Issue with Document Uploads to Vector Stores - "Unable to Process Document" Error

API
,
1

Hello,

We’re using the OpenAI Assistants API and uploading documents to vector stores, but we’re facing an intermittent issue. Occasionally, after uploading a document and asking for information from said document, the model responds with:

“I’m currently unable to process the document directly due to a technical issue.”

From what we can tell, the model itself seems to be functioning properly, and we suspect the issue might stem from no document chunks being pulled from the vector store into the LLM. This could potentially be due to a database timeout or a similar issue. Unfortunately, since the vector store is managed by OpenAI, it’s difficult for us to debug and confirm what’s going on under the hood.

Has anyone else encountered this problem? Or is this perhaps a known issue with vector store integration? Any insights or workarounds would be greatly appreciated.

Thanks in advance!

1 Like
2

i am seeing intermittent issues with associating a successfuly uploaded file with a vector store

  • starting at ~ Nov 7th to present
  • ~60% failure rate
  • files used are utf-8 encoded text files containing scraped content
  • the uploaded files all have a status of ‘processed’
  • sometimes, a previously failed attempt at associating/embedding the file to the vector store will work after 3 or 4 attempts - but in most instances it will fail regardless