3 ms·

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

• arbuge 3y ago
  Ah, I missed that. Thanks.
• OkGoDoIt 3y ago
  That’s clever. Do other LLM API’s do that?
  • dygd 3y ago
    Yesterday there was a "Launch HN" thread for credal.ai [0] and I noticed that they use the same openai.api_base trick [1]. [0] https://news.ycombinator.com/item?id=36326525 https://news.ycombinator.com/item?id=36326525 [1] https://credalai.notion.site/Drop-In-APIs-3a45d32405c347e8bfd0a570afb61e8b https://credalai.notion.site/Drop-In-APIs-3a45d32405c347e8bf...
  • anonzzzies 3y ago
    It would take you (or gpt) 3 seconds to write an openai compatible wrapper; the inference api is trivial for all LLMs.