file retrieval has been working fine for me over the last week. Just as of an hour or less, my assistant is reporting errors access my uploaded files, which I uploaded using the API. The playground seems to be OK, but I’m getting the assistant reporting a technical error back from the API. I have not changed the contents of the files for days. Is OpenAI having a partial outage? Anyone else having similar issues?
1 Like
I’m having the same issue, content retrieval against files uploaded via the API (not playground) has been broken for some time. It’s frustrating to have such a glaring issue go without acknowledgment for so long.
I’m having a similar issue, when I upload a file the output seems to suggest that only part of the file was accessible or uploaded. When I check the file size it appears to be correct. For example if I upload a C source file and analyze it, the messages coming back seem to suggest that only a fragment was uploaded or accessible. Is this due to some file size limiting I’m not aware of?
Same issue here. Just set up the whole upload file, then attach to message when creating a run…and all it does is search the entire vector_store already attached to the assistant, instead of just the file attached to the message.
The result is super expensive, and less than ideal.
It will hallucinate the contents of the file, until I mentioned the title of the document attached, and then magically it knows all about it.
From the docs it would appear that adding a message to a thread with attachments magically creates the vector store, parses the file, and then has it ready when streaming … unless maybe thats too much to ask that it be ready right away?
Same issue. And also the content policy censorship seems to have not only become worse, but irrational as well.
For writers it starts to becone usleless unless you write for very young children. The level of mentioning of sex premitted is less than a Nordic 6 year old knows. And I am not reading erotica in any way. However, writing thrillers, darkness will emerge in the text.
1 Like
Same issue. Happening both using playground and API.
Yesterday it seemed to work for a while, but now it keeps me returning that “that seems to be a persistent issue accessing the file information”. I switched from .pdf to .txt and used different files of no more than 10KB but with no success.
Anyone have solve this issue?