ProtonBase Demo

This repository contains the demo files for ProtonBase (Tacnode), a unified multi-modal database platform that combines relational, JSON, text, geospatial, and vector data types in a single system.

ProtonBase (Tacnode) is a next-generation database platform that eliminates data silos by providing a unified solution for all your data needs:

• Instant Lakehouse - Real-time performance at scale with sub-second analytics
• Online Retrieval - Real-time vector and semantic search with zero-latency updates
• PostgreSQL Compatibility - High compatibility with existing tools and extensions
• Cloud-Native Design - Instant elasticity for dynamic workloads
• Bring AI to Your Data - Extract knowledge and discover insights with AI

ProtonBase supports multiple data types in a single platform:

• Relational data - Traditional structured data with ACID transactions
• JSON data - Flexible schema design for semi-structured data
• Text data - Full-text search with ranking and highlighting
• Geospatial data - Location-based queries and spatial analysis
• Vector data - Similarity search for AI and machine learning applications

• Single Point of Truth - One database for all your data needs
• Simplified Architecture - No need to integrate multiple specialized databases
• Lower Latency - No cross-database joins or data movement
• Consistent Data - No synchronization issues between systems
• Reduced Costs - Fewer systems to maintain and operate

• 01_setup_schema.sql - Creates the database schema and tables
• 02_insert_data.sql - Loads sample property data into the database
• 03_unified_query.sql - Basic unified multi-modal query examples
• 03_unified_query_enhanced.sql - Advanced unified multi-modal query examples
• 04_cleanup.sql - Cleans up the database after the demo

• presentation.md - Slide deck for presenting the ProtonBase demo
• vector_search_explanation.md - Detailed explanation of vector search functionality

• sample_properties.json - Sample property data used in the demo

This demo relies on the PostGIS and pgvector extensions being available on your Postgres/ProtonBase instance:

• CREATE EXTENSION postgis; - for the GEOMETRY column and spatial functions (ST_Distance, ST_DWithin, ST_Contains, etc.)
• CREATE EXTENSION vector; - for the VECTOR column and similarity search (<=> cosine distance operator)

Most managed Postgres-compatible services (including ProtonBase) ship both extensions, but they may need to be explicitly enabled before running 01_setup_schema.sql, which issues both CREATE EXTENSION IF NOT EXISTS statements for you.

1. Install ProtonBase following the instructions at tacnode.io
2. Clone this repository
3. Run the setup script: psql -f scripts/01_setup_schema.sql
4. Load the sample data: psql -f scripts/02_insert_data.sql
5. Run the example queries: psql -f scripts/03_unified_query.sql
6. Try the advanced queries: psql -f scripts/03_unified_query_enhanced.sql
7. Clean up when finished: psql -f scripts/04_cleanup.sql

The demo shows how to create a unified data model that combines multiple data types:

CREATE TABLE property_data.unified_properties (
    -- Relational data
    id SERIAL PRIMARY KEY,
    title VARCHAR(255) NOT NULL,
    price DECIMAL(10, 2) NOT NULL,
    
    -- JSON data for flexible attributes
    amenities JSONB NOT NULL,
    features JSONB NOT NULL,
    
    -- Text data with vector for search
    description TEXT NOT NULL,
    description_tsv TSVECTOR GENERATED ALWAYS AS 
        (to_tsvector('english', description)) STORED,
    
    -- Geospatial data
    location GEOMETRY(POINT, 4326) NOT NULL,
    
    -- Vector data for similarity search
    embedding VECTOR(384) NOT NULL
);

The demo includes examples of multi-modal queries that combine different data types:

SELECT 
    -- Relational data
    p.id, p.title, p.price,
    
    -- JSON data
    p.amenities->'indoor' AS indoor_amenities,
    
    -- Full-text search with highlighting
    ts_headline(p.description, query) AS highlighted_text,
    
    -- Geospatial data
    ST_Distance(p.location::geography, point) / 1609.344 AS miles_from_downtown,
    
    -- Vector similarity
    1 - (p.embedding <=> vector) AS style_match_score
FROM 
    property_data.unified_properties p
WHERE 
    -- Text search condition
    p.description_tsv @@ query
    
    -- JSON condition
    AND p.amenities->'indoor' ? 'Smart Home System'
    
    -- Geospatial condition
    AND ST_DWithin(p.location::geography, point, 16093.4)
    
    -- Vector similarity condition
    AND (p.embedding <=> vector) < 0.5;

The demo shows how to create personalized property recommendations using vector similarity search:

-- Get the reference property the client loves
WITH user_interest AS (
    SELECT id, embedding, amenities, features, location
    FROM property_data.unified_properties
    WHERE id = 1  -- The property the client is viewing
)

SELECT 
    p.id, p.title,
    1 - (p.embedding <=> ui.embedding) AS style_similarity
FROM 
    property_data.unified_properties p,
    user_interest ui
WHERE 
    p.id != ui.id  -- Exclude the reference property
ORDER BY 
    style_similarity DESC
LIMIT 3;  -- Show the top 3 recommendations

• Simplified Architecture - One database instead of multiple specialized systems
• No Data Synchronization - All data types in a single platform
• Lower Infrastructure Costs - Fewer systems to maintain and scale
• Faster Development - No need to integrate multiple databases
• Better User Experience - Faster, more intuitive search and recommendations

ProtonBase (Tacnode) excels in various use cases:

• AI Applications - Bring AI to your data with vector embeddings and similarity search
• Real-time Analytics - Process and analyze data in real-time with sub-second performance
• Personalization - Create personalized experiences based on user preferences and behavior
• Fraud Detection - Identify suspicious patterns across multiple data dimensions
• Content Discovery - Help users find relevant content through multi-modal search

This demo is provided for educational purposes only. ProtonBase (Tacnode) is a commercial product.

For more information about ProtonBase (Tacnode), visit tacnode.io or contact info@tacnode.io.