Hey, I’m confused on what file types are actually supported for the assistant’s file retrieval/search. This is the current official supported file list. I could not find any other list, that might be more accurate.
As you can see, it clearly says .csv and .xlsx are supported. But when I try to upload such files to a vector store, I get this:
Uploading it through the API also fails. And regarding the .csv, there’s also a sentence in the docs, saying that .csv isn’t suported, which contradicts this list.
So, does anyone know which files are actually supported? Or is there just something else I’m missing?
I have encountered the same issue where the documentation indicates support for xlsx, but when I upload through the API, it prompts that xlsx is not supported
You need to upload / attach the file in question directly to the code interpreter (with the code interpreter enabled) and not the file search/vector store. It should work then.
Based on your error message it looks like you are still trying to add the file to file search and not the code interpreter. The code interpreter has a different error message for unsupported file types (see below):
Yes, that’s confusing and should be clarified in the documentation. Currently, it states that XLSX and CSV files are supported for file searches, but the process is actually different. Although it seems logical that the code interpreter needs to be enabled to work with tabular data, this distinction is not clearly made.
I don’t exactly see how the code interpreter is relevant. My goal is to have files like CSVs, XLSXs, and such, ready for a proper file search. With code interpreter, it just writes and executes some code, that then reads through the provided files. But once you got a decent amount of files, it’s not really a proper solution, and it’s not comparable to the file-search feature.
With my original question, I was just curious if there was some information that I was missing, or anything like that. Sure, in the end, you can convert the files to another format, but if that extra step is avoidable, I’d appreciate it.
