Python infrastructure that turns enterprise data into a graph-based context layer with ontologies, reasoning, provenance and decision audit trails for AI systems.
14.1K1.7KPythonMITUpdated 19 hours ago
666ghj/MiroFish
A multi-agent swarm simulation engine that builds a digital world from seed material to predict outcomes of news, policy, or financial events.
- Stars
- 80,574
- Forks
- 12,228
- Open issues
- 156
- Last commit
- 4 days ago
What is it?
MiroFish is a prediction engine that uses multi-agent simulation. You upload seed material such as news articles, policy drafts, financial signals, or even a novel, and describe what you want predicted in natural language. The system extracts entities and relationships to build a knowledge graph, generates agents with distinct personalities and memory, and runs a dual-platform simulation where they interact over many rounds.
During a run, users can inject new variables into the simulated environment to test how different conditions change the trajectory. Once the simulation finishes, a ReportAgent produces a prediction report and lets you chat with individual agents or the report agent to explore the results.
The backend is Python and the frontend runs on Node.js. It is built on the OASIS agent social simulation framework from CAMEL-AI and uses Zep Cloud for memory, with any OpenAI-compatible LLM API supported for model calls. Docker Compose deployment is also provided.
Key features
- 01Builds a knowledge graph and agent personas from uploaded seed material.
- 02Runs multi-round simulations where agents with memory interact on two platforms.
- 03Allows injecting variables into the running simulation to test different scenarios.
- 04Generates a prediction report via a ReportAgent that can be queried interactively.
- 05Lets users chat directly with individual simulated agents.
- 06Supports any OpenAI-compatible LLM API and offers source or Docker deployment.
Use cases
- Rehearsing how public opinion might evolve around a news event or policy before releasing it.
- Exploring alternative plot endings or scenarios for a novel or story.
- Running what-if simulations on financial or market signals.
- Studying how individual agent interactions produce collective outcomes in social systems.
Quick start
- 1
Copy the example environment file
cp .env.example .env - 2
Fill in required API keys in .env (LLM and Zep)
- 3
Install all dependencies
npm run setup:all - 4
Start frontend and backend (frontend at localhost:3000, backend at localhost:5001)
npm run dev
Things to know
- Requires an LLM API key and a Zep Cloud API key, and simulations can be expensive in tokens.
- Requires Python 3.11 to 3.12 and Node.js 18+ plus the uv package manager for source deployment.
- Licensed under AGPL-3.0, which requires sharing source code of modified versions offered over a network.
Topics
FAQ
- What is MiroFish?
- A multi-agent swarm simulation engine that builds a digital world from seed material to predict outcomes of news, policy, or financial events. MiroFish is a prediction engine that uses multi-agent simulation. You upload seed material such as news articles, policy drafts, financial signals, or even a novel, and describe what you want predicted in natural language. The system extracts entities and relationships to build a knowledge graph, generates agents with distinct personalities and memory, and runs a dual-platform simulation where they interact over many rounds.
- Is MiroFish free to use? What license does it have?
- MiroFish is open source under the AGPL-3.0 license. Check the license terms before using it in commercial products.
- Is MiroFish still maintained?
- Actively maintained. The last commit to MiroFish was on Oct 16, 2026.
- What can I use MiroFish for?
- Rehearsing how public opinion might evolve around a news event or policy before releasing it. Exploring alternative plot endings or scenarios for a novel or story. Running what-if simulations on financial or market signals. Studying how individual agent interactions produce collective outcomes in social systems.
- How do I install MiroFish?
- Copy the example environment file: cp .env.example .env Fill in required API keys in .env (LLM and Zep) Install all dependencies: npm run setup:all Start frontend and backend (frontend at localhost:3000, backend at localhost:5001): npm run dev
- Where can I get MiroFish?
- The source code and installation instructions are on GitHub: https://github.com/666ghj/MiroFish
Summary generated by AI from the project's README. GitHub stats refresh daily.
Related projects
A Python library that constrains LLM outputs during generation so they match a given type, regex, JSON schema, or grammar.
16K907PythonApache-2.0Updated 28 days ago
A self-hosted Rust knowledge base that builds a bitemporal knowledge graph from documents, with RAG search, ontology-driven reasoning and MCP access for agents.
8.4K1.2KRustApache-2.0Updated 22 hours ago
Turns unstructured text into an interactive knowledge graph by using an LLM to extract subject-predicate-object triples.
3.1K434PythonApache-2.0Updated last month
A local knowledge graph tool where a locally run model reasons step by step and the reasoning is stored as an RDF graph, all on your own machine via Ollama.
58451PythonOtherUpdated 2 months ago
A learning repository of 30 runnable Jupyter notebooks covering agent memory techniques for LLM agents, from conversation buffers to Mem0, Letta, Zep and Graphiti.
1.1K142Jupyter NotebookApache-2.0Updated 3 days ago