Hybrid Search

CrateDB supports full-text indexing and search directly within SQL queries. Use the MATCH() function to find relevant documents, log entries, or descriptions instantly. No need for an external search system.

Example use cases:

• Search sensor logs for specific error messages
• Filter IoT events containing certain keywords
• Combine text search with numerical or temporal filters

Why it matters: You get the power of a search engine, with the simplicity of SQL and the scalability of a distributed database.

CrateDB brings vector similarity search natively into its query engine. Using the KNN_MATCH() function, you can find semantically similar results based on embeddings, ideal for AI and machine learning use cases.

Use cases include:

• Recommendation engines based on user embeddings
• Semantic document or image retrieval
• Intelligent anomaly detection in multidimensional data

Highlights:

• Store and query vector data alongside relational or JSON data
• Combine similarity search with SQL filters for full context
• Query results in milliseconds, even at high dimensionality

What sets CrateDB apart is the ability to combine MATCH() and KNN_MATCH() in the same query, alongside standard SQL filters, joins, or aggregations. This enables complex, hybrid search scenarios where textual, numerical, and semantic dimensions interact seamlessly.

SELECT id, title, score
FROM articles
WHERE MATCH(content, 'renewable energy')
ORDER BY KNN_MATCH(embedding, [0.23, 0.41, 0.67]) LIMIT 10;

This example blends keyword relevance with semantic similarity to surface the most meaningful results; instantly and at scale.

Because CrateDB continuously indexes incoming data, your search results always reflect the latest state, even as new records are ingested. No manual refreshes, no sync lag between systems.

Key benefits:

• Real-time visibility into logs, documents, or sensor data
• Fresh and accurate search results within seconds of ingestion
• Unified query layer for analytics, AI, and discovery

Traditional search systems	CrateDB unified engine
Separate engines for text, vector, and SQL	One system for all search types
Complex data synchronization	Automatic indexing and consistency
Minutes to update search indexes	Query fresh data in milliseconds