Sahayak — PAN Card Voice Assistant

Overview

Sahayak is a voice assistant that helps people apply for a PAN card in India. Instead of reading through confusing government portal instructions, you just talk to it — ask what documents you need, how much the fee is, which form to fill — and it answers back in speech, in English or Hindi. It also remembers which step you were on if you come back later.

Demo

▶️ Watch the demo video

Features

Ask questions by voice, get spoken answers back (English or Hindi)
Answers are grounded in a small knowledge base of PAN facts, not made up by the LLM
Tracks which stage of the application you're on (eligibility → form → documents → photo → payment → submission → tracking)
Remembers your progress — reload the page and it picks up where you left off
Warns you if you accidentally type/say a real PAN or Aadhaar number
Falls back to a safe pre-written answer if the LLM or TTS service is down, instead of crashing

Tech Stack

Python, Flask
Qdrant (vector database — stores PAN facts + session memory)
Groq (LLaMA 3.3 70B) for answering questions
Rime for text-to-speech
Browser's built-in Web Speech API for voice input (Chrome only)
sentence-transformers for embeddings, falls back to TF-IDF if not installed

Project Structure

sahayak/
├── backend/
│   ├── main.py            # Flask app + routes
│   ├── agent.py            # orchestrates a conversation turn
│   ├── config.py           # env vars, voice config, stage constants
│   ├── knowledge_base.py   # PAN facts
│   ├── embeddings.py       # text -> vector
│   ├── qdrant_store.py     # Qdrant collections + retrieval + session memory
│   ├── llm_service.py      # Groq reasoning
│   ├── voice.py            # Rime TTS
│   ├── validation.py       # PAN/Aadhaar/age checks
│   └── requirements.txt
├── frontend/
│   ├── index.html
│   └── diag.html
├── media/
└── docs/

How to Run

1. Clone the repository

git clone https://github.com/devasheeshrana/sahayak-pan-assistant.git
cd sahayak-pan-assistant/backend

2. Create virtual environment

python -m venv venv
venv\Scripts\activate

3. Install dependencies

pip install -r requirements.txt

4. Add your API keys

copy .env.example .env

Fill in RIME_API_KEY and GROQ_API_KEY in .env.

5. Run the server

python main.py

6. Open it

Go to http://localhost:5000 in Chrome (Web Speech API is Chrome-only).

Screenshots

Greeting screenAnswering a question in English

Answering a question in Hindi

Notes

Qdrant runs in-memory by default, so session memory resets when the server restarts (set QDRANT_URL + QDRANT_API_KEY for a persistent Qdrant Cloud instance)
This is a guidance layer, not an official government service — always double check on onlineservices.nsdl.com or pan.utiitsl.com