Semantic Search, Zero Configuration Required
We're excited to announce one of our most requested features yet: automatic semantic similarity search across any database column. This capability fundamentally changes how WAII handles natural language queries, making them smarter and more intuitive than ever before.
What's New?
- Automatic Column Detection: WAII now intelligently identifies columns that benefit from similarity search during knowledge graph generation
- Zero-Config Vector Indexing: Automatic indexing of relevant columns in our vector database with no additional setup required
- Smart Query Enhancement: Seamless integration of similarity search into natural language queries
- High-Performance Execution: Federated execution through virtual table functions
Technical Deep Dive
The magic happens through our new WAII_SEMANTIC_SEARCH function, which seamlessly integrates with your existing queries. For example:
SELECT
p.*,
m.matched_text,
m.similarity_score
FROM products p
JOIN TABLE(WAII_SEMANTIC_SEARCH(
'womens shoes',
'category',
'product_id')) m
ON p.product_id = m.matched_id;
Our bespoke compiler can resolve these queries either at runtime or materialize them for fixed values:
SELECT
p.*,
m.matched_text,
m.similarity_score
FROM products p
JOIN (
VALUES
('womens running shoes', 0.92, 'PROD123'),
('womens tennis shoes', 0.85, 'PROD456'),
('womens walking shoes', 0.78, 'PROD789')
) as m(matched_text,
similarity_score,
matched_id)
ON p.product_id = m.matched_id;
Why This Matters
- Enhanced Accuracy: Find the right matches even when users don't use exact database terminology
- Better User Experience: Natural language queries now understand semantic similarity, not just exact matches
- Zero Overhead: No additional configuration or maintenance required
- Universal Applicability: Works across enums, strings, and categorical values
Flexible Configuration Options
While our automatic detection works great out of the box, we understand that different use cases require different approaches. That's why we've also included:
- Manual column selection for fine-grained control
- Customizable similarity thresholds
- Performance optimization options for different data scales
Real-World Impact
From our extensive testing and early user feedback, this feature has proven transformative for:
- E-commerce product search
- Content management systems
- Customer support tools
- Data analytics platforms
This release represents a significant step forward in making text-to-SQL more intelligent and accessible. We're excited to see how our users leverage this capability to build even more powerful applications.
Ready to try it out? Update to Waii v1.22 today and experience the difference for yourself. As always, we're here to help if you have any questions!