Plz

[r0man]

Hi @ahyatt,

here's another PR with some small bug fixes and improvements.

• Update plz-media-type.el (this is only indentation).
• Really set reponse body to nil (I had files in different repos and messed this up).
• Add llm-request-plz-connect-timeout. I have to actually tweak this to get a reliable behaviour.
• Set llm-request-plz-timeout to nil. I think this should actually be the default, and if users know better they should figure this out themselves. I had now multiple the times Gemini outputing a long response which I told it to compute, and then timing out in the middle of the response. For slower LLM I think this timeout is mostly annoying.

[r0man]

Btw, I now created seperate repos for my code. I update the files in the llm library with this script.

#!/usr/bin/env bash
set -e
set -o pipefail
wget -q -O plz.el https://raw.githubusercontent.com/r0man/plz.el/plz-stream/plz.el
wget -q -O plz-media-type.el https://raw.githubusercontent.com/r0man/plz-media-type/main/plz-media-type.el
wget -q -O plz-event-source.el https://raw.githubusercontent.com/r0man/plz-event-source/main/plz-event-source.el

[ahyatt]

Thanks! For the timeout, I've seen some instances where ollama will just keep on generating forever, so it's important to have some timeout at least for ollama. The ones actually run server-side presumably would never do that.