Release v0.1.17
What's Changed
GPT-OSS support! Improved K2 support!
*LLxprt-Specific New Features
Multi-Provider Architecture Enhancements
Profile Save/Load Model Synchronization Fix - Profile saves now correctly capture the current model when switching between providers
Generic Model Parameters Support - All providers (OpenAI, Anthropic, etc.) now support provider-specific model parameters that persist in profiles
Enhanced Provider Switching - Clear ephemeral settings and model parameters when switching providers for cleaner transitions
Tool Dialog for Disabled Tools - New interactive dialog for managing disabled tools via /tools command
Tool Execution Improvements
Unrestricted Parallel Tool Execution Improvements - supporting K2 and Gemini took some changes to how parallel and queues are handled.
Better Tool Error Handling - Improved error reporting and handling for tool execution failures
Finergrained control over tool outputs including new "max-prompt-tokens" and prompt-trimming when exceeded
You can now disable or enable tools selectively and save them in the profile. Some models really shouldn't be given ReadManyFiles. Also K2 has no sense it will try and just read the whole filesystem so the new limits help keep it in line.
Developer Experience
Git Pre-commit Hooks - Automatic code quality enforcement with lint, typecheck, and format checks
Windows Compatibility - Full Windows support for UTF-8 encoding, shell commands, and integration tests
Improved Debug Logging - Debug statements now properly guarded behind DEBUG=1 environment variable
Documentation & Branding
LLXPRT.md Support - Full support for LLXPRT.md memory files (alongside CLAUDE.md compatibility)
Provider Documentation - Added comprehensive documentation for xAI (Grok) provider
Profile System Documentation - Enhanced documentation for settings and profiles
Features from Upstream gemini-cli v0.1.17
Multi-Directory Workspace Support
Configure multiple directories in settings.json - You can now work with multiple project directories simultaneously by configuring them in your settings file
Improved workspace management for complex projects
Shell Command History Search
Reverse search for shell commands - Press Ctrl+R to search through your shell command history, similar to standard terminal behavior
Quickly find and reuse previous commands
Enhanced @ Command Parsing
Better punctuation handling - @ commands now properly handle punctuation marks (commas, periods, etc.)
Example: @file.txt, then do something now correctly parses the file reference
IDE Mode Improvements
New openDiff tool - Compare files side-by-side in your IDE when using IDE mode
Improved error messages for better troubleshooting
Cleaner user experience with flag cleanup
MCP (Model Context Protocol) Enhancements
Connect to MCP servers that only provide prompts (no tools required)
Smarter browser tab management - prevents opening browser when no MCP servers are configured
Bug Fixes
LLxprt-Specific Fixes
Profile Model Synchronization - Fixed issue where saved profiles weren't capturing the provider's current model
Tool Queue Race Conditions - Resolved race conditions in tool scheduling for streaming models
Windows UTF-8 Support - Fixed encoding issues in Windows for shell commands and file operations
Provider Authentication - Fixed auth credentials not loading correctly from profiles
From Upstream
Conversation History Preserved - Chat history now persists when switching authentication methods using /auth
Model Configuration - Model selection can now be properly set from settings.json
Help Command - Fixed /help command flickering and made it respect the clear shortcut (Ctrl+L)
Cloud Shell Logging - Resolved Cloud Shell logging issues
Security & Privacy
DEBUG and DEBUG_MODE environment variables are now excluded from .env files by default
Removed "local modifications" string from bug reports for cleaner submissions
Telemetry re-enabled for local file logging only (no network transmission)
UI/UX Enhancements
Added trailing space after @ command completions for smoother typing
New setting to disable version update notifications
Updated theme documentation with new color keys
Enhanced model selection dialog with alphabetical sorting and search functionality
Other Improvements
Added NOTICES.TXT generation for better license compliance
Fixed self-reference issues in build scripts
Improved CI/CD workflows with Node 24.x support
Documentation updates and typo fixes
Full Changelog: v0.1.16...v0.1.17