0.14.0
Added
- Judge/Evaluation System - Compare model responses with automated judge model evaluation, scoring, and reasoning
- Custom Request Parameters - User-defined request parameters with multi-select support for advanced API configuration
- Usage Tracking with Timing Metrics - Comprehensive performance insights including prompt tokens, cached tokens, and timing data
- Judge Response Management - Delete judge responses from evaluation comparisons
Changed
- Message ID Protocol - Unified to use UUIDs consistently for assistant responses across frontend and backend
- OpenAI API Compatibility - Updated response_format parameter handling (moved to text.format for compatibility)
- Judge Response Format - Enhanced judge evaluation response structure for better display and usability
Fixed
- Custom Parameters UI - Improved width consistency in custom request parameter popup items
- Message ID Handling - Resolved issue where frontend mixed sequential (integer-based) and UUID formats in judge requests
Full Changelog: v0.13.2...v0.14.0