Windows-Use v0.6.8

Fixes

• Removed the Informative nodes which consuming more tokens
• Handled the dialog window part correctly
• Fixed the locking of the foreground window after switching the windows

Windows-Use 0.6.6.7

Migration from:

• langchain
• langgraph

Used to improve speed of operation

Windows-Use v0.6.5

Features

• Unified the App-related tools into a single tool called App Tool
• Included support to access the element via label
• Add support to switch between apps
• Added support for multi-select and multi-edit via tools
• Added support to open-router (Thanks @mrtfff )

Fixes

• Fixed the issue of the taskbar not being recognised in a multi-screen display.
• Simplified the switching between apps.
• IOU based bounding box to trim the ends of the UI elements that are offscreen.
• Injected the high-priority focus to the dialogue windows while traversing.

🪟Windows-Use v0.6.1

Features

• Auto-minimise the VS Code/IDE after submitting the query to the agent.

Fixes

• Fixed a bug while pruning the tree that prevented the app from minimising/closing/maximising.

pip install windows-use==0.6.1

🪟Windows-Use v0.6.0

Features

• Added Resize Tool to the list quickly resize any applications.
• Added the default language of the system added to the system prompt.
• Shifted from native agent loop to agent loop using langgraph for more organised.

Fixes

• Resolved the issues of the chat history growing rapidly and filling context window.
• Updated the system prompt.
• Fixed the issue while sending image to the vllms (Thanks @nauyiahc)

pip install windows-use==0.6.0

🪟Windows-Use v0.5.0

Feature

• Applied the updated Grounding from windows-mcp
• Set default browser for the agent
• Random point generation within the bounding box.

Fixes

• #2 Added comprehensive unit testing @richard-devbot
• Removed relevant elements from the context window of the agent
• Proper separation between GUI and DOM

🪟Windows-Use v0.3.5

Features:

• Better Grounding of the UI elements using Ally tree.
• Fixed the issues when it comes to clicking certain buttons like on notepad Save As, Save, ..etc
• Improved the speed of grounding to 1.7- 2.5s
• Added web scraping capability to the agent.
• Launching apps on the start menu without visiting the start menu.
• Fixed issues in scrolling specific sections of a window or the webpage.
• Support for both VLLMs and LLMs.
• Added human-like cursor movement.

Install the package using pip install windows-use

🪟Windows-Use v0.2.0

Features

• Added new tool (Switch Tool) to switch between opened apps
• Added watch_cursor functionality to live inspect elements under the cursor.
• Updated the existing system prompt with more instruction-oriented.
• Added print_response method to show the result in markdown in the console.