#Responses are bundled together in Discord

1 messages · Page 1 of 1 (latest)

wheat sand
#

When the agent generates a response before doing any work, e.g. in a pattern like this:

  • "Got it, I'll do ..."

  • Does work

  • "Ok, I've done.."

Then the first message does not arrive until the turn is fully complete.
In other words, it often takes 5 minutes+ before the agent even acknowledges the request. It's as if all responses get bundled together and sent upon completion

Discord is my only communication channel atm so I can't tell whether this is a Discord issue or gateway/setup. Has anyone figured out a way to improve this?

timid gorgeBOT
#

To help others find answers, you can mark your question as solved via Right click solution message -> Apps -> âś… Mark Solution

sage hillBOT
#

What you’re seeing is basically “expected” with the default Discord settings: the model may generate that initial “Got it…” text early, but Discord won’t see anything until OpenClaw flushes the message at the end of the turn (unless you enable one of the streaming modes).

The fix is to turn on Discord preview streaming:

{
  channels: {
    discord: {
      streaming: "partial" // or "block"
    }
  }
}
  • partial = sends one “draft” message quickly, then edits it in-place as more text arrives (good default).
  • block = similar, but updates in chunkier steps (often nicer / less edit spam); tune with draftChunk if needed.
#

Docs:

If you instead want multiple real messages to appear as the assistant writes (not a single edited draft), that’s block streaming (separate setting) — but it’s noisier in Discord, so most people start with channels.discord.streaming: "partial".

After changing config, restart the gateway for it to take effect.