Adds your gateway's models to Copilot Chat (ask, edit and agent mode). Every request
uses a fresh credential from your vkm sign-in and carries the labels
(such as customer and project) of the workspace folder you are working in.
Nothing is stored in VS Code.
Requirements
VS Code 1.104 or newer with GitHub Copilot Chat. Your organization's Copilot
policy "Bring Your Own Language Model Key" must allow model provider extensions.
The vkm CLI on the same machine as the extension host (inside the
container or SSH host for remote windows), with a signed-in profile:
vkm login.
A value for every required label in each workspace folder (vkm repo show
must succeed there). While one is missing, requests fail before anything is sent.
See the labels reference.
Settings
vkm.command: how to run the CLI, e.g. ["/path/to/venv/bin/vkm"]
or ["python", "-m", "virtual_key_mapper"]. Default: ["vkm"] from PATH.
vkm.profile: the profile; empty for the CLI's default.
vkm.maxInputTokens, vkm.maxOutputTokens, vkm.imageInput:
what Copilot is told about the models. These are not discovered from the gateway.
Use
Choose a "Virtual Key Mapper" model in the Copilot Chat model picker. The status bar
shows the label values of the current folder in order, joined by · (e.g.
acme · webshop), missing required labels in warning style, and no labels when
none are defined; its tooltip lists each label with its description and source. In
multi-root workspaces the active editor's folder wins; folders with equal label
values count as one. The status bar lets you choose a folder or assign labels:
"Assign labels" opens a terminal running vkm repo bind, which asks for each missing
required label and saves the answers for you only. "Virtual Key Mapper: Sign in"
opens a terminal running vkm login.
Not covered: inline completions, next edit suggestions, embeddings and some agent
actions always use GitHub's own models, which this extension cannot attribute.
