Menu

1 / 2

fix(coding-agent): use the cheapest vision model for image descriptions by gcp · Pull Request #17196 · can1357/oh-my-pi

GitHub Public Timeline Feed·gcp·1 day ago
#CIZy91pT
Reading 0:00
15s threshold
Selected the least expensive known vision model only when no configured or active vision model is available, avoiding catalog-order surprises.

Added image-description regression coverage for unpriced candidates and explicit role precedence.

Fixes can1357#16755

roboomp

chatgpt-codex-connector[bot]

Rows without a rate card rank last, declared free rows rank at 0, and prices come from getEffectiveTokenCost so dated rates apply.

@H4vC