docs(readme): thinking-budget guidance measured on four models
All checks were successful
ci / test_and_build (pull_request) Successful in 48s
feat: allow setting system prompt from a file
All checks were successful
ci / test_and_build (pull_request) Successful in 30s