OpenHands

Was Deepseek-v4-Flash on litellm proxy provider changed from something else, to `openrouter/` ? 

cc <@U08H0CRU1V4> It seems we have it for weeks <https://github.com/OpenHands/software-agent-sdk/pull/3638#issuecomment-4673256360|https://github.com/OpenHands/software-agent-sdk/pull/3638#issuecomment-4673256360> 

But it was working, so I assume it wasn’t served from OpenRouter? I’ll try to see if we can detect OpenRouter… or just drop reasoning effort

Yeah we recently flipped to use openrouter :facepalm:

I'm about to remove `reasoning effort` for it… though OpenHands says that litellm knows it accepts this parameter on some providers, but it fails on OpenRouter

The trouble is, we kinda don’t know it’s openrouter and litellm doesn’t know

We are currently migrating a some models to open router, to simplify endpoint maintenance.

Will check on DeepSeek-v4-flash and the others 

working on a fix now :slightly_smiling_face:

i think just adding them to allowed paragram should be ok

<https://github.com/All-Hands-AI/infra/pull/1384> <@U08H0CRU1V4> this should fix it :pray:

<@U06QJHRBLQG> <@U06PYTVAES1> endpoints are now updated daily, and issue is assigned to me if there is a regression <https://openhands-llm-support-tracker.vercel.app/>