Spice Cloud Update: August 2026

August was a busy month across Spice Cloud and the Spice runtime.

The Spice.ai v2.3.0 runtime is now available in the Preview Channel and includes Cloud Connect to bring self-hosted runtimes into Spice Cloud for BYOC management, so data and compute stay on your infrastructure. High-throughput replication is now available for MySQL tables and PostgreSQL catalogs, and any Debezium source plugin can now send changes directly to Spice. And vector and full-text searches are now faster with a warm in-memory tier.

Let's get into it!

New in Spice Cloud

Use case onboarding templates

Spice Cloud template selection screen Figure 1: Spice Cloud project templates.

New projects in Spice Cloud can now start from a template, including Real-Time Analytics Replicas, AI Data Warehouses, and Datalake Accelerators. Have a project deployed and queryable in a few clicks, or start from scratch and configure it yourself.

Deploy an analytics replica in Spice Cloud

Spice Cloud analytics replica configuration screen Figure 2: Analytics replica setup.

Connect your PostgreSQL, MySQL, MongoDB, or DynamoDB instance, choose which tables to replicate, and an isolated, real-time analytics replica deploys in minutes. High-throughput replication applies every committed insert, update, and delete from the source's change log into a Spice acceleration for fast, secure analytical query.

August Runtime Updates

Spice Cloud Connect (BYOC)

Spice Cloud standalone runtime instances screen Figure 3: Standalone runtimes for BYOC.

Self-hosted runtimes can now connect to Spice Cloud for BYOC (Bring Your Own Cloud) management and observability. Run spice cloud link to enroll a standalone runtime.

Spice Cloud Connect runtime enrollment form Figure 4: Cloud Connect enrollment.

Get started with the Cloud Connect on a Development Machine recipe. BYOC support for Kubernetes clusters through the Spice Kubernetes Operator will be released in September.

Spice Cloud Connect connected runtime dashboard Figure 5: Cloud Connect dashboard.

Cloud Connect docs.

MySQL high-throughput replication

Spice MySQL refresh mode configuration Figure 6: MySQL CDC configuration.

The MySQL Data Connector now supports real-time replication; set refresh_mode: changes and the runtime loads an initial snapshot of the table and then applies inserts, updates, and deletes from the MySQL binlog.

MySQL Binlog Replication docs.

PostgreSQL catalog high-throughput replication

Spice PostgreSQL catalog refresh mode configuration Figure 7: PostgreSQL catalog CDC configuration.

The PostgreSQL Catalog Connector is now beta and can stream changes by setting refresh_mode: changes. One configuration replicates every table that the include patterns match; queries then read fresh PostgreSQL data with no per-table setup. Catalog CDC acceleration support is at Alpha.

PostgreSQL Catalog Logical WAL Replication docs.

Debezium CDC without Kafka

Spice Debezium change event ingestion screen Figure 8: Debezium CDC.

Any Debezium source plugin can now stream change events directly into Spice, with no Kafka bus in between. A dataset with from: cdc: ... accepts change events at POST /v1/datasets/{name}/cdc in JSON or Avro.

Debezium docs.

Faster search

Spice search results with vector and full-text indexes Figure 9: Hybrid search in Spice.

Support for a warm in-memory tier for vector and full-text indexes is now available. The runtime writes each change to the warm tier and to the durable store together; searches read the warm tier first with fall back to the durable store.

Search docs.

Additional runtime updates

Search improvements. cosine_distance uses SIMD instructions, full-text search applies stemming by default, and full-text search pushes SQL filters down into the Tantivy index.
Operations and observability. Query timeout improvements, trace IDs added to logs, improved Grafana and Datadog dashboards, and dynamic Kubernetes CPU sizing.
Data Connectors. Fixes for Oracle DATE, Snowflake NUMBER precision, and catalog exclude patterns. SQL Server Kerberos authentication now works on Unix.
SDK updates: spice.js v3.2.0 adds query cancellation, mTLS, HTTP fallback, and typed parameter binding; spicepy v4.0.0 adds streaming, NSQL, search, and query cancellation; spice-rs v3.2.0 adds search, NSQL, async queries, query cancellation, and mTLS.

Check out the release notes for more details.

From the Spice blog

The Analytics Replica Pattern: The Shortest Path to Data-Grounded AI

Analytics replica architecture diagram Figure 10: Simplify real-time analytics on operational data.

We've mentioned analytics replicas a few times in this update. In this blog, learn how the analytics replica pattern keeps operational systems isolated while delivering real-time analytical queries for AI agents and dashboards.

Read the analytics replica blog.

Migrating Off Dremio: A Phased Guide for Data Teams

Dremio and Spice migration concepts Figure 11: Dremio to Spice migration guide.

This guide walks through a phased migration off Dremio with no forced cutover: run Spice side-by-side federating through Dremio's Arrow Flight endpoint, move datasets and Reflections one at a time, and then decommission. The guide includes a Dremio -> Spice concept map and a function mapping table for Dremio-specific SQL.

Read the Dremio migration guide.

That's it for now; we'll be back at the end of September with the next Spice Cloud Update. Join us in the Spice Community Slack to follow along with the latest news or if you have any questions!