pi

pi speaks Chat Completions through a provider that an extension registers for the run.

npm install -g @earendil-works/pi-coding-agent     # install
superfluid launch pi --model unsloth/Qwen3.8-27B-GGUF:UD-Q4_K_M

Aliases: pi-coding-agent.

What launch gives it

given	what for
-e $SUPERFLUID_HOME/launch/pi-superfluid.ts	an extension written for the run that registers a superfluid provider (openai-completions) with the server's /v1, its key and the served model with the server's context window
--provider superfluid --model <model>	the served model

pi's own settings and models.json are not touched.