MFP — Mathematical Flavor Profile Engine

Deterministic, auditable flavor intelligence in 20 dimensions

Overview

MFP (Mathematical Flavor Profile) is a computational engine for analyzing and optimizing recipes using a 20-dimensional flavor space. It provides deterministic, auditable scoring of dishes against cuisine-specific targets, along with actionable recommendations for improvement.

Features

• 20-dimensional flavor space — UMAMI, SALT, SWEET, SOUR, BITTER, HEAT_PEPPER, WARM_SPICE, SMOKE, ROASTED, FAT_RICH, CREAMY, HERBAL, CITRUS, ALLIUM, FERMENT, EARTHY, NUTTY, FLORAL, TEXTURE_CRISP, TEXTURE_TENDER
• Role-based normalization — PRIMARY / SUPPORT / FINISH categories with dish-type-specific constants
• Method kernels — 12 cooking methods with heat/volatility transforms (GRILL, ROAST, SAUTE, etc.)
• Scoring system — Similarity, balance, structural coverage, clash penalty with weighted final formula
• Recommendations — Add-ins, fixes, substitutions, method adjustments
• Cultural libraries — AA Foodways (Midwest + Gulf States), with more to come

Packages

| Package | Description | |---------|-------------| | @studio46/mfp-core | Types, math primitives, and scoring algorithms | | @studio46/mfp-data | Ingredient libraries and cuisine style targets | | @studio46/mfp-engine | Complete MFP engine with recommendations |

Installation

npm install @studio46/mfp-engine

Or install individual packages:

npm install @studio46/mfp-core @studio46/mfp-data

Quick Start

import {
  runMFPEngine,
  getIngredient,
  CookingMethod,
  DishType,
} from '@studio46/mfp-engine';

const result = runMFPEngine({
  ingredients: [
    { card: getIngredient("chicken_breast"), quantity: 200 },
    { card: getIngredient("garlic"), quantity: 10 },
    { card: getIngredient("olive_oil"), quantity: 30 },
    { card: getIngredient("lemon_juice"), quantity: 15 },
    { card: getIngredient("basil"), quantity: 8 },
  ],
  method: CookingMethod.SAUTE,
  heatLevel: 0.7,
  dishType: DishType.COMPLETE_PLATE,
  styleTargetId: "italian",
});

console.log("Final Score:", result.scored.score.toFixed(2));
console.log("Components:", result.scored.components);
console.log("Top Recommendations:", result.recommendations.slice(0, 3));

Style Targets

| ID | Name | Characteristics | |----|------|-----------------| | italian | Italian (Classic) | Umami, herbal, allium, fat-rich | | japanese | Japanese (Washoku) | Umami, ferment, delicate | | mexican | Mexican (Traditional) | Heat, citrus, earthy | | thai | Thai | Sweet-sour-heat balance | | french | French (Classic) | Fat, cream, roasted | | indian | Indian (North) | Warm spice, heat, allium | | bbq | American BBQ | Smoke, sweet, fat, tender | | mediterranean | Mediterranean | Herbal, citrus, olive oil | | soul_food | Soul Food (AA Core) | Smoke, bitter greens, fat, tender | | gulf_creole | Gulf Creole (Louisiana) | Heat, roux, seafood, allium |

Scoring Formula

Score = 0.45×S_sim + 0.35×S_bal + 0.20×S_struct - 0.40×P_clash

Where:
  S_sim    = cosine similarity to target profile
  S_bal    = weighted balance score
  S_struct = structural role coverage
  P_clash  = ingredient class clash penalty

Calibration Status

Ingredient vectors have calibration status indicators:

| Status | Meaning | Count | |--------|---------|-------| | A | Sensory-validated (blind tasting) | 0 | | B | Database-grounded | ~37 | | C | Prior estimate (needs calibration) | ~100 |

Documentation

• Specification (MFP v1.1)
• API Reference
• Calibration Guide
• Contributing

Development

# Install dependencies
pnpm install

# Build all packages
pnpm build

# Run tests
pnpm test

# Type check
pnpm typecheck

License

MIT © Studio46-Go