Why lowered paste summary?

[jgordijn]

Question

I noticed in the v1.0.26 changelog: "Lowered paste summary threshold to trigger on shorter content." I'm curious about the reasoning behind this change.

I frequently use Whispr Flow for input, where the transcription is immediately pasted in when I finish speaking. It occasionally makes small transcription errors that I need to correct. With a lower threshold, even brief dictations now appear as [Pasted ~1 lines], which makes it harder for me to verify what was actually pasted.

I can definitely see the value of paste summarization when working with large documents in the CLI. It keeps things manageable. However, for use with dictation software like Whispr Flow, it adds more friction than help.

Could you share the thinking behind lowering this threshold? Are there specific use cases this change is meant to address?

[rekram1-node]

I think @thdxr would know better here but if you wanna disable the paste summaries you should be able to set the following:

{
  "$schema": "https://opencode.ai/config.json",
  "experimental": {
    "disable_paste_summary": true
  }
}

in your opencode.json

I think there was some issue with certain pastes causing multiple messages to be submitted or something like that

[jgordijn]

Thanks for the prompt reply! I will try that.

[zenyr]

Side note: When pasting images/text with summary enabled, CJK graphemes cause rendering issues. Disabling paste summary resolved this for me.

Sends broken payload: