IndexCache, vLLM, and upstream

[nejch]

First, thanks for this promising patch and all the research @bys0318! I just stumbled across this while researching other GLM-5 stuff.

Seems like the focus here is more on slang (which would explain some teething problems I've had with GLM-5 on vLLM), but I was wondering if you've done any tests with vLLM or were considering upstreaming this (to vLLM or slang) already?

Thanks again 🙇

[nejch]

Ah, I see this was already upstreamed in chutesai's fork of sglang, hopefully also upstream and vLLM soon :)

[bys0318]

Thanks for your suggestion. We've added vLLM support.