v-compress

To reduce session context bloat, compress completed phases into concise summaries that preserve core decisions. This keeps the active conversation lightweight while detailed content is stored to disk.

You should compress session context during phase transitions or when conversations become too long. This prevents context limits from interrupting your workflow by summarizing completed work and storing full details to disk.

Yes, you can manually trigger context compression for a specific phase by running the /v-compress command followed by the phase number. This allows you to selectively compress completed work without affecting active tasks.

Phase compression does not delete original detailed conversation content. It stores the detailed phase content to disk files and only replaces the active context with a concise summary to preserve essential information for future tasks.

The best way to extend AI agent sessions without losing core decisions is to use automatic phase compression. It converts completed phases into summaries during transitions, ensuring essential details are preserved while extending the session length.