🌮 SazónBot - AI-Powered Mexican Recipe Chatbot

An intelligent conversational chatbot that shares authentic Mexican recipes from the García family collection. Built with LangChain agents, OpenAI GPT-4o-mini, and a warm, bilingual personality.

✨ Features

• 🤖 Intelligent Agent: Powered by LangChain with 11+ specialized tools for recipe search, scaling, substitutions, and more
• 🌎 Bilingual Support: Natural English/Spanish conversation with code-switching
• 🔍 Semantic Recipe Search: Vector-based search using FAISS and OpenAI embeddings
• 📹 Rich Media: Embedded YouTube cooking videos and food images
• 📏 Recipe Scaling: Automatically adjust ingredients for any serving size
• 🔄 Ingredient Substitutions: Smart alternatives for dietary restrictions and missing ingredients
• 📱 Responsive Design: Beautiful mobile-first UI with smooth animations
• 💾 Session Memory: Maintains conversation context per user
• 🎨 Mexican-Themed UI: Custom color palette inspired by Mexican culture

🚀 Live Demo

Frontend: https://sazonbot.vercel.app
API: https://mexican-ai-chatbot.onrender.com/docs

⚠️ Note: First request may take 20-30 seconds on Render's free tier due to cold starts.

🛠️ Tech Stack

Backend

• Framework: FastAPI (Python)
• AI/ML: LangChain, OpenAI GPT-4o-mini
• Vector Store: FAISS with OpenAI embeddings
• Tools: Google Serper API (web search), Pushover (notifications)
• PDF Processing: PyPDF for recipe extraction

Frontend

• Framework: Next.js 15 (React 19)
• Language: TypeScript
• Styling: Tailwind CSS
• HTTP Client: Axios

📋 Prerequisites

• Python 3.11+
• Node.js 18+
• OpenAI API key
• Serper API key (optional, for web search)
• Pushover keys (optional, for feedback notifications)

🔧 Local Development Setup

Backend Setup

1. Clone the repository

git clone https://github.com/yourusername/mexican-recipe-chatbot.git
cd mexican-recipe-chatbot/backend

2. Create virtual environment (using uv)

uv venv
source .venv/bin/activate  # On Windows: .venv\Scripts\activate

3. Install dependencies

uv pip install -r requirements.txt

4. Create .env file

cp .env.example .env

Add your API keys:

OPENAI_API_KEY=sk-...
SERPER_API_KEY=...  # Optional
PUSHOVER_USER_KEY=...  # Optional
PUSHOVER_API_TOKEN=...  # Optional

5. Create vector store

python -m app.vector_store
# Choose option 1 to create new vector store

6. Run the server

python -m uvicorn app.main:app --reload

Backend will be available at http://localhost:8000

API docs at http://localhost:8000/docs

Frontend Setup

1. Navigate to frontend directory

cd ../frontend

2. Install dependencies

npm install

3. Create .env.local file

echo "NEXT_PUBLIC_API_URL=http://localhost:8000" > .env.local

4. Run development server

npm run dev

Frontend will be available at http://localhost:3000

📁 Project Structure

mexican-recipe-chatbot/
├── backend/
│   ├── app/
│   │   ├── __init__.py
│   │   ├── agent.py           # LangChain agent with 11 tools
│   │   ├── config.py          # Environment configuration
│   │   ├── main.py            # FastAPI application
│   │   ├── models.py          # Pydantic models
│   │   ├── tools.py           # Agent tools (search, scale, etc.)
│   │   ├── vector_store.py    # FAISS vector store management
│   │   └── utils/
│   │       ├── recipe_parser.py    # Recipe scaling logic
│   │       └── safety.py           # Prompt injection protection
│   ├── data/
│   │   ├── recipes.pdf        # García family recipes
│   │   └── recipe_vectors/    # FAISS index (generated)
│   ├── requirements.txt
│   └── .env
├── frontend/
│   ├── pages/
│   │   ├── index.tsx          # Main chat interface
│   │   ├── _app.tsx
│   │   └── _document.tsx
│   ├── lib/
│   │   ├── api.ts             # API client
│   │   └── translations.ts    # i18n translations
│   ├── styles/
│   │   └── globals.css
│   ├── public/
│   │   └── logo.png
│   ├── package.json
│   └── .env.local
└── README.md

🔌 API Endpoints

Health Check

GET /health

Agent Chat

POST /agent-chat
Body: { "message": "How do I make pozole?" }
Response: { "response": "...", "sources_used": [...], "session_id": "..." }

Clear Conversation

POST /agent-chat/clear/{session_id}

Recipe Search

POST /search-recipes
Body: { "query": "chicken", "limit": 3 }

Full API documentation available at /docs when server is running.

