Executing a plan under context constraints
A user running the Qwen 3.6 35B-A3B model locally encountered high context window usage while executing a refactoring plan. The model reached 92.6% context window utilization before auto-compaction occurred. The user is seeking advice on how to manage context window pressure during plan execution to prevent such issues, suggesting methods like starting a new session with the previous plan pasted in. AI
IMPACT Users may need strategies to manage context window limitations when executing complex plans with local LLMs.