AI API with self hosted cube server

[kevinleeTCA]

Problem

1. In regarding of proof of concept, Is there a way to test the AI API against our self-hosted cube server ？

I tried to connect to our host with the cubejs-api/v1/ai/query/completions, it gives 404

I also tried using cube cloud via following:

Using my git account and, select our cube repo, added some ENV VARs under setting, But can't make it connect to either my local cubejs server or our dev environment cubejs server, as it keeps failing with:

I guess it is due to the fact that it doesn't connect to our cube server, instead, it connects the one created by cube cloud:

Proof of this can be found in the running log:

Which is different from the version we are using Cube API server (1.1.18).

We really would like to invest on this if this API looks promising with our cube data and query requirement

2. If we decide to use a different LLM, e.g., OpenAI, train it with our own cube meta data, would it work ? any guidance on this ?

[igorlukanin]

Hi @kevinleeTCA 👋

1. AI API is part of Cube Cloud, there's no way to use it in Cube Core.
2. You can actually "bring your own LLM" to the AI API: https://cube.dev/blog/make-your-ai-smarter-turbocharging-cubes-ai-api-with-value-search-byollm-and#byollm

I think you might want to talk to @djayatillake about the AI API.

[djayatillake]

Hi, @kevinleeTCA, you can actually try this feature on a pay-as-you-go starter plan. No need to commit to long-term use.