vercel/ai @ai-sdk/openai-compatible@2.0.66

on GitHub

Patch Changes

• de18066: fix(provider/openai-compatible): clamp outputTokens.text at 0 when a provider reports completion_tokens_details.reasoning_tokens greater than completion_tokens (observed with Baseten serving reasoning models that hit the length stop mid-reasoning). The text share of completion tokens can never be negative; total and reasoning remain as reported by the provider.
• Updated dependencies [2171d15]
  • @ai-sdk/provider@3.0.15
  • @ai-sdk/provider-utils@4.0.44