ZIP 
It also says.ZIP is supported in the error here, but it’s not
Is there no update on this ever since? they just removed csv/xlsx and never added support back? Especially being a M$ backed company, I’d think they’d like their format to be supported, especially because XLSX is just Zipped XML after all, and CSV is just another variation of TXT
Vector store works on chunking documents without content awareness, and then using semantic search.
Tables of CSV data, where the first operation is to chop off the header line with a key to the data, and then huge chunks of arbitrarily split lines, fed to an embeddings engine, simply will not work well for amending AI knowledge.
Blocking can easily be justified, because anyone attempting to upload CSV (or Excel, even harder to parse your merged cells and pivot table pages) would be expecting query-like retrieval, or any usefulness at all, which cannot be done.
You’d write your own function for that.
Here is a direct link to “Responses”, File Search, and the file types. There is no confusion about the purpose, and vector stores are universal to both Responses and Assistants endpoints, so any file type there that cannot be sent or that is denied by false detections is a bug, an error. https://platform.openai.com/docs/guides/tools-file-search
Point taken. I see in other threads people saying “Code Interpreter” does support both formats so if one doesn’t require searching through the data in those CSV/XLSX documents, I suppose using the other tool would make sense, but if all you’re looking to do is extract meaningful information from the structured data in those files in a similar way to what you’d do in TXT (ie: delimited or tab/space separated), JSON, XML etc. I believe having to convert CSV/XSLX to an alternative format, text or binary like PDF, in order to have the vector store index it for search seems like an unnecessary workaround.