Skip to main content

Roo Code 3.10 Release Notes (2025-03-25)

This release cycle focused on enhancing context management with partial file reads and suggested responses, improved OpenRouter integration, and various fixes for stability and user experience.

Enhanced Context Management

  • Reading Large Files in Chunks: Roo Code can now read large files in chunks, allowing for more efficient processing and context utilization. (thanks samhvw8!)
  • Suggested Responses to Questions: When Roo asks a question, it now provides suggested responses to streamline user interaction. (thanks samhvw8!)
  • More Consistent @-mention Lookups: Improved lookup consistency for files and folders using @-mentions.
  • Consolidated Code Actions: Code actions are now grouped into a submenu for a cleaner interface. (thanks samhvw8!)

Provider & API Enhancements

  • Dynamically Fetched Instructions: Instructions for creating/editing custom modes and MCP servers are now fetched dynamically. (thanks diarmidmackenzie!)
  • Gemini 2.5 Pro Added: The Gemini 2.5 Pro model has been added to the Google Gemini provider. (thanks samsilveira!)
  • OpenRouter Token Tracking: Switched to a more direct method of tracking OpenRouter token usage and spend.
  • Control Over Partial Read Chunk Size: Users now have more control over the chunk size for partial file reads.
  • R1 Support for OpenAI Compatible: Added an R1 support checkbox to the OpenAI compatible provider for QWQ. (thanks teddyOOXX!)
  • Bedrock Application Inference Profile: Added Bedrock support for application-inference-profile. (thanks maekawataiki!)
  • Task Created Event in API: Added a taskCreated event to the API. (thanks wkordalski!)

QOL Improvements

  • Updated Welcome Page: The welcome page now provides 1-click OAuth flows with LLM routers. (thanks dtrugman!)
  • New Task Command: Added a "New Task" command in the Command Palette. (thanks qdaxb!)
  • UX for Chat Text Area: Updated the user experience for the chat text area. (thanks chadgauth!)
  • Custom Storage Path for Tasks: Supported a custom storage path for tasks. (thanks Chenjiayuan195!)

Bug Fixes

  • Fixed MCP error logging (thanks aheizi!).
  • Improved search_files tool formatting and logic (thanks KJ7LNW!).
  • Fixed changelog formatting in GitHub Releases (thanks pdecat!).
  • Fixed a bug causing task history to be lost when using WSL.
  • Fixed issues where questions and suggestions weren’t showing for non-streaming models or were hard to read.
  • Fixed opacity of drop-down menus in settings (thanks KJ7LNW!).
  • Fixed bugs with reading and mentioning binary files like PDFs.
  • Fixed pricing information for OpenRouter free models (thanks Jdo300!).
  • Fixed an issue with unit tests on Windows (thanks diarmidmackenzie!).
  • Fixed a maxTokens issue for the Outbound provider (thanks pugazhendhi-m!).
  • Fixed a line number issue with partial file reads (thanks samhvw8!).
  • Fixed context mentions on Windows.
  • Fixed German translations (thanks cannuri!).
  • Fixed telemetry banner internationalization.
  • Corrected max output tokens for Sonnet 3.7 non-thinking models.
  • Fixed logic for firing task completion events.
  • Updated value of max tokens for gemini-2.5-pro-03-25 (thanks linegel!).

Misc Improvements

  • Added fake provider for integration tests (thanks franekp!).
  • Reflected Cross-region inference option in ap-xx region (thanks Yoshino-Yukitaro!).
  • Made partial file reads backwards-compatible with custom system prompts.
  • A variety of fixes and improvements to experimental multi-block diff (thanks KJ7LNW!).
  • Renamed and migrated global MCP and modes files (thanks StevenTCramer!).
  • Added watchPaths option to McpHub for file change detection (thanks 01Rian!).
  • Enabled reading of image responses from MCP calls (thanks nevermorec!).
  • Fixed numeric formatting suffix internationalization (thanks feifei325!).
  • Fixed open tab support in context mention suggestions (thanks aheizi!).
  • Better display of OpenRouter “overloaded” error messages.
  • Fixed browser tool visibility in system prompt preview (thanks cannuri!).
  • Fixed supportsPromptCache value for OpenAI models (thanks PeterDaveHello!).
  • Fixed readme links to docs (thanks kvokka!).
  • Ran ‘npm audit fix’ on libraries.
  • Made suggested responses optional to not break overridden system prompts.