Connect a Rozalia AI proxy deployment to VS
Code's chat and agent model picker.
Rozalia exposes an OpenAI-compatible API in front of self-hosted models
(llama.cpp, Ollama, vLLM, LM Studio, ...). This extension lets you point VS Code
at a Rozalia server and use its models directly from Copilot Chat / agent mode.
Features
Configure a connection — set the server URL and API token once; the token
is stored in VS Code's secret storage, never in settings.
Auto-discovery — models available on your Rozalia deployment are fetched
from GET /v1/models and registered in the chat model picker automatically.
The list is refreshed on startup and then periodically (every 15 minutes by
default, configurable via rozalia.refreshIntervalMinutes), so newly
published models appear without any action on your part.
Streaming responses and tool calling — chat requests are relayed to
Rozalia's /v1/chat/completions endpoint, including agent-mode tool calls.
Getting started
Run Rozalia: Configure Connection from the Command Palette.
Enter your Rozalia server URL (e.g. https://ai.example.org) and an API
token from that account's Tokens page.
Open the chat/agent model picker and pick a Rozalia model.
Use Rozalia: Refresh Models to re-fetch the model list at any time, or the
gear icon next to the Rozalia entry in the model picker to reconfigure the
connection.
