Been checking out the new ai assistant stuff. Is there possibility for custom retrieval coming as part of the knowledge retrieval. So we can have our own vector db or endpoints to return the docs we would want to use based on the summarized context window with the latest question?
On a separate note, will the knowledge retrieval assistant ever use multiple documents with annotations, I saw that it will use entire documents if it’s small enough or the chunked vector data to find the best answer but didn’t see anything that would imply support for multiple documents having the answer to the thread. If not are there ways to tweak the vector db retrieval etc.
Are there any guarantees or policies on the uploaded files for security concerns?