Connect Roo Code to PrivateDeskAI
Roo Code (the VS Code coding assistant) has an "OpenAI Compatible" API provider built into its settings panel. This page gives the exact values for this gateway's DeepSeek model, taken live from this gateway's own /v1/models endpoint.
1. Get a key
Get your API key first - see Getting started for the two-minute walk-through (sign up, get a key, copy it once).
2. Open Roo Code's settings
- Click the settings icon to open the Roo Code settings panel.
- Set API Provider: to
OpenAI Compatible.
3. Fill in the fields
| Field | Value |
|---|---|
| Base URL: | https://api.privatedeskai.com/v1 |
| API Key: | YOUR_SAGG_KEY (from step 1) |
| Model: | deepseek-ai/DeepSeek-V4-Flash-0731 |
Under Model Configuration, Roo Code also lets you set the context window and max output tokens by hand; this gateway's own values (context_length and max_tokens) are listed live at /v1/models.
4. Test
Start a task to confirm it answers.
Model id and alias
The model id above, deepseek-ai/DeepSeek-V4-Flash-0731, and its short alias deepseek-v4-flash-fast, are both listed live at /v1/models - check that page if a name there ever differs from this one. Current prices are on the pricing page (also sourced live from /v1/models).
If something does not work
Re-check the Base URL character by character (it should end in /v1, with no trailing /chat/completions) and that the API key was pasted without extra spaces. For what request parameters this API accepts, see Compatibility; for current uptime and speed, see Status.
Source for the Roo Code steps above: Roo Code's own documentation, "OpenAI Compatible" provider page (docs.roocode.com), checked 2026-11-15. Roo Code may change its own settings UI at any time.