aaclause
Conflict between OpenAI and AI Content Describer Add-ons on NVDA Beta Version
Conflict between OpenAI and AI Content Describer Add-ons on NVDA Beta Version
Environment
- NVDA Version: alpha-31030,421fe4ef
- OpenAI Add-on Version: 24.02.11
- AI Content Describer Add-on Version: 2023.11.20
- Operating System: Windows 11 23H2 (10.0.22631) workstation AMD64
- Python Version: 3.11.6
- Comtypes Version: 1.2.0
- Configobj Version: 5.1.0 with validate version 1.0.1
Description
When using the OpenAI add-on in conjunction with the AI Content Describer add-on (version 2023.11.20) on the NVDA beta version, an ImportError is encountered. This issue specifically arises when executing a script bound to the gesture 'NVDA+o', which is associated with an object recognition feature in the OpenAI add-on.
Although this really happens when I want to interact with the graphical interface of the complement, that is, by pressing the Nvda more G, NVDA more E and NVDA more or as I mentioned above
The only solution I have found is to uninstall one of both supplements.It is also worth mentioning that in the stable version of NVDA this does not happen.
Error Message
error executing script: <bound method GlobalPlugin.script_recognizeObject of GlobalPlugin ('globalPlugins.openai')> with gesture 'NVDA+o'
Traceback (most recent call last):
File "scriptHandler.pyc", line 295, in executeScript
...
File "C:\Users\angel\AppData\Roaming\nvda\addons\AI Content Describer\globalPlugins\AIContentDescriber\PIL\Image.py", line 100, in
Steps to Reproduce
- Install the OpenAI add-on and the AI Content Describer add-on (version 2023.11.20) on the NVDA beta version.
- Activate the 'NVDA+o' gesture to initiate the object recognition feature in the OpenAI add-on.
- Observe the ImportError in the NVDA log.
Expected Behavior
The OpenAI add-on's object recognition feature should work seamlessly without causing import errors, even when used in conjunction with other add-ons like AI Content Describer.
Actual Behavior
An ImportError occurs, indicating a conflict or an incorrect dependency resolution between the OpenAI and AI Content Describer add-ons, specifically related to the PIL (Python Imaging Library) module.
Possible Solutions
- Investigate the dependency management within both add-ons to ensure they do not interfere with each other's libraries.
- Consider isolating or encapsulating dependencies to prevent such conflicts.
Additional Information
This issue suggests a broader problem with how add-ons manage and resolve their dependencies, particularly when multiple add-ons rely on the same external libraries.
Thank you for looking into this issue.
Hello, Thanks for the report. I'll try to investigate ASAP. However I'm unable to reproduce the issue with NVDA alpha-31154,91f4f145, OpenAI 24.02.11 and AI Content Describer 2023.11.20.
