Something u guys could improve the documentation on:
explaining the retrievers module, recently i started trying new retrievers on the documentation but there is 0 explanation what is it for, why is it good etc.
for example,
auto retrieval,bm25, as someone who is not a background from data science i have no idea what does it do, how does it differ from normal rag retrievers and what kind of scenario its good for
yea fair. BM25 is good for more keyword-focused lookup. But requires all nodes to be in memory. Usually you combine this with vector retrieval, but it can be used standalone as well
AutoRetrieval is just getting the LLM to write the top-k and metadata filters needed to answer a query. Good if you have tagged all your data with predictable metadata, but can be slow (an extra LLM call) and unstable (some LLMs may predict incorrect results)
