Generate a Kernel for FlagGems or vLLM project

Generate a Kernel for FlagGems or vLLM project#

Prerequisites#

Before generating a kernel, make sure you read the prerequisites and accomplish the pre-installation steps in this section.

Preinstall FlagGems or vLLM from source:

  • KernelGen Skills support FlagGems project, see the next Preinstall FlagGems section.

  • KernelGen Skills support vLLM project, see vLLM user guide.

Preinstall FlagGems#

For installation information, see FlagGems Documentation. The FlagTree is installed while installing the requirement text file.

Generate a kernel#

Using VSCode (and Copilot), Claude Code, or OpenClaw to generate an operator for the FlagGems or vLLM project follows a similar general process (including connecting to the KernelGen Operator Development MCP Toolkit and load skills) in KernelGen Skills User Guide. You only need to add “Integrate the kernel into FlagGems” additionally to the prompt documented in the “Generate an operator generally” section. KernelGen automatically detects if FlagGems is installed and submits the output files to the project’s experimental directory.