Skip to content

LOTTO v2.0.0-rc.4 — AI Configuration refresh

Pre-release
Pre-release

Choose a tag to compare

View all tags
@MattGeiger MattGeiger released this 15 Sep 05:49
· 16 commits to main since this release
v2.0.0-rc.4
5c2c72b

AI Configuration refresh

Ports FEED 1.8.0's current OpenAI, Anthropic, and Google catalog into LOTTO's existing six-step wizard. Twelve current presets, model-specific request parameters, and lifecycle/replacement guidance are available; existing saved models, keys, and custom reference prices remain intact.

LOTTO retains its simplified experience. Model Pricing shows reference input/output rates, with provider portals remaining the place to check spending. No built-in cost tracking or budget dashboard was added. Test key and model checks the selected model without billed generation and clearly describes that limit.

Validation

  • 963 tests passed (135 files), 1 skipped; TypeScript, lint, production build, and the 48-chunk legacy scan passed.
  • Authenticated WTH production creation/edit/deletion, blank-key preservation, zero-valued top-p, custom prices, selected-model success/failure, and legacy recovery passed. The disposable inactive record was removed; original configurations and public state were preserved.
  • Live structured generation passed on 11 of 12 models. Representative 100-row batches and single-text checks passed on all three providers. Conservative test-charge bound: $1.53315, below the approved $5 budget; provider portals determine actual charges.
  • iPad mini 4 Simulator on the actual iOS 15.4 runtime passed built/hosted login interactivity checks. Desktop responsive wizard checks and hosted hydration smoke passed.

Known validation limits

Claude Opus 5 returned explicit refusals with the test key, including a minimal food fixture; successful generation with that model remains unverified. Full authenticated legacy touch/wizard testing and physical iPad acceptance remain unverified. These limits are recorded rather than inferred from modern-browser tests.

Deployment and evidence

This prerelease is deployed directly to the WTH Vercel project at williamtemple.app. St. Johns remains on its existing deployment. The shared main branch was not advanced; reconcile the release branch before any future main-based WTH deployment.

The attached verification JSON records the exact tagged commit, final Vercel deployment, rollback target, and post-deployment checks. No schema, environment, or dependency changes are required.

Stable-v2 realtime acceptance remains separate.

Assets 3
Loading