Skip to content

v0.42.1 - Fix for token art generation

Hotfix on top of v0.42.0.

Bugs

  • [x] AI token-art generation with a user-level Gemini (or Grok) key failed with No API key available for provider "gemini". The v0.35.0 fix injected the user's image-gen key only on the model-listing call (/api/ai/image-models), so the provider appeared in the picker — but the actual generation calls (POST /api/modules/genai/generate, and POST /api/modules/genai/extend for "Extend This Image") reached the genai module with no key. In prod those providers are user-level (no server env var), so every generate/extend 400'd. useImageGenAuthFetch now also injects the key on the genai generate/extend endpoints (still excluding the public /local-models/ weights, least privilege). Platform-only, client-side fix — genai already reads the header and needs no release/modules.hosted.json bump.

Related, still open: the x-ai-provider / x-ai-key headers are singular, so a user with user-level keys for two or more image providers still only sees the first by priority. Tracked in the Multi-provider image-gen keys entry under Next Version: v0.48.0 in CHANGELOG.md.