Using LiteLLM with JetBrains AI Assistant

Question

Is it possible to set up an "LLM Proxy" for JetBrains AI Assistant, integrating LiteLLM in the flow as follows: JetBrains AI Assistant -> LiteLLM -> JetBrains AI Service? Additionally, can we receive an API key for https://brreg.jetbrains-ide-services.com?

Answer

The flow you suggested (IDE -> LiteLLM -> IDE Services) is not supported, but a near-equivalent setup can be achieved. According to the Custom Models page, while logged into IDE Services, the server determines the provider selection, and the BYOK picker in AI Assistant is disabled.

What is supported is the following flow:

IDE -> IDE Services -> LiteLLM -> upstream LLM

Here, LiteLLM is configured as an "OpenAI Compatible" provider inside AI Enterprise. This configuration ensures that guardrails and logging occur at the LiteLLM level, specifically on the IDE-Services-to-LiteLLM segment.

Please note that IDE Services does not expose its own OpenAI-compatible endpoint. Therefore, API keys for https://brreg.jetbrains-ide-services.com are not available. Instead, an admin token is used by IDE Services to authenticate outbound to LiteLLM, which also supports optional header-based per-user identity passthrough. However, per-user API keys are not supported at this time.