DocsEditors

Roo Code

Use Roo Code in VS Code with your Osarai GPU through the OpenAI Compatible provider, plus codebase indexing with your own embedding model.

Roo Code is a coding agent for VS Code with modes for coding, planning and debugging. It reaches Osarai through its OpenAI Compatible provider.

Set up

Install Roo Code from the extensions view, open its panel and press the gear icon.

Set API Provider to OpenAI Compatible and fill in:

Under Model Configuration:

Setting	Value for `qwen3.8-27b` on an RTX 5090
Context Window	`131072`
Max Output Tokens	`16384`
Image Support	On only if your instance takes image input (Models)
Input / Output Price	`0`: you pay for the GPU by the hour

To set how long the model thinks, turn on Reasoning effort and pick a level; see reasoning effort.

Save the setup as a configuration profile so you can switch between Osarai and other providers per mode.

Roo Code can index your project for semantic search. Run an embedding model on a second instance (qwen3-embedding-8b) and point the indexer at it:

With two instances running different models, each request goes to the one that serves its model. See Embeddings.

Symptom	Fix
`401`	The key field holds something other than your `qck_…` key
`404`	The Base URL must end with `/v1`
`404 model_not_found`	Two instances run different models and the `model` matches neither: copy the id from the instance page
`503 … not ready yet`	The instance is still starting; wait for Ready on the instance page