Hi there!
I’m new here. However, building a RAG system for my small company I ran into kind of a wall.
Currently I’m using an Ollama set up on my Mac Studio, running gpt-oss:20b as a chat model and qllama/multilingual-e5-large:q4_k_m as a multilingual embedding model.
I’m port-forwarding both of them via ngrok with my custom domain.
And it just works like a charm with Private AI plug-in on desktops, laptops and even Windows tablets of my team. Letting my stuff to use all the power of RAG system. However, Privat AI is not there for both iPad and iOS devices. So I dug through all of AI plug-ins out there in existence. And guess what? None of them works with iPad. Copilot is the one getting very close. But anyways, for whatever reason on iPad it is using not ETT GET, but ETT OPTIONS requests. Getting declined by Ollama Rest API.
So, my request is as follows:
Please someone, implement a version/fork of Privat AI plug in compartible with both desktops/laptops and iPad OS / iOS as well! Letting users to utilise the RAG capabilities, while using server-based Ollama.
P.S.: it would be great to have an option there, letting the user to link a Supabase instance. It would be extremely helpful to have an option to use a dedicated vector DB to be used by embedding model automatically while indexing and re-indexing. In fact, while having an extensive and frequently updated vault, it is crucial to have an option not to rebuild an entire index. But to make some kind of a diff and put changes only there. And it would be also helpful to have an option to use a server-based vector database for all of the synced devices.