You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
[Feature]: OpenAI Compatible Embeddings for the Recommender #400
I already have a dedicated LXC that handles my AI models and is located on a separate machine in my home lab. All my apps access my models using the OpenAI compatible endpoint v1.
By implementing this capability, users can keep their AI and model setups and use pre-defined endpoints for the recommender without having to setup network mounts, or redownload and thus, duplicate models
What would you like to see?
Support in the recommender for using the /v1/embeddings endpoint rather than requiring models be stored in the environment and have support for setting the endpoint and API key as environment variables
You know what, I was kinda thinking about this yesterday lmao 😂 But I don't really know this side: is it expensive? Is it easy? How does it work? Is there any documentation? I have a lot of questions in mind, so I'd love to learn more if you can help me! If there are any docs or other resources you can link me, that would be great.
It's pretty straightforward imo. I haven't looked at your repo backend, but generally you store the embeddings data as per usual. All you really need is the API key, and the endpoint. The LocalAI docs do contain some good info on the API
Activity
Priveetee commentedon Nov 12, 2026
You know what, I was kinda thinking about this yesterday lmao 😂 But I don't really know this side: is it expensive? Is it easy? How does it work? Is there any documentation? I have a lot of questions in mind, so I'd love to learn more if you can help me! If there are any docs or other resources you can link me, that would be great.
shopsD commentedon Nov 12, 2026
@Priveetee
LocalAI have some good documentation on where to start.
It's pretty straightforward imo. I haven't looked at your repo backend, but generally you store the embeddings data as per usual. All you really need is the API key, and the endpoint. The LocalAI docs do contain some good info on the API