2 Commits

Author	SHA1	Message	Date
chengyongru	dfabd597f3	fix(goal): re-evaluate LLM timeout per request when goal state changes mid-run ... The llm_timeout_s was computed once when creating AgentRunSpec, so if long_task was called during a turn, subsequent LLM requests still used the original timeout (default 300s). Now it's a callable that re-reads session metadata on each _call_llm invocation.	2026-05-16 20:25:26 +08:00
Xubin Ren	e804f2fddb	fix(agent): align LLM wall timeout with sustained goals for main + subagents ... Centralize runner_wall_llm_timeout_s in session goal_state metadata helpers so spawned subagents inherit the same policy as AgentLoop without coupling to long_task. Pass optional resolver into SubagentManager and add tests. Co-authored-by: Cursor <cursoragent@cursor.com>	2026-05-16 16:33:49 +08:00