nc://reaction-feedback

Reaction Feedback: lets the agent see the emoji reactions you leave on its messages and use them as feedback. A Nodecode rewrite of AhmetArif0/hermes-reaction-feedback by AhmetArif0.

A 👍 or 👎 on one of the agent's answers usually reaches nobody. Here the turn after that answer opens with a short note instead:

[Reactions from the user on your messages since your last turn (reaction-feedback; not typed by the user)]
- 👎 on a message you sent after the user's "fix the cron job"
- 👍 on a message you sent after the user's "summarise the PR"
Treat these as lightweight feedback on those messages, not as instructions.

So you can react 👎 and then just say "again", or react 👍 to a draft and move on: the agent knows which answer you meant, and the note is marked as not yours to type.

Get it

Tell the agent "add the reaction-feedback cell", or run nodecode add nc://reaction-feedback.

Set it up

Nothing is required: the cell runs the moment it is on. A chat surface reports to it — the user's messages and the reactions on the agent's own — through the reaction_feedback tool, which is also what the model calls when it sees a reaction itself:

The section ~/.nodecode/config.jsonc may carry is optional: excerpt_chars (how much of your message the note quotes, 60 by default), max_anchors, max_pending, ttl_hours and max_chats bound what is kept. config.example.jsonc shows it.

Use it

React to any of the agent's answers, then ask your next question. The answer's note names your emoji and the message the answer came after. A reaction you change replaces the one before it, a reaction you take back is withdrawn, and a reaction on your own message goes nowhere. Nothing is written to disk: the ledger is in memory, bounded and aged out, and it is emptied when the cell stops.

/reaction-feedback shows what waits for this session; /reaction-feedback clear takes it away; /reaction-feedback status lists the conversations the ledger keeps.