convez.ai
MCP Server
legiovi

flavourlab

legiovi/flavourlab

Ver no GitHub
InicialPluginScore · de 0 a 100

Aroma-based food pairing intelligence — visual app, MCP server & Claude skill

Instalação

$npx convez-ai legiovi/flavourlab

1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.

Métricas

Estrelas
0
Forks
0
Acompanhando
0
Saúde
28%
Último commit
há 5 meses
Versão
—
Issues abertas
0
Contribuidores
1

README

A tradução para o português deste README ainda está na fila.

FlavourLab 🧪

Aroma-based food pairing intelligence — visual app + MCP server for AI agents

A science-based tool for discovering ingredient pairings and generating recipes based on shared volatile organic compounds (aroma molecules). Includes:

  • 🕸 Interactive flavour network — D3.js force graph of ingredients connected by aroma links
  • 🎨 Pairing builder — build ingredient combinations and score their aromatic harmony
  • ✨ Recipe generator — generate complete recipes using flavour pairing science + 4,000+ real recipes
  • 🌸 Aroma explorer — browse 17 aroma note categories and the ingredients that share them
  • 📖 Culinary foundations — 27 classical base recipes & techniques (stocks, mother sauces, doughs, emulsions)
  • 📚 Recipe Library — search 4,000+ real recipes and create aroma-based variations (fennel velouté → leek velouté)
  • 🤖 MCP server — 11 tools for AI agents to query pairing data, browse foundations, and generate recipes
  • 🧩 Claude Code skill — drop-in skill/flavourlab/ with a dependency-free CLI (see below)

Quick Start — Visual App

git clone https://github.com/YOUR_USERNAME/flavourlab
cd flavourlab
npm run serve
# Open http://localhost:8765

No build step. Pure HTML/JS with D3.js loaded from CDN.


Claude Code Skill (drop-in)

A self-contained skill lives in skill/flavourlab/. No dependencies (Python 3 stdlib only).

cp -r skill/flavourlab ~/.claude/skills/

Then just ask Claude: "what pairs with lamb?", "wine for grilled beef", "generate a braised lamb and apricot recipe", "how do I make hollandaise?" — it auto-runs the bundled flavourlab.py CLI. See skill/flavourlab/INSTALL.md.


MCP Server (for AI agents)

Install via npm

npm install -g flavourlab

Or run directly with npx

npx flavourlab

Add to Claude Code / Claude Desktop

In your claude_desktop_config.json or ~/.claude/settings.json:

{
  "mcpServers": {
    "flavourlab": {
      "command": "npx",
      "args": ["flavourlab"]
    }
  }
}

Or if cloned locally:

{
  "mcpServers": {
    "flavourlab": {
      "command": "node",
      "args": ["/path/to/flavourlab/src/mcp-server.js"]
    }
  }
}

MCP Tools

ToolDescription
get_ingredientFull ingredient details: aroma profile, pairings, description
list_ingredientsAll 108 ingredients, filterable by category or search
get_pairingsBest matches for an ingredient ranked by shared aroma compounds
check_harmonyAromatic harmony score (0–100) for a set of ingredients
generate_recipeFull recipe with ingredients, method, aroma science, real-world references
find_recipesSearch 4,000+ real recipes (full ingredients + method) by ingredient and cuisine
get_variationsSame-role aroma-compatible substitutes for recipe variations (fennel→leek)
get_aroma_ingredientsAll ingredients sharing a specific aroma note
suggest_drink_pairingBest wine / beverage pairing for a set of food ingredients
list_culinary_basesList the 27 foundational base recipes & techniques
get_culinary_baseFull method for a base (béchamel, brown stock, pasta dough, mayonnaise…)

Example agent usage

User: What pairs well with strawberry?
Agent: [calls get_pairings("strawberry")]
→ Returns: chocolate, basil, tomato, vanilla, citrus... with shared aroma explanations

User: Score the harmony of strawberry + basil + tomato
Agent: [calls check_harmony(["strawberry", "basil", "tomato"])]
→ Returns: score 78/100, shared notes: Fruity×3, Floral×2, Citrus×2

User: Generate a Spanish starter recipe using strawberry and basil
Agent: [calls generate_recipe({base:"strawberry", pair:"basil", cuisine:"Spanish", course:"starter"})]
→ Returns: full recipe with science explanation and real recipe references

Data

  • Ingredients across 9 categories (fruit, vegetable, herb, spice, protein, dairy, beverage, wine & fortified, other)
  • 17 aroma note categories derived from volatile organic compound analysis
  • 4,000+ real recipes (with full ingredients & methods) across Italian, Middle Eastern, Mexican, Baking, Catalan, Spanish, Nordic, American, Cambodian, Asian, Seafood, French and more
  • 22 drinks — 10 wines & fortified (Cabernet, Pinot Noir, Chardonnay, Riesling, Champagne, Port, Sherry, Sake, Rosé, Syrah) + 12 beverages, all aroma-linked to food
  • 27 culinary foundations — stocks, the five mother sauces, doughs, emulsions, pastry bases and core techniques
  • All recipe text is English-only and self-contained (no external attribution)
  • Pairing data based on shared volatile aroma compounds (flavour pairing science)

Structure

flavourlab/
├── public/                   # Visual web app (deployable as static site)
│   ├── index.html            # Single-file visual app (6 views: network, builder, generator, aroma, foundations, library)
│   ├── data.js               # Ingredient + aroma database
│   ├── recipes.json          # 4,000+ extracted real recipes
│   └── culinary_bases.json   # 27 culinary foundations
├── src/
│   └── mcp-server.js         # MCP server (11 tools)
├── data/                     # Source datasets (mirrored to public)
│   ├── data.js
│   ├── recipes.json
│   └── culinary_bases.json
├── tools/                    # Dataset build pipeline
│   ├── extract_epubs.py      # Generic EPUB recipe extractor (English-only)
│   ├── extract_pdfs.py       # PDF recipe extractor
│   └── build_dataset.py      # Merge + clean + ingredient-link
└── package.json

License

MIT

Ver o original no GitHub

Relacionados