DurinDoorRoo
Point Roo at DurinDoor with the Ollama provider and a /v1 base URL.
Use the Roo guide card with a running gateway, a DurinDoor key, and a chat-capable model or combo. Roo is a guide card and does not write settings on your behalf.
Configure Roo
Open CLI Tools → Roo, select the endpoint, key, and model, then follow the copied steps in Roo:
Open the Roo Settings panel.
Choose API Provider, then Ollama.
Paste the base URL. The card substitutes {{baseUrl}} and appends /v1 if it is missing.
Paste the API key.
Select the DurinDoor model id, or type a combo name.
Manual config
In Roo, API Provider Ollama:
Base URL: http://localhost:20128/v1
API key: YOUR_DURINDOOR_API_KEY
Model: provider/model-idIf a Roo build also offers an OpenAI-compatible custom provider, those three values work there too. Combo names such as coding-default are valid in the model field if you created that combo.
The copied base URL must include /v1. Roo's Ollama provider typically posts to {base}/api/chat, which becomes POST /v1/api/chat on DurinDoor. DurinDoor returns Ollama-compatible NDJSON on that endpoint. Curl by hand still uses /v1/chat/completions or /v1/messages.
Check it works
Send a one-line prompt in Roo. Confirm the row on Usage.
| Symptom | Check |
|---|---|
| Connection refused | Base URL missing /v1, or the gateway is not on port 20128 |
| Invalid API key | Paste the DurinDoor secret, not an upstream Ollama or OpenAI key |
| Model not found | Use an id from GET /v1/models or a combo name |
| Tools fail after a simple chat | Every combo member must support the tool-call shape Roo sends |