Ygg666ghj/MiroFish
↗ GitHubA Simple and Universal Swarm Intelligence Engine, Predicting Anything. 简洁通用的群体智能引擎，预测万物
47,539
Stars
6,833
Forks
295
Watchers
197
Open Issues
Safety Rating A
No hardcoded secrets, malicious code patterns, dependency vulnerabilities, or prompt injection attempts were detected. API keys are externalized via a .env file pattern (users copy .env.example and fill in their own keys). The repository is a legitimate, well-documented open source simulation engine with institutional backing (Shanda Group) and clear academic/research intent.
ℹAI-assisted review, not a professional security audit.
AI Analysis
MiroFish is a multi-agent swarm intelligence simulation and prediction engine built in Python. It ingests real-world seed materials (news, policy documents, financial signals, or narrative fiction) to automatically construct a high-fidelity parallel digital world. Thousands of AI agents with distinct personas, long-term memory (via Zep Cloud), and behavioral logic interact and socially evolve within this simulation. Users can inject variables from a 'god's-eye view' to forecast outcomes, receiving detailed prediction reports and an interactive simulated world. The engine is powered by the OASIS simulation framework from CAMEL-AI and uses GraphRAG for knowledge graph construction. Use cases span public opinion analysis, financial forecasting, political event prediction, and creative fiction exploration.
Use Cases
- Public opinion and sentiment forecasting for news events
- Financial market trend prediction using seed signals
- Policy impact simulation and decision-support
- Social dynamics and group behavior modeling
- Creative fiction outcome exploration (e.g., predicting lost novel endings)
- Multi-agent social simulation research
Tags
Project Connections
skyclaw
→MiroFish provides swarm-intelligence simulation and prediction while skyclaw (Tem) provides persistent autonomous agent runtimes with swarm capabilities ('Many Tems'). Tem agents could act as orchestrators or observers within MiroFish simulations.
clawvault
→ClawVault's structured markdown-native memory and knowledge graph primitives could complement MiroFish's per-agent memory and GraphRAG layer, providing richer persistent context management for simulation agents.
OpenJarvis
→Both are multi-agent AI frameworks that orchestrate large numbers of LLM-backed agents. OpenJarvis focuses on on-device personal AI agents, while MiroFish focuses on large-scale social simulation and prediction; they serve overlapping but distinct use cases.
code-review-graph
→code-review-graph's knowledge graph and semantic search capabilities (built on MCP) could be used to enrich MiroFish's GraphRAG knowledge base construction step with structured code or document graphs.