Hi team, it seems like the models we're adding for the natural language search are not persisting; I'm following the instructions outlined in the docs here. Response we're getting from the POST request:

{
  "api_key": "**********************************",
  "id": "gemini-model",
  "max_bytes": 16000,
  "model_name": "google/gemini-2.5-flash",
  "temperature": 0
}

It seems like after a couple of hours, when I make a GET request to

/nl_search_models

, it will return an empty array (and we are unable to access the model for any natural language search requests). Is this expected? Thanks!

Hey @Alan Chen, Do you have the time when you made the creation and the time when you verified that it was no longer there?

@Alan Martini I don't have the exact timestamps - I probably created ~2 weeks ago? And then had issues accessing yesterday. And yep we had a configuration change yesterday

Alan, On reindexing, if the collection can't connect to OpenAI API, this kind of issue can arrise. I can see it timeouted right in the reindexing of your collection. Could you check OpenAI on your part? if the ApiKey is still valid and if there isnt billing issues?

Hm I just checked our billing with OAI and it doesn't seem like there are any issues - no limits reached or any declines. Are you saying that if there was an issue with the OAI generations, it would also clear our Gemini models on the natural language search?

Sorry Alan, my bad, I mixed the words, I usually work with issues on openai. I mean if you could check on the Google billing for it?

No issues on Google Billing