- Refactored logging configuration to support dynamic log levels via command-line arguments, allowing for better control over log verbosity.
- Introduced a new signal handler for graceful shutdown of environments and processes, improving robustness during termination.
- Added functionality to save command-line arguments to a JSON file for better traceability of execution parameters.
- Maintained existing logic while enhancing the overall structure and error handling capabilities of the script.
- Added a new update entry for the introduction of **OSWorld-Verified** highlighting major updates and community fixes.
- Corrected the spelling of "VirtualBox" in the environment refactor entry.
- Enhanced clarity in the Docker section title for better readability.
- Introduced a new script `run_multienv_o3.py` to facilitate end-to-end evaluation across multiple environments.
- Implemented command-line argument parsing for various configurations including environment settings, logging levels, and AWS parameters.
- Integrated signal handling for graceful shutdown of environments and processes.
- Enhanced logging capabilities for better traceability during execution.
- Maintained existing logic from previous scripts while introducing new functionalities for improved evaluation processes.
- Modified the Flask application to allow the port to be set via the `FLASK_PORT` environment variable, defaulting to 8080 if not specified.
- Ensured existing application logic remains unchanged while enhancing configurability for deployment environments.
- Introduced `--client_password` argument in `run_multienv_aguvis.py`, `run_multienv_claude.py`, and `run_multienv_gta1.py` for enhanced security and flexibility.
- Updated agent classes (`PromptAgent`, `AguvisAgent`, `GTA1Agent`) to accept and utilize `client_password` for improved configuration.
- Modified evaluation guidelines to reflect the new client password requirement.
- Ensured existing logic remains intact while enhancing functionality for better user experience.
- Added checks for the presence of "toc.ncx" and "content.opf" in the EPUB file before attempting to process them.
- Introduced debug logging to notify when these files are not found, enhancing error handling and traceability.
- Maintained existing logic while improving robustness of the EPUB processing function.
- Added DummyAgent class to facilitate coordinate generation and action assignment.
- Updated GTA1Agent to utilize DummyAgent for improved planning and execution.
- Increased max_steps and N_SEQ parameters for better performance.
- Enhanced logging for planning and execution processes.
- Maintained existing logic while integrating new functionality.
- Refactor _fix_pyautogui_less_than_bug to improve handling of press('<') and typewrite calls.
- Introduce Unicode escape decoding for typewrite content to ensure proper '<' character processing.
- Maintain existing logic while enhancing functionality for better compatibility.
- Fix delete_vm() method to accept region and **kwargs parameters
- Fix occupy_vm() method to accept pid, region and **kwargs parameters
- Ensures consistency with base VMManager interface and other providers
- Resolves runtime argument mismatch errors when calling these methods
This maintains backward compatibility while fixing the interface contract.
Enables GPT models to use custom API endpoints through OPENAI_BASE_URL environment variable. This addresses the limitation where only Azure OpenAI supported custom endpoints while standard GPT models were hardcoded to api.openai.com.
- Add intelligent URL handling to avoid duplicate /v1 paths
- Maintain backward compatibility with default OpenAI API
- Update README with configuration instructions
- Non-breaking change preserving existing functionality
Fixes API integration issues for users with custom OpenAI-compatible services.
- Add command-line argument --provider_name to support flexible provider selection
- Default provider remains vmware for backward compatibility
- Fix Docker provider controller initialization issue with delayed setup
- Add safety checks for controller existence in error handling
This enables users to specify different virtualization providers directly
from the command line and resolves Docker container lifecycle issues.
- Introduced signal handling for graceful shutdown of environments and processes.
- Enhanced logging configuration to support dynamic log levels and structured output.
- Updated argument parsing to include new parameters for model selection and task execution.
- Refactored task distribution logic to streamline environment task management.
- Improved error handling during task execution and environment cleanup.
- Adjusted Qwen25VLAgent initialization to support new model and thought prefix options.
- Reduced max tries for LLM calls to optimize performance.
Add support for HuggingFace mirror (hf-mirror.com) to improve download
speeds in regions where huggingface.co access is slow.
- Support HF_ENDPOINT environment variable detection
- Automatically switch to hf-mirror.com when HF_ENDPOINT is set
- Apply to Docker, VMware, and VirtualBox providers
- Maintain backward compatibility with default huggingface.co URLs
Users can now set HF_ENDPOINT=https://hf-mirror.com to use the mirror.
