LLM Capabilities

Manage the LLM models and reusable AI capabilities (instruction classes) your workspace uses, all from one sidebar panel.

What you can do

• See every instruction class your workspace knows about — role, temperature, and token limit — in one list. Three general-purpose ones ship built in: risk-assessor, validator, and reviewer.
• Add a local or remote instruction class, then edit, delete, or inspect the ones you created.
• See your available models (local and remote) together and mark which one is active.
• Add a local or remote model, activate it, edit it, view its details, or remove it.
• Refresh to re-scan for newly registered classes and models.

Getting started

Open the LLM Capabilities panel from the activity bar. Adding models and capabilities relies on companion extensions — install the ones you need:

• Remote models and remote instruction classes: the AI Core (Service), Instruction Class (Service), and LLM Services extensions.
• Local GGUF or ONNX models: the matching GGUF Runtime or ONNX Runtime extension.
• Inspecting and testing a class: the LLM Capabilities Viewer.

If a companion isn't installed, the action tells you instead of failing silently.