Open AI compatible LLM "Method not allowed" – why?

[benfrain]

We use a proxy at our workplace, which is Open AI compatible. This proxy works fine via Curl and in VSCode.

I am trying to get it working in Zed. I am adding an Open AI compatible provider in the UI, and adding my details.

However, when I enter a message I get a HTTP response error "405 Method not allowed – "method not allowed\n"

The proxy expects a POST and the only thing we have in the proxy regarding this is:

if r.Method != http.MethodPost {
		http.Error(w, "method not allowed", http.StatusMethodNotAllowed)
		return
	}

Can anyone point out what I am doing wrong here? Feels like there is some piece of config that Zed needs I am not providing.

[janfer95]

I get the a very similar error with an Open AI compatible model as well:

Error in model response stream: HTTP response error from My Company's GPT's API: status 405 Method Not Allowed - "{\"detail\":\"Method Not Allowed\"}"

More specifically, it does not work for a GPT 5.2 codex model (I had to turn off chat completions), but it does work for other models, so I assume it is linked to the request that Zed is making.

Is there a way to inspect the request made by an LLM? It would make debugging a lot easier!

@benfrain, did you manage to resolve this issue?

[benfrain]

Sadly not. Ended up going back to VSCode and RooCode. 😮