Summary

MiroFish is a novel AI prediction engine that utilizes multi-agent swarm intelligence to construct high-fidelity digital simulations of reality. It allows users to input real-world data and natural language queries to predict future outcomes, serving as a "digital sandbox" for both serious decision-making and creative exploration. The engine aims to predict "anything" by simulating emergent group behavior from individual agent interactions.

How It Works

The engine extracts "seed information" from real-world data (news, policy, financial signals) to construct parallel digital worlds populated by thousands of agents with independent personalities, long-term memories, and behavioral logic. Using a GraphRAG approach combined with entity extraction and persona generation, it simulates agent interactions and societal evolution within a configured environment. Users can inject variables from a "God's eye view" to dynamically steer the simulation and observe emergent outcomes. A dedicated ReportAgent, equipped with a rich toolset, analyzes the simulation results to generate detailed predictions and facilitates deep interaction with the simulated world.

Quick Start & Requirements

• Installation: Requires Node.js (18+), Python (3.11+), and the uv package manager. Dependencies are installed via npm run setup:all or separate npm run setup (Node) and npm run setup:backend (Python) commands. Services are started with npm run dev.
• Prerequisites: API keys for an LLM (OpenAI SDK compatible, e.g., Aliyun qwen-plus) and Zep Cloud are mandatory. Environment variables must be configured in a .env file.
• Environment: Developed and tested on macOS; Windows compatibility is explicitly stated as unknown and currently under testing.
• Links: Frontend runs at http://localhost:3000, Backend API at http://localhost:5001.

Highlighted Details

• The core simulation engine is powered by OASIS, a high-performance social media simulation framework developed by the CAMEL-AI team, designed to support millions of agent interactions for emergent behavior.
• Demonstrates predictive capabilities through video examples, including simulating the ending of "Dream of the Red Chamber" and financial market predictions.
• Offers deep interactivity, allowing users to converse with individual simulated agents and the ReportAgent for deeper analysis.

Maintenance & Community

MiroFish receives strategic support and incubation from Shanda Group. The project relies heavily on the OASIS framework developed by the CAMEL-AI team. No direct community links (e.g., Discord, Slack) or roadmap are provided in the README.

Licensing & Compatibility

The project is licensed under the MIT license. This license generally permits commercial use and integration into closed-source projects, but users should verify specific terms.

Limitations & Caveats

Windows compatibility is explicitly stated as unknown and currently under testing, potentially posing an adoption blocker for non-macOS users. The project is at version v0.1.0, indicating early-stage development and potential instability or missing features.