- Added a static method to validate image responses for PNG and JPEG formats using magic bytes.
- Improved error handling in the get_screenshot method to log invalid payloads and retry attempts.
- Updated the requests call to include a timeout for better reliability.
- Included a new section to acknowledge institutions and students who contributed feedback and participated in fixes.
- Enhanced recognition of collaborative efforts in the project while maintaining the existing structure of the README.
- Updated the proxy configuration section to specify that some tasks may require proxy settings to function properly, depending on website defenses.
- Enhanced user guidance by clarifying the importance of proper proxy configuration for task execution.
- Maintained existing content while improving clarity and user understanding of configuration requirements.
- Added a section detailing essential configuration requirements for Google Account Tasks and proxy settings.
- Highlighted the impact of missing configurations on task execution and evaluation scores.
- Maintained existing content while enhancing user guidance and clarity in setup instructions.
- Updated the AWS support section to emphasize the benefits of using cloud services for parallel evaluation, including potential time reductions.
- Improved clarity in the username and password information for virtual machines, ensuring security measures are highlighted.
- Maintained existing content while enhancing the overall readability and user guidance in the documentation.
- Added a new section for Local Evaluation, clarifying the import process for `run_multienv.py`.
- Introduced a Public Evaluation section detailing the process for verifying results on the leaderboard and requirements for sharing agent implementations.
- Included links to the Public Evaluation Guideline for user reference.
- Maintained existing content while enhancing clarity and providing additional resources for users.
- Expanded the OSWorld-Verified update entry to include new model results and a comparison with previous benchmarks.
- Added a new section on AWS support, detailing the benefits of using cloud services for parallel evaluation and providing links to setup guides.
- Corrected the baseline agent command example to reflect the updated model name and added a new example for parallel execution.
- Clarified the username and password information for virtual machines, emphasizing security measures for cloud services.
- Maintained existing content while enhancing clarity and providing additional resources for users.
- 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.
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.
docs: add note about cleaning up residual docker containers
Add note in README about cleaning up residual docker containers after abnormal experiment interruption to prevent performance issues
* Initailize aws support
* Add README for the VM server
* Refactor OSWorld for supporting more cloud services.
* Initialize vmware and aws implementation v1, waiting for verification
* Initlize files for azure, gcp and virtualbox support
* Debug on the VMware provider
* Fix on aws interface mapping
* Fix instance type
* Refactor
* Clean
* hk region; debug
* Fix lock
* Remove print
* Remove key_name requirements when allocating aws vm
* Clean README
---------
Co-authored-by: XinyuanWangCS <xywang626@gmail.com>
