Gemini CLI

Why Gemini CLI can’t target Valkyrie, and what to use instead.

You can’t point Gemini CLI at a Valkyrie deployment.

Gemini CLI only authenticates against Google (Login with Google, a Gemini API key, or Vertex AI). It does have a GOOGLE_GEMINI_BASE_URL override, but that only changes where it sends Gemini-protocol (generateContent) requests. It has no OpenAI-compatible or Anthropic-compatible mode, so it can’t call Valkyrie’s /v1/chat/completions or /v1/messages endpoints.

Use Qwen Code instead

Qwen Code is a fork of Gemini CLI that adds OpenAI-compatible support. If you want the Gemini CLI experience against a Valkyrie model, that’s the way to do it.