Documentation · Connect a tool
Set up JetBrains AI Assistant
JetBrains AI Assistant can use an OpenAI-compatible provider. The endpoint field is the one thing that differs between IDE versions, so check the path your IDE actually calls.
Add the provider
Create a key in the Console first. Then add an OpenAI-compatible provider in the AI Assistant settings. JetBrains describes the steps for each IDE version in its custom models guide.
- Provider type
- OpenAI-compatible
- API key
- Your
sk-dc-key. Enter it only in the key field. - Model
chat-35b-moe
Use a synthetic or non-sensitive project to try it. Your IDE sends the code and conversation it uses for a request to a node.
This is for chat. Tenvor serves the chat completions endpoint and no text completions one, so it is not offered for inline completion, fill-in-the-middle or edit prediction.
Choose the root or /v1
Tenvor answers at two paths: https://orchestrator.tenvor.io/v1/models lists the model, and https://orchestrator.tenvor.io/v1/chat/completions answers a chat. Which address belongs in the endpoint field depends on what your IDE adds to it.
| If the IDE adds | Enter |
|---|---|
/v1/models and /v1/chat/completions | https://orchestrator.tenvor.io |
/models and /chat/completions | https://orchestrator.tenvor.io/v1 |
Do not enter the full chat address. It cannot also serve the model list.
A 404 usually means a doubled /v1
If the model list or a chat gives HTTP 404, the IDE may have called /v1/v1/models. Check only the path the IDE called, and keep your key and your prompt out of anything you share. No single value of the endpoint field is right for every IDE and plugin version.
Check one real chat
Test the connection
Use Test Connection in the provider settings, save, and choose the Tenvor model. Turn off any automatic provider or model fallback for this check.
Send one short chat
A passing connection test, or a curl request that works, does not show that the IDE's chat works. Send one real chat message.
Find the charge
Open the Usage page in the Console. The request is there as a charge.
If your IDE sends an empty tools list on its own when tool calling is off, Tenvor treats it as no tools. An empty list sent with a tool_choice other than "none", or with parallel_tool_calls set to true, is refused with a 400 (see tools). When you do use tools, your IDE runs them, never Tenvor.
Something wrong? See troubleshooting.
Content checked . Live limits and prices can change.