QUICKSTART.md

🚀 Antigravity Notebook - Quick Start Guide

Get up and running with Antigravity Notebook in 5 minutes!

Step 1: Prerequisites

Ensure you have:

• ✅ Python 3.9 or higher
• ✅ Docker & Docker Compose
• ✅ CUDA GPU (recommended, 16GB+ VRAM) or CPU (slower)
• ✅ ~20GB free disk space (for model + data)

Step 2: Installation

Clone & Install

# Clone the repository
git clone <your-repo-url>
cd antigravity-notebook

# Install Python dependencies
pip install -r requirements.txt

Configure Environment

# Copy environment template
cp .env.example .env

# (Optional) Edit .env if you want custom settings
# Default settings work out of the box!

Step 3: Start PostgreSQL

# Start PostgreSQL with Docker Compose
docker-compose up -d

# Verify it's running
docker ps

You should see a container named antigravity_postgres running.

Step 4: Initialize Database

# Create database tables
python -m backend.database

You should see: ✅ Database initialized successfully!

Step 5: Start Backend API

# Start the FastAPI backend
python -m backend.main

Wait for:

✅ CLaRa model loaded!
✅ Antigravity Notebook is ready!
📍 API: http://0.0.0.0:8000
📚 Docs: http://0.0.0.0:8000/docs

Note: First startup takes ~5 minutes to download CLaRa-7B model (14GB).

Step 6: Start Frontend UI

Open a new terminal and run:

# Start Streamlit UI
streamlit run frontend/app_notebook.py

Your browser should automatically open to http://localhost:8501

Step 7: Create Your First Notebook

In the Streamlit UI:

1. Create a Notebook

   • Click "Create New Notebook" in the sidebar
   • Name it "My First Notebook"
   • Add a description (optional)
   • Click "Create Notebook"

2. Add a Source

   • Choose one of:
     • PDF: Upload a PDF file
     • URL: Paste a webpage URL (e.g., Wikipedia article)
     • Text: Paste some text content

   Example URL to try: https://en.wikipedia.org/wiki/Artificial_intelligence

3. Wait for Processing

   • The source will be compressed into latent tensors
   • PDF (50 pages): ~30 seconds
   • URL: ~20 seconds
   • Text: ~10 seconds

4. Ask a Question

   • Type a query in the chat box
   • Example: "What are the main points discussed in this document?"
   • Press Enter
   • Wait ~10 seconds for response

5. View Sources

   • Click "Sources" under the response to see which sources were cited
   • Check the Memory Usage gauge to see context utilization

🎉 You're Done!

You now have a working NotebookLM clone!

📚 Next Steps

Try These Examples:

Example 1: Research Assistant

1. Create a notebook called "AI Research"
2. Add these Wikipedia URLs:
   • https://en.wikipedia.org/wiki/Artificial_intelligence
   • https://en.wikipedia.org/wiki/Machine_learning
   • https://en.wikipedia.org/wiki/Deep_learning
3. Ask: "Compare and contrast AI, ML, and Deep Learning"

Example 2: Document Analysis

1. Create a notebook called "Company Docs"
2. Upload 3-5 PDF reports or documents
3. Ask: "Summarize the key findings across all documents"

Example 3: Web Research

1. Create a notebook called "Topic Research"
2. Add 5-10 URLs about a topic you're interested in
3. Ask questions that require synthesizing information across sources

🛠️ Common Issues

Issue: "Database connection failed"

Solution: Ensure PostgreSQL is running

docker-compose up -d
docker ps  # Check if container is running

Issue: "CUDA out of memory"

Solution: Use CPU mode

# Edit .env
DEVICE=cpu

Issue: "Model download is slow"

Solution: Be patient! CLaRa-7B is 14GB. It only downloads once. Check progress at: ./model_cache/

Issue: "PDF extraction failed"

Solution: Ensure PDF has extractable text (not scanned images)

Issue: "URL scraping failed"

Solution: Some websites block scraping. Try a different URL.

🔧 Configuration Tips

For CPU-Only Systems

Edit .env:

DEVICE=cpu

For Limited Memory

Reduce context window in .env:

MAX_CONTEXT_TOKENS=16384  # Half the default

For Production Use

1. Change database password:

   POSTGRES_PASSWORD=<secure-password>
   

2. Set secret key:

   SECRET_KEY=<random-secure-key>
   

3. Configure CORS in backend/main.py:

   allow_origins=["https://your-frontend-domain.com"]
   

📊 Monitoring

Check API Health

curl http://localhost:8000/health

View Storage Stats

curl http://localhost:8000/stats

API Documentation

Open: http://localhost:8000/docs

Try the interactive API explorer!

🧪 Testing the API Directly

Create a Notebook

curl -X POST http://localhost:8000/notebooks/ \
  -H "Content-Type: application/json" \
  -d '{"name": "Test Notebook", "description": "API test"}'

List Notebooks

curl http://localhost:8000/notebooks/

Add Text Source

curl -X POST http://localhost:8000/sources/notebooks/<NOTEBOOK_ID>/sources/text \
  -H "Content-Type: application/json" \
  -d '{
    "title": "Test Document",
    "content": "This is a test document with some content to analyze."
  }'

Query Notebook

curl -X POST http://localhost:8000/chat/notebooks/<NOTEBOOK_ID>/chat \
  -H "Content-Type: application/json" \
  -d '{"query": "What is this document about?"}'

🎓 Learning More

• Read the Plan: See ANTIGRAVITY_PLAN.md for architecture details
• Explore the Code: Check out:
  • backend/services/context_manager.py - The "brain"
  • backend/models/clara.py - CLaRa wrapper
  • backend/services/ingestion.py - Multi-modal processing

💡 Pro Tips

1. Start Small: Begin with 1-2 sources to understand the system
2. Check Memory Usage: Watch the gauge to see when you hit limits
3. Use Descriptive Titles: Makes it easier to understand citations
4. Mix Source Types: PDFs + URLs + Text work great together
5. Ask Synthesis Questions: The AI excels at combining information

🆘 Need Help?

• Documentation: See README.md
• API Docs: http://localhost:8000/docs
• Issues: Open a GitHub issue
• Logs: Check terminal output for debugging

---

Happy NotebookLM-ing! 🚀