Roo Code 3.23.7 Release Notes (2025-01-13)
This release improves codebase indexing reliability, adds new provider options, and enhances the user interface.
Provider Updates
- Devstral Medium: Added support for Mistral's Devstral Medium model with 131,000 token context window and image support capabilities
- Claude Code: Added configurable maximum output tokens setting (8,000-64,000 tokens) for complex code generation tasks
- Vertex AI: Expanded region configuration to include all available GCP Vertex AI regions for better compliance and access to third-party models like Anthropic Claude (thanks shubhamgupta731!) (#5557)
Bug Fixes
- Context Token Display: Fixed inflated context token count displays by excluding cache tokens from calculation, showing accurate context window usage (thanks daniel-lxs!) (#5603)
- Codebase Indexing Errors: Fixed cryptic error messages when configuring codebase indexing - users now see clear, understandable error messages instead of translation keys (thanks daniel-lxs, nikhil-swamix!) (#5574)
- Embedding Model Switching: Fixed issues when switching between embedding models with different vector dimensions, allowing use of models beyond 1536 dimensions like Google Gemini's text-embedding-004 (thanks daniel-lxs, mkdir700!) (#5565)
- Mermaid Diagrams: Fixed Mermaid diagram rendering issues in architect mode by preventing common syntax errors (thanks MuriloFP, joshmouch!) (#5530)
- UI Button Overlap: Fixed overlapping enhancement and text-to-speech buttons, making both features accessible when text-to-speech is enabled (thanks sensei-woo!) (#5586)
QOL Improvements
- Marketplace Access: Added Marketplace buttons to both MCP and modes tabs for easier discovery and access
- MCP Interface: Improved MCP UI with toggle switches instead of eye buttons for enabling/disabling tools and servers, providing clearer visual feedback
- Codebase Search: Cleaner and more readable codebase search results with improved visual styling and better internationalization
- Error Messages: Friendlier error messages when API configuration profiles aren't compatible with organization settings
- Architect Mode: Intelligently selects the best available tools for context discovery, including semantic search when available
Misc Improvements
- Error Telemetry: Improved debugging capabilities for better reliability and faster issue resolution, with enhanced privacy protection through hashed file paths (thanks daniel-lxs!) (#5595)
Documentation Updates
- Typo Fixes: Improved accuracy and clarity of documentation and code comments across multiple files (thanks noritaka1166!) (#5569)