AI-powered legal assistant grounded in Indian law. Every response is constrained to the knowledge base or the uploaded document. Zero hallucinations by design.
![Nyay AI Chat Interface]
Nyay AI is a full-stack legal assistant that answers questions about Indian law in plain language. It covers IPC sections, CrPC procedures, tenant rights, employment law, and contract review.
Most people cannot afford a lawyer for everyday legal doubts. Nyay fills that gap by giving clear, grounded answers without hallucinating legal citations that do not exist.
The core design decision: the AI is not allowed to answer from general knowledge alone. Every response is constrained to the legal knowledge base or the specific document the user uploads. If the answer is not there, the system says so.
Ask questions about Indian law in plain language. The system loads the full conversation history from PostgreSQL before every response, so context carries across multiple sessions. You never have to repeat yourself.
Covers:
- Indian Penal Code (IPC) sections and offences
- Code of Criminal Procedure (CrPC)
- Tenant rights and rental disputes
- Employment law and wrongful termination
- Contract review and clause explanation
- FIR generation with correct section references
- IPC section lookup by number or keyword
- PDF export of any consultation
Upload a contract, legal notice, or any PDF. The system extracts the content using PyMuPDF and constrains the AI to answer strictly from that document. General knowledge is explicitly blocked in the prompt.
Use this for:
- Understanding what a rental agreement actually says
- Checking if a job offer letter has unusual clauses
- Reviewing a legal notice before responding to it
The prompt injection for document mode:
Answer only using the content from the document provided below.
Do not use any general knowledge. If the answer is not in the
document, say: "I could not find this in the document."
This matters in a legal context. A hallucinated IPC section number or a fabricated contract clause can cause real harm.
The primary model is LLaMA 3.3 70B via Groq API. If it hits a rate limit, the system silently falls back to LLaMA 3.1 8B, then Gemma 2 9B. The user never sees an error. This was a deliberate production decision built into the Groq client wrapper.
Request arrives
|
v
Try LLaMA 3.3 70B
|
success? --> return response
|
rate limit?
|
v
Try LLaMA 3.1 8B
|
success? --> return response
|
rate limit?
|
v
Try Gemma 2 9B --> return response
The full conversation history is loaded from PostgreSQL before every API call. Most demo chatbots reset on every message. Legal consultations span multiple sessions. Nyay keeps the entire thread intact.
Local development uses Django's default SQLite for zero-setup. Production on Render uses PostgreSQL. The switch is handled entirely via the DATABASE_URL environment variable and dj-database-url, so no code changes are needed between environments.
| Layer | Technology |
|---|---|
| Backend | Python 3.10, Django |
| AI inference | Groq API (LLaMA 3.3 70B + LLaMA 3.1 8B + Gemma 2 9B) |
| Document RAG | PyMuPDF (text extraction from uploaded PDFs) |
| Database | PostgreSQL (production), SQLite (local dev) |
| Frontend | HTML, CSS, JavaScript |
| Auth | Django built-in authentication |
| Deployment | Render |
nyay_ai_legal_advisor/
|
+-- manage.py
+-- requirements.txt
+-- .env # never commit this
+-- .env.example # safe to commit, no real keys
|
+-- nyay_project/
| +-- settings.py
| +-- urls.py
| +-- wsgi.py
|
+-- chat/
+-- models.py # Conversation and Message models
+-- views.py # Chat logic, Groq API calls, fallback chain
+-- urls.py
+-- templates/chat/
+-- index.html # Main chat UI
+-- login.html
+-- signup.html
# Clone
git clone https://github.com/balaji049/nyay_ai_legal_advisor.git
cd nyay_ai_legal_advisor
# Create and activate virtual environment
python -m venv venv
source venv/bin/activate # macOS and Linux
.\venv\Scripts\activate # Windows PowerShell
# Install dependencies
pip install -r requirements.txtCreate a .env file in the root folder:
SECRET_KEY=your_django_secret_key_here
GROQ_API_KEY=gsk_your_groq_key_here
For local dev, SQLite is used automatically. No DATABASE_URL needed.
# Run migrations
python manage.py makemigrations chat
python manage.py migrate
# Start the server
python manage.py runserverOpen http://127.0.0.1:8000 in your browser.
- Go to console.groq.com and sign up (free)
- Click API Keys, then Create API Key
- Copy the key immediately. It starts with
gsk_and is shown only once - Paste it into your
.envfile
- Push the repo to GitHub
- Create a new Web Service on Render, connect the repo
- Set build command:
pip install -r requirements.txt - Set start command:
gunicorn nyay_project.wsgi - Add environment variables in Render dashboard:
SECRET_KEY=your_production_secret_key
GROQ_API_KEY=gsk_your_groq_key
DATABASE_URL=postgresql://... # from Render PostgreSQL add-on
Render's PostgreSQL add-on provides the DATABASE_URL automatically when you attach it to the service.
User message
|
v
Load full conversation history from PostgreSQL
|
v
Document uploaded? -- YES --> PyMuPDF extracts text --> inject as context
-- NO --> use general legal knowledge prompt
|
v
Call Groq API (LLaMA 3.3 70B)
|
v
Rate limit hit? -- YES --> LLaMA 3.1 8B --> Gemma 2 9B
-- NO --> use response
|
v
Save new messages to PostgreSQL
|
v
Render in chat UI
- Hindi, Telugu, and Tamil language support. Most people with legal questions in Telangana are not thinking in English.
- Source highlighting: show the exact sentence in the uploaded document that the answer came from.
- Voice input for accessibility.
- Find a nearby advocate using location and Bar Council data.
Nyay AI provides general legal information, not legal advice. For anything serious, consult a lawyer registered with the Bar Council of India.
Balaji Bhairwad B.Tech, Computer Science and Engineering (AI and ML) Malla Reddy Engineering College, Hyderabad
Built because most people cannot afford a lawyer for every small legal doubt.