User file uploads with RBAC

[placerda]

Why are we doing this?
Users want to upload files through the chat UI and immediately ask questions about those files together with the content indexed in the main RAG index.

What does it do?
The Web UI allows users to upload documents. Each document is linked to the user ID and conversation ID. The document is then chunked and stored in a shared conversation index (there is no per-conversation index). The agent can retrieve these chunks to ground the user’s questions.

Technical guidelines
The solution design still needs refinement.
The system should query the conversation file index only when files have been uploaded for that conversation, avoiding unnecessary double queries on every user request.

Reference:
https://github.com/Azure-Samples/azure-search-openai-demo/blob/09dd0ef6918c41d35a1e95406d7a2384f6fa889c/app/backend/app.py#L370