ForumEngine

Multi-Agent Simulation Engine for Operational Decisions

Run the scenario before it runs you.

ForumEngine is a multi-agent simulation engine designed for mission planning and operational decision-making. Feed any briefing document — policy drafts, incident reports, strategic proposals — and ForumEngine will automatically spin up a high-fidelity simulation environment where autonomous agents with independent personalities, long-term memory, and behavioral logic interact freely.

Observe how stakeholders react, identify emerging patterns, and stress-test decisions before committing to them in the real world.

Input: Upload mission documents (analysis reports, briefing materials) and describe what you want to simulate in natural language.

Output: A detailed prediction report and a fully interactive simulation world you can explore.

ForumEngine runs a 5-step pipeline:

1. Graph Building — Extracts entities and relationships from seed documents, injects individual and group memory into a knowledge graph (GraphRAG)
2. Environment Setup — Generates agent profiles, extracts entity relationships, and configures simulation parameters
3. Simulation — Runs parallel multi-agent simulation with automatic scenario parsing and dynamic memory updates
4. Report Generation — A Report Agent with specialized tools analyzes the simulation results and produces a comprehensive report
5. Deep Interaction — Chat with any agent in the simulated world or query the Report Agent for deeper analysis

cp .env.example .env

Edit .env with your API keys:

# LLM API (any OpenAI SDK-compatible endpoint)
LLM_API_KEY=your_api_key
LLM_BASE_URL=https://your-llm-provider/v1
LLM_MODEL_NAME=your-model

# Zep Cloud (knowledge graph memory)
# Free tier is sufficient for basic usage: https://app.getzep.com/
ZEP_API_KEY=your_zep_api_key

# Optional: Boost LLM (faster model for parallel tasks)
# Only add these if you want to use a separate accelerated model
# LLM_BOOST_API_KEY=your_api_key
# LLM_BOOST_BASE_URL=your_base_url
# LLM_BOOST_MODEL_NAME=your_model_name

# Install everything (root + frontend + backend)
npm run setup:all

Or step by step:

# Node dependencies (root + frontend)
npm run setup

# Python dependencies (backend, auto-creates virtualenv)
npm run setup:backend

# Start both frontend and backend
npm run dev

Start individually:

npm run frontend   # Frontend only
npm run backend    # Backend only

cp .env.example .env
# Edit .env with your keys, then:
docker compose up -d

Exposes ports 3000 (frontend) and 5001 (backend). Reads .env from the project root.

ForumEngine's simulation engine is powered by OASIS.