chinthamthanuja/Kitchen_Secrets

Kitchen Secrets – Preserving Culinary Heritage

1. Team Information

• —Team Name: Innovault
• —Team Members: Ch. Thanuja – Project Lead V. Rishi – AI Engineer Nitin Sain – Frontend Developer Chandra Harsha – UX Designer M. Sai Kiran – Data Scientist

2. Project Overview

Kitchen Secrets is a Streamlit-based open-source platform that enables users to share and preserve traditional Indian recipes, particularly those tied to festivals, families, and local cultures. Recipes are enriched with geo-coordinates, contributor details, media (image, video, audio), and metadata such as category, title, and description. The app ensures only logged-in users can contribute, ensuring traceability and community trust.

This project contributes to building a diverse, multilingual recipe corpus that can serve future AI, cultural, and educational purposes.

3. Key Features

• —Secure login system – No anonymous users
• —Geo-coordinates – Automatically capture or submit contributor location
• —User detail tracking – Track submissions by authenticated user
• —Corpus categorization – Festival, seasonal, snack/main/sweet, etc.
• —Title & description – Contextual metadata for each recipe
• —Media uploads – Add images, videos, and audio
• —Map visualization – View recipe locations across India
• —Reactions & comments – Community interaction
• —Leaderboard – Sorted by contribution count

4. Technical Architecture

Frontend

• —Developed in Streamlit
• —Responsive layout with step-by-step form
• —Uses Streamlit session state for login and interaction
• —Includes dropdowns, map, media uploader, and live leaderboard

Backend

• —User and recipe data stored in users.json and recipes.json
• —Media files saved locally (or base64-encoded)
• —Location fetched using IP-based lookup (ipinfo.io)
• —All data is structured for exportable corpus use

AI Layer (Planned / Optional)

• —Use langdetect or FastText to detect recipe language
• —Normalize ingredients and instructions using IndicTrans2
• —Semantic search using SentenceTransformers
• —Future: Speech-to-text for voice inputs

5. Project Structure

kitchen-secrets/
├── app.py                  # Main Streamlit app
├── users.json              # Stores registered user info
├── recipes.json            # Stores submitted recipes
├── requirements.txt        # Python dependencies
├── README.md               # Project documentation
├── assets/
│   └── logo.png            # Logo or branding assets
├── media/
│   └── ...                 # Uploaded images, audio, video files
├── .streamlit/
│   └── config.toml         # Streamlit configuration
└── utils/
    ├── auth.py             # Login/signup functionality
    └── helpers.py          # Geo, media, leaderboard utilities

6. User Feedback & Improvements

Testing Approach

• —Conducted with 10+ users across Telegram & WhatsApp
• —Asked users to share 1 traditional recipe with media and context
• —Monitored engagement and completion rate

Feedback Summary

7. Corpus Impact

• —300+ recipes submitted in 5 languages
• —Location metadata collected from 18+ states
• —Dishes include Pongal, Diwali sweets, Biryani variants, tribal dishes
• —Used in 3 regional food awareness events
• —Exploring regional speech corpus collection via recipes

8. Sustainability & Future Use

• —Datasets will be open-sourced via Hugging Face or GitHub
• —Will support export in .json, .csv, .txt
• —Planned outreach to food bloggers & NGOs
• —Possible publication as an open digital cookbook
• —AI fine-tuning on collected corpus (NLP & speech models)

9. Team Contributions

Acknowledgement

This project is part of the Viswam.ai Open Source Fellowship, aimed at building inclusive and culturally rich AI datasets and applications for India’s diverse regions. ---

📄 License

This project is licensed under the MIT License.