Hacker News Reader

they point openai.api_base to their server that implements the same API

That’s clever. Do other LLM API’s do that?

It would take you (or gpt) 3 seconds to write an openai compatible wrapper; the inference api is trivial for all LLMs.

Yesterday there was a "Launch HN" thread for credal.ai [0] and I noticed that they use the same openai.api_base trick [1].

Ah, I missed that. Thanks.