Can we expect, with time, that the new Gemini Cache feature will be available in LI?

How does one turn a Q&A RAG into a multi-turn chat experience?

1. Initial question retrieves context
2. Subsequent questions should continue with List[Message] like a chatbot, using the template & context as the system message?

If I'm finding that the default prompt of the RetieverQueryEngine to need customization, is there a way to customize just that part, or do I need to be looking at a custom QueryPipeline?

Is there any work on text+video RAG? Is that something llama-index can handle today, if the right model is found? LLAVA or something like that?

After integrating Langfuse with LlamaIndex, I noticed that certain steps did not have their own spans, notably the embedding of the user's question and the subsequent lookup of docs in Qdrant.

Is there a way to get these to show up? Am I missing a config or option somewhere?

Difficulty in retrieving documents from a word...

When using Qdrant and indexing documents from SimpleDirectoryReader, how can I set a metadata value for the documents? (same for all) ... related, how can I set per document metadata?

Are there any LI guides on Multi-modal RAG? I have

1. A text blurb / explanation
2. Meta data as JSON
3. A graph as an image

They are not within a single document, pdf, or page, but separately index entities. I'm not sure the best way to calc the embedding for something like this, as one thing or as multiple parts that are then just retrieved together?

Any guidance or insights from the community would be greatly appreciated