feat: add Valkey vector store handler

[daric93]

Description

Closes: #12457

Adds a new Valkey Vector Store Handler to MindsDB, enabling Valkey (with Search module) as a vector database backend. Uses the valkey-glide async client with GLIDE's Batch pipeline API for efficient bulk operations, wrapped in synchronous methods to implement the VectorStoreHandler interface.

Fixes #AEA-484

What's included

• Full CRUD operations (create_table, insert, select, delete, drop_table)
• KNN vector similarity search via Valkey Search (FT.CREATE / FT.SEARCH)
• Filter expressions for metadata-based queries
• Configurable vector dimensions, distance metric (COSINE/L2/IP), and index algorithm (FLAT/HNSW)
• TLS support and configurable request timeout
• GLIDE Batch (pipeline) for efficient batched reads
• Handler registration with MindsDB plugin system

Files Added (9 files, +2,001 lines)

File	Purpose
valkey_handler.py	Core handler implementation
connection_args.py	Connection parameter definitions
__init__.py	Plugin registration
__about__.py	Package metadata
requirements.txt	Dependencies (valkey-glide, numpy)
icon.svg	Handler icon
README.md	Usage documentation with examples
tests/test_valkey_handler.py	Unit + integration tests
tests/__init__.py	Test package marker

Type of change

• ⚡ New feature (non-breaking change which adds functionality)

Verification Process

• Test Location: mindsdb/integrations/handlers/valkey_handler/tests/
• Verification Steps:

1. Start Valkey with Search module:

   docker run -d --name valkey-test -p 6379:6379 valkey/valkey-bundle:9.1

2. Run unit tests (no external deps):

   python3 -m pytest mindsdb/integrations/handlers/valkey_handler/tests/ -v -k "Unit"

3. Run integration tests:

   VALKEY_HOST=localhost VALKEY_PORT=6379 python3 -m pytest mindsdb/integrations/handlers/valkey_handler/tests/ -v -k "Integration"

Test Results

• Unit Tests: 30 passed ✅
• Integration Tests: 15 passed ✅ (against valkey/valkey-bundle:9.1, search module v1.2.0)
• Linting: ruff check + ruff format --check — clean

Configuration

SQL Usage

CREATE DATABASE valkey_store
WITH ENGINE = 'valkey',
PARAMETERS = {
  "host": "localhost",
  "port": 6379,
  "index_algorithm": "HNSW",
  "distance_metric": "COSINE",
  "vector_dimension": 384
};

CREATE TABLE valkey_store.my_embeddings (
  SELECT embeddings, content, metadata FROM model_output
);

Connection Parameters

Parameter	Default	Description
host	localhost	Valkey server hostname
port	6379	Valkey server port
password	None	Authentication password
db	0	Database number
vector_dimension	384	Vector dimension for indexes
distance_metric	COSINE	COSINE, L2, or IP
index_algorithm	HNSW	HNSW or FLAT
prefix	doc:	Key prefix for document hashes
use_tls	False	Enable TLS/SSL
request_timeout	5000	Request timeout in ms

Checklist

• My code follows the style guidelines (PEP 8) of MindsDB.
• I have appropriately commented on my code, especially in complex areas.
• Necessary documentation updates are either made or tracked in issues.
• Relevant unit and integration tests are updated or added.

Related

• No existing files modified — purely additive change

[github-actions]

All contributors have signed the CLA ✍️ ✅
_{Posted by the CLA Assistant Lite bot.}

[daric93]

I have read the CLA Document and I hereby sign the CLA

[daric93]

recheck

[egoriklok]

Public no-secret MCP Buyer-Agent Readiness Snapshot for mindsdb:

• Public signal: Public MCP candidate mindsdb: Query Engine for AI - The only MCP Server you'll ever need. Matched R1 terms: agent, analytics, ci, database, github, mcp, postgres, server.
• R1 fit: agent, analytics, ci, database, github, mcp, postgres, server.
• Readiness status: public evidence review needed before an autonomous buyer-agent should rely on this surface.
• Blind spot 1: explicit auth scopes and delegated-permission boundary.
• Blind spot 2: spend/API cost cap plus approval semantics before paid actions.
• Blind spot 3: receipt, audit-log, revocation, or dispute evidence for safe buyer-agent use.
• Single next question: For mindsdb, is there already a documented policy for agent spend/auth limits, receipt evidence, and revocation before a buyer-agent can invoke it?

No secrets, invoice, payment link, delivery link, private endpoint, paid call, or wallet signature; this is only a free public snapshot.