Repository navigation
Replies: 1 comment
|
+1, especially for the second option (one lookup across all running Herdr sessions). I ran into the same thing: the native session ID is the only handle that stays the same after the agent exits or gets resumed by hand, so it's the natural key for scripts. One small addition that would help: when no running pane has that ID, return a clear "not running" result instead of a generic not-found or a timeout. Then a script can decide to resume the conversation itself ( Disclosure: I'm the author of Herduck, a small open-source layer on top of Herdr (it adds to Herdr, it doesn't replace it). It tracks each conversation by the agent's native session ID for the same reason you describe: when you pick up a piece of work, it goes back to the matching live pane, or resumes the original conversation if nothing is running. A built-in lookup like this would make that simpler for scripts and tools alike. |
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
idea / problem
When I want to reach an agent in Herdr, for example to send it a prompt or read its output, I first need to know which pane it runs in. I know the agent by its native session ID, but the only way I've found to get from that ID to a pane is to list the panes of every running Herdr session and compare each pane's
agent_session.valuewith the ID myself:That works, but every script that needs it repeats the same loop, including handling a session whose server doesn't answer.
requested change
Support looking up an agent by its native session ID, for example:
herdr agent get,agent promptandagent readwork with the ID alone.The first would cover most uses within one Herdr session; the second would also tell a script which Herdr session the agent is in.
why you want this
I keep the native session ID of each Claude Code and Codex conversation I work with, because that ID stays the same when the conversation is resumed. Herdr agent names (the targets used by
herdr agent prompt) don't work as a stable handle for this: Herdr clears an agent's name when the agent process exits, so restarting the agent by hand, even to resume the same conversation, leaves it unnamed. With a session-ID lookup, scripts and agents could address each other by the conversation they belong to.Herdr 0.9.5 on macOS.
All reactions