Tenvor Compute

Tenvor documentation

JetBrains AI Assistant setup

JetBrains AI Assistant can use an OpenAI-compatible provider. The endpoint field depends on the IDE and plugin version, so verify the resulting request path before treating the setup as working.

Content checked . Live limits and prices can change.

Configure your provider

Start with the account and balance steps and a synthetic project. Create a programmatic key in Console and enter it only in the trusted provider's API-key field.

JetBrains AI Assistant 2026.2 Help describes Settings → Tools → AI Assistant → Providers & API keys, an OpenAI-compatible provider, an endpoint, an API key and a Tool calling option. Your installed IDE or plugin may use different labels.

Follow JetBrains' provider setup documentation for your installed version. Enable Tool calling only when you intend to test a supported tool workflow and retain your IDE's approval controls.

Choose the root or /v1 for your version

The actual Tenvor endpoints are https://orchestrator.tenvor.io/v1/models for discovery and https://orchestrator.tenvor.io/v1/chat/completions for chat.

  • If your provider appends /v1/models and /v1/chat/completions, configure the host root: https://orchestrator.tenvor.io.
  • If it appends /models and /chat/completions, configure the API base: https://orchestrator.tenvor.io/v1.
  • Do not use the full chat endpoint as the provider URL; it cannot also serve model discovery.

A discovery or chat 404 can come from a duplicated /v1/v1 path. Check only the sanitized request path or the version's provider documentation, without exposing the key or content. There is no single endpoint-field value qualified here for every GoLand or AI Assistant version.

Test the actual IDE workflow

  1. Use Test Connection, apply the settings, then explicitly select the current Tenvor model. For this check, disable automatic provider/model fallback and note your IDE and plugin versions.
  2. Send one short synthetic chat request. A connection test or a successful curl request alone does not qualify the IDE workflow.
  3. For tools, approve only an expected read of a small synthetic project file, return the tool result through the client and verify the final assistant continuation.

After completion, check the matching time window, model, usage and charge in Console Activity. A configured endpoint or an answer claiming to be Tenvor is not proof of a completed Tenvor request. If the account evidence is missing or still pending, record the outcome as unverified and reconcile it before submitting again.

Tool requests can stream text in batches; clients must retain complete tool-call and successful terminal boundaries. Stream closure alone is not success. This chat interface is not advertised for inline completion, fill-in-the-middle or edit prediction.

Troubleshoot a failed or incomplete request