Apppublic

Jatin-Kumar/Bhasha_Stories

sourceHugging Faceupdated 3mo agoView on Hugging Face

0likes

App README

📚 Bhasha Stories – Preserve India's Linguistic Heritage

An open-source, AI-powered platform to collect, transcribe, and share regional stories from across India.

🎯 Project Overview

Bhasha Stories is a professional-grade Streamlit web app built to capture India’s oral and written storytelling traditions in native languages and dialects. Users can upload audio, write text, and generate shareable story cards — all enhanced by modern AI tools like OpenAI Whisper, Sentence Transformers, and Groq's LLaMA 3.1 via API.

🔗 Live Demo

Access the application on Hugging Face Spaces:

![Live on Hugging Face Spaces](https://huggingface.co/spaces/Jatin-Kumar/Bhasha_Stories)

📍 URL: https://huggingface.co/spaces/Jatin-Kumar/Bhasha_Stories

🌟 Key Features

—🎤 Multilingual Voice Recording – Upload stories in any Indian language.
—✍️ Text Story Submission – Submit stories directly in text using native scripts.
—🔉 Whisper-Based Transcription – Automatic speech-to-text with OpenAI Whisper.
—🤖 Groq-Powered AI Assistant – LLaMA 3.1 (8B Instant) via Groq API for natural language querying.
—🔍 Semantic Search with RAG – Enhanced Retrieval-Augmented Generation using sentence-transformers.
—🎨 Story Card Generator – Export visually appealing HTML-based story cards.
—📊 Analytics Dashboard – Real-time insights on story count, language coverage, and embedding health.
—☁️ Supabase Integration – PostgreSQL/JSONB-backed persistent storage with semantic embedding storage.

🧠 AI Integrations

Component	Technology Used
Transcription	OpenAI Whisper (Tiny, CPU-optimized)
Semantic Embedding	Sentence Transformers (`MiniLM`)
RAG + Chat	Groq API with LLaMA 3.1 8B Instant
Search Embeddings	Supabase + cosine similarity

📦 Tech Stack

Layer	Technology
Frontend	Streamlit (Custom dark UI)
Backend	Python 3.10+
Transcription	OpenAI Whisper (tiny)
Embeddings	Sentence Transformers (`MiniLM`)
AI Chat	Groq API (LLaMA 3.1 8B)
Storage	Supabase (PostgreSQL + JSONB)
Hosting	Hugging Face Spaces / Streamlit Cloud

🚀 Getting Started

🛠️ Local Development Setup

1.Clone the repository

bash

   git clone https://gitlab.com/<your-group>/bhasha-stories.git
   cd bhasha-stories

1.Install Python dependencies

bash

   pip install -r requirements.txt

1.Add Supabase and Groq secrets

Create a .streamlit/secrets.toml file:

toml

   SUPABASE_URL = "https://<your-project>.supabase.co"
   SUPABASE_ANON_KEY = "<your-anon-key>"

Set your Groq API key in the Streamlit sidebar during runtime.

1.Run the app

bash

   streamlit run app.py

📂 Directory Structure

bhasha-stories/
├── app.py                    # Main Streamlit app
├── requirements.txt          # Python packages
├── .streamlit/
│   └── secrets.toml          # Secrets (Supabase credentials)
├── CONTRIBUTING.md           # Contribution guide
├── LICENSE                   # MIT License
├── stories_data/             # Local backup of stories
└── README.md                 # This file

📊 AI Analytics Dashboard

—Total stories & AI embedding coverage
—Language distribution of stories
—Average story length
—Embedding quality indicators
—Missing embedding alerts & auto-repair

🧲 Who It's For

Audience	Use Case
Cultural Enthusiasts	Record oral folktales and regional lore
Students/Researchers	Collect linguistic data for NLP & studies
Families	Preserve elders' memories in voice/text
Creators/Educators	Convert stories into educational material

📈 Growth Strategy

Channel	Plan
Social Media	Share auto-generated story cards
Academia	Collaborate with language departments
NGOs & Orgs	Conduct story collection drives
Local Competitions	Host storytelling events in schools/villages
Influencers	Promote through regional YouTube/Instagram

📚 Example Use Cases

—Archive endangered languages
—Personal memory preservation
—Research-grade multilingual corpora
—Story-of-the-day social feeds

🚧 Planned Enhancements

—🌐 AI-based translation between Indian languages
—📱 Mobile-first PWA interface
—🔈 Voice synthesis / playback of stories
—🧠 Advanced cross-lingual search + clustering

🤝 Contributing

We welcome contributions! See CONTRIBUTING.md for instructions.

You can contribute by:

—Adding folktales and sharing the platform with friends and family.

📄 License

MIT License See LICENSE for more information.

🙏 Acknowledgments

—OpenAI Whisper – STT Engine
—Sentence Transformers – Embedding model
—Groq – LLaMA 3.1 8B Inference API
—Streamlit – Frontend Framework
—Supabase – Cloud DB
—And the storytellers of India ❤️

Made with 💙 by the Viswam.ai Internship Team – Empowering linguistic diversity through AI.