FR: Support for custom cloud service with VectorDB and AI-models

[Nibot1]

Describe the feature you'd like to request

It would be nice to have support for custom cloud services as an alternative to AppAPI with usage of a local GPU.

I hope I point this enhancement to the right repo, or should I add it to the backend repo?

Sadly, I don't have any experience with coding a Nextcloud app.

Describe the solution you'd like

The cloud service I am thinking about is a German service hosted by IONOS AI Model Hub
This service provides an OpenAI-compliant API as well as a hosted VectorDB to store document metadata.
The developer documentation can be found here: https://docs.ionos.com/cloud/ai/ai-model-hub/

The service is 100% compliant with GDPR (info from their website).

Describe alternatives you've considered

[kyteinsky]

hello,
the IONOS AI models can directly be used with the integration_openai app in Nextcloud. Context chat will use the text2text (generate text) provider set for the answer generation. We're working on making this possible to use the embedding models from them (OpenAI-compatible APIs) as well.
for the vector db though, a new wrapper needs to be written for IONOS that implements all the functions and ACLs that we currently support with pgvector. Although if you have a hosted postgresql, installing the pgvector extension there should also make it usable as an external db.
IONOS vector db offering: https://docs.ionos.com/cloud/ai/ai-model-hub/tutorials/document-collections
Context chat's vector db implementation with pgvector: https://github.com/nextcloud/context_chat_backend/blob/25b894d10796c8df508547d28f9d2ca26c477462/context_chat_backend/vectordb/pgvector.py

I hope I point this enhancement to the right repo, or should I add it to the backend repo?

yeah it belongs more in the backend repo, I'll move it there, don't worry.