Skip to content

QUICKSTART_VIDEO.md

Latest commit

 

History

History
168 lines (124 loc) · 4.45 KB

QUICKSTART_VIDEO.md

File metadata and controls

168 lines (124 loc) · 4.45 KB

🎬 Quick Start: Video Demo

Want to see Vectro+ in action? Here's how to run the demo:

Option 1: Enhanced Interactive Demo (Recommended)

./demo_enhanced.sh

This script demonstrates:

  • ✨ Streaming compression with progress indicators
  • 🗜️ Quantization (75% size reduction)
  • 🔍 Semantic search across themed data
  • 🌐 Web UI with real-time search
  • 📊 Beautiful terminal output

Perfect for: Screen recording, presentations, video demos

Option 2: Simple Demo

./demo.sh

Classic demo with all core features.

Option 3: Web UI Only

# Generate sample data
python3 scripts/generate_themed_embeddings.py --count 1000 --theme products > products.jsonl

# Compress it
cargo run --release -p vectro_cli -- compress products.jsonl products.bin

# Start server
cargo run --release -p vectro_cli -- serve --port 8080

# Open browser to http://localhost:8080

📹 Recording a Video Demo?

See VIDEO_DEMO.md for:

  • Complete recording guide (5-7 min demo)
  • Script with timestamps
  • Visual tips and best practices
  • Post-production editing guide

🎨 Sample Datasets

Generate realistic themed embeddings:

# Products (electronics, clothing, food, etc.)
python3 scripts/generate_themed_embeddings.py --count 5000 --dim 384 --theme products > products.jsonl

# Movies (by genre)
python3 scripts/generate_themed_embeddings.py --count 5000 --dim 384 --theme movies > movies.jsonl

# Documents (by topic)
python3 scripts/generate_themed_embeddings.py --count 5000 --dim 384 --theme documents > docs.jsonl

# Mixed dataset
python3 scripts/generate_themed_embeddings.py --count 10000 --dim 768 --theme mixed > mixed.jsonl

🚀 Web Server Features

The new web interface includes:

📊 Real-time Dashboard

  • Live embedding count
  • Dimension info
  • Index status

🔍 Interactive Search

  • Visual query input
  • Top-K results
  • Similarity scores
  • Query time metrics

📤 Dataset Management

  • Upload embeddings via UI
  • Load from file
  • REST API for automation

API Endpoints

# Health check
curl http://localhost:8080/health

# Get stats
curl http://localhost:8080/api/stats

# Search (POST JSON)
curl -X POST http://localhost:8080/api/search \
  -H "Content-Type: application/json" \
  -d '{"query": [0.1, 0.2, 0.3, ...], "k": 10}'

# Upload embeddings
curl -X POST http://localhost:8080/api/upload \
  -H "Content-Type: application/json" \
  -d '{"embeddings": [{"id": "doc1", "vector": [...]}, ...]}'

# Load dataset
curl "http://localhost:8080/api/load?path=./dataset.bin"

💡 Demo Tips

For Best Visual Impact:

  1. Use themed data (products, movies) for semantic meaning
  2. Show compression ratio with ls -lh
  3. Highlight query speed (< 1ms)
  4. Demonstrate the web UI for interactivity

Terminal Setup:

# Increase font size
# Set color scheme to dark theme
# Use larger terminal window (1280x720 or 1920x1080)

Talking Points:

  • Streaming compression - No memory limits
  • Quantization - 75% smaller with minimal accuracy loss
  • Fast search - Microsecond query times
  • Beautiful UI - Production-ready web interface
  • Rust - Safety and performance

🎯 Demo Flow (3 minutes)

# 1. Generate data (10 seconds)
python3 scripts/generate_themed_embeddings.py --count 1000 --theme products > data.jsonl

# 2. Compress (15 seconds)
cargo run --release -p vectro_cli -- compress data.jsonl data.bin
cargo run --release -p vectro_cli -- compress data.jsonl data_q.bin --quantize
ls -lh data*.bin  # Show size comparison

# 3. Search (30 seconds)
cargo run --release -p vectro_cli -- search "0.1,0.2,..." --top-k 5 --dataset data.bin

# 4. Web UI (2 minutes)
cargo run --release -p vectro_cli -- serve --port 8080
# Open browser, load dataset, run searches

🎥 Example Video Structure

  1. Intro (15s): "Vectro+ - High-performance embedding search in Rust"
  2. Data (20s): Show generating themed embeddings
  3. Compression (30s): Regular + quantized, show size reduction
  4. Search (45s): CLI search with multiple queries
  5. Web UI (90s): Load data, interactive search, show metrics
  6. Outro (20s): Key benefits, GitHub link

Total: ~3-4 minutes (perfect for social media, demos, presentations)

Ready to demo? Pick an option above and get started! 🚀

For more details, see: