Skip to content

fix(ai): default the assistant's Gemini models to 3.1 Pro and 3.8 Flash - #27

Merged
vlidholt merged 1 commit into
serverpod:mainfrom
vlidholt:fix/gemini-models
Oct 20, 2026
Merged

vlidholt merged 1 commit into
serverpod:mainfrom
vlidholt:fix/gemini-models

Conversation

@vlidholt

Copy link
Copy Markdown
Contributor

Problem

The AI assistant didn't work with a valid Gemini key. Its default models, gemini-2.5-pro and gemini-2.5-flash, are no longer served to new API keys (they return 404). The Bugs & improvements board had already moved to newer models; the assistant's defaults in AiModels had not.

Change

  • AiModels.geminiPro → gemini-3.1-pro-preview (Best and Balanced), AiModels.geminiFlash → gemini-3.8-flash (Fast and inline completions), the same models the board uses.
  • Updated the example config in ai_config.dart and config/development.yaml, and the defaults in specs/10-ai-assistant.md.
  • wonky: ai: models: gemini: still overrides the defaults.

Testing

  • dart analyze clean; dart test test/unit/ai passes.
  • Not yet tried against the live Gemini API.

🤖 Generated with Claude Code

gemini-2.5-pro and gemini-2.5-flash are no longer served to new API keys
(404), so the assistant failed with a valid Gemini key. Use the same models
as the Bugs & improvements board: gemini-3.1-pro-preview and
gemini-3.8-flash. `ai.models.gemini` still overrides them.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@vlidholt
vlidholt merged commit fdc82d4 into serverpod:main Oct 20, 2026
41 checks passed
@github-actions github-actions Bot locked and limited conversation to collaborators Oct 20, 2026
Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant