Just making REST calls against the predict endpoint is simple enough. Finding the right example document in the documentation was a mess. Didn't get a correct generated client for Elixir from the client generators. But this curl example got me there with minimal problems. Aside from the plentiful problems of auth and access on GCP.
You might need to do the song and dance of generating short-lived tokens. It is a whole thing. But the API endpoint itself has worked fine for what I needed. Eventually.
https://cloud.google.com/vertex-ai/docs/generative-ai/text/t...
You might need to do the song and dance of generating short-lived tokens. It is a whole thing. But the API endpoint itself has worked fine for what I needed. Eventually.
OpenAI was much easier of course. So much easier.