version/2026.8.2 v2.3.0
high risk
HKUDS/Vibe-TradingSep 9, 2026 → Sep 9, 20260 days between releases
What's New
Adds BigQuery/GitHub support and migrates Google models to Vertex AI
Breaking Changes
- Google models now use Vertex AI, requiring migration for deployments using Google AI Studio credentials
New Features
- SQL Federation Improvements including function translation, NULL handling, and timezone declarations
- BigQuery Federation supporting temporal expressions, recursive CTEs, and window aggregates
- Caching Accelerator Limits for `caching_max_size` and `caching_max_items`
- Stale-While-Revalidate Across Refreshes for cached SQL results
- GitHub Data Connector with review, release, and repository tables
Improvements
- Performance improvements include cache reuse and reduced memory use for cached SQL results
- Broader query federation capabilities
- Expanded data connector capabilities
Bug Fixes
- Fixes for function translation, NULL handling, and timezone declarations in SQL federation
Upgrade Notes
Migrate Google AI Studio credentials to Vertex AI as Google models now authenticate differently. Configure new `caching_max_size` and `caching_max_items` limits for the caching accelerator.
The Two Releases
v2.3.0
Spice v2.3.0 (September 9, 2026)
Sep 9, 2026· 1 month ago
This release introduces broader query federation capabilities, expanded data connectors (specifically for BigQuery and GitHub), and new explicit limits for the caching accelerator. It also migrates Google models to Vertex AI.
Full release notes