Merge branch 'feat/aws-provider-support' of https://github.com/xlang-ai/OSWorld into feat/aws-provider-support

ACCOUNT_GUIDELINE.md

@@ -113,6 +113,62 @@ To configure the OAuth2.0 screen for the created GCP. Go to page [OAuth consent
   <img src="assets/authorization.png" width="45%" alt="Authorization">
 </p>
 
+
+## Generating `credentials.json` for Public Eval (Optional)
+
+If you are using **Public Eval** for evaluation, you need to complete the OAuth2 authorization **locally**, and then upload the generated `credentials.json` file to your AWS Host instance.
+
+Please follow the steps below:
+
+1. Run the following Python script on your **local machine** (make sure you have installed `google-auth-oauthlib` and `google-auth`):
+
+   ```python
+   from google_auth_oauthlib.flow import InstalledAppFlow
+   import json
+   
+   SCOPES = ['https://www.googleapis.com/auth/drive']
+   
+   flow = InstalledAppFlow.from_client_secrets_file(
+       'client_secrets.json',
+       scopes=SCOPES
+   )
+   
+   credentials = flow.run_local_server(port=0)
+   
+   with open('credentials.json', 'w') as f:
+       f.write(credentials.to_json())
+   
+   print("OAuth2 credentials have been saved to credentials.json.")
+   ```
+
+   > ⚠️ **Note**: This script will open a browser window for you to log in and authorize access using the configured Google account.
+
+   After successful authorization, a `credentials.json` file will be generated on your local machine.
+
+2. Upload this file to your AWS Host instance and place it at the following path:
+
+   ```
+   OSWorld/evaluation_examples/settings/googledrive/credentials.json
+   ```
+
+After the setup, your directory structure on the AWS Host should look like this:
+
+```
+evaluation_examples/
+├── examples/
+├── examples_windows/
+├── settings/
+    ├── google/
+    │   ├── settings.json
+    │   ├── settings.json.template
+    ├── googledrive/
+    │   ├── client_secrets.json
+    │   ├── credentials.json
+    │   ├── settings.yml
+    └── thunderbird/
+```
+
+
 ## Potential Issues
 
 Due to strict check by Google safety teams, even if we shut down the 2-step verification, Google still detects potential risks of your account, especially __when you frequently change the login device__. You may encounter the following issues:

mm_agents/openai_cua_agent.py

@@ -300,7 +300,7 @@ class OpenAICUAAgent:
         Raises:
             requests.exceptions.RequestException: If the API request fails
         """
-        MAX_RETRIES = 100
+        MAX_RETRIES = 200
         retry_count = 0
         while retry_count < MAX_RETRIES:
             try:
@@ -338,7 +338,7 @@ class OpenAICUAAgent:
                     logger.warning("Unknown message structure, cannot update screenshot")
                 
                 retry_count += 1
-                time.sleep(min(30, 2 ** retry_count))
+                time.sleep(5)
         logger.critical("Max retries exceeded for OpenAI API")
         raise RuntimeError("OpenAI API failed too many times")