Skip to main content

Roo Code 3.24.0 Release Notes (2025-07-26)

This release introduces Hugging Face provider with open source model support, cross-tool AI coding standards, enhanced terminal security controls, and improved diagnostic management.

Terminal Command Permissions UI​

Managing terminal command permissions is now easier with our new interactive UI (thanks hannesrudolph!) (#5798):

  • Visual Management: See and manage command patterns directly in the chat interface
  • Pattern Suggestions: Get intelligent pattern recommendations based on commands
  • Toggle Controls: Easily switch between allowed and denied states for command patterns

Hugging Face Provider​

We've added support for Hugging Face as a new provider, bringing access to thousands of open source models (thanks TGlide, daniel-lxs!) (#6134, #6190, #6228, #6236):

  • Open Source Models: Access a vast library of community models directly from Hugging Face
  • Flexible Integration: Use models hosted on Hugging Face's infrastructure
  • Easy Configuration: Simple setup process to get started with your preferred models and providers

This opens up Roo Code to the entire Hugging Face ecosystem of open source AI models.

Diagnostic Controls​

Take control of how many diagnostic messages appear in your context with new settings (thanks hannesrudolph!) (#5582):

  • Limit Errors and Warnings: Prevent overwhelming amounts of diagnostics from filling up the model's context window
  • Improved Performance: Reduce slowdowns caused by processing too many diagnostic messages
  • Legacy Code Support: Especially helpful when working with codebases that temporarily have many errors during development

This feature is essential for preserving the context window when working with large or legacy codebases that have many diagnostic messages.

Agent Rules Standard Support​

Roo Code now supports the Agent Rules standard through AGENTS.md files (thanks sgryphon!) (#5969):

  • Cross-Tool Compatibility: Share natural language guidelines across Roo Code, Aider, Cline, and other compatible AI tools
  • Single Source of Truth: Maintain one set of coding standards, security practices, and workflow rules
  • Automatic Detection: Roo Code automatically finds and applies AGENTS.md files in your project

This solves the fragmentation problem where developers using multiple AI tools had to maintain separate configuration files.

QOL Improvements​

  • Apply Diff Guidance: Added efficiency warnings to guide better use of the apply_diff tool (thanks KJ7LNW!) (#6055)
  • Error Boundaries: Better error handling prevents complete UI crashes, showing helpful messages instead (thanks KJ7LNW, elianiva!) (#5085)

Bug Fixes​

  • Todo List Toggle: Fixed the todo list toggle that wasn't responding to clicks (thanks chrarnoldus!) (#6103)
  • Markdown List Styles: Restored proper formatting for ordered and unordered lists in chat (thanks village-way!) (#6095)
  • Ollama URL Handling: Fixed API URL normalization issues with trailing slashes (thanks Naam!) (#6079)
  • Large File Protection: Respects maxReadFileLine setting to prevent context exhaustion (thanks sebinseban!) (#6073)
  • Auto-Approve Safety: Fixed critical issue where auto-approve checkbox became unresponsive (thanks KJ7LNW!) (#6061)
  • Git Checkpoint Warning: Added clear warning when Git is not installed for checkpoints (thanks MuriloFP!) (#5920)
  • Bash Command Parsing: Fixed crashes with complex bash syntax and substitutions (thanks daniel-lxs, KJ7LNW!) (#5743)

Misc Improvements​

  • Merge Resolver Mode: Added internal mode for intelligent Git conflict resolution to improve PR Fixer capabilities (thanks daniel-lxs!) (#6090)