desktop-v0.0.14
ClineSummary
This release introduces native macOS notifications and voice input for enhanced background operation and dictation. Key improvements include streaming command output, inline image generation, and a redesigned agent decision card for a more interactive experience. Several bug fixes address issues with command execution, session prompts, and custom base URL support.
New Features
- Added native macOS notifications for task completion or input required, allowing users to leave the app running in the background.
- Introduced voice input for dictating directly into the composer, transcribing speech in real-time using configured providers and models.
- Enabled image generation for models that support it, with images rendering inline within the transcript.
- Added a separate "Cline Code Beta" app for tracking the experimental branch, clearly identified as beta in the UI.
- Introduced support for Gemini custom base URLs, including host-root and proxy values.
Bug Fixes
- Fixed an issue where turns would get stuck on a streaming shimmer without final output, only healing after sending another message.
- Corrected sessions being assigned the Yolo-mode system prompt when auto-approve was on, ensuring auto-approval is an independent tool policy.
- Resolved an issue where `/skill` and `/workflow` commands would dump the entire skill body into the chat.
- Fixed command execution breaking for entire sessions when a model emitted a full command line without separate arguments.
- Ensured restoring a checkpoint now trims the saved transcript to prevent showing reverted file changes.
- Fixed Gemini custom base URLs, including host-root and proxy values, which were previously failing.
- Restored correct LiteLLM input token limits reported by the server, preventing a 128K default override.
- Fixed misaligned columns in the Usage table.
- Fixed routine dialog dropdowns not responding to mouse clicks.
Improvements
- Commands now stream their output into the transcript as they run, preserving terminal colors and allowing scrolling without being yanked to the bottom.
- Added a "Proceed while running" option for long-running commands, allowing the agent to move on while the command finishes.
- Finished agent runs now collapse into a concise summary, keeping the final answer visible.
- Added animations to the opening and closing of reasoning traces and tool rows, respecting reduced-motion settings.
- Redesigned the question card for agent decisions, with explicit option selection, button submission, support for multiple-choice, and keyboard shortcuts.
- Updated the Web search toggle explanation in Settings to clarify provider requirements and show readiness.
- Refreshed assistant markdown rendering, including headings, code blocks with hover copy, and table cards, for a consistent look across desktop and cloud.
- Message hover actions now float over the transcript, improving conversation density.
- Restyled session hover cards for immediate opening, removal of duplicated information, and removal of animation.
- Added a "See More" link to the full usage dashboard from the Usage table.