context-budgeting

Optimize OpenClaw context window usage by partitioning information into objectives, short-term history, and decision logs, then scheduling pre-compression checkpoints to summarize older data and reduce token costs.

Pre-compression checkpointing is a memory management process that creates a memory checkpoint and runs a cleanup script to summarize older data before the context limit is exceeded.

Run a context checkpoint by explicitly updating the HOT_MEMORY.md file and executing the gc_and_checkpoint.sh script to pre-compress memory and preserve recent dialogue.

Yes, you can automate garbage collection triggers by integrating with a Heartbeat mechanism that automatically initiates cleanup when context usage exceeds 80%.

Context partitioning works for long-running OpenClaw sessions and memory-constrained deployments by gating long-term memory and preserving recent dialogue while summarizing older data.

Manual context checkpointing requires explicit HOT_MEMORY.md updates and executing the gc_and_checkpoint.sh script, meaning missed updates can lead to unmanaged context overflow and lost data.