The following process and specifications need to be followed to develop plug-ins for G-Assist:
Development readiness:
- Visit the official NVIDIA GitHub repository forPlug-in Development Kits (PDK)
- Ensure that you have a Python 3.10+ environment and basic JSON knowledge
- Read the documentation for API interface specifications and data format requirements
Core development steps:
- Generated using tools such as ChatGPTPlugin description file (.json)Defining Functional Metadata
- compilePython Execution ScriptRealization of specific logic
- Verify functional stability in a local test environment
- Place the file package into the G-Assist plugin directory (default path is Program Files/NVIDIA/plugins)
- Restart NVIDIA App to load new plugin
Caveats:
- Plugin permissions are limited to a local sandboxed environment with no access to sensitive system areas
- The size of each plug-in should be limited to 10MB or less
- Dependency libraries need to be explicitly declared and labeled in the documentation.
- Complex plug-ins are recommended to provide sample commands for easy invocation by users
Typical application scenarios include: live streaming status query, hardware Benchmark test, specific game macro commands and so on. Successful cases include Twitch live broadcast detection plug-in, which can query the online status of the anchor by voice.
This answer comes from the articleProject G-Assist: an AI assistant that uses voice and text to optimize computer performanceThe
May not be reproduced without permission:AI productivity tools " How do developers create custom plugins for G-Assist? What are the do's and don'ts?
English 
简体中文 
日本語 
Deutsch 
Português do Brasil