bytedance/UI-TARS-desktop v0.1.1

UI-TARS-v0.1.1

on GitHub

We are excited to announce the release of UI-TARS-Desktop v0.1.1!

This update primarily focuses on fixing known issues and improving overall system stability. Below are the key changes:

Key Changes

• Significantly enhanced the stability of Browser Operator functionality.
• Optimized ErrorMessage handling and display, and enhanced the logging system functionality.
• Fixes for various known stability issues.

Highlights

Name Changes

To better align with product positioning, we have renamed Browser Use and Computer Use to Browser Operator and Computer Operator. Additionally, we’ve added detailed explanations on the homepage to help users better understand these features. (#571)

Browser Operator Enhancements

This update greatly improves the stability of Browser Operator.

• Browser Compatibility: We now support Chrome, Edge, and Firefox, along with their sub-versions (Beta, Dev, Canary). The system will sequentially detect local browsers in the order of Chrome → Edge → Firefox, resolving issues where the target browser could not be found. (#537, #541, #547)
• Default Search Engine Configuration: Users can now configure their default search engine, ensuring smoother usage of Browser Operator, even in cases of network issues during initialization. (#553)

• Cross-Platform Shortcuts: Added support for common shortcuts across different OS platforms and browsers (e.g., Select All, Copy, Paste). (#530, #560)
• Screenshot Fixes: Fixed an issue where the browser page would flicker during screenshots. (#551)

ErrorMessage Optimization

We’ve refined the error-handling mechanisms throughout the app:

• Error Classification: Reorganized and detailed error states across different stages of the app, and refactored the GUIAgentError type for clearer issue identification. (#534)
• Log Optimization: Added support for persisting recent history logs, making it easier to troubleshoot and identify past issues. (#548)
• UI Improvements: Enhanced the UI for ErrorMessage, enabling users and developers to locate issues more efficiently. (#571)

Other Updates and Fixes

• Report Enhancements: Reports in HTML format now display model and conversation information. (#574)

• Bug Fixes:
  • Fixed an issue where manually closing the browser in Browser Operator mode prevented relaunching. (#582)
  • Resolved a white-screen issue caused by empty action_type. (#526)
  • Fixed a white-screen issue when unsupported shortcuts were used with action_type. (#560)
  • Fixed a black-screen background issue when closing the app in full-screen mode. (#575)

Thank you for your continued support! 🎉

What's Changed

• release(apps): ui-tars-desktop support UI-TARS-1.5 model by @ZhaoHeh in #519
• fix(document): docs/quick-start.md VLM Base URL by @quicksandznzn in #524
• chore(ci): release pkgs by @ycjcl868 in #525
• fix(ui-tars): handle empty action_type to prevent white page by @skychx in #526
• fix(agent-tars): set highlight div backgroundColor to transparent. by @youngjuning in #500
• fix(ui-tars): add arrow hotkey actions for operators (#528) by @ZhaoHeh in #530
• feat(ui-tars): browser-finder support chrome and edge by @skychx in #537
• chore(ci): update bug_report_ui_tars_desktop.yml by @helio9cn in #540
• refactor(browser): refactor chrome-paths error by @skychx in #541
• fix(agent-tars): implicitly chat session by @knoxnoe in #494
• feat(browser): add firefox-paths and browser-use support firefox by @skychx in #547
• feat(ui-tars): refact log files management by @ZhaoHeh in #548
• fixed page blinking caused by viewport changes when executing screenshot in Puppeteer by @skychx in #551
• fix setting.md by @laoguodong in #550
• feat(ui-tars): format error status and messages by @ZhaoHeh in #534
• docs(readme): add ask deepwiki by @ycjcl868 in #555
• feat(action-parser): add support for format in action parser by @ZhaoHeh in #556
• fix(ui-tars): make error message expandable by @ZhaoHeh in #554
• feat(ui-tars): support costomize use's search enging preference at Br… by @ZhaoHeh in #553
• chore(mcp-client): remote pkg type module by @ycjcl868 in #546
• fix(agent-tars): share reporter not work by @ycjcl868 in #558
• fix(browser): add shortcut key support to the browser. by @skychx in #560
• docs: readme github-trending by @ycjcl868 in #562
• fix(browser-mcp): element index validation to properly handle zero index in browser tools by @falconlee236 in #567
• feat(ui-tars): add operator desc in the welcome page and update the ErrorMessage UI by @skychx in #571
• feat(visualizer): report html show model detail and actions by @skychx in #574
• fix(ui-tars): where closing an Electron window in fullscreen mode leaves a black window by @skychx in #575
• fix(browser): where the browser does not relaunch after being manually closed by @skychx in #582

New Contributors

• @quicksandznzn made their first contribution in #524
• @laoguodong made their first contribution in #550
• @falconlee236 made their first contribution in #567

Full Changelog: v0.1.0...v0.1.1