🤖 Agent Tools

The chatbot uses 11 specialized tools:

1. recipe_search_tool - Semantic search in recipe database
2. recipe_list_by_type_tool - Browse recipes by category
3. get_full_recipe_tool - Get complete recipe details
4. web_search_tool - Search web for cooking information
5. recipe_scale_tool - Scale recipes for different servings
6. ingredient_substitution_tool - Find ingredient alternatives
7. cooking_technique_tool - Explain cooking methods
8. recipe_filter_by_criteria_tool - Complex recipe filtering
9. video_search_tool - Find YouTube cooking tutorials
10. image_search_tool - Find food images
11. record_unknown_question_tool - Log unanswered questions

🎨 Color Palette

The UI uses a Mexican-inspired color scheme:

• Turkey Red (#a81007) - Primary actions
• Oxford Blue (#0d082d) - Sidebar background
• Fulvous (#dc8300) - Accents and highlights
• Cornsilk (#f9edcc) - Message bubbles
• Ghost White (#f8f4f9) - Main background

🌐 Deployment

Deploy Backend to Render

1. Create new Web Service on Render
2. Connect your GitHub repo
3. Configure:
   • Root Directory: backend
   • Build Command: pip install -r requirements.txt
   • Start Command: uvicorn app.main:app --host 0.0.0.0 --port $PORT
4. Add environment variables (API keys)
5. Deploy

Deploy Frontend to Vercel

1. Import project on Vercel
2. Configure:
   • Root Directory: frontend
   • Framework: Next.js
3. Add environment variable:
   • NEXT_PUBLIC_API_URL: Your Render backend URL
4. Deploy

Update CORS

After deployment, update backend/app/main.py CORS settings with your Vercel URL.

📝 Environment Variables

Backend (.env)

OPENAI_API_KEY=sk-...           # Required
SERPER_API_KEY=...              # Optional (for web search)
PUSHOVER_USER_KEY=...           # Optional (for feedback)
PUSHOVER_API_TOKEN=...          # Optional (for feedback)

Frontend (.env.local)

NEXT_PUBLIC_API_URL=http://localhost:8000  # Backend URL

🧪 Testing

Test Backend

cd backend
python -m app.vector_store  # Option 2: Test vector store
python -m app.agent         # Test agent (uncomment test code)

Test API Endpoints

Visit http://localhost:8000/docs for interactive API testing.

🚀 Performance Optimizations

• Model: GPT-4o-mini for 5-10x faster responses vs GPT-4
• Streaming: Real-time response streaming (ready for SSE implementation)
• Caching: Session-based memory for context retention
• Vector Search: FAISS for sub-second recipe lookups
• Responsive UI: Smooth animations and optimistic updates

🔒 Security Features

• Prompt injection protection
• Off-topic query filtering
• Session isolation per user
• CORS configuration
• Environment variable management

📚 Recipe Collection

The chatbot has access to authentic Mexican recipes including:

• Pozole Blanco
• Fajitas a la Vizcaína
• Tortitas de Atún
• And many more traditional dishes!

Recipes are stored in backend/data/recipes.pdf and processed into a vector database.

🤝 Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

1. Fork the repository
2. Create your feature branch (git checkout -b feature/AmazingFeature)
3. Commit your changes (git commit -m 'Add some AmazingFeature')
4. Push to the branch (git push origin feature/AmazingFeature)
5. Open a Pull Request

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.

🙏 Acknowledgments

• García family for the authentic Mexican recipes
• OpenAI for GPT-4o-mini
• LangChain for the agent framework
• The open-source community

📧 Contact

Cristina Rodriguez - LinkedIn

Project Link: https://github.com/Yosolita1978/mexican-ai-chatbot

---

¡Buen provecho! 🌮🇲🇽