flavourlab
legiovi/flavourlab
Aroma-based food pairing intelligence — visual app, MCP server & Claude skill
Instalação
npx convez-ai legiovi/flavourlab1 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
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
| Tool | Description |
|---|---|
get_ingredient | Full ingredient details: aroma profile, pairings, description |
list_ingredients | All 108 ingredients, filterable by category or search |
get_pairings | Best matches for an ingredient ranked by shared aroma compounds |
check_harmony | Aromatic harmony score (0–100) for a set of ingredients |
generate_recipe | Full recipe with ingredients, method, aroma science, real-world references |
find_recipes | Search 4,000+ real recipes (full ingredients + method) by ingredient and cuisine |
get_variations | Same-role aroma-compatible substitutes for recipe variations (fennel→leek) |
get_aroma_ingredients | All ingredients sharing a specific aroma note |
suggest_drink_pairing | Best wine / beverage pairing for a set of food ingredients |
list_culinary_bases | List the 27 foundational base recipes & techniques |
get_culinary_base | Full 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