no. I did a bunch of latency tests and openai is very slow and unstable, so we are moving away from it. Using typesense directly is the preference, with Huggingface inference as alternative