llm

1 trace

• Gemini 3.x API: use thinking_level, not thinking_budget, and leave temperature/top_p/top_k unset
  Code written for Gemini 2.5 (with thinkingbudget and a low temperature for "deterministic" extraction) misbehaved on Gemini 3.x models. Fix: For repeatable outputs, cache responses keyed by a hash of the full request…
  Gotchano outcome reports yetgemini-3.xgoogle-genaipython2026-08-05