optX

High-grade strategy, without the consulting bill.

optX is an AI-powered business simulator that turns raw company data into structured, auditable scenario intelligence. Ingest your financials, build scenarios on a node-graph canvas, and watch six specialized AI agents debate, simulate, and converge on ranked recommendations — all without a six-figure consulting engagement.

---

Demo Flow

1. Ingest — Upload financials (balance sheet, income statement, cash flow) and layer in contextual signals: interest rates, market sentiment, workforce data, supply chain, and more
2. Simulate — The math pipeline constructs a unified variable universe, runs 10,000+ Monte Carlo iterations with copula-preserved correlations, and builds a Bayesian causal DAG
3. Explore — Drop into the n8n-style node graph to wire up custom scenario logic, connect nodes, and watch the canvas update live
4. Debate — Six AI agents (Market, Financial, Growth, Risk, Brand, Operations) analyze the simulation output, challenge each other, and converge on a ranked set of strategic recommendations
5. Report — Fan charts, tornado charts, waterfall P&L breakdowns, and a calibrated Accuracy Report you can actually trust

---

Architecture

Layer	Stack	Role
Frontend	Next.js · React · React Flow · Recharts · Tailwind	Dark-mode node canvas, data ingestion UI, real-time charts
State	Zustand	Graph state, simulation runs, agent outputs
AI	OpenAI (gpt-4o / gpt-4o-mini)	Agents, scenario generation, AI chat, custom data extraction
Backend	FastAPI · Python	Math pipeline — Monte Carlo, Bayesian networks, sensitivity, backtesting
Database	Supabase (Postgres)	User auth, persisted scenarios, simulation results

flowchart TD
    subgraph Frontend["Frontend · Next.js + React Flow"]
        DI[Data Ingestion]
        Canvas[Scenario Graph Canvas]
        Chat[AI Chat Panel]
        Report[Report Page]
    end

    subgraph Agents["Six AI Agents"]
        A1[Market]
        A2[Financial]
        A3[Growth]
        A4[Risk]
        A5[Brand]
        A6[Operations]
    end

    subgraph Pipeline["Math Pipeline · FastAPI + Python"]
        VU[Variable Universe]
        MC[Monte Carlo\n10,000+ iterations]
        BN[Bayesian Network\nCausal DAG]
        SA[Sensitivity Analysis\nSobol + Morris]
        BT[Backtesting\nWalk-forward validation]
    end

    subgraph DB["Supabase"]
        S[(Scenarios)]
        R[(Results)]
        U[(Users)]
    end

    DI -- POST /api/simulate --> VU
    VU --> MC --> BN --> SA --> BT
    BT --> Agents
    A1 & A2 & A3 & A4 & A5 & A6 --> Report
    Canvas -- CRUD --> S
    Report --> R

Loading

---

Math Pipeline

The intellectual core of optX. Five stages run in sequence on every simulation:

Stage	Method	Output
Variable Universe	Empirical distributions, copulas	Joint probability model of all inputs
Monte Carlo	10,000+ iterations, correlated sampling	P5–P95 outcome distributions for every metric
Bayesian Network	Directed acyclic graph, posterior updates	Causal map: which variables drive which outcomes
Sensitivity Analysis	Sobol indices, Morris method, SHAP attribution	"Your net income is 40% driven by raw material costs"
Backtesting	Walk-forward validation, Brier scores, calibration	Accuracy Report — model predicted within ±X% for Y% of test periods

---

Quickstart

# Frontend
npm install
npm run dev

# Backend (separate terminal)
cd python
pip install -r requirements.txt
python -m uvicorn main:app --reload --port 8000

Required env vars (.env.local):

• NEXT_PUBLIC_SUPABASE_URL
• NEXT_PUBLIC_SUPABASE_ANON_KEY
• SUPABASE_SERVICE_ROLE_KEY
• PYTHON_API_URL=http://localhost:8000
• OPENAI_API_KEY (also add to python/.env)

Open http://localhost:3000. Both services must be running.

---

Key API Endpoints

GET  /health                → service status
POST /api/simulate          → run full math pipeline on ingested data
POST /api/scenario          → create / update scenario graph
POST /api/chat              → AI chat with scenario context
GET  /api/results/{id}      → fetch simulation results + accuracy report

---

Tech Stack

Layer	Technology
Framework	Next.js 16 (App Router, TypeScript)
Graph Editor	React Flow (@xyflow/react)
Charts	Recharts
State	Zustand
Styling	Tailwind 4 + ShadCN
AI	OpenAI API (gpt-4o / gpt-4o-mini)
Backend	Python · FastAPI
Math	Monte Carlo · Bayesian Networks · Sobol · Walk-forward validation
Database	Supabase (Postgres)

---

Team

Built by Krish Garg.

---

optX — scenario intelligence for strategic decisions.