2 months ago
4fedb49
Remove the gpt-5.3-codex-spark summary default and filter stale summary settings against the effective Codex app-server model before turn/start. Add registry and app-server fallback regressions that reject any Spark summary parameter.
Parent483d9ae
