compact-context

To save AI session state before compaction, capture active tasks, modified files, decisions, and open questions into a checkpoint file. This preserves essential context and creates a resumption point before reclaiming context window space.

Context compaction is the process of reducing conversation history to reclaim context window space. It allows for longer or more focused interactions by summarizing or removing older messages while preserving key session details.

Free up context window space by creating a structured checkpoint file of active tasks and modified files, then compacting the conversation history. This approach retains essential information while reducing the active context footprint.

Checkpoint your conversation history when the context window approaches its limit. Creating a checkpoint file ensures that active tasks, key decisions, and open questions are safely stored before reducing the conversation history.

You can resume an AI coding session after compacting by loading the previously generated checkpoint file. This file contains the structured session state, enabling seamless continuation of work in a new or compacted session.

Context compaction does not re-read modified files. It gathers active tasks, key decisions, and open questions directly from the existing conversation context to build the checkpoint file efficiently.