🍽️ Mood-Based Food Recommender Setup Guide

Quick Start

1. Install Dependencies:

   pip install -r requirements.txt

2. Get Google API Key:

   • Visit https://ai.google.dev/
   • Sign in with your Google account
   • Create a new API key for Gemini
   • Keep your API key secure

3. Run the Application:

   streamlit run mood_food_recommender.py

4. Access the Web Interface:

   • Open your browser to http://localhost:8501
   • Enter your Google API key in the sidebar
   • Start getting personalized food recommendations!

File Structure

your-project/
├── mood_food_recommender.py    # Main Streamlit application
├── user_profile.txt            # User preferences and history
├── contextual_data.txt         # Current context (time, weather, etc.)
├── requirements.txt            # Python dependencies
└── README.md                   # This file

How It Works

Algorithm Overview

The system uses a sophisticated 4-phase scoring algorithm:

1. Mood Analysis (40% weight) - Analyzes your text input for emotional state
2. Historical Behavior (30% weight) - Considers your past food preferences
3. Contextual Integration (20% weight) - Factors in time, weather, location
4. Demographic Personalization (10% weight) - Adjusts for age, income, preferences

Input Requirements

1. User Input (GUI): Your current mood/feelings as text
2. User Profile (File): Your preferences, history, demographics
3. Contextual Data (File): Current time, weather, location info

Technology Stack

• Frontend: Streamlit (Web Interface)
• LLM: Google Gemini 2.0 Flash
• Framework: LangChain for LLM orchestration
• Data Validation: Pydantic for structured outputs
• Language: Python 3.8+

Customization

Modifying User Profile

Edit user_profile.txt to include:

• Your actual food preferences
• Order history
• Dietary restrictions
• Budget preferences
• Demographics

Updating Context Data

Edit contextual_data.txt to reflect:

• Current time and date
• Local weather conditions
• Your location
• Current situation/stress levels

Algorithm Tweaks

The algorithm weights can be adjusted in the prompt template within mood_food_recommender.py:

• Mood Analysis: Currently 40%
• Historical Behavior: Currently 30%
• Contextual Integration: Currently 20%
• Demographics: Currently 10%

Troubleshooting

Common Issues:

1. API Key Error:

   • Ensure your Google API key is correct
   • Check if you have Gemini API access enabled
   • Verify billing is set up if required

2. File Not Found:

   • Ensure user_profile.txt and contextual_data.txt exist
   • Check file permissions and encoding (UTF-8)

3. Import Errors:

   • Run pip install -r requirements.txt
   • Check Python version (3.8+ required)
   • Consider using a virtual environment

4. Rate Limiting:

   • Google API has usage limits
   • Wait a few minutes between requests if hitting limits
   • Check your API quota in Google Cloud Console

Advanced Features

Structured Output

The system uses Pydantic models to ensure consistent response formatting:

• RecommendationResponse - Main output structure
• FoodRecommendation - Individual recommendation format

LangChain Integration

• Uses ChatGoogleGenerativeAI for Gemini access
• Implements PromptTemplate for consistent prompting
• Leverages with_structured_output() for JSON mode

Safety & Privacy

• API keys are handled securely through Streamlit's sidebar
• No data is stored permanently
• All processing happens in real-time

Contributing

Feel free to enhance the algorithm by:

1. Adding more sophisticated mood detection
2. Implementing user feedback loops
3. Adding more contextual factors
4. Improving the UI/UX

License

This project is open source and available under the MIT License.