A voice-first WhatsApp assistant, initially serving Telugu and English speakers to discover government welfare schemes. Powered by AI/NLU, it converts a single voice or text message into a personalized profile, matches it to relevant programs, and delivers guidance. Our vision: to one day support any language and context globally and operate seamlessly on a modular smartphone platform.
- Phase 1: Launch on WhatsApp for Telugu + English.
- Phase 2: Improve AI‑powered understanding (NLU).
- Phase 3: Future hardware support: build a modular phone with the assistant pre-installed.
- Phase 4: Expand to universal, language-agnostic voice assistant for civic services worldwide.
- Voice + Text Input – via WhatsApp using Whisper for Telugu and English.
- AI‑Powered NLU – extracts structured user data: age, income, intent, gender, caste, etc.
- Intelligent Matching – selects relevant welfare schemes from a verified database.
- Seamless Workflow – sends document requirements and next steps via WhatsApp.
| Component | Technology |
|---|---|
| Backend | Python 3.10+, Flask |
| NLU & Embeddings | OpenAI/Google Gemini/Fine‑tuned Whisper |
| Vector Search | LangChain + pgvector |
| Messaging | Twilio WhatsApp API |
| Database | SQLite / JSON session store |
| Logging & Testing | logging_config.py, pytest |
git clone https://github.com/your-org/safesound-assistant.git
cd safesound-assistant
python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt
cp .env.example .env
# Add your TWILIO_* and GEMINI_API_KEY
python app.pyUse ngrok to expose your bot and update your Twilio webhook accordingly:
ngrok http 5000safesound/
├── app.py # Webhook + dialog flow
├── nlu.py # NLU & schema extraction
├── utils.py # Telugu-parsers, number handling
├── logging_config.py # Structured logging
├── session_manager.py # Tracks user conversations
├── welfare_schemes.db # Sample scheme database
├── user_states.json # Persistent session storage
├── requirements.txt # Installed dependencies
├── .env.example # Env vars template
└── README.md # This file
- Minimal & Modular: Logic split into clear, single-responsibility modules.
- Type-safe & Tested: Type hints across codebase + unit tests.
- Localisation-first: Built for Telugu-English; designed to scale across languages.
- Security-minded: Validates Twilio webhook signatures, secures keys with
.env. - Extensible: Easy to plug in new states, languages, or hardware platform.
- Improve NLU accuracy using Gemini or fine-tuned models.
- Add language detection for seamless user experience.
- Build on modular phone prototype once NLU scales.
- Scale pilot with nonprofits and AP state government; then expand globally.
We welcome contributions—whether through code, documentation, or issue reports. Please open an issue or submit a PR, and include tests for new features.
Founder & Engineer: Prudvish Korrapati — [email protected]
This project is released under the MIT License.