Kabir Dohas

A spiritual wisdom application featuring the timeless verses (dohas) of Sant Kabir, with AI-powered search using local embeddings and LLM synthesis.

Features

• 🔍 Semantic Search - Find relevant dohas using natural language queries
• 🤖 AI Synthesis - Groq LLM provides contextual interpretations
• 🏠 Local Embeddings - Runs Transformers.js locally (no external API needed)
• 📱 Mobile-First Design - Responsive, serene UI
• ⚡ Smart Caching - LRU cache for repeated queries

Tech Stack

• Framework: Next.js 16 (Pages Router + App Router for APIs)
• Embeddings: Transformers.js (multilingual-e5-small)
• LLM: Groq (llama-3.3-70b-versatile)
• Auth: Firebase
• Styling: TailwindCSS v4

Quick Start

# Install dependencies
npm install

# Set up environment variables
cp .env.example .env.local
# Edit .env.local with your API keys

# Run development server
npm run dev

Open http://localhost:3000

Environment Variables

GROQ_API_KEY=your_groq_api_key
FIREBASE_PROJECT_ID=your_project_id
FIREBASE_CLIENT_EMAIL=your_client_email
FIREBASE_PRIVATE_KEY=your_private_key

Project Structure

src/
├── app/api/          # API routes (App Router)
├── pages/            # Page components (Pages Router)
├── components/
│   ├── ui/           # UI Kit (Input, Card, Chip, etc.)
│   ├── Layouts/      # Page layouts
│   └── ...           # Feature components
├── hooks/            # Custom React hooks
├── lib/              # Core logic (embeddings, search, config)
└── types/            # TypeScript definitions

Deploy to Google Cloud Run

Prerequisites

• Google Cloud SDK installed
• A GCP project with billing enabled
• .env.production file with your environment variables

Deployment

# 1. Authenticate with GCP (if not already)
gcloud auth login

# 2. Set your project
gcloud config set project YOUR_PROJECT_ID

# 3. Create .env.production with your secrets
cp .env.example .env.production
# Edit .env.production with your API keys

# 4. Deploy using the provided script
./deploy_cloud_run.sh

The script will:

• Extract environment variables from .env.production
• Build the Docker image with the embedding model pre-loaded
• Deploy to Cloud Run in us-central1

Manual Deployment

If you prefer to deploy manually:

gcloud run deploy kabir-dohas \
  --source . \
  --region us-central1 \
  --allow-unauthenticated \
  --memory 1Gi \
  --port 8080

Documentation

See the docs/ folder for detailed documentation:

• ARCHITECTURE.md - System design
• DEVELOPMENT.md - Development guide
• DECISION_LOG.md - Technical decisions
• CODEBASE_CONTEXT.md - AI-friendly codebase map

License

MIT