supports dynamic interaction: users can chat with any agent in the simulated world or with the ReportAgent to explore scenarios.
MiroFish's simulation engine is built on OASIS (Open Agent Social Interaction Simulations) from the CAMEL-AI team. The backend uses Python with FastAPI; the frontend is Node.js with React. It supports deployment via Docker or source code.
The project positions itself at two levels: as a rehearsal laboratory for decision-makers to test policies and public relations at zero risk, and as a creative sandbox for individual users exploring imaginative scenarios.