* ver Jun17th
updating annotations
* ver Jun17th
corrected annotation of 1d17
added check for cell merge
* ver Jun17th
updated several annotations
* ver Jun20th
fixed set-up config of 2bd59342-0664-4ccb-ba87-79379096cc08
* fix: Enhance instructions in LibreOffice Calc examples for clarity and specificity, including details on using Pivot Tables, column placements, and revenue calculations.
* ver Jun21st
updating calc evals
* ver Jun22nd
fixed an impress task
* ver Jun22ndv2
adjusted several calc tasks
* Clean scalfolds
* ver Jul18th
added two try-excepts to handle possible formula parsing and calculation
failures
* ver Jul19th
added supports for cellIs and some other new types of conditional
formatting for calc evaluation
---------
Co-authored-by: BowenBryanWang <bryanwang.nlp@connect.hku.hk>
Co-authored-by: yuanmengqi <yuanmengqi@mail.ustc.edu.cn>
- Introduced a new function `read_x11_image` to read and convert X11 (XWD) format images to PIL Image, supporting both 24-bit and 32-bit formats.
- Enhanced the `compare_image_text` function to include checks for X11 image formats, with multiple conversion attempts using PIL, a custom reader, and netpbm tools.
- Improved error handling and logging for OCR processing, providing detailed feedback on conversion attempts and potential issues with X11 images.
- Maintained existing logic while expanding functionality for better image processing reliability.
- Introduced a new function `safe_open_image_with_retry` to handle image file opening with retries for truncated or corrupted files.
- Enhanced error handling and logging for image processing in `check_palette_and_structure_sim`.
- Updated the logic to safely open both source and target images, ensuring robust evaluation without altering existing functionality.
These changes improve the reliability of image handling in the GIMP evaluator while maintaining the original code logic.
- Added a new function to ensure URLs have a scheme, defaulting to 'http://' if missing.
- Integrated tldextract to normalize URLs by extracting domain parts and handling 'www' subdomains.
- Updated the compare_urls function to include logging for better traceability during URL comparisons.
- Added tldextract to requirements.txt to support the new functionality.
- Updated the AWS manager with a new AMI ID for the specified resolution.
- Modified Chrome desktop launcher to include --remote-debugging-port=1337 for GUI debugging support.
These changes improve the robustness of URL handling and enable consistent Chrome debugging capabilities without altering existing logic.
- Updated the timeout for the page load state from 10 seconds to 60 seconds to ensure better stability during page processing.
- Removed redundant retry mechanisms from the active tab checks to streamline the code while maintaining existing functionality.
- Enhanced logging to provide clearer insights into the page loading process.
These changes aim to improve the reliability of the Chrome evaluator without altering the core logic.
* ver Jun17th
updating annotations
* ver Jun17th
corrected annotation of 1d17
added check for cell merge
* ver Jun17th
updated several annotations
* ver Jun20th
fixed set-up config of 2bd59342-0664-4ccb-ba87-79379096cc08
* fix: Enhance instructions in LibreOffice Calc examples for clarity and specificity, including details on using Pivot Tables, column placements, and revenue calculations.
* ver Jun21st
updating calc evals
* ver Jun22nd
fixed an impress task
* ver Jun22ndv2
adjusted several calc tasks
* Clean scalfolds
* ver Jul18th
added two try-excepts to handle possible formula parsing and calculation
failures
---------
Co-authored-by: BowenBryanWang <bryanwang.nlp@connect.hku.hk>
Co-authored-by: yuanmengqi <yuanmengqi@mail.ustc.edu.cn>
* feat&style: add task status configuration and clear cache functionality; enhance UI styles
* feat&refactor: enhance current configuration API and improve cache clearing logic
* refactor&style: simplify task status update logic and improve page refresh mechanism
* refactor&feat: streamline default configuration retrieval and enhance cache initialization logic
* feat&refactor: add caching to default configuration retrieval and streamline task status logic
* feat&style: add collapsible section for additional model parameters and enhance styling for config items
* refactor&style: remove floating action button and clean up related styles
* fix: update video and screenshot sources to include action space, observation type, and model name parameters
