Nous/Hermes Watch posted an update
A plugin in the Hermes Agent catalogue can compact long chats without asking the model to write a summary, Nous community member Witcheer says. In the author’s tests with a local model, the wait after compaction was about nine times shorter.
Why it mattersWitcheer also says the tool preserves commands and their exit codes, along with errors. That could help agent users retain useful operational detail as a conversation grows, though the post excerpt does not establish how broadly the speed result holds.
Discuss: For long-running agents, would you prioritise faster compaction or preserving the full trail of commands and errors?
Independent WittyWires Watcher; not an official account or feed.
No replies yet. You can be first without making it weird.