Is there an existing integration?
Use Case
Valkey is a high-performance, open-source in-memory data store (community-driven fork of Redis) that includes a native Search module with vector similarity search capabilities. Adding Valkey as a vector store integration would allow MindsDB users to:
- Store and query embeddings directly in Valkey for RAG (Retrieval-Augmented Generation) pipelines
- Perform KNN vector similarity search using COSINE, L2, or Inner Product distance metrics
- Leverage Valkey's sub-millisecond latency for real-time semantic search workloads
- Use metadata filtering alongside vector search for hybrid queries
- Build AI applications backed by a fully open-source vector database with no licensing restrictions
This is particularly valuable for users who already run Valkey/Redis infrastructure and want to avoid adding a separate vector DB to their stack.
Motivation
Imagine a MindsDB user who wants to build a semantic search system. With a Valkey vector store integration, they could:
-- Connect to their Valkey instance
CREATE DATABASE valkey_store
WITH ENGINE = 'valkey',
PARAMETERS = {
"host": "localhost",
"port": 6379
};
-- Store embeddings from a model
CREATE TABLE valkey_store.my_embeddings (
SELECT embeddings, content, metadata FROM model_output
);
-- Query by vector similarity
SELECT * FROM valkey_store.my_embeddings
WHERE search_vector = (SELECT embeddings FROM model WHERE text = 'search query')
LIMIT 10;The integration would implement the VectorStoreHandler interface, supporting full CRUD operations (create_table, insert, select, delete, drop_table) and KNN search via Valkey's FT.CREATE / FT.SEARCH commands.
Implementation
The handler would use the valkey-glide client library (official Valkey client) and support:
- Connection parameters: host, port, password, SSL/TLS options
- Index configuration: vector dimensions, distance metric (COSINE/L2/IP), index algorithm (FLAT/HNSW)
- Operations: Full VectorStoreHandler interface — create_table, insert, select (with KNN), delete, drop_table, get_tables, get_columns
- Filter expressions: Metadata-based filtering combined with vector search
- Data storage: Hash-based document storage with JSON-serialized metadata fields
The handler would be registered as engine type 'valkey' in the MindsDB plugin system.
Anything else?
Is there an existing integration?
Use Case
Valkey is a high-performance, open-source in-memory data store (community-driven fork of Redis) that includes a native Search module with vector similarity search capabilities. Adding Valkey as a vector store integration would allow MindsDB users to:
This is particularly valuable for users who already run Valkey/Redis infrastructure and want to avoid adding a separate vector DB to their stack.
Motivation
Imagine a MindsDB user who wants to build a semantic search system. With a Valkey vector store integration, they could:
The integration would implement the
VectorStoreHandlerinterface, supporting full CRUD operations (create_table, insert, select, delete, drop_table) and KNN search via Valkey'sFT.CREATE/FT.SEARCHcommands.Implementation
The handler would use the
valkey-glideclient library (official Valkey client) and support:The handler would be registered as engine type
'valkey'in the MindsDB plugin system.Anything else?
valkey-glideclient: https://github.com/valkey-io/valkey-glide