How does vector similarity work? I have a case in which I have a group of documents, and I have sent an exact sentence found in one of the documents, but when I checked the returned nodes, none of them contains the sentence I have sent; any clarification why this may happen?

I have upgraded my llama index to 0.6.20, and I used the compact mode as a response mode, but it's not working as what is written in the documentation; the way it works is the same as the refine mode work; Does anyone knows why this happened?
https://gpt-index.readthedocs.io/en/latest/reference/query/response_synthesizer.html#llama_index.indices.response.type.ResponseMode

I have a case in which I have set the input size to 3500, and the chunk size is 750. When I query my index with top_k =4, it always does refining. Any idea why this happens??
Note: qa prompt size is 93 and the query around 50 tokens

When I send the same query multiple times, different responses are returned. Can anyone explain why this happened?

Hello guys, I need to use Redis as the vector store with the llama. What is the best way to do that, I did not find any documentation regarding customizing the vector store

Is there a way to disable the refine prompt?

So not all the data is included in the prompt; how can the model provide accurate results?

How to customize the tokenizer used for embedding??

hello is there a way to create custom embeddings and provide it as an embedding model when using chromadb

hello, is there a way to summarize the answer provided by QA system using llama

i have already try to use response.source_nodes and response.get_formatted_sources() but still one context returned