Tantrium Boundary Engine

Find the exact point where your system breaks — not a probability, not a range. The boundary.

A production-grade boundary analysis web application. Upload or auto-fetch real operational data, run the Tantrium six-step methodology, and receive a certified, evidence-hashed Boundary Report identifying the safe operating envelope, the first break boundary, the primary failure driver, and a concrete stabilization path.

---

What It Does

Most operational analytics tools tell you that a system is failing. Tantrium tells you where it breaks.

Given any operational dataset with a measurable target metric, Tantrium computes:

Output	Description
Safe Operating Envelope	The parameter ranges where the system runs without failure — computed from real records
First Break Boundary	The exact value at which the system transitions from stable to unstable
First Obstruction	The primary variable driving the boundary breach, ranked by Boundary Pressure Score
Closure Path	A concrete sequence of operational changes to return to the safe zone

Every certified analysis generates a cryptographic evidence hash. No synthetic outputs. No black boxes.

---

Live Demo

The app ships with auto-fetch access to real public datasets across 8 sectors. On the Live Reports page, the engine automatically fetches and analyzes:

• Server Thermal Failure Boundary — 22,695 rows of server temperature data (NAB dataset)
• Insurance Claims Severity Boundary — 1,338 policyholder records with charges
• Process Quality Degradation Boundary — Red wine physicochemical parameters vs. quality score
• Product Pricing Anomaly Boundary — 53,940 diamond records, price vs. physical attributes

All outputs are labeled REAL DATA when actual rows are processed, or INSUFFICIENT DATA if the dataset does not meet the minimum row threshold.

---

Pages

Route	Purpose
/	Executive home page
/hunt	Master Problem Hunt — guided workflow: pick template → load data → configure → run → get report
/reports	Live Boundary Reports — auto-runs 4 real datasets, shows full certified reports
/core	Tantrium Core methodology — six-step workflow explained for technical audiences
/demo	Interactive synthetic demo
/datasets	Full dataset registry (15 entries, 8 sectors)
/pricing	Pricing tiers and analyst report request form

---

Sectors Covered

• Data Centers & Hosting
• Plastics / Polymer / Manufacturing
• General Manufacturing & OEE
• Pharma / Biotech / Toxicity
• Finance & Insurance Risk
• Energy & HVAC
• Logistics
• Cyber & SLA Risk

---

Tech Stack

Layer	Technology
Frontend	React 19 + Vite 7 + TypeScript 5.9
Routing	Wouter
UI Components	shadcn/ui + Tailwind CSS
Charts	Recharts
CSV Parsing	PapaParse
Animation	Framer Motion
Package Manager	pnpm workspaces
Backend	Express 5 (API server)
Database	PostgreSQL + Drizzle ORM

All boundary analysis runs client-side in the browser — no server round-trip required for computation.

---

Getting Started

Prerequisites

• Node.js 20+
• pnpm 9+
• PostgreSQL (for API server features)

Install & Run

# Install dependencies
pnpm install

# Run the web app (development)
pnpm --filter @workspace/tantrium run dev

# Run the API server
pnpm --filter @workspace/api-server run dev

# Typecheck all packages
pnpm run typecheck

The web app will be available at http://localhost:PORT (port assigned by the runtime).

---

How the Analysis Engine Works

The Tantrium engine (src/lib/analysis-engine.ts) runs six labeled steps:

DATA_INGEST      → Parse and validate all rows; detect numeric/categorical columns
STABLE_REGION    → Classify each row as stable or unstable against the target threshold
BREAK_BOUNDARY   → Detect stable-to-unstable transition points in sorted driver space
FIRST_OBSTRUCTION → Score all drivers by Boundary Pressure Index (Pearson + crossing strength + proximity + monotonicity)
CLOSURE_PATH     → Compute recommended operational changes to return to the safe envelope
EVIDENCE_HASH    → Generate SHA-256 fingerprint of dataset + parameters + results

Output type: TantriumObject — a fully typed result containing all six step outputs, mode label, certification status, timestamp, and evidence hash.

Mode labels:

• REAL DATA — at least 10 valid rows processed, engine ran to completion
• DEMO — synthetic or placeholder data
• INSUFFICIENT DATA — fewer than 10 valid rows after filtering

---

Dataset Registry

15 real public datasets pre-configured with suggested target columns, thresholds, and directions. 9 are auto-fetchable directly from public GitHub raw URLs. 6 require manual download due to licensing or size.

See /datasets in the app or src/lib/dataset-registry.ts for the full registry.

---

Project Structure

artifacts/
  tantrium/          # React web application
    src/
      lib/
        analysis-engine.ts   # Core Tantrium computation engine
        dataset-registry.ts  # 15 real public datasets
        evidence-log.ts      # User analysis history (localStorage)
        report-store.ts      # Auto-computed live reports (localStorage)
        tantrium-core.ts     # Methodology explanations and glossary
        prospect-mode.ts     # LinkedIn buyer persona intelligence
      pages/
        home.tsx             # Executive landing page
        hunt.tsx             # Master Problem Hunt workflow
        reports.tsx          # Live Boundary Reports
        core.tsx             # Methodology page
        demo.tsx             # Synthetic demo
        datasets.tsx         # Dataset registry
        pricing.tsx          # Pricing and contact
  api-server/        # Express API backend
lib/                 # Shared workspace libraries
scripts/             # Utility scripts

---

Business Context

Tantrium Boundary Engine is a sales and demonstration tool for the Tantrium Boundary Analysis consultancy. It serves as:

1. A live proof-of-capability — showing real analysis on real public data
2. A prospect onboarding tool — LinkedIn targets can self-serve the Problem Hunt
3. A pricing gateway — the /pricing page converts interested prospects to paid engagements

Pricing tiers:

• Quick Screen: $5,000 / 5 business days
• Boundary Report: $25,000 / 3 weeks (standard)
• Enterprise Program: from $100,000 / ongoing monitoring

---

Evidence & Integrity

Every certified analysis produces:

• evidenceHash — SHA-256 of datasetName + targetCol + threshold + direction + rowCount + topDriver + evidenceNonce
• mode — REAL DATA / DEMO / INSUFFICIENT DATA
• certified: boolean — true only when mode is REAL DATA

The hash can be shared as proof that the analysis was run on a specific dataset with specific parameters. It does not expose the underlying data.

---

License

Proprietary. All rights reserved. This codebase is not open-source.