Skip to content

SETUP.md

Latest commit

 

History

History
143 lines (101 loc) · 3.13 KB

SETUP.md

File metadata and controls

143 lines (101 loc) · 3.13 KB

🛠️ SaralPolicy Setup Guide

This guide covers the complete setup process for SaralPolicy, including environment configuration, dependency installation, and local AI setup.

📋 Prerequisites

  • OS: Windows, Linux, or macOS
  • Python: 3.9 or higher
  • RAM: Minimum 8GB (16GB recommended for optimal performance)
  • Disk Space: ~10GB (for models and virtual environment)

🚀 Quick Start

1. Clone Repository

git clone https://github.com/VIKAS9793/SaralPolicy.git
cd SaralPolicy

2. Environment Setup

Create and activate a virtual environment to keep dependencies isolated.

Windows:

python -m venv venv
venv\Scripts\activate

Linux/Mac:

python3 -m venv venv
source venv/bin/activate

3. Install Dependencies

cd backend
pip install -r requirements.txt

4. Local AI Setup (Ollama)

SaralPolicy uses Ollama for privacy-first, local AI processing.

Detailed Guide: See Ollama Setup Guide for full instructions.

Summary:

  1. Download Ollama
  2. Pull the required model: 
    ollama pull gemma2:2b
  3. Pull the embedding model: 
    ollama pull nomic-embed-text
  4. Start the server (keep running in background): 
    ollama serve

5. Environment Configuration (Optional)

For high-quality Hindi TTS using Indic Parler-TTS:

# Copy environment template
copy .env.example .env  # Windows
# cp .env.example .env  # Linux/Mac

Edit backend/.env and add your HuggingFace token:

HF_TOKEN=hf_your_token_here

Get a token from: https://huggingface.co/settings/tokens

Note: This is optional. Without the token, TTS falls back to gTTS (instant but lower quality).

6. Initialize Knowledge Base

Index the IRDAI regulatory documents into the local vector database.

# Ensure you are in the 'backend' directory
python scripts/index_irdai_knowledge.py

7. Run Application

python main.py

Visit http://localhost:8000 in your browser.

🔊 Optional: High-Quality Hindi TTS

SaralPolicy supports Indic Parler-TTS for high-quality Hindi speech synthesis.

Requirements

  • HuggingFace token (free)
  • ~4GB RAM during inference
  • ~2GB disk for model cache

Setup

  1. Get token from https://huggingface.co/settings/tokens
  2. Add to backend/.env: 
    HF_TOKEN=hf_your_token_here

Performance Notes

  • CPU: 2-5 minutes per generation (expected)
  • GPU (CUDA): 5-15 seconds per generation
  • Falls back to gTTS automatically if unavailable

🧪 Verifying Installation

To ensure everything is working correctly, you can run the integration tests:

python -m pytest tests/test_rag_citations.py
python -m pytest tests/test_translation_offline.py

📝 Configuration

Configuration is managed via environment variables and main.py constants.

  • Port: Default 8000. Change in main.py.
  • Allowed Origins: Configured in app/dependencies.py.
  • Ollama Host: Default localhost:11434.

See System Architecture for deeper technical details.