Skip to content

refactor(console): proxy every keyless free Zen model to new inference - #53967

Merged
vimtor merged 1 commit into
devfrom
zen-free-all
Oct 8, 2026
Merged

vimtor merged 1 commit into
devfrom
zen-free-all

Conversation

@vimtor

@vimtor vimtor commented Oct 8, 2026

Copy link
Copy Markdown
Collaborator

Issue for this PR

N/A

Type of change

  • Bug fix
  • New feature
  • Refactor / code improvement
  • Documentation

What does this PR do?

Every busy free model already serves keyless requests directly from new inference (#53717, #53959, #53963). This replaces the hardcoded model list with the model's existing allowAnonymous flag in ZEN_MODELS, so a new free model is proxied without a code change.

The only models this adds are deepseek-v4-flash-free, mimo-v2.5-free, and keyless jev-1.13. All three already route to new inference through inf.* providers and already fail there (missing from the registry, deprecated, and paid respectively), so their outcome is unchanged; callers now see new inference's error directly.

Keyed requests and Go are unchanged.

How did you verify your code works?

bun typecheck in packages/console/app. The previous batches were smoke-tested in production with keyless requests through opencode.ai/zen/v1.

Screenshots / recordings

N/A

Checklist

  • I have tested my changes locally (typecheck only; see above)
  • I have not included unrelated changes in this PR

@vimtor
vimtor merged commit 687664c into dev Oct 8, 2026
10 checks passed
@vimtor
vimtor deleted the zen-free-all branch October 8, 2026 14:22
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant