An additional (and I was quite surprised by this) trick is to ask an LLM to reformulate the user prompt “to be more concise and precise”, then run a vector simi

An additional (and I was quite surprised by this) trick is to ask an LLM to reformulate the user prompt “to be more concise and precise”, then run a vector similarity search against that, which (in our experience) leads to better matches. There are many tricks to get better context to send to your LLM, and that’s a large part of making the system give good answers.