v2.84.0

New Features

  • OpenRouter AI Provider: Added comprehensive integration with OpenRouter's unified API, providing access to 100+ AI models from various providers (OpenAI, Anthropic, Google, Meta, Mistral, etc.) with automatic routing and failover capabilities
  • Mistral Codestral 2508 Model: New high-performance code generation model with multi-vendor support through both Mistral and OpenRouter inference providers
  • Enhanced React Test Harness: Improved component linting capabilities and better test execution for React components

Improvements

  • RunView Processing Architecture: Refactored pre/post processing flow to ensure consistent handling of RunView/RunViews calls regardless of entry point, moving processing logic to provider level (SQLServer/GraphQL) for better reliability
  • AI Model Prioritization: Updated priority rankings for GPT-5 class models with OpenAI as preferred provider and Azure as secondary option
  • AI Metadata Cleanup: Fixed model naming conventions (GPT-5 → GPT 5) and improved UI defaults for Flow/Loop agent types
  • OpenAI Provider Flexibility: Enhanced constructor flexibility to support sub-classing for OpenAI-compatible providers

Bug Fixes

  • BaseEntity Conversion: Resolved double conversion issue in RunView post-processing that could cause data inconsistencies
  • AI Model Metadata: Fixed GPT-5 model metadata and effort level mapping for OpenAI models
  • Package Dependencies: Updated package-lock.json with latest dependency versions and security patches