Feature
Workspace provider toggles
Workspace admins can now enable, disable, and prioritize LLM providers per workspace. Configure provider preferences from the workspace settings page.
Changelog
Recent changes to the a21e platform, API, and documentation.
Feature
StylePrint extraction API
New /v1/styleprint endpoint extracts design tokens (colors, fonts, spacing, borders, shadows) from any URL. Extracted tokens can be exported as CSS variables.
Improvement
Adaptive memory health scoring
Memory health scores now factor in time-based decay, contradiction detection, and usage frequency. Stale or conflicting memories are automatically deprioritized.
Breaking
Credit model tier multipliers
Model tier multipliers are now applied to credit costs: Economy 1x, Standard 6x, Premium 25x. Existing plans are unaffected until the next billing cycle.
Improvement
OpenAI shim streaming support
The /v1/chat/completions endpoint now supports streaming responses via server-sent events. Pass stream: true in your request body.
Feature
Prompt Lab rollback
The Prompt Lab now supports rolling back a promoted candidate to the previous version. Use 2024-01/lab.rollback with a reason to revert.
Fix
Fix composite key parsing for xAI provider
Resolved an issue where composite keys containing the xAI provider identifier were not parsed correctly, causing authentication failures.
Fix
Intent clarification timeout extended
Intents with needs_clarification status now remain active for 72 hours (previously 24 hours), giving users more time to respond.