Spice.ai v2.0-rc.2 is now available in Spice Cloud!
v2.0 brings major Spice Cayenne upgrades, a new ADBC Data Connector with BigQuery support, new Catalog Connectors, and a range of developer experience improvements across the Spice Cloud Portal.
Let's get into it.
New in Spice Cloud!
Update Channel for Preview Releases
Spice Cloud users can now select Preview or Stable update channels, with Spice apps automatically upgraded to preview versions (including v2.0-rc.2). Head to Settings -> Runtime to try v2.0 before it's GA!
api.spice.ai Management API
Programmatic control over your entire Spice Cloud footprint (apps, deployments, secrets, API keys, and organization members) via the api.spice.ai REST API and a new Terraform provider for infrastructure-as-code workflows. The API supports OAuth clients for external applications that need to authenticate on behalf of your organization, with granular scopes controlling access per resource type.
Visit the management API docs for more info
Audit Log
Track every change to your Spice Cloud organization - Apps, deployments, configuration updates, and secret management - with a full audit trail of who changed what and when. Available on all plans, with downloadable logs and retention periods of up to 60 days on the Enterprise plan.
Continue reading for the overview of v2.0-rc.2, or check out the full release notes at spiceai.org/releases/v2.0-rc.2
Major v2.0-rc.2 Features
Spice Cayenne Reaches RC
Spice Cayenne, the Spice data accelerator built on the Vortex columnar format, has been promoted to release candidate status. Cayenne delivers 1.4x faster queries than DuckDB with 3x lower memory usage on TPC-H SF100 benchmarks.
New in RC2:
MERGE INTO: Upsert-style insert/update/delete operations on Cayenne catalog tables, fully distributed across executors.PARTITION BY: Define Cayenne table partitioning directly in SQL; partition metadata is persisted and reloaded on restart.- Reliability fixes: Correct target file size enforcement (default 128MB), and proper primary key and conflict resolution handling for Cayenne tables.
ADBC Data Connector & Catalog
Spice now offers an ADBC data connector and catalog for connectivity to any data system with an ADBC Driver, such as BigQuery. It supports full query federation, automatic schema, and table discovery.
New Catalog Connectors
Catalog connectors for PostgreSQL, MySQL, MS SQL, and Snowflake make it straightforward to expose external database metadata through Spice for cross-system federation, using each system's native metadata catalog for schema and table discovery.
Apache DataFusion v52 Upgrade
DataFusion has been upgraded to v52.4.0, with updates across arrow-rs, datafusion-federation, and datafusion-table-providers. This release adds strict overflow handling (try_cast_to now returns errors on overflow rather than silently producing NULL values), along with a federation fix for Inexact filter pushdown with aliases and improved partial aggregation performance for FlightSQLExec.
Additional Updates
- Delta Lake Column Mapping: Name and Id column mapping modes are now supported for Delta Lake datasets.
- HTTP Pagination: The HTTP data connector now supports paginated API endpoints.
- JSON Ingestion: Added support for single-object JSON documents, soda format support (Socrata Open Data), json_pointer extraction for nested payloads, and improved auto-detection across arrays, objects, JSONL, and BOM-prefixed input.
- Per-Model Rate-Limited AI UDF Execution: Control concurrent AI function invocations on a per-model basis.
- Caching Retention: Added retention policies for cached query results.
- DynamoDB DML: INSERT, UPDATE, and DELETE support for the DynamoDB table provider.
- mTLS Client Auth: Added mTLS client authentication to the spice sql REPL.
...and much more! Check out the full release notes for details:
Related Deep Dives
Apache Iceberg at Spice AI: Query, Accelerate, and Write to Open Table Formats
Learn how Spice AI integrates Apache Iceberg for zero-ETL federation, sub-second query acceleration, and ACID-compliant writes to open table formats stored in object storage.
Vortex at Spice AI: The Columnar Format for Data-Intensive Workloads
Spice AI uses the Vortex columnar format in Cayenne to speed data-intensive workloads, reduce memory use, and scale concurrent ingestion and query execution. This blog reviews the implementation and how we've extended Vortex.
Apache Ballista at Spice AI: Distributed Query Execution Without the Operational Tax
This blog explores how we built distributed query execution, replacing Ballista's single-scheduler design with a multi-active HA architecture on top of object storage.
Join the Spice Community on Slack
Interested in discussing the release or learning more about the platform? Connect with fellow Spice users, swap tips, ask questions, and get a front-row seat to product updates and roadmap insights in the Spice Community Slack channel!
