Docs / source reviewed
fast-jev-compaction
Trim old tool records without rewriting the rest
How this project uses Jev
- Input
- Conversation, tool calls, and result references
- Jev decides
- Which historical tool records remain useful
- Code executes
- Drop or truncate stale records and preserve the remaining text
EDITOR’S NOTES
Where to start
Who it is for
For coding-assistant users with long conversations and large tool histories who want to reduce context usage.
Setup requirements
Requires the supported Claude Code plugin environment and Jev configuration. Test in a disposable session rather than replacing the only history of an important task.
Suggested first experiment
Use a conversation containing useful clues and obsolete output. Ask the same questions before and after compaction, and inspect dropped or truncated records.
Before adopting
Keeping retained text verbatim does not preserve every piece of information. Check whether constraints, error evidence or file paths were lost, in addition to measuring size reduction.
Editorial guidance based on linked sources; not an independent runtime test.
Evidence and limitations
Avoiding rewrites does not make filtering lossless. Important context may still be removed; inspect results in real conversations.
This project has not been run independently here. Author-reported results are not independently verified results.