MiroFish: An Open-Source Multi-Agent Prediction Engine for Simulating the Future

Forum topic · ✨步子哥 · 2025-12-23

Summary

MiroFish is an open-source, general-purpose swarm intelligence engine built on multi-agent technology, positioned as a next-generation AI prediction engine. It extracts seed information from the real world to automatically construct a high-fidelity parallel digital world in which thousands of agents with independent personalities, memories, and reasoning logic interact, allowing users to rehearse possible futures. Core features include knowledge graph construction (reality seed extraction, memory injection, GraphRAG), environment setup (entity-relation extraction, persona generation, parameter injection), dual-platform parallel social evolution simulation with dynamic temporal memory updates, report generation via a ReportAgent, and deep interaction through conversations with any agent. Application scenarios span macro decision-making (zero-risk digital sandbox testing for policy and PR) and micro creativity (individual users exploring story endings and ideas). The workflow is simple: upload a seed, describe the requirement, and get a report. The project uses a Python backend, Vue frontend, and the OASIS engine, and is available on GitHub at 666ghj/MiroFish.

MiroFish: A Simple, General-Purpose Swarm Intelligence Engine — "Predict Everything"

MiroFish is an open-source project introduced on zhichai.net. It is a new-generation AI prediction engine based on multi-agent (swarm intelligence) technology.

What It Does

MiroFish extracts seed information from the real world and automatically builds a high-fidelity parallel digital world. Thousands of agents — each with independent personality, memory, and logic — interact inside this world, helping users rehearse possible futures.

Core Features

Graph construction: Reality seed extraction, memory injection, and GraphRAG building.
Environment setup: Entity-relation extraction, persona generation, and parameter injection.
Social evolution simulation (highlight): Dual-platform parallel simulation with dynamically updated temporal memory for accurate extrapolation of future trends.
Report generation: A ReportAgent deeply interacts with the environment to produce detailed reports.
Deep interaction: Chat with any agent to explore details of the digital world.

Application Scenarios

Macro decision-making: A rehearsal lab for decision-makers — test policies and PR campaigns in a zero-risk digital sandbox.
Micro creativity: A creative sandbox for individuals — simulate novel endings and explore unlimited ideas.

Workflow

1. Upload a seed 2. Describe your requirement 3. Get a report

Tech Stack & Links

Python backend
Vue frontend
OASIS engine
GitHub: https://github.com/666ghj/MiroFish

*Note: The original post is a promotional poster page (HTML/CSS); the details above are translated from its content.*