v0.10.12
0.10.12 (2026-08-07)
- fix(providers): pass max reasoning effort through on the Responses API for probe-verified models (bc978f770f0d5041491b6b73ed608130ffd583a7)
- docs(llm): align caching option docs with the zero-config defaults; surface prompt caching in the README (d59bac5edc60b9527cdc8da9bc0cce594d5bf738)