Results for “lideke”

Picks the model each subagent runs on from a per-project table, steps up a model when a task is retried, and logs every choice so an agent can tune the table